config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file.
[official-gcc.git] / gcc / ChangeLog
blob3711cf8ed9ab204e962699943811e031d385dbb8
1 2018-12-09  John David Anglin  <danglin@gcc.gnu.org>
3         * config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file.  Define
4         d_target_objs.
5         * config/pa/pa-protos.h (pa_d_target_versions): New prototype.
6         * config/pa/pa.h (TARGET_D_CPU_VERSIONS): Define.
7         * config/pa/pa-d.c: New file.
8         * config/pa/t-pa: New file.
10 2018-12-08  Bin Cheng  <bin.cheng@linux.alibaba.com>
12         * ipa-cp.c (update_profiling_info): Call adjust_for_ipa_scaling for
13         zero profile count.
15 2018-12-08  Jakub Jelinek  <jakub@redhat.com>
17         PR fortran/88304
18         * tree-nested.c (convert_local_reference_stmt): Handle clobbers where
19         lhs is not a decl normally, don't call use_pointer_in_frame on that
20         lhs.
22 2018-12-08  Eric Botcazou  <ebotcazou@adacore.com>
24         PR rtl-optimization/88390
25         * dwarf2cfi.c (struct dw_cfi_row): Add window_save field.
26         (cfi_row_equal_p): Compare it.
27         (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter.
28         If FAKE is false, set window_save of the current row.
29         (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above
30         function with FAKE parameter set to true.
31         <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false.
32         (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary.
34 2018-12-07  Peter Bergner  <bergner@linux.ibm.com>
36         PR target/87496
37         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
38         -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
39         Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
40         * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
41         require -mlong-double-128.
43 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
45         * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add
46         sve_width entry.
47         (aarch64_parse_sve_width_string): Define.
49 2018-12-07  Jeff Law  <law@redhat.com>
51         PR middle-end/87813
52         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add
53         m_update_global_ranges member.  Add corresponding argument to ctor.
54         * gimple-ssa-evrp-analyze.c
55         (evrp_range_analyzer::evrp_range_analyzer): Add new argument and
56         initialize m_update_global_ranges.
57         (evrp_range_analyzer::set_ssa_range_info): Assert that we are
58         updating global ranges.
59         (evrp_range_analyzer::record_ranges_from_incoming_edge): Only
60         update global ranges if explicitly requested.
61         (evrp_range_analyzer::record_ranges_from_phis): Similarly.
62         (evrp_range_analyzer::record_ranges_from_stmt): Similarly.
63         * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to
64         evrp_range_analyzer ctor.
65         * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly.
66         * tree-ssa-dom.c (dom_opt_dom_walker): Similarly.
68 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
70         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum):
71         Add SVE_NOT_IMPLEMENTED value.
72         * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width
73         field.
74         * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings,
75         cortexa53_tunings, cortexa57_tunings, cortexa72_tunings,
76         cortexa73_tunings, exynosm1_tunings, thunderx_tunings,
77         thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings,
78         saphira_tunings, thunderx2t99_tunings, emag_tunings):
79         Specify sve_width.
80         (aarch64_estimated_poly_value): Define.
81         (TARGET_ESTIMATED_POLY_VALUE): Define.
83 2018-12-07  Paul A. Clarke  <pc@us.ibm.com>
85         PR target/88408
86         * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector".
88 2018-12-07  Vladimir Makarov  <vmakarov@redhat.com>
90         PR rtl-optimization/88349
91         * ira-costs.c (record_operand_costs): Check bigger reg class on
92         NO_REGS.
94 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
96         * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3):
97         Split the patterns after reload if we don't need the predicate
98         operand.
99         (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns.
101 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
103         * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code
104         iterator.
105         (sve_fp_op): Handle minus and mult.
106         * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3)
107         (*mul<mode>3): Split the patterns after reload if we don't
108         need the predicate operand.
109         (*post_ra_<sve_fp_op><mode>3): New pattern.
111 2018-12-07  Bin Cheng  <bin.cheng@linux.alibaba.com>
113         * profile-count.h (profile_count::oeprator>=): Fix typo by inverting
114         return condition when *this is precise zero.
116 2018-12-07  Jakub Jelinek  <jakub@redhat.com>
118         PR target/85593
119         * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
120         for functions with naked attribute.
122         PR c/88367
123         * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR
124         with -fno-delete-null-pointer-checks, set_nonnull only if the pointer
125         is non-NULL and offset is known to have most significant bit clear.
126         * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR
127         of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with
128         most significant bit clear.  If offset does have most significant bit
129         set and -fno-delete-null-pointer-checks, don't return true even if
130         the base pointer is non-NULL.
132 2018-12-06  Alexandre Oliva  <aoliva@redhat.com>
134         * cselib.c (cselib_record_sets): Skip strict low part sets
135         with NULL src_elt.
137 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
139         PR target/88316
140         * config/rs6000/smmintrin.h: New file.
141         * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*.
143 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
145         PR target/88316
146         * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
147         (_mm_unpacklo_pi8): Likewise.
148         (_mm_mulhi_pi16): Likewise.
149         (_mm_packs_pi16): Fix for big-endian. Use preferred API.
150         (_mm_packs_pi32): Likewise.
151         (_mm_packs_pu16): Likewise.
152         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
153         (_mm_cvtss_si64): Likewise.
154         (_mm_cvtpi32x2_ps): Likewise.
155         (_mm_shuffle_ps): Likewise.
156         (_mm_movemask_pi8): Likewise.
157         (_mm_mulhi_pu16): Likewise.
158         (_mm_sad_pu8): Likewise.
159         (_mm_sad_pu8): Likewise.
160         (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
161         (_mm_cvtpu8_ps): Likewise.
162         (_mm_movemask_ps): Better #else case for big-endian (no functional
163         change).
164         (_mm_shuffle_pi16): Likewise.
165         * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
166         Better #else case for big-endian (no functional change).
167         (_mm_movemask_epi8): Likewise.
168         (_mm_shufflehi_epi16): Likewise.
169         (_mm_shufflelo_epi16): Likewise.
170         (_mm_shuffle_epi32): Likewise.
171         (_mm_mul_epu32): Fix for big-endian.
172         (_mm_bsrli_si128): Likewise.
173         (_mm_cvtps_pd): Better #else case for big endian.
174         (_mm_mulhi_epi16): Likewise.
175         (_mm_mul_epu32): Likewise.
176         (_mm_slli_si128): Likewise.
177         (_mm_sll_epi16): Likewise.
178         (_mm_sll_epi32): Likewise.
179         (_mm_sra_epi16): Likewise.
180         (_mm_sra_epi32): Likewise.
181         (_mm_srl_epi16): Likewise.
182         (_mm_srl_epi32): Likewise.
183         (_mm_mulhi_epu16): Likewise.
184         (_mm_sad_epu8): Likewise.
185         * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
186         (_mm_sub_ps): Likewise.
187         * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
188         * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN
189         macros consistently (no functional changes).
190         (_mm_alignr_pi8): Likewise.
192 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
194         PR c++/87380
195         * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
196         default.
197         * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
199 2018-12-06  Vladimir Makarov  <vmakarov@redhat.com>
201         PR target/88282
202         * ira.c (ira_init_register_move_cost): Use info from
203         hard_regno_mode_ok instead of contains_reg_of_mode.
204         * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger
205         hard register class for some fixed hard registers.
207 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
209         * doc/extend.texi (Using Assembly Language with C): Document asm inline.
210         (Size of an asm): Fix typo.  Document asm inline.
211         * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
212         * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
213         (gimple_asm_set_volatile): Fix typo.
214         (gimple_asm_inline_p): New.
215         (gimple_asm_set_inline): New.
216         * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
217         tree to gimple.
218         * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
219         gimple_asm_inline_p flag, too.
220         * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
221         in an ASM_EXPR.
222         * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
223         a minimum size for an asm.
224         * tree.h (ASM_INLINE_P): New.
226 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
228         PR inline-asm/55681
229         * doc/extend.texi (Basic Asm): Update grammar.
230         (Extended Asm): Update grammar.
232 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
234         PR target/78444
235         * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
236         * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
237         stack alignment in non-leaf functions.
239 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
241         PR target/87598
242         * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
243         call output_operand_lossage on VOIDmode CONST_INTs.  After
244         output_operand_lossage do return false.
246 2018-12-06  Richard Biener  <rguenther@suse.de>
248         * df-problems.c (df_rd_local_compute): Use bitmap_release.
249         (df_live_free): Likewise.
250         (df_md_local_compute): Likewise.
251         (df_md_free): Release df_md_scratch bitmap.
252         * loop-invariant.c (calculate_loop_reg_pressure): Use
253         bitmap_release.
254         * sched-deps.c (true_dependency_cache, output_dependency_cache,
255         anti_dependency_cache, control_dependency_cache,
256         spec_dependency_cache): Use bitmap instead of bitmap_head *.
257         * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY
258         dont_calc_deps as bitmap allocated from obstack not GC.
259         (schedule_ebbs_finish): Use bitmap_release.
260         * sched-rgn.c (schedule_insns): Initialize non-GTY
261         not_in_df as bitmap allocated from obstack not GC.
262         Use bitmap_release.
263         * sel-sched.c (_forced_ebb_heads): Remove premature optimization.
264         (sel_region_init): Allocate forced_ebb_heads.
265         (sel_region_finish): Free forced_ebb_heads.
267 2018-12-06  Richard Biener  <rguenther@suse.de>
269         * bitmap.c (bitmap_head::crashme): Define.
270         * bitmap.h (bitmap_head): Add constexpr default constructor
271         poisoning the obstack member.
272         (bitmap_head::crashme): Declare.
273         (bitmap_release): New function clearing a bitmap and poisoning
274         the obstack member.
275         * gengtype.c (main): Make it recognize CONSTEXPR.
276         * lra-constraints.c (lra_inheritance): Use bitmap_release
277         instead of bitmap_clear.
278         * ira.c (ira): Work around class-memaccess warning.
279         * regrename.c (create_new_chain): Likewise.
281 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
283         PR tree-optimization/85726
284         * generic-match-head.c (optimize_successive_divisions_p): New function.
285         * gimple-match-head.c (optimize_successive_divisions_p): Likewise.
286         * match.pd: Don't combine successive divisions if they aren't exact
287         and optimize_successive_divisions_p is false.
289         PR fortran/88304
290         * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers
291         for non-local automatic decls.
293 2018-12-05  David Edelsohn  <dje.gcc@gmail.com>
295         * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor
296         default. 
298 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
300         * dwarf2out.c (field_byte_offset): Remove
301         #ifdef PCC_BITFIELD_TYPE_MATTERS check.
303 2018-12-05  Richard Biener  <rguenther@suse.de>
304             Martin Sebor  <msebor@redhat.com>
306         PR c/87028
307         * calls.c (get_attr_nonstring_decl): Avoid setting *REF to
308         SSA_NAME_VAR.
309         * gcc/gimple-low.c (lower_stmt): Fold builtin calls here.
310         * gimplify (maybe_fold_stmt): Avoid folding builtin calls.
312 2018-12-05  Iain Sandoe  <iain@sandoe.co.uk>
314         * configure.ac (gcc_cv_otool): Set.
315         * configure: Regenerate.
317 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
319         * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove.
320         (GROUP_BITFIELDS_BY_ALIGN): Ditto.
322 2018-12-05  Richard Sandiford  <richard.sandiford@arm.com>
324         PR tree-optimization/88064
325         * tree-vect-patterns.c (vect_convert_input): Convert the result of
326         an existing cast if it has the right width but the wrong sign.
327         Do not test the signedness of the required result when
328         considering whether to split an existing cast; instead split to
329         a type with the same signedness as the source of the cast, then
330         convert it to the opposite signedness where necessary.
331         (vect_recog_over_widening_pattern): Handle sign changes between
332         the final PLUS_EXPR and the RSHIFT_EXPR.
333         (vect_recog_average_pattern): Use an unsigned operation when
334         truncating an addition, subtraction or multiplication.  Cast the
335         result back to the "real" signedness before promoting.
337 2018-12-05  Richard Biener  <rguenther@suse.de>
339         PR tree-optimization/86637
340         * tree-vectorizer.c (pass_slp_vectorize::execute): Reset
341         vect_location at the end.
343 2018-12-05  Jakub Jelinek  <jakub@redhat.com>
345         PR tree-optimization/87360
346         * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
347         data dependencies, don't return false, just continue.  Formatting
348         fixes.
349         (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p,
350         fuse_loops): Formatting fixes.
352         PR sanitizer/88333
353         * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with
354         aligning frame offset to ASAN_RED_ZONE_SIZE bytes.
356 2018-12-05  Xianmiao Qu  <xianmiao_qu@c-sky.com>
358         * config.gcc (csky-*-linux-gnu*): Force .init_array support.
360 2018-12-05  Segher Boessenkool  <segher@kernel.crashing.org>
362         * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special
363         aggregate handling on actual AIX, not on somewhat similar systems.
364         (rs6000_arg_partial_bytes): Ditto.
366 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
368         PR rtl-optimization/88317
369         * lra-constraints.c (split_reg): Don't set up check_only_regs if
370         we are outside the inheritance pass.
372 2018-12-04  Jan Hubicka  <hubicka@ucw.cz>
374         * ipa-prop.c (jump_function_useful_p): New.
375         (ipa_read_node_info): Do not allocated useless jump functions.
377 2018-12-04  David Malcolm  <dmalcolm@redhat.com>
379         PR c/82967
380         * spellcheck.c (selftest::test_suggestions): Add another assertion.
382 2018-12-04  Martin Sebor  <msebor@redhat.com>
384         * doc/extend.texi (__builtin_has_attribute): Fix typo.
386 2018-12-04  Jakub Jelinek  <jakub@redhat.com>
388         PR target/88188
389         * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion
390         failures.
391         (print_operand): Use REG_P instead of GET_CODE == REG.
392         <case 'D'>: Also check CR_REGNO_P (REGNO (x)).
393         <case 't'>: Likewise.  Remove GET_MODE check.
394         * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't
395         return -1.
397 2018-12-04  Martin Sebor  <msebor@redhat.com>
399         * doc/extend.texi (attribute aligned): Expand.
401 2018-12-04  David Edelsohn  <dje.gcc@gmail.com>
403         PR target/61976
404         * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
405         in FPRs on AIX.
406         (rs6000_arg_partial_bytes): Same.
408 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
410         PR target/88282
411         * ira-costs.c (exec): Try bigger class to use smaller register
412         move cost.
414 2018-12-04  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
416         PR ipa/88297
417         * ipa-cp.c (create_specialized_node): Track clone counters by
418         node assembler names.
419         (ipcp_driver): Change type of clone_num_suffixes key to const char*.
421 2018-12-04  Richard Biener  <rguenther@suse.de>
423         * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations
424         for epilogue vectorization.
426 2018-12-04  Richard Biener  <rguenther@suse.de>
428         PR tree-optimization/88301
429         * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion
430         issues in last commit.
432 2018-12-04  Richard Biener  <rguenther@suse.de>
434         PR tree-optimization/88315
435         * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify
436         and fix initialization vector for SAD and DOT_PROD SLP reductions.
438 2018-12-03  Sandra Loosemore  <sandra@codesourcery.com>
440         PR c/59039
441         * doc/extend.texi (Nonlocal gotos): New section.
443 2018-12-03  Uros Bizjak  <ubizjak@gmail.com>
445         * config/i386/sse.md (vec_concatv2df): Change (v,xm,C)
446         alternative to (v,vm,c).
448 2018-12-03  Segher Boessenkool  <segher@kernel.crashing.org>
450         * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*.
451         (powerpc*-*-linux*): Don't handle it here.
453 2018-12-03  Jeff Law  <law@redhat.com>
455         * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep
456         warnings quiet.
458         * config/h8300/h8300.md (low byte masking peepholes): Only accept
459         nonimmediate_operand for the to-be-masked operand.
461 2018-12-03  Richard Biener  <rguenther@suse.de>
463         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
464         allow expected function-pointer cast re-instantiation.
466 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
468         * common/config/s390/s390-common.c (s390_option_init_struct):
469         Use section anchors by default.
470         * config/s390/s390.c (s390_check_symref_alignment): Handle
471         anchors.
472         (TARGET_MAX_ANCHOR_OFFSET): Use short displacement.
473         * output.h (assemble_align): Pass `align' as unsigned int, so
474         that the value 0x80000000, which corresponds to `aligned(1 <<
475         28)', would pass the `align > BITS_PER_UNIT' check.
476         * varasm.c (assemble_align): Likewise.
478 2018-12-03  Julian Brown  <julian@codesourcery.com>
480         * tree-pretty-print.c (dump_omp_clause): Make default case
481         gcc_unreachable.
483 2018-12-03  Richard Biener  <rguenther@suse.de>
485         PR tree-optimization/88301
486         * tree-vrp.c (register_edge_assert_for_2): Handle conversions
487         that do not change the value by registering the same assert
488         for the operand.
490 2018-12-03  Richard Biener  <rguenther@suse.de>
492         * tree-vect-loop.c (vect_transform_loop): Properly compute
493         upper bound for the epilogue when doing epilogue vectorization.
495 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
497         PR target/80080
498         * cfgcleanup.c (class pass_postreload_jump): New pass.
499         (pass_postreload_jump::execute): Likewise.
500         (make_pass_postreload_jump): Likewise.
501         * passes.def: Add pass_postreload_jump before
502         pass_postreload_cse.
503         * tree-pass.h (make_pass_postreload_jump): New pass.
505 2018-12-02  Uros Bizjak  <ubizjak@gmail.com>
507         * config/i386/i386.c (inline_memory_move_cost): Correctly
508         compute number of TFmode GPR moves.
510 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
512         PR target/88278
513         * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns.
515 2018-12-02  Jeff Law  <law@redhat.com>
517         * config/h8300/h8300.md (call, call_value): Drop mode from
518         operand containing number of bytes pushed.
519         (call_insn, call_value_insn): Use mode iterator for number
520         of bytes pushed operand.  Make pattern name vary appropriately.
522 2018-12-02  Jan Hubicka  <jh@suse.cz>
524         * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges):
525         trust precise profiles.
527 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
529         * tree-nested.c (convert_nonlocal_omp_clauses,
530         convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION,
531         OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses.
532         (convert_nonlocal_reference_stmt, convert_local_reference_stmt):
533         Convert clauses for GIMPLE_OMP_TASKGROUP.
535         * omp-low.c (check_omp_nesting_restrictions): Allow cancel or
536         cancellation point with taskgroup clause inside of taskloop.  Consider
537         a taskloop construct without nogroup clause as implicit taskgroup for
538         diagnostics if cancel/cancellation point with taskgroup clause is
539         closely nested inside of taskgroup region.
541         PR sanitizer/88291
542         * asan.c (asan_clear_shadow): Move assert that len is multiple of 4
543         to the start of the function.
544         (asan_emit_stack_protection): When emitting clearing sequence for
545         epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity,
546         add last_size_aligned which is last_size padded to multiples of
547         ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte
548         boundaries.
550 2018-12-01  Jakub Jelinek  <jakub@redhat.com>
552         PR sanitizer/88289
553         * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up
554         an off-by-one for BYTES_BIG_ENDIAN.
556         PR target/54589
557         * combine.c (find_split_point): For invalid memory address
558         nonobj + obj + const, if reg + obj + const is valid addressing
559         mode, split at nonobj.  Use if rather than else if for the
560         fallback.  Comment fixes.
562 2018-11-30  Indu Bhagat  <indu.bhagat@oracle.com>
564         * coverage.c (get_coverage_counts): Use from_function_decl for precise
565         function location.
566         * profile-count.c (profile_count::dump): Add handling for precise
567         profile quality.
568         * profile.c (compute_branch_probabilities): Rely on exec_counts instead
569         of profile_info to set profile_status of function.
570         (branch_prob): Do not set profile_status of function based on
571         profile_info. Done above based on exec_counts.
573 2018-11-30  Jakub Jelinek  <jakub@redhat.com>
575         PR tree-optimization/88274
576         * tree-ssa-reassoc.c (optimize_range_tests_xor,
577         optimize_range_tests_diff): If type has smaller precision than
578         corresponding mode or if it has non-standard min/max, compute
579         everything in a standard type for the precision.
581         PR testsuite/85368
582         * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
583         * tree-ssa-ifcombine.c (ifcombine_ifandif): If
584         --param logical-op-non-short-circuit is present, override
585         LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
586         * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
588 2018-11-30  Jeff Law  <law@redhat.com>
590         * optabs.c (expand_binop): Use "machine_mode" rather than
591         "enum machine mode" in most recent change.
593 2018-11-30  Wilco Dijkstra  <wdijkstr@arm.com>
595         PR middle-end/64242
596         * builtins.c (expand_builtin_longjmp): Use a temporary when restoring
597         the frame pointer.
598         (expand_builtin_nonlocal_goto): Likewise.
600 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
602         * diagnostic-core.h (emit_diagnostic): New decl.
603         * diagnostic.c (emit_diagnostic): New overload, taking a
604         rich_location *.
606 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
608         * pretty-print.c (class selftest::test_pretty_printer): New
609         subclass of pretty_printer.
610         (selftest::test_prefixes_and_wrapping): New test.
611         (selftest::pretty_print_c_tests): Call it.
613 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
615         Minimize clone counter memory usage in create_virtual_clone.
616         * cgraph.h (clone_function_name): Add a variant that takes a
617         tree decl.
618         * cgraph.h (cgraph_node::create_virtual_clone): Add a new
619         argument: num_suffix.
620         * cgraphclones.c (cgraph_node::create_virtual_clone): Pass
621         num_suffix to clone_function_name.
622         (clone_function_name): Add a variant that takes a tree decl.
623         * ipa-cp.c (create_specialized_node): Keep track of clone
624         counters in clone_num_suffixes hash map.
625         (ipcp_driver): Free the counter hash map.
626         * ipa-hsa.c (process_hsa_functions): Creates at most one hsa
627         clone per function.
629 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
630         Make function assembly more independent.
632         This is achieved by having clone_function_name assign unique clone
633         numbers for each function independently.
635         * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids;
636         hash map.
637         (clone_function_name_numbered): Use clone_fn_ids.
639 2018-11-30  Vladimir Makarov  <vmakarov@redhat.com>
641         PR rtl-optimization/88179
642         * lra-constraints.c (address_eliminator::address_eliminator):
643         Don't eleminate regs in illegitimate address.
645 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
647         PR preprocessor/88257
648         * input.c (get_substring_ranges_for_loc): Fix indentation.
649         Bulletproof against getting a different files back from
650         linemap_resolve_location and expand_location_to_spelling_point.
652 2018-11-30  Alexander Monakov  <amonakov@ispras.ru>
654         PR gcov-profile/88279
655         * tree-profile.c (init_ic_make_global_vars): Call
656         decl_default_tls_model for variable and not it's type.
658 2018-11-30  Sam Tebbs  <sam.tebbs@arm.com>
660         * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint.
662 2018-11-30  Martin Liska  <mliska@suse.cz>
664         PR sanitizer/81715
665         * asan.c (asan_shadow_cst): Remove, partially transform
666         into flush_redzone_payload.
667         (RZ_BUFFER_SIZE): New.
668         (struct asan_redzone_buffer): New.
669         (asan_redzone_buffer::emit_redzone_byte): Likewise.
670         (asan_redzone_buffer::flush_redzone_payload): Likewise.
671         (asan_redzone_buffer::flush_if_full): Likewise.
672         (asan_emit_stack_protection): Use asan_redzone_buffer class
673         that is responsible for proper aligned stores and flushing
674         of shadow memory payload.
675         * asan.h (ASAN_MIN_RED_ZONE_SIZE): New.
676         (asan_var_and_redzone_size): Likewise.
677         * cfgexpand.c (expand_stack_vars): Use smaller alignment
678         (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory
679         for automatic variables more compact.
681 2018-11-30  Alan Modra  <amodra@gmail.com>
683         * config/rs6000/predicates.md (easy_fp_constant): Avoid long
684         dependent insn sequences.
685         * config/rs6000/rs6000.c (num_insns_constant): Support long
686         double constants.
687         * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust
688         length attribute.
690 2018-11-30  Alan Modra  <amodra@gmail.com>
692         * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from
693         num_insns_constant_wide.  Make static.  Revise comment.
694         (num_insns_constant_multi): New function.
695         (num_insns_constant): Formatting.  Correct CONST_WIDE_INT
696         calculation.  Simplify and extract code common to both
697         CONST_INT and CONST_DOUBLE.  Add gcc_unreachable for unhandled
698         const_double modes.
699         * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete.
701 2018-11-30  Richard Biener  <rguenther@suse.de>
703         PR tree-optimization/88274
704         * tree-vrp.c (ranges_from_anti_range): Fix handling of
705         TYPE_MIN/MAX_VALUE.
707 2018-11-29  Andi Kleen  <ak@linux.intel.com>
709         * config/i386/i386.c (current_fentry_section): Add.
710         (x86_function_profiler): Handle fentry section.
711         (ix86_attribute_table): Add fentry section.
712         * config/i386/i386.opt: Add -mfentry-section.
713         * doc/extend.texi: Document fentry_section attribute.
714         * doc/invoke.texi: Document -mfentry-section.
716 2018-11-29  Andi Kleen  <ak@linux.intel.com>
718         * config/i386/i386.c (x86_print_call_or_nop): Handle nop name.
719         (current_fentry_name): Add.
720         (x86_function_profiler): Handle fentry_name attribute.
721         (ix86_handle_fentry_name): Add.
722         (ix86_attribute_table): Add fentry_name.
723         * config/i386/i386.opt: Add -mfentry-name
724         * doc/extend.texi: Document fentry_name.
725         * doc/invoke.texi: Document minstrument-return.
727 2018-11-29  Andi Kleen  <ak@linux.intel.com>
729         * config/i386/i386-opts.h (enum instrument_return): Add.
730         * config/i386/i386.c (output_return_instrumentation): Add.
731         (ix86_output_function_return): Call output_return_instrumentation.
732         (ix86_output_call_insn): Call output_return_instrumentation.
733         * config/i386/i386.opt: Add -minstrument-return=.
734         * doc/invoke.texi (-minstrument-return): Document.
736 2018-11-29  Eric Botcazou  <ebotcazou@adacore.com>
738         PR target/87807
739         * config/sparc/sparc-modes.def: Minor tweak.
740         * config/sparc/sparc.c: Minor reordering.
741         (sparc_pass_by_reference): Move around.
742         (traverse_record_type): Change offset from HOST_WIDE_INT to int.
743         (classify_registers): Likewise for bitpos.
744         (function_arg_slotno): Remove dead test and tweak comments.
745         <MODE_RANDOM>: Remove useless assertion and test whether the
746         parameter is named in order to pass it in FP registers.  Return
747         the regno for floating-point vector types.
748         (compute_int_layout): Change bitpos from HOST_WIDE_INT to int.
749         (compute_fp_layout): Likewise.
750         (count_registers): Likewise.
751         (assign_int_registers): Likewise.
752         (assign_fp_registers): Likewise.
753         (assign_registers): Likewise.
754         (function_arg_record_value): Change size from HOST_WIDE_INT to int
755         and use CEIL_NWORDS to compute the number of registers.
756         (function_arg_union_value): Minor tweaks.
757         (function_arg_vector_value): Add slotno and named parameters, use
758         CEIL_NWORDS to compute the number of registers.
759         (sparc_function_arg_1): Rework handling of vector types.  Change
760         size from HOST_WIDE_INT to int.
761         (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal
762         with vector types for the 64-bt ABI.
763         (sparc_function_arg_advance): Likewise.
764         (sparc_return_in_memory): Add reference to -fpcc-struct-return.
765         (sparc_struct_value_rtx): Return NULL_RTX instead of 0.
766         (sparc_function_value_1): Rework handling of vector types.  Change
767         size from HOST_WIDE_INT to int.
769 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
771         PR target/88152
772         * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt,
773         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt,
774         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift,
775         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift,
776         *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New
777         define_insn_and_split patterns.
779         PR target/54700
780         * config/i386/sse.md
781         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
782         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
783         *<sse4_1_avx2>_pblendvb_lt): Change define_insn into
784         define_insn_and_split.
786 2018-11-29  Peter Bergner  <bergner@linux.ibm.com>
788         PR target/87496
789         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
790         -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
792 2018-11-29  Uros Bizjak  <ubizjak@gmail.com>
794         * config/i386/i386.c (inline_memory_move_cost):
795         Check "in" for 2 in MMX_CLASS_P case.
796         * config/i386/mmx.md (*mov<mode>_internal): Correct
797         TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC
798         alternatives in preferred_for_speed attribute calculation.
800 2018-11-29  Martin Sebor  <msebor@redhat.com>
802         PR c/88172
803         PR testsuite/88208
804         * doc/extend.texi (attribute constructor): Clarify.
806 2018-11-29  Martin Liska  <mliska@suse.cz>
808         PR middle-end/88246
809         * builtins.c (expand_movstr): Fix thinko introduced
810         when switching to the new enum.
812 2018-11-29  qing zhao  <qing.zhao@oracle.com>
814         * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
815         * common.opt: Add -flive-patching flag.
816         * doc/invoke.texi: Document -flive-patching.
817         * flag-types.h (enum live_patching_level): New enum.
818         * ipa-inline.c (can_inline_edge_p): Disable external functions from
819         inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
820         * opts.c (control_options_for_live_patching): New function.
821         (finish_options): Make flag_live_patching incompatible with flag_lto.
822         Control IPA optimizations based on different levels of
823         flag_live_patching.
825 2018-11-29  Giuliano Belinassi  <giuliano.belinassi@usp.br>
827         * match.pd (sinh (atanh (x))): New simplification rules.
828         (cosh (atanh (x))): Likewise.
830 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
832         PR target/54700
833         * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
834         (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
835         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
836         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
837         *<sse4_1_avx2>_pblendvb_lt): New define_insns.
839         PR target/88152
840         * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
841         * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
842         * match.pd (define_predicates): Add uniform_integer_cst_p.
843         (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
844         so that it works also for vector comparisons with uniform constants
845         with INTEGER_CST element.
847         PR target/88234
848         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
849         vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
850         in unsigned_type_for instead of vector integral type where overflow
851         doesn't wrap.
853 2018-11-29  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
855         There can be at most one .resolver clone per function
856         * config/rs6000/rs6000.c (make_resolver_func): Generate
857         resolver symbol with clone_function_name instead of
858         clone_function_name_numbered.
860 2018-11-29  Richard Biener  <rguenther@suse.de>
862         PR tree-optimization/88243
863         * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
864         type of all pattern-sequence stmts to vect_internal_def.
866 2018-11-29  Andre Vieira  <andre.simoesdiasvieira@arm.com>
868         PR target/88224
869         * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
870         (cortex-r7, cortex-r8): Update default and add new configuration.
871         * doc/invoke.texi (armv7-r): Add two new vfp options.
872         (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
873         support this option.
875 2018-11-29  Alan Modra  <amodra@gmail.com>
877         * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
878         double split for targets other than Darwin.
880 2018-11-29  Alan Modra  <amodra@gmail.com>
882         * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
884 2018-11-29  Alan Modra  <amodra@gmail.com>
886         * config/rs6000/predicates.md (easy_fp_constant): Remove code
887         dealing with integers in const_double.  Assert on unexpected
888         modes.  Delete superfluous ABI_V4 test.
889         * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
890         Don't call easy_fp_constant for Pmode.
891         (rs6000_hash_constant): Remove code dealing with integers in
892         const_double.
893         (rs6000_legitimate_constant_p): Likewise.
894         (output_toc): Formatting, use CONST_DOUBLE_P.
896 2018-11-29  Alan Modra  <amodra@gmail.com>
898         * config/rs6000/constraints.md (G, H): Comment on purpose of
899         constraint.  Correct mode comments and "G" mode test.
900         * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
901         alternative handling "F".  Add length attr.  Formatting.
902         (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
903         alternative, and move "F" to separate alternative.  Correct
904         insn lengths.
906 2018-11-29  Alan Modra  <amodra@gmail.com>
908         * config/rs6000/predicates.md (easy_vector_constant): Don't call
909         easy_fp_constant.
911 2018-11-29  Alan Modra  <amodra@gmail.com>
913         * config.in (HAVE_AS_PLTSEQ): Add.
914         * config/rs6000/predicates.md (indirect_call_operand): New.
915         * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
916         (rs6000_sibcall_sysv): Declare.
917         * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
918         CALL_LONG for -fno-plt.
919         (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
920         (rs6000_indirect_call_template_1): Emit .reloc directives for
921         UNSPEC_PLTSEQ calls.
922         (rs6000_pltseq_template): New function.
923         (rs6000_longcall_ref): Add arg parameter.  Use PLT16 insns if
924         relocs supported by assembler.  Move SYMBOL_REF test to callers.
925         (rs6000_call_aix): Adjust rs6000_longcall_ref call.  Package
926         insns in UNSPEC_PLTSEQ, preserving original func_desc.
927         (rs6000_call_sysv): Likewise.
928         (rs6000_sibcall_sysv): New function.
929         * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
930         * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
931         UNSPEC_PLT16_LO): New.
932         (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
933         (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
934         cookie in constraints.  Test explicitly for flags in length attr.
935         Handle unspec operand 1.
936         (call_value_indirect_nonlocal_sysv): Likewise.
937         (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
938         (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
939         (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
940         (sibcall_indirect_nonlocal_sysv): New pattern.
941         (sibcall_value_indirect_nonlocal_sysv): Likewise.
942         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
943         call alternatives.
944         * configure.ac: Check for gas plt sequence marker support.
945         * configure: Regenerate.
947 2018-11-29  Alan Modra  <amodra@gmail.com>
949         * config/rs6000/predicates.md (unspec_tls): New.
950         * config/rs6000/rs6000-protos.h (rs6000_call_template),
951         (rs6000_sibcall_template): Update prototype.
952         (rs6000_longcall_ref): Delete.
953         (rs6000_call_sysv): Declare.
954         * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
955         (global_tlsarg): New variable.
956         (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
957         handling.
958         (print_operand): Extract UNSPEC_TLSGD address operand.
959         (rs6000_call_template, rs6000_sibcall_template): Remove arg
960         parameter, extract from second call operand instead.
961         (rs6000_longcall_ref): Make static, localize vars.
962         (rs6000_call_aix): Rename parameter to reflect new usage.  Take
963         tlsarg from global_tlsarg.  Don't create unused rtl or nop insns.
964         (rs6000_sibcall_aix): Rename parameter to reflect new usage.  Take
965         tlsarg from global_tlsarg.
966         (rs6000_call_sysv): New function.
967         * config/rs6000/rs6000.md: Adjust rs6000_call_template and
968         rs6000_sibcall_template throughout.
969         (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
970         (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
971         (tls_gdld_nomark): New insn.
972         (tls_gd): Swap operand order.  Simplify mode selection.
973         (tls_gd_high, tls_gd_low): Swap operand order.
974         (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
975         Simplify mode selection.
976         (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
977         (call, call_value): Don't assert for second call operand.
978         Use rs6000_call_sysv.
980 2018-11-29  Alan Modra  <amodra@gmail.com>
982         * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
983         (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
984         (call_value_nonlocal_darwin64): Remove constraints from second call
985         arg, the rounded_stack_size_rtx arg.
986         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
987         (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
988         (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
989         (call_value_local64, call_indirect_nonlocal_sysv),
990         (call_nonlocal_sysv, call_nonlocal_sysv_secure),
991         (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
992         (call_value_nonlocal_sysv_secure, call_local_aix),
993         (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
994         (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
995         (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
996         (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
997         (sibcall_value_aix): Likewise.
999 2018-11-29  Alan Modra  <amodra@gmail.com>
1001         * config/rs6000/rs6000.md (TLSmode): Delete mode iterator.  Replace
1002         with P throughout except for call mems which should use SI.
1003         (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
1004         attributes.  Replace with bits, mode and ptrload respectively.
1006 2018-11-29  Alan Modra  <amodra@gmail.com>
1008         * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
1009         (rs6000_indirect_sibcall_template): Declare.
1010         * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
1011         (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
1012         New functions.
1013         * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
1014         (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
1015         (call_indirect_aix, call_value_indirect_aix): Use
1016         rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
1017         call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
1018         handle both speculation and non-speculation cases.
1019         (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
1020         (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
1022 2018-11-29  Alan Modra  <amodra@gmail.com>
1024         * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
1025         (rs6000_sibcall_template): Declare.
1026         (macho_call_template): Rename from output_call.
1027         * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
1028         (rs6000_call_template, rs6000_sibcall_template): Likewise.
1029         (macho_call_template): Rename from output_call.
1030         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
1031         (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
1032         (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
1033         (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
1034         (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
1035         (call_value_nonlocal_aix): Use rs6000_call_template and update
1036         occurrences of output_call to macho_call_template.
1037         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
1038         (sibcall_value_aix): Use rs6000_sibcall_template.
1040 2018-11-28  Aaron Sawdey  <acsawdey@linux.ibm.com>
1042         * config/rs6000/rs6000-string.c (expand_block_clear): Change how
1043         we determine if unaligned vsx is ok.
1045 2018-11-28  Jan Hubicka  <jh@suse.cz>
1047         * predict.c (determine_unlikely_bbs): Forward declare; also determine
1048         edges that are always known to be taken because it is only likely
1049         edge out of given BB.
1050         (tree_estimate_probability): Call before profile guessing to get bit
1051         of extra precision.
1053 2018-11-28  Jan Hubicka  <jh@suse.cz>
1055         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
1056         profile_probability::always better.
1058 2018-11-28  Jan Hubicka  <jh@suse.cz>
1060         * profile-count.h (profile_count::split): Give better result when
1061         splitting profile_probability::always.
1063 2018-11-28  Vladimir Makarov  <vmakarov@redhat.com>
1065         PR target/88207
1066         * ira-costs.c (scan_one_insn): Process subregs when updating costs
1067         for pseudos and allocnos from insn.
1069 2018-11-28  David Edelsohn  <dje.gcc@gmail.com>
1071         * config/rs6000/aix72.h: Update to match aix71.h changes.
1073 2018-11-28  Jeff Law  <law@redhat.com>
1075         * config/h8300/constraints.md: Add "C" constraint for call insns.
1076         * config/h8300/h8300.md (call, call_value): Turn into a define_expand
1077         and define_insn pair.  Move invalid call targets into a register in
1078         the expander and fix constraints in the matching pattern.
1079         * config/h8300/predicates.md (call_expander_operand): Renamed from
1080         call_insn_operand.  Reject things we shouldn't be trying to handle.
1081         (call_insn_operand): New predicate for use by the call/call_value
1082         insns.
1083         (small_call_insn_operand): Update appropriately.
1085 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
1087         * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
1088         calls to strtok with strtok_r.
1090 2018-11-28  Richard Biener  <rguenther@suse.de>
1092         PR tree-optimization/88223
1093         * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
1094         over a stored-same value may-alias store make sure to consider
1095         partial overlaps which are valid when TBAA reasonings do not
1096         apply and byte-granular overlaps are possible at all.
1098 2018-11-28  Richard Biener  <rguenther@suse.de>
1100         PR tree-optimization/88217
1101         * vr-values.c (vr_values::extract_range_from_phi_node): Make
1102         sure to handle results > +INF and < -INF correctly when
1103         trying to drop down to +INF - 1 or -INF + 1.
1105 2018-11-28  Alan Modra  <amodra@gmail.com>
1107         * xcoffout.c (do_block): Signed/unsigned warning fix.
1109 2018-11-28  Richard Biener  <rguenther@suse.de>
1111         PR tree-optimization/79351
1112         * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
1113         empty CONSTRUCTORs ensure the store is at a constant position.
1115 2018-11-28  Richard Biener  <rguenther@suse.de>
1117         PR tree-optimization/88229
1118         * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
1119         is an INTEGER_CST before accessing it so.
1121 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
1123         * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
1125 2018-11-28  Jakub Jelinek  <jakub@redhat.com>
1127         PR target/88189
1128         * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
1129         SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1.  Formatting fixes.
1130         * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
1132 2018-11-27  Martin Liska  <mliska@suse.cz>
1134         * gcov.c (generate_results): Append current_working_directory
1135         only when exists.
1137 2018-11-27  Mike Gulick  <mgulick@mathworks.com>
1139         PR preprocessor/83173
1140         * input.c (dump_location_info): Dump reason and included_from
1141         fields from line_map_ordinary struct.  Fix indentation when
1142         location > 5 digits.
1143         * diagnostic-show-locus.c (num_digits, num_digits): Move to
1144         diagnostic.c to allow it to be utilized by input.c.
1145         * diagnostic.c (num_digits, selftest::test_num_digits): Moved
1146         here.
1147         (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
1148         * diagnostic.h (num_digits): Add extern definition.
1150 2018-11-27  Fredrik Noring  <noring@nocrew.org>
1152         * config/mips/mips.c (mips_reorg_process_insns)
1153         (mips_option_override): Handle `-mfix-r5900'.
1154         * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
1155         `mno-fix-r5900'.
1156         * config/mips/mips.opt (mfix-r5900): New option.
1157         * doc/invoke.texi: Document the `r5900' processor name, and
1158         `-mfix-r5900' and `-mno-fix-r5900' options.
1160 2018-11-27  Jakub Jelinek  <jakub@redhat.com>
1162         PR target/88188
1163         * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
1164         output_operand_lossage instead of gcc_assert.
1165         <case 't'>: Likewise.
1166         <case 'z'>: Likewise.
1167         <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
1169 2018-11-27  Jeff Law  <law@redhat.com>
1171         * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
1172         in call to move_by_pieces.
1174         * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
1175         to move_by_pieces.
1177         * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
1178         call to move_by_pieces.
1180         * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
1181         in call to move_by_pieces.
1183         * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
1184         RETURN_BEGIN in call to move_by_pieces.
1185         (microblaze_expand_block_move): Likewise.
1187 2018-11-27  Tamar Christina  <tamar.christina@arm.com>
1189         * doc/md.texi (xorsign): Document it.
1191 2018-11-27  Martin Liska  <mliska@suse.cz>
1193         * config/gnu-user.h (TARGET_F951_OPTIONS): New.
1194         * gcc.c (find_fortran_preinclude_file): New function
1195         to handle Fortran pre-include.
1197 2018-11-27  Martin Liska  <mliska@suse.cz>
1199         * asan.c (asan_emit_stack_protection): Use new enum values
1200         instead of int constants.
1201         * builtins.c (expand_builtin_memory_copy_args): Replace int
1202         type with memop_ret enum type.
1203         (expand_builtin_mempcpy_args): Likewise.
1204         (expand_builtin_memcpy): Use new enum values
1205         instead of int constants. Likewise.
1206         (expand_builtin_mempcpy): Likewise.
1207         (expand_movstr): Likewise.
1208         (expand_builtin_strcpy_args): Likewise.
1209         (expand_builtin_stpcpy_1): Likewise.
1210         (expand_builtin_strncpy): Likewise.
1211         (expand_builtin_memset_args): Likewise.
1212         * expr.c (move_by_pieces_d::finish_endp): Rename to ...
1213         (move_by_pieces_d::finish_retmode): ... this.
1214         (move_by_pieces): Change last argument type to memop_ret.
1215         (store_by_pieces): Use new enum values
1216         instead of int constants.
1217         (emit_block_move_hints): Likewise.
1218         (emit_push_insn): Likewise.
1219         (store_expr): Likewise.
1220         * expr.h (store_by_pieces): Change int to newly added enum
1221         type.
1222         * rtl.h (enum memop_ret): Define.
1223         (move_by_pieces): Use the enum type.
1225 2018-11-27  Alan Modra  <amodra@gmail.com>
1227         * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
1228         cpu here.
1229         (ASM_CPU_SPEC): Do so here.  Rewrite using if .. else if .. specs
1230         form.  Error on missing -mcpu case.
1231         * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
1232         (asm_names <!_AIX>): Add missing cpus.  Update NULL case.  Apply
1233         PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
1234         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
1235         else if .. specs form.  Error on missing -mcpu case.  Don't output
1236         duplicate -maltivec.  Apply PR63177 fix for -mcpu=powerpc64le.
1238 2018-11-26  David Malcolm  <dmalcolm@redhat.com>
1240         * dump-context.h (dump_context::dump_loc): Convert 1st param from
1241         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
1242         const dump_location_t & to const dump_user_location_t &.
1243         (dump_context::dump_loc_immediate): Convert 2nd param from
1244         const dump_location_t & to const dump_user_location_t &.
1245         (dump_context::dump_gimple_stmt): Convert 1st param from
1246         dump_flags_t to const dump_metadata_t &.
1247         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
1248         2nd param from const dump_location_t & to
1249         const dump_user_location_t &.
1250         (dump_context::dump_gimple_expr): Convert 1st param from
1251         dump_flags_t to const dump_metadata_t &.
1252         (dump_context::dump_gimple_expr_loc): Likewise; convert
1253         2nd param from const dump_location_t & to
1254         const dump_user_location_t &.
1255         (dump_context::dump_generic_expr): Convert 1st param from
1256         dump_flags_t to const dump_metadata_t &.
1257         (dump_context::dump_generic_expr_loc): Likewise; convert
1258         2nd param from const dump_location_t & to
1259         const dump_user_location_t &.
1260         (dump_context::dump_printf_va): Convert 1st param from
1261         dump_flags_t to const dump_metadata_t &.
1262         (dump_context::dump_printf_loc_va): Likewise; convert
1263         2nd param from const dump_location_t & to
1264         const dump_user_location_t &.
1265         (dump_context::dump_dec): Convert 1st param from
1266         dump_flags_t to const dump_metadata_t &.
1267         (dump_context::dump_symtab_node): Likewise.
1268         (dump_context::begin_scope): Split out 2nd param into
1269         user and impl locations.
1270         (dump_context::ensure_pending_optinfo): Add metadata param.
1271         (dump_context::begin_next_optinfo): Replace dump_location_t param
1272         with metadata and user location.
1273         * dumpfile.c (dump_context::dump_loc): Convert 1st param from
1274         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
1275         const dump_location_t & to const dump_user_location_t &.
1276         (dump_context::dump_loc_immediate): Convert 2nd param from
1277         const dump_location_t & to const dump_user_location_t &.
1278         (dump_context::dump_gimple_stmt): Convert 1st param from
1279         dump_flags_t to const dump_metadata_t &.
1280         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
1281         2nd param from const dump_location_t & to
1282         const dump_user_location_t &.
1283         (dump_context::dump_gimple_expr): Convert 1st param from
1284         dump_flags_t to const dump_metadata_t &.
1285         (dump_context::dump_gimple_expr_loc): Likewise; convert
1286         2nd param from const dump_location_t & to
1287         const dump_user_location_t &.
1288         (dump_context::dump_generic_expr): Convert 1st param from
1289         dump_flags_t to const dump_metadata_t &.
1290         (dump_context::dump_generic_expr_loc): Likewise; convert
1291         2nd param from const dump_location_t & to
1292         const dump_user_location_t &.
1293         (dump_context::dump_printf_va): Convert 1st param from
1294         dump_flags_t to const dump_metadata_t &.
1295         (dump_context::dump_printf_loc_va): Likewise; convert
1296         2nd param from const dump_location_t & to
1297         const dump_user_location_t &.
1298         (dump_context::dump_dec): Convert 1st param from
1299         dump_flags_t to const dump_metadata_t &.
1300         (dump_context::dump_symtab_node): Likewise.
1301         (dump_context::begin_scope): Split out 2nd param into
1302         user and impl locations.
1303         (dump_context::ensure_pending_optinfo): Add metadata param.
1304         (dump_context::begin_next_optinfo): Replace dump_location_t param
1305         with metadata and user location.
1306         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
1307         const dump_metadata_t &.
1308         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
1309         const dump_location_t & to const dump_user_location_t &.
1310         (dump_gimple_expr): Convert 1st param from dump_flags_t to
1311         const dump_metadata_t &.
1312         (dump_gimple_expr_loc): Likewise; convert 2nd param from
1313         const dump_location_t & to const dump_user_location_t &.
1314         (dump_generic_expr): Convert 1st param from dump_flags_t to
1315         const dump_metadata_t &.
1316         (dump_generic_expr_loc): Likewise; convert 2nd param from
1317         const dump_location_t & to const dump_user_location_t &.
1318         (dump_printf): Convert 1st param from dump_flags_t to
1319         const dump_metadata_t &.
1320         (dump_printf_loc): Likewise; convert 2nd param from
1321         const dump_location_t & to const dump_user_location_t &.
1322         (dump_dec): Convert 1st param from dump_flags_t to
1323         const dump_metadata_t &.
1324         (dump_symtab_node): Likewise.
1325         (dump_begin_scope): Split out 2nd param into user and impl
1326         locations.
1327         (selftest::assert_impl_location_eq): New function.
1328         (ASSERT_IMPL_LOCATION_EQ): New macro.
1329         (selftest::test_impl_location): Update to use
1330         ASSERT_IMPL_LOCATION_EQ.
1331         (selftest::test_capture_of_dump_calls): Convert "loc" to
1332         dump_user_location_t.  Add ASSERT_IMPL_LOCATION_EQ throughout,
1333         verifying line numbers of dump emissions.
1334         * dumpfile.h (class dump_metadata_t): New class.
1335         (dump_printf): Convert 1st param from dump_flags_t to
1336         const dump_metadata_t &.
1337         (dump_printf_loc): Likewise; convert 2nd param from
1338         const dump_location_t & to const dump_user_location_t &.
1339         (dump_generic_expr_loc): Likewise.
1340         (dump_generic_expr): Convert 1st param from dump_flags_t to
1341         const dump_metadata_t &.
1342         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
1343         const dump_location_t & to const dump_user_location_t &.
1344         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
1345         const dump_metadata_t &.
1346         (dump_gimple_expr_loc): Likewise; convert 2nd param from
1347         const dump_location_t & to const dump_user_location_t &.
1348         (dump_gimple_expr): Convert 1st param from dump_flags_t to
1349         const dump_metadata_t &.
1350         (dump_symtab_node): Likewise.
1351         (dump_dec): Likewise.
1352         (dump_begin_scope): Split out 2nd param into user and impl
1353         locations.
1354         (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
1355         location and impl_location, and capture the impl_location.
1356         (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
1357         * loop-unroll.c (report_unroll): Update for changes to
1358         dump_printf_loc and dump_printf.
1359         * opt-problem.cc (opt_problem::opt_problem): Update for change to
1360         dump_loc.
1361         * optinfo-emit-json.cc
1362         (selftest::test_building_json_from_dump_calls): Convert "loc" from
1363         dump_location_t to dump_user_location_t.
1364         * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
1365         dump_loc_immediate.
1366         * profile.c (compute_branch_probabilities): Update for change to
1367         dump_printf_loc.
1368         * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
1369         * tree-vect-slp.c (vect_print_slp_tree): Update for change to
1370         dump_printf_loc.
1372 2018-11-27  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1374         * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
1375         and check_effective_target_ptr_eq_long.
1377 2018-11-27  Alan Modra  <amodra@gmail.com>
1379         * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
1380         (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
1381         (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
1382         (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
1383         (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
1384         * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
1385         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
1386         (ASM_APP_ON, ASM_APP_OFF): Don't define.
1387         * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
1388         * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
1389         * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1390         * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
1391         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
1392         * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1393         * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
1394         (GNU_USER_TARGET_CRTN): Redefine.
1395         (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
1396         (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
1397         (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
1398         (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
1399         (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
1400         (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
1401         (LINK_EH_SPEC): Don't define.
1403 2018-11-26  Martin Sebor  <msebor@redhat.com>
1405         PR 87756
1406         * expr.c (string_constant): Handle top-level decls of all character
1407         types and subobjects of narrow character type.
1409 2018-11-27  Alan Modra  <amodra@gmail.com>
1411         * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
1412         * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
1413         GNU_USER_TARGET_OS_CPP_BUILTINS.
1414         (RS6000_ABI_NAME): Define.
1415         * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
1416         GNU_USER_TARGET_OS_CPP_BUILTINS.
1417         (MUSL_DYNAMIC_LINKER32): Undef before defining.
1418         (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
1419         (CHOOSE_DYNAMIC_LINKER): Don't define.
1420         (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
1421         * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
1422         (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
1423         not already defined.
1424         (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
1425         TARGET_OS_CPP_BUILTINS.
1427 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1429         * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
1431 2018-11-26  Jakub Jelinek  <jakub@redhat.com>
1433         PR target/88195
1434         * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
1435         and !TARGET_64BIT, return NULL_TREE.
1437         PR c++/86900
1438         * dwarf2out.c (secname_for_decl): For functions with
1439         DECL_SECTION_NAME if in_cold_section_p, try to return
1440         current_function_section's name if it is a named section.
1442 2018-11-26  Maya Rashish  <coypu@sdf.org>
1444         PR target/58397
1445         * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
1446         * config/x-netbsd: New file.
1447         * config/host-netbsd: New file.
1449 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1451         * doc/sourcebuild.texi: Document check_effective_target_size20plus.
1452         Clarify documentation for check_effective_target_size32plus.
1454 2018-11-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
1456         * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
1457         and modify expand_strncmp_vec_sequence.
1458         (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
1459         (generate_6432_conversion): New function.
1460         (expand_block_compare): Add support for vsx.
1461         (expand_block_compare_gpr): New function.
1462         * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
1463         default limit to 63 because of more compact vsx code.
1465 2018-11-26  Uros Bizjak  <ubizjak@gmail.com>
1467         PR target/88178
1468         * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
1469         for registers for which no debug information can be generated.
1470         Use INVALID_REGNUM for non-existent registers.
1471         (dbx64_register_map): Ditto.
1472         (svr4_dbx_register_map): Ditto.
1474 2018-11-26  Sandra Loosemore  <sandra@codesourcery.com>
1476         PR c/57182
1477         * doc/invoke.texi (Optimize Options): Document that
1478         -freorder-functions also uses "hot" and "cold" attributes.
1480 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1482         * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
1484 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1486         * config/arm/types.md (mul64): Rename to...
1487         (widen_mul64): ... This.
1488         * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
1489         * config/arm/cortex-a15.md: Likewise.
1490         * config/arm/cortex-a5.md: Likewise.
1491         * config/arm/cortex-a53.md: Likewise.
1492         * config/arm/cortex-a57.md: Likewise.
1493         * config/arm/cortex-a7.md: Likewise.
1494         * config/arm/cortex-m4.md: Likewise.
1495         * config/arm/exynos-m1.md: Likewise.
1496         * config/arm/marvell-pj4.md: Likewise.
1497         * config/arm/xgene1.md: Likewise.
1499 2018-11-26  Richard Biener  <rguenther@suse.de>
1501         PR tree-optimization/88182
1502         * tree-vect-loop.c (vectorizable_reduction): Pick up single
1503         correct reduc_def_info.
1504         * tree-vect-slp.c (vect_analyze_slp_instance): Set
1505         STMT_VINFO_REDUC_DEF of the first stmt.
1507 2018-11-26  Andreas Krebbel  <krebbel@linux.ibm.com>
1509         * doc/invoke.texi: Document z14/arch12 -march option.
1511 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
1513         PR c/57166
1514         * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
1515         (Warning Options): Likewise.
1517 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
1519         PR web/79738
1520         * doc/extend.texi (Common Function Attributes): Clarify that
1521         functions with "const" attribute can read const global variables.
1523 2018-11-26  Alan Modra  <amodra@gmail.com>
1525         * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
1527 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
1529         PR other/54265
1530         * doc/extend.texi (Common Variable Attributes): Use preferred
1531         placement of type attributes in examples, plus whitespace fixes.
1532         (Type Attributes): Clarify why placement of attributes
1533         immediately after struct/union/enum keyword is preferred.
1534         (Common Type Attributes): Use preferred placement of type
1535         attributes in examples, plus more whitespace fixes.
1537 2018-11-25  Paul Koning  <ni1d@arrl.net>
1539         * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
1541 2018-11-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1543         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
1544         attributes): Document d_runtime.
1546 2018-11-25  Vladimir Makarov  <vmakarov@redhat.com>
1548         PR bootstrap/88157
1549         * ira-costs.c (record_operand_costs): Initiate register move cost
1550         for mode.
1552 2018-11-23  Jeff Law  <law@redhat.com>
1554         PR rtl-optimization/87468
1555         * tree-ssa-threadupdate.c (create_block_for_threading): Clear
1556         EDGE_IGNORE on all outgoing edges of the duplicate block.
1558 2018-11-23  Vladimir Makarov  <vmakarov@redhat.com>
1560         PR bootstrap/88157
1561         * ira-costs.c (record_operand_costs): Use bigger hard reg class if
1562         its mode does not fit to the original class.
1564 2018-11-23  Martin Sebor  <msebor@redhat.com>
1566         PR tree-optimization/87756
1567         * expr.c (string_constant): Synthesize a string literal from
1568         the address of a constant character.
1569         * tree.c (build_string_literal): Add an argument.
1570         * tree.h (build_string_literal): Same.
1572 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
1574         * config/aarch64/aarch64-cores.def: Define emag.
1575         * config/aarch64/aarch64-tune.md: Regenerated with emag.
1576         * config/aarch64/aarch64.c (emag_tunings): New struct.
1577         * doc/invoke.texi: Document mtune value.
1579 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
1580             Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
1582         * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
1583         for GCC 9.
1584         * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
1585         prefetch tunings.
1586         * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
1587         costs.
1588         * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
1589         for XGene1.
1591 2018-11-23  Richard Biener  <rguenther@suse.de>
1593         PR tree-optimization/88149
1594         * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
1595         the case where there are two different def types for the
1596         same operand at different operand position in the same stmt.
1598 2018-23-11  Mihail Ionescu  <mihail.ionescu@arm.com>
1600         * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
1601         gen_atomic_compare_swap_1.
1602         (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
1603         (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
1604         (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
1605         (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
1606         (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
1607         checks.
1608         * config/arm/arm.md (neon_vext<mode>),
1609         (neon_vrev64<mode>, neon_vrev32<mode>),
1610         (neon_vrev16<mode>, neon_vtrn<mode>_internal),
1611         (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
1612         before the pattern name.
1613         * config/arm/sync.md:
1614         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
1615         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
1617 2018-11-23  Jakub Jelinek  <jakub@redhat.com>
1619         PR tree-optimization/86614
1620         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
1621         if TREE_NO_WARNING is set on ref.ref.
1623 2018-11-22  Sandra Loosemore  <sandra@codesourcery.com>
1624             Alan Coopersmith  <alan.coopersmith@oracle.com>
1626         PR c/53608
1627         * doc/extend.texi (Designated Inits): Clarify handling of multiple
1628         initializers for unions.
1630 2018-11-22  Jan Hubicka  <jh@suse.cz>
1632         PR lto/88142
1633         * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
1634         parameters; do not warn here.
1635         (odr_subtypes_equivalent_p): Likewise.
1636         (warn_odr): Fix typo.
1637         (warn_types_mismatch): Do not output confused warnings on integer types.
1638         (odr_types_equivalent_p): Update.
1640 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
1642         * config/i386/i386.c (ix86_option_override_internal,
1643         ix86_can_inline_p, classify_argument, construct_container,
1644         ix86_expand_prologue, ix86_expand_split_stack_prologue,
1645         ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
1646         expand_setmem_epilogue_via_loop, promote_duplicated_reg,
1647         ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
1648         Formatting fixes.
1650         * config/i386/i386.c (ix86_option_override_internal): For
1651         stack_protector_guard related options, use opts_set->x_ instead
1652         of global_options_set. and prefix options with opts->x_ .  Move
1653         defaults for offset and reg into else block.
1655 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
1657         * config/i386/i386.c (ix86_check_avx_upper_register):
1658         Return true for all SSE registers with mode bitsize > 128.
1660 2018-11-22  Vladimir Makarov  <vmakarov@redhat.com>
1662         PR rtl-optimization/87718
1663         * ira-costs.c: Remove trailing white-spaces.
1664         (record_operand_costs): Add a special treatment for moves
1665         involving a hard register.
1667 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
1669         * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
1670         (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
1672 2018-11-22  Martin Liska  <mliska@suse.cz>
1674         * common/config/i386/i386-common.c (processor_names): Add
1675         static assert and add missing "znver2".
1676         (ix86_get_valid_option_values): Add checking assert for null
1677         values and add "native" value if feasible.
1678         * config/i386/i386.h: Do not declare size of processor_names.
1679         * common/config/i386/i386-common.c:
1680         * config/i386/i386.c: Add static assert for size
1681         of processor_cost_table.
1683 2018-11-22  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
1685         * target-insns.def (stack_protect_combined_set): Define new standard
1686         pattern name.
1687         (stack_protect_combined_test): Likewise.
1688         * cfgexpand.c (stack_protect_prologue): Try new
1689         stack_protect_combined_set pattern first.
1690         * function.c (stack_protect_epilogue): Try new
1691         stack_protect_combined_test pattern first.
1692         * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
1693         parameters to control which register to use as PIC register and force
1694         reloading PIC register respectively.  Insert in the stream of insns if
1695         possible.
1696         (legitimize_pic_address): Expose above new parameters in prototype and
1697         adapt recursive calls accordingly.  Use pic_reg if non null instead of
1698         cached one.
1699         (arm_load_pic_register): Add pic_reg parameter and use it if non null.
1700         (arm_legitimize_address): Adapt to new legitimize_pic_address
1701         prototype.
1702         (thumb_legitimize_address): Likewise.
1703         (arm_emit_call_insn): Adapt to require_pic_register prototype change.
1704         (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
1705         (thumb1_expand_prologue): Likewise.
1706         * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
1707         change.
1708         (arm_load_pic_register): Likewise.
1709         * config/arm/predicated.md (guard_addr_operand): New predicate.
1710         (guard_operand): New predicate.
1711         * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
1712         prototype change.
1713         (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
1714         prototype change.
1715         (stack_protect_combined_set): New expander..
1716         (stack_protect_combined_set_insn): New insn_and_split pattern.
1717         (stack_protect_set_insn): New insn pattern.
1718         (stack_protect_combined_test): New expander.
1719         (stack_protect_combined_test_insn): New insn_and_split pattern.
1720         (arm_stack_protect_test_insn): New insn pattern.
1721         * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
1722         * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
1723         (UNSPEC_SP_TEST): Likewise.
1724         * doc/md.texi (stack_protect_combined_set): Document new standard
1725         pattern name.
1726         (stack_protect_set): Clarify that the operand for guard's address is
1727         legal.
1728         (stack_protect_combined_test): Document new standard pattern name.
1729         (stack_protect_test): Clarify that the operand for guard's address is
1730         legal.
1732 2018-11-22  Richard Biener  <rguenther@suse.de>
1734         PR tree-optimization/88148
1735         * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
1736         valueization callback parameter and handle it.
1737         * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
1738         * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
1740 2018-11-22  Richard Biener  <rguenther@suse.de>
1742         PR lto/87229
1743         PR lto/88112
1744         * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
1745         which can appear in size expressions.
1746         * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
1747         CALL_EXPR_BY_DESCRIPTOR.
1748         (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
1749         * tree-streamer-out.c (pack_ts_base_value_fields): Stream
1750         CALL_EXPR_BY_DESCRIPTOR.
1751         (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
1753         Revert
1754         PR lto/87229
1755         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
1756         sizepos values.
1758 2018-11-22  Richard Biener  <rguenther@suse.de>
1760         PR tree-optimization/88069
1761         * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
1762         apply to default defs.
1764 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
1766         PR target/85644
1767         PR target/86832
1768         * config/i386/i386.c (ix86_option_override_internal): Default
1769         ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
1770         is defined.
1771         * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
1772         stack_protect_test, stack_protect_test_<mode>): Use empty condition
1773         instead of TARGET_SSP_TLS_GUARD.
1775 2018-11-22  Martin Liska  <mliska@suse.cz>
1777         PR sanitizer/88017
1778         * config/i386/i386.c (ix86_option_override_internal): Error
1779         about usage -mabi=ms and -fsanitize=thread.
1781 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
1783         Revert the revert:
1784         2013-10-26  Vladimir Makarov  <vmakarov@redhat.com>
1786         Revert:
1787         2013-10-25  Vladimir Makarov  <vmakarov@redhat.com>
1789         * lra-spills.c (lra_final_code_change): Remove useless move insns.
1791 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
1793         PR middle-end/88129
1794         * function.c (expand_function_end): Do not emit extra blockage insn.
1796 2018-11-21  Lokesh Janghel  <lokeshjanghel91@gmail.com>
1798         PR target/85667
1799         * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
1800         of FIRST_SSE_REG for 4 or 8 byte modes.
1802 2018-11-21  Jan Hubicka  <jh@suse.cz>
1804         PR lto/87957
1805         * tree.c (fld_decl_context): Break out from ...
1806         (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
1807         DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
1808         (fld_incomplete_type_of): Build copy of TYP_DECL.
1809         * ipa-devirt.c (free_enum_values): Rename to ...
1810         (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
1811         and TREE_TYPEs of TYPE_DECLs.
1812         (get_odr_type): Initialize odr_vtable_hash if needed.
1814 2018-11-21  Alexandre Oliva <oliva@adacore.com>
1816         * final.c (compute_discriminator): Declare.  Renamed from...
1817         (maybe_set_discriminator): ... this.  Set and return a local.
1818         (override_discriminator): New.
1819         (final_scan_insn_1): Set it.
1820         (notice_source_line): Adjust.  Always set discriminator.
1822 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
1824         PR target/87839
1825         * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
1826         rIJ constraint for aarch64_plus_operand rather than rn.
1828 2018-11-21  Renlin Li  <renlin.li@arm.com>
1830         PR middle-end/84877
1831         * explow.h (get_dynamic_stack_size): Declare it as external.
1832         * explow.c (record_new_stack_level): Remove function static attribute.
1833         * function.c (assign_stack_local_1): Dynamically align the stack slot
1834         addr for parameter copy on the stack.
1836 2018-11-21  Richard Biener  <rguenther@suse.de>
1838         PR bootstrap/88133
1839         * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
1840         * Makefile.in (bitmap.o-warn): Remove again.
1842 2018-11-20  Jeff Law  <law@redhat.com>
1844         PR tree-optimization/88069
1845         * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
1846         degenerate virtual PHIs.
1848 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
1850         PR target/87317
1851         * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
1852         nonimmediate_operand with register_operand.
1853         (avx2_<code>v8qiv8si2<mask_name>): Likewise.
1854         (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
1855         (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
1856         (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
1857         (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
1858         (avx2_<code>v4qiv4di2<mask_name>): Likewise.
1859         (avx2_<code>v4hiv4di2<mask_name>): Likewise.
1860         (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
1861         (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
1862         (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
1863         (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
1864         (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
1865         (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
1866         (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
1867         (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
1868         (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
1869         (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
1870         (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
1871         (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
1872         (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
1873         (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
1874         (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
1875         (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
1876         (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
1877         (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
1878         (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
1879         (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
1881 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
1883         * read-rtl.c (apply_subst_iterator): Handle define_split and
1884         define_insn_and_split.
1886 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
1888         PR rtl-optimization/87817
1889         * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
1890         *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
1891         in the pattern to avoid triggering UB when operands[2] is zero.
1892         (tbm_bextri_<mode>): New expander.  Renamed the old define_insn to ...
1893         (*tbm_bextri_<mode>): ... this.
1895 2018-11-21  Tom de Vries  <tdevries@suse.de>
1897         PR driver/79855
1898         * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
1899         period.
1901 2018-11-21  Andreas Krebbel  <krebbel@linux.ibm.com>
1903         * configure.ac: Add check for Binutils to determine whether vector
1904         load/store alignments hints are being supported.
1905         * config.in: Regenerate.
1906         * configure: Regenerate.
1907         * config/s390/s390.c (print_operand): Support new output
1908         modifier A.
1909         * config/s390/s390.md ("movti"): Append alignment hint output
1910         using the new output modifier 'A'.
1911         * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
1912         ("*vec_ti_to_v1ti"): Likewise.
1914 2018-11-20  Martin Sebor  <msebor@redhat.com>
1916         * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
1918 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1920         PR lto/84044
1921         * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
1922         compare ENUM values.
1924 2018-11-20  Jakub Jelinek  <jakub@redhat.com>
1926         PR tree-optimization/87895
1927         * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
1928         or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
1929         (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
1930         redirects edges to EXIT to edges to incr_bb, iterate while EXIT
1931         has any preds and always use EDGE_PRED (, 0).
1933 2018-11-20  Uros Bizjak  <ubizjak@gmail.com>
1935         PR target/88070
1936         * mode-switching.c (create_pre_exit): After reload, always split the
1937         fallthrough edge to the exit block.
1939 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1941         * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
1942         * tree.c (build_array_type_1): Forward declare.
1943         (fld_type_variant_equal_p): Add INNER_TYPE parameter.
1944         (fld_type_variant): Likewise.
1945         (fld_simplified_types): New hash.
1946         (fld_process_array_type): New function.
1947         (fld_incomplete_type_of): Handle array and enumeration types.
1948         (fld_simplified_type): Handle simplification of arrays.
1949         (free_lang_data): Allocate and free simplified types hash.
1951 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1953         PR lto/87957
1954         * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
1955         when sybtype already violates ODR.
1956         (get_odr_type): Do not ICE when insert is false and type duplicate
1957         is not registered yet.
1958         (register_odr_type): Be sure to register subtypes first.
1960 2018-11-20  Andreas Krebbel  <krebbel@linux.ibm.com>
1962         * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
1963         DImode parts of the target operand.
1965 2018-11-20  Nathan Sidwell  <nathan@acm.org>
1967         PR 87926
1968         * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
1970 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1972         PR lto/87997
1973         * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
1974         type checking to...
1975         (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
1977 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1979         PR ipa/87706
1980         * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
1981         * ipa.c (possible_inline_candidate_p): Break out from ..
1982         (process_references): ... here ; drop before_inlining_p;
1983         cleanup handling of alises.
1984         (walk_polymorphic_call_targets): Likewise.
1985         (symbol_table::remove_unreachable_nodes): Likewise.
1986         * passes.c (pass_data_ipa_remove_symbols): New structure.
1987         (pass_ipa_remove_symbols): New pass.
1988         (make_pass_ipa_remove_symbols): New function.
1989         * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
1990         * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
1992 2018-11-20  Richard Biener  <rguenther@suse.de>
1994         * tree-vect-stmts.c (vectorizable_condition): Do not get
1995         at else_clause vect def for EXTRACT_LAST_REDUCTION.  Remove
1996         pointless vect_is_simple_use calls.
1998 2018-11-20  Richard Biener  <rguenther@suse.de>
2000         PR tree-optimization/88074
2001         * tree-ssa-dom.c (pass_dominator::execute): Do not walk
2002         backedges.
2004 2018-11-20  Richard Biener  <rguenther@suse.de>
2006         PR tree-optimization/88069
2007         * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
2008         virtual PHI arguments.
2010 2018-11-20  Ilya Leoshkevich  <iii@linux.ibm.com>
2012         PR target/88083
2013         * config/s390/s390.md: Skip LT(G) peephole when literal pool is
2014         involved.
2015         * rtl.h (contains_constant_pool_address_p): New function.
2016         * rtlanal.c (contains_constant_pool_address_p): Likewise.
2018 2018-11-20  Richard Biener  <rguenther@suse.de>
2020         PR middle-end/83215
2021         * alias.c (component_uses_parent_alias_set_from): Remove
2022         alias-set zero and TYPE_TYPELESS_STORAGE case both already
2023         handled in other ways.
2025 2018-11-20  Richard Biener  <rguenther@suse.de>
2027         PR tree-optimization/88087
2028         * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
2029         call fntype.
2030         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
2031         call fntype.
2033 2018-11-20  Richard Biener  <rguenther@suse.de>
2035         PR middle-end/88089
2036         * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
2038 2018-11-19  Jan Hubicka  <hubicka@ucw.cz>
2040         PR lto/87957
2041         * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
2043 2018-11-19  Sandra Loosemore  <sandra@codesourcery.com>
2045         PR driver/50250
2046         * doc/invoke.texi (Link Options): Mention shared libraries
2047         in documentation for the -l option.  Simplify discussion and
2048         point to the system linker documentation for details.
2050 2018-11-19  Peter Bergner  <bergner@linux.ibm.com>
2052         PR rtl-optimization/88033
2053         * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
2054         to itself.  Use HARD_REGISTER_NUM_P.
2056 2018-11-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2058         * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
2059         of the loop if vf is non-constant.
2061 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
2063         PR tree-optimization/87025
2064         * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
2065         immediately after creating the scope optinfo.
2066         (selftest::test_pr87025): New function.
2067         (selftest::dumpfile_c_tests): Call it.
2068         * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
2069         that we're not popping the top-level records array.
2070         * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
2071         not changing the kind of a "scope" optinfo.
2073 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
2075         PR tree-optimization/87025
2076         * dump-context.h: Include "optinfo.h".
2077         (class optrecord_json_writer): New forward decl.
2078         (dump_context::forcibly_enable_optinfo_p): Delete.
2079         (dump_context::optinfo_enabled_p): New member function.
2080         (dump_context::optimization_records_enabled_p): New member
2081         function.
2082         (dump_context::set_json_writer): New member function.
2083         (dump_context::emit_optinfo): New member function.
2084         (dump_context::m_forcibly_enable_optinfo): Delete.
2085         (dump_context::m_json_writer): New member data.
2086         * dumpfile.c (dump_context::set_json_writer): New member function.
2087         (dump_context::finish_any_json_writer): New member function.
2088         (dump_context::end_scope): Replace call to
2089         optimization_records_maybe_pop_dump_scope with call to
2090         m_json_writer->pop_scope.
2091         (dump_context::optinfo_enabled_p): New member function.
2092         (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
2093         to dump_context::emit_optinfo.
2094         (dump_context::emit_optinfo): New member function.
2095         (temp_dump_context::temp_dump_context): Replace
2096         m_forcibly_enable_optinfo with call to set_json_writer.
2097         (temp_dump_context::~temp_dump_context): Clean up any json writer.
2098         * optinfo-emit-json.cc (class optrecord_json_writer): Move to
2099         optinfo-emit-json.h
2100         (the_json_writer): Delete.
2101         (optimization_records_start): Delete.
2102         (optimization_records_finish): Delete.
2103         (optimization_records_enabled_p): Delete, in favor of
2104         dump_context::optimization_records_enabled_p.
2105         (optimization_records_maybe_record_optinfo): Delete.
2106         (optimization_records_maybe_pop_dump_scope): Delete.
2107         * optinfo-emit-json.h: Include "json.h".  Delete forward
2108         decl of opt_pass.
2109         (optimization_records_start): Delete.
2110         (optimization_records_finish): Delete.
2111         (optimization_records_enabled_p): Delete.
2112         (optimization_records_maybe_record_optinfo): Delete.
2113         (optimization_records_maybe_pop_dump_scope): Delete.
2114         (class optrecord_json_writer): Move here from
2115         optinfo-emit-json.cc.
2116         * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
2117         to optinfo::emit with call to dump_context::emit_optinfo.
2118         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
2119         (optinfo_enabled_p): Delete, in favor of
2120         dump_context::optinfo_enabled_p.
2121         (optinfo_wants_inlining_info_p): Update for conversion o
2122         optimization_records_enabled_p to a member function of
2123         dump_context.
2124         * optinfo.h (optinfo_enabled_p): Delete, in favor of
2125         dump_context::optinfo_enabled_p.
2126         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
2127         * toplev.c: Include "dump-context.h".
2128         (compile_file): Replace call to optimization_records_finish with
2129         dump_context::finish_any_json_writer.
2130         (do_compile): Replace call to optimization_records_start with
2131         conditionally creating a optrecord_json_writer for the
2132         dump_context.
2134 2018-11-19  Jakub Jelinek  <jakub@redhat.com>
2136         PR tree-optimization/88071
2137         * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
2138         of false as last argument to gsi_remove.
2139         * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
2140         false as last argument to gsi_replace.
2142         PR debug/87039
2143         * omp-expand.c: Don't include debug.h.
2144         (adjust_context_and_scope): Add REGION argument.  Find DECL_CONTEXT
2145         from innermost outer parallel, task, teams or target that has a
2146         child_fn set, or, if there is no such outer region, use
2147         current_function_decl.  Do the DECL_CONTEXT adjustment regardless of
2148         whether a suitable BLOCK is found or not.
2149         (expand_parallel_call, expand_teams_call): Don't call
2150         adjust_context_and_scope here.
2151         (grid_expand_target_grid_body): Revert 2017-01-25 changes.
2152         (expand_omp_taskreg, expand_omp_target): Likewise.  Call
2153         adjust_context_and_scope.
2154         * dwarf2out.c (dwarf2out_early_global_decl): For
2155         decl_function_context recurse instead of calling dwarf2out_decl.
2157 2018-11-20  Eric Botcazou  <ebotcazou@adacore.com>
2159         PR rtl-optimization/85925
2160         * rtl.h (word_register_operation_p): New predicate.
2161         * combine.c (record_dead_and_set_regs_1): Only apply specific handling
2162         for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
2163         * rtlanal.c (nonzero_bits1): Likewise.  Adjust couple of comments.
2164         (num_sign_bit_copies1): Likewise.
2166 2018-11-19  Richard Biener  <rguenther@suse.de>
2168         PR lto/87229
2169         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
2170         sizepos values.
2172 2018-11-19  Eric Botcazou  <ebotcazou@adacore.com>
2174         * compare-elim.c (struct comparison): Add not_in_a field.
2175         (is_not): New static function.
2176         (strip_not): Likewise.
2177         (conforming_compare): Handle a NOT in the first operand.
2178         (can_eliminate_compare): Likewise.
2179         (find_comparison_dom_walker::before_dom_children): Likewise.
2180         (try_eliminate_compare): Likewise.
2181         * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
2182         (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
2184 2018-11-19  Jonathan Wakely  <jwakely@redhat.com>
2186         * doc/extend.texi (Common Type Attributes): Fix typos.
2188 2018-11-19  Martin Liska  <mliska@suse.cz>
2190         PR gcov-profile/88045
2191         * coverage.c (coverage_begin_function): Add assert.
2193 2018-11-18  Sandra Loosemore  <sandra@codesourcery.com>
2195         PR other/40498
2196         * doc/extend.texi (Common Function Attributes): Document that
2197         no_instrument_function applies to -p and -pg, too.
2198         * doc/invoke.texi (Instrumentation Options): Add cross-references
2199         to docs for -p, -pg, and -finstrument-functions.
2201 2018-11-18  Alan Modra  <amodra@gmail.com>
2203         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
2204         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
2205         (GNU_USER_TARGET_CRTN): Define.
2206         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
2207         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
2209 2018-11-18  Alan Modra  <amodra@gmail.com>
2211         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
2212         !HAVE_LD_PIE variant.
2213         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
2215 2018-11-17  Nick Clifton  <nickc@redhat.com>
2216             Sandra Loosemore  <sandra@codesourcery.com>
2218         PR driver/31357
2219         * doc/invoke.texi (Overall Options): Document that --help and
2220         --help=<value> options cannot be combined.
2222 2018-11-17  Jim Wilson  <jimw@sifive.com>
2224         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
2225         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
2226         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
2227         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
2228         to step2 if frame_pointer_needed and step1 is 0.
2230 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
2232         PR c++/4225
2233         * doc/cpp.texi (System Headers): Add note about implicit
2234         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
2236 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
2238         PR c++/4025
2239         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
2240         -fno-implicit-templates.
2242 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
2244         PR tree-optimization/87546
2245         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
2246         min_precision variable, initially set it to orig_precision, only does
2247         something if op_type's precision is <= min_precision and update
2248         min_precision whenever calling set_op.
2250 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
2252         PR ipa/87957
2253         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
2255 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
2257         * doc/invoke.texi (Option Summary): Fix whitespace and line
2258         breaks in @gccoptlist environments.
2259         (Warning Options): Likewise.
2260         (Optimize Options): Likewise.
2261         (PowerPC SPE Options): Likewise.
2262         (RS/6000 and PowerPC Options): Likewise.
2264 2018-11-16  Jeff Law  <law@redhat.com>
2266         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
2267         gcc_assert.
2269 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
2271         PR middle-end/23197
2272         * doc/invoke.texi (Optimize Options): Update options enabled by
2273         fprofile-generate, -fprofile-use, and -fauto-profile.
2275 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2277         PR target/87927
2278         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
2279         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
2280         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
2281         the byte_op hook.
2282         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
2283         * targhooks.c (default_print_patchable_function_entry): Assert
2284         asm_int_op does not return a NULL string.
2285         * varasm.c (integer_asm_op): Return the op for a partial int type
2286         when the requested size does not correspond to an integer type.
2287         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
2288         * doc/tm.texi: Regenerate.
2290 2018-11-12  Jason Merrill  <jason@redhat.com>
2292         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
2294 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
2296         * config/rs6000/constraints.md (wF constraint): Remove power9
2297         fusion documentation.  Just document wF's use for power8 fusion.
2298         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
2299         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
2300         fusion support to require the upper 12 bits to be all 0's or all
2301         1's.
2302         (fusion_gpr_mem_load): Add comment.
2303         (fusion_addis_mem_combo_load): Remove power9 fusion support.
2304         (fusion_addis_mem_combo_store): Delete.
2305         (fusion_offsettable_mem_operand): Delete.
2306         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
2307         power8 fusion here.
2308         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
2309         (POWERPC_MASKS): Delete power9 fusion.
2310         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
2311         (fusion_p9_p): Delete.
2312         (expand_fusion_p9_load): Delete.
2313         (expand_fusion_p9_store): Delete.
2314         (emit_fusion_p9_load): Delete.
2315         (emit_fusion_p9_store): Delete.
2316         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
2317         fusion support.
2318         (rs6000_option_override_internal): Set power8 fusion based on
2319         whether we are tuning for power8.  Delete power9 fusion support.
2320         (rs6000_opt_masks): Delete -mpower9-fusion switch.
2321         (emit_fusion_load): Rename emit_fusion_load_store to
2322         emit_fusion_load, and drop fusion store support.  Update callers.
2323         (emit_fusion_load_store): Likewise.
2324         (emit_fusion_gpr_load): Likewise.
2325         (fusion_p9_p): Delete.
2326         (expand_fusion_p9_load): Delete.
2327         (expand_fusion_p9_store): Delete.
2328         (emit_fusion_p9_load): Delete.
2329         (emit_fusion_p9_store): Delete.
2330         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
2331         (GPR_FUSION): Delete.
2332         (FPR_FUSION): Delete.
2333         (power9 fusion peephole2s): Delete.
2334         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
2335         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
2336         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
2337         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
2338         (fusion_p9_<mode>_constant): Delete.
2339         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
2340         power9 fusion switch.
2341         * doc/md.texi (PowerPC constraints): Update wF constraint
2342         documentation for power8 fusion only.
2344 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
2346         PR rtl-optimization/87475
2347         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
2348         CROSSING_JUMP_P insns.
2349         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
2351 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
2353         PR target/88051
2354         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
2355         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
2356         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
2358 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
2360         PR middle-end/88032
2361         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
2362         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
2364 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
2366         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
2367         location_t parameter.
2368         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2369         (switch_decision_tree::do_jump_if_equal): Likewise.
2370         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
2371         of switch statement to emit_case_nodes.
2372         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
2373         set it on the newly built GIMPLE comparison statement.
2374         (switch_decision_tree::do_jump_if_equal): Likewise.
2375         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
2376         in calls to do_jump_if_equal as well as recursive calls.
2378 2018-11-16  Andi Kleen  <ak@linux.intel.com>
2380         * config/i386/i386.md: Allow memory operands to ptwrite.
2382 2018-11-16  Michael Matz  <matz@suse.de>
2384         * system.h (PRsa): New macro.
2385         (SIZE_AMOUNT): Cast number to uint64_t.
2386         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
2387         (pool_usage::dump_footer): Likewise and also use PRIu64.
2388         * bitmap.h (bitmap_usage::dump): Likewise.
2389         * ggc-common.c (ggc_usage::dump): Likewise.
2390         * ggc-page.c (ggc_print_statistics): Likewise.
2391         * input.c (dump_line_table_statistics): Likewise.
2392         * mem-stats.h (mem_usage::dump): Likewise.
2393         (mem_usage::dump_footer): Likewise.
2394         * rtl.c (dump_rtx_statistics): Likewise.
2395         * tree-cfg.c (dump_cfg_stats): Likewise.
2396         * tree-dfa.c (dump_dfa_stats): Likewise.
2397         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2398         * tree-ssanames (ssanames_print_statistics): Likewise.
2399         * vec.c (vec_usage::dump): Likewise.
2400         (vec_usage::dump_footer): Likewise.
2402 2018-11-16  Richard Biener  <rguenther@suse.de>
2404         PR tree-optimization/88011
2405         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
2406         replacing set_value_range_to_undefined and
2407         set_value_range_to_varying with method calls.
2409 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
2411         * config/s390/s390.md
2412         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
2413         assembler.
2414         (*r<noxa>sbg_<mode>_sll): Likewise.
2415         (*r<noxa>sbg_<mode>_srl): Likewise.
2416         (*r<noxa>sbg_sidi_srl): New pattern.
2418 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
2420         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
2421         Default to TARGET_VXWORKS_RTP.
2422         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
2423         of TARGET_VXWORKS_RTP.
2424         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
2425         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
2426         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
2428 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
2430         PR c++/25759
2431         * doc/extend.texi (Common Type Attributes): Make it explicit
2432         that attribute "packed" can apply to C++ classes.
2434 2018-11-15  Martin Sebor  <msebor@redhat.com>
2436         PR c++/87541
2437         PR c++/87542
2438         * tree.c (type_argument_type): New function.
2439         * tree.h (type_argument_type): Declare it.
2440         * gcc/doc/extend.texi (alloc_align): Update and clarify.
2441         (alloc_size, nonnull, sentinel): Same.
2443 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
2444             Kwok Cheung Yeung  <kcy@codesourcery.com>
2446         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
2447         int_mode_for_size fails.
2448         (vectorizable_load): Likewise.
2450 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
2452         * doc/ux.texi (Group logically-related diagnostics): Move
2453         discussion of auto_diagnostic_group into this new subsection.
2454         Give an example of where this grouping is used.
2456 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
2458         PR other/19165
2459         * Makefile.in (OBJS): Move json.o to...
2460         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
2461         * common.opt (fdiagnostics-format=): New option.
2462         (diagnostics_output_format): New enum.
2463         * diagnostic-format-json.cc: New file.
2464         * diagnostic.c (default_diagnostic_final_cb): New function, taken
2465         from start of diagnostic_finish.
2466         (diagnostic_initialize): Initialize final_cb to
2467         default_diagnostic_final_cb.
2468         (diagnostic_finish): Move "being treated as errors" messages to
2469         default_diagnostic_final_cb.  Call any final_cb.
2470         (default_diagnostic_finalizer): Add diagnostic_t param.
2471         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
2472         diagnostic_finalizer callback.
2473         * diagnostic.h (enum diagnostics_output_format): New enum.
2474         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
2475         (struct diagnostic_context): Add "final_cb".
2476         (default_diagnostic_finalizer): Add diagnostic_t param.
2477         (diagnostic_output_format_init): New decl.
2478         * doc/invoke.texi (-fdiagnostics-format): New option.
2479         * dwarf2out.c (gen_producer_string): Ignore
2480         OPT_fdiagnostics_format_.
2481         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
2482         * lto-wrapper.c (append_diag_options): Ignore it.
2483         * opts.c (common_handle_option): Handle it.
2485 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
2487         PR tree-optimization/88015
2488         * graphite-isl-ast-to-gimple.c
2489         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
2490         for dump_enabled_p.
2491         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
2493 2018-11-15  Richard Biener  <rguenther@suse.de>
2495         PR middle-end/88029
2496         * gimple.c (gimple_call_flags): Union flags from decl, type
2497         and call fntype.
2498         * trans-mem.c (is_tm_pure_call): Simplify.
2500 2018-11-15  Richard Biener  <rguenther@suse.de>
2502         PR tree-optimization/88031
2503         * tree-vect-loop.c (vectorizable_reduction): Move check
2504         for multiple types earlier so we get the expected dump.
2505         Simplify calls to vectorizable_condition.
2506         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
2507         * tree-vect-stmts.c (vectorizable_condition): Instead of
2508         reduc_def and reduc_index take just a flag.  Simplify
2509         code-generation now that we can rely on the defs being set up.
2510         (vectorizable_comparison): Remove unused argument.
2512 2018-11-15  Richard Biener  <rguenther@suse.de>
2514         PR tree-optimization/88030
2515         * tree-complex.c (need_eh_cleanup): New global.
2516         (update_complex_assignment): Mark blocks that need EH update.
2517         (expand_complex_comparison): Likewise.
2518         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
2519         perform EH cleanup and schedule CFG cleanup if that did anything.
2521 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
2523         PR rtl-optimization/88018
2524         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
2525         on the presence of fallthru edge, rather than if it is a USE or not.
2527 2018-11-15  Richard Biener  <rguenther@suse.de>
2529         PR middle-end/87917
2530         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
2531         analyze_subscript_affine_affine properly.
2533 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2535         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
2536         (GLIBC_DYNAMIC_LINKER): Define.
2537         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
2539 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
2541         PR tree-optimization/84648
2542         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
2543         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
2544         by calling adjust_cond_for_loop_until_wrap.
2546 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
2548         PR other/56334
2549         * doc/extend.texi (Common Function Attributes): Clarify linker
2550         restrictions on "aligned" attribute.
2551         (Common Variable Attributes): Likewise.  Mention that linker
2552         restrictions don't apply to stack-allocated variables.
2554 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
2556         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
2557         as an lvalue.
2559 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
2561         PR lto/55102
2562         PR lto/56700
2563         * doc/invoke.texi (Optimize Options): Remove bad example about
2564         interaction between -flto and -O.  Replace it with a note that
2565         you need to compile with -O and not just link.  Copy-edit -flto
2566         discussion to reduce verbiage and improve flow.
2568 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
2570         PR middle-end/59658
2571         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
2572         also suppress many optimizations.  Alphabetize option lists for
2573         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
2574         correct documentation for those options to say that.
2575         * opts.c (default_options_table): Sort table by level and option
2576         name, to make it easier to correlate to the manual.
2578 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
2580         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
2581         printing of the names of segment registers.
2583 2018-11-14  Nathan Sidwell  <nathan@acm.org>
2585         PR debug/88006
2586         PR debug/87462
2587         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
2588         type list.
2590 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
2592         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
2594 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
2596         P1236R1 - Signed integers are two's complement
2597         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
2599         PR bootstrap/86739
2600         * hash-map.h (hash_map::iterator::reference_pair): New class.
2601         (hash_map::iterator::operator*): Return it rather than std::pair.
2603 2018-11-14  Jeff Law  <law@redhat.com>
2605         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
2606         iff the operand is a constant.
2608 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
2610         * gimple-ssa-evrp-analyze.c
2611         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
2612         ignore_equivs_equal_p to equal_p.
2613         * ipa-cp.c (meet_with_1): Use equal_p instead of
2614         ignore_equivs_equal_p.
2615         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
2616         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
2617         (value_range::operator==): Remove.
2618         (value_range::operator!=): Remove.
2619         (vrp_prop::visit_stmt): Use equal_p.
2620         * tree-vrp.h (value_range): Remove operator==, operator!=,
2621         ignore_equivs_equal_p.
2622         * vr-values.c (update_value_range): Use equal_p.
2624 2018-11-14  Michael Matz  <matz@suse.de>
2626         PR middle-end/86575
2627         * gimplify.c (collect_fallthrough_labels): Add new argument,
2628         return location via that, don't modify statements.
2629         (warn_implicit_fallthrough_r): Adjust call, don't use
2630         statement location directly.
2632 2018-11-14  Richard Biener  <rguenther@suse.de>
2634         PR middle-end/87985
2635         * tree-data-ref.c (split_constant_offset): Add wrapper
2636         allocating a cache hash-map.
2637         (split_constant_offset_1): Cache results of expanding
2638         expressions from SSA def stmts.
2640 2018-11-14  Richard Biener  <rguenther@suse.de>
2642         PR middle-end/88021
2643         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
2644         to lambda_int.
2645         (lambda_vector_mult_const): Likewise.
2646         (lambda_matrix_right_hermite): Use lambda_int temporaries.
2648 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
2649             Jackson Woodruff  <jackson.woodruff@arm.com>
2651         PR 71026/tree-optimization
2652         * match.pd: Simplify floating point comparisons.
2654 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
2656         PR rtl-optimization/87817
2657         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
2658         last argument has low 8 bits clear, fold to 0.
2660 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
2662         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
2663         MAX_OFILE_ALIGNMENT as needed.
2665 2018-11-14  Richard Biener  <rguenther@suse.de>
2667         PR tree-optimization/88019
2668         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
2669         COND_EXPR arguments to non-trapping overflow.
2671 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
2673         * config/arc/arc.h (reg_class): Reorder registers classes, remove
2674         unused register classes.
2675         (REG_CLASS_NAMES): Likewise.
2676         (REG_CLASS_CONTENTS): Likewise.
2677         (FIXED_REGISTERS): Make lp_count fixed.
2678         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
2679         (PROGRAM_COUNTER_REGNO): Remove.
2680         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
2681         register classes, use constants for register numbers, remove
2682         reg_class_contents references.
2683         (arc_process_double_reg_moves): Add asserts.
2684         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
2685         lra_in_progress predicate.
2686         (arc_init_reg_tables): Remove unused register classes.
2687         (arc_register_move_cost): Likewise.
2688         (arc_preferred_reload_class): Likewise.
2689         (hwloop_optimize): Update rtx patterns involving lp_count register.
2690         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
2691         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
2692         CHEAP_CORE_REGS.  Former one will be used for LRA.
2693         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
2694         one will be used for LRA.
2695         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
2696         one will be used for LRA.
2697         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
2698         Former one will be used for LRA.
2699         ("f"): Delete constraint.
2700         ("k"): Likewise.
2701         ("e"): Likewise.
2702         ("l"): Change it from register constraint to constraint.
2703         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
2704         (movhi_insn): Likewise.
2705         (movsi_insn): Update pattern.
2706         (arc_lp): Likewise.
2707         (dbnz): Likewise.
2708         (stack_tie): Remove 'b' constraint letter.
2709         (R4_REG): Define.
2710         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
2711         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
2712         (R57_REG, R59_REG, PCL_REG): Likewise.
2713         (ILINK1_REGNUM): Renamed to ILINK1_REG.
2714         (ILINK2_REGNUM): Renamed to ILINK2_REG.
2715         (Rgp): Remove.
2716         (SP_REGS): Likewise.
2717         (Rcw): Remove unused reg classes.
2718         * config/arc/predicates.md (dest_reg_operand): Just default on
2719         register_operand predicate.
2720         (mpy_dest_reg_operand): Likewise.
2721         (move_dest_operand): Use macros instead of constants.
2722         * config/arc/arc.opt (mlra): Switch to lra as default.
2724 2018-11-14  Richard Biener  <rguenther@suse.de>
2726         PR tree-optimization/87974
2727         * tree-vect-loop.c (vectorizable_reduction): When computing
2728         the vectorized reduction PHI vector type ignore constant
2729         and external defs.
2731 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
2733         PR tree-optimization/87977
2734         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
2735         stmt, build a new one and replace the old one with it.  Formatting fix.
2736         Call release_ssa_name (x) if !has_other_use and !delete_div.
2737         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
2738         verify lhs of stmt is still def.
2740 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
2742         PR rtl-optimization/87507
2743         * lower-subreg.c (operand_for_swap_move_operator): New function.
2744         (simple_move): Strip simple operators.
2745         (find_pseudo_copy): Likewise.
2746         (resolve_operand_for_swap_move_operator): New function.
2747         (resolve_simple_move): Strip simple operators and swap operands.
2749 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
2751         PR tree-optimization/87898
2752         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
2753         (ipa_simd_modify_function_body): Remove debug stmts where the first
2754         argument was changed into a non-decl.
2756 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
2758         PR rtl-optimization/87899
2759         * lra-lives.c (start_living): Update white space in comment.
2760         (enum point_type): New.
2761         (sparseset_contains_pseudos_p): New function.
2762         (update_pseudo_point): Likewise.
2763         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
2764         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
2765         handling.  Move early exit after adding conflicts.
2766         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
2767         if regno is already live.  Remove all handling of program points.
2768         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
2769         after adding conflicts.  Remove all handling of program points and
2770         ignore_reg_for_conflicts.
2771         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
2772         and do not guard call to mark_pseudo_live.
2773         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
2774         and do not guard call to mark_pseudo_dead.
2775         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
2776         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
2777         Use new function update_pseudo_point.  Handle register copies by
2778         removing the source register from the live set.  Handle INOUT operands.
2779         Update to the next program point using the unused_set, dead_set and
2780         start_dying sets.
2781         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
2783 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2785         * builtins.c: Replace "source_location" with "location_t".
2786         * diagnostic-show-locus.c: Likewise.
2787         * diagnostic.c: Likewise.
2788         * dumpfile.c: Likewise.
2789         * gcc-rich-location.h: Likewise.
2790         * genmatch.c: Likewise.
2791         * gimple.h: Likewise.
2792         * gimplify.c: Likewise.
2793         * input.c: Likewise.
2794         * input.h: Likewise.  Eliminate the typedef.
2795         * omp-expand.c: Likewise.
2796         * selftest.h: Likewise.
2797         * substring-locations.h (get_source_location_for_substring):
2798         Rename to...
2799         (get_location_within_string): ...this.
2800         * tree-cfg.c: Replace "source_location" with "location_t".
2801         * tree-cfgcleanup.c: Likewise.
2802         * tree-diagnostic.c: Likewise.
2803         * tree-into-ssa.c: Likewise.
2804         * tree-outof-ssa.c: Likewise.
2805         * tree-parloops.c: Likewise.
2806         * tree-phinodes.c: Likewise.
2807         * tree-phinodes.h: Likewise.
2808         * tree-ssa-loop-ivopts.c: Likewise.
2809         * tree-ssa-loop-manip.c: Likewise.
2810         * tree-ssa-phiopt.c: Likewise.
2811         * tree-ssa-phiprop.c: Likewise.
2812         * tree-ssa-threadupdate.c: Likewise.
2813         * tree-ssa.c: Likewise.
2814         * tree-ssa.h: Likewise.
2815         * tree-vect-loop-manip.c: Likewise.
2817 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2819         * doc/extend.texi: Fix typo in the weakref description.
2821 2018-11-13  Richard Biener  <rguenther@suse.de>
2823         PR tree-optimization/86991
2824         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
2825         group building until we have successfully detected the SLP
2826         reduction.
2827         (vect_is_simple_reduction): Remove fixup code here.
2829 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2831         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
2832         (dump_gimple_stmt): Use it.
2833         (dump_gimple_stmt_loc): Likewise.
2834         (dump_gimple_expr): Likewise.
2835         (dump_gimple_expr_loc): Likewise.
2836         (dump_generic_expr): Likewise.
2837         (dump_generic_expr_loc): Likewise.
2838         (dump_printf): Likewise.
2839         (dump_printf_loc): Likewise.
2840         (dump_dec): Likewise.
2841         (dump_dec): Likewise.
2842         (dump_hex): Likewise.
2843         (dump_symtab_node): Likewise.
2844         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
2845         Guard dump call with dump_enabled_p.
2846         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
2847         * graphite-optimize-isl.c (optimize_isl): Likewise.
2848         * graphite.c (graphite_transform_loops): Likewise.
2849         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
2850         * tree-parloops.c (parallelize_loops): Likewise.
2851         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
2852         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
2853         (vect_prune_runtime_alias_test_list): Likewise.
2854         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
2855         (vect_estimate_min_profitable_iters): Likewise.
2856         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
2857         (vect_build_slp_tree_2): Likewise.
2858         (vect_supported_load_permutation_p): Likewise.
2859         (vect_slp_analyze_operations): Likewise.
2860         (vect_slp_analyze_bb_1): Likewise.
2861         (vect_slp_bb): Likewise.
2862         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
2863         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
2864         (pass_slp_vectorize::execute): Likewise.
2865         (increase_alignment): Likewise.
2867 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2869         PR ipa/87955
2870         * ipa-inline.c (report_inline_failed_reason): Guard calls to
2871         cl_target_option_print_diff and cl_optimization_print_diff with
2872         if (dump_file).
2874 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2876         * doc/invoke.texi (-fsave-optimization-record): Note that the
2877         output is compressed.
2878         * optinfo-emit-json.cc: Include <zlib.h>.
2879         (optrecord_json_writer::write): Compress the output.
2881 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
2883         * tree-vrp.c (value_range_base::dump): Dump type.
2884         Do not use INF nomenclature for 1-bit types.
2885         (dump_value_range): Group all variants to common dumping code.
2886         (debug): New overloaded functions for value_ranges.
2887         (value_range_base::dump): Remove no argument version.
2888         (value_range::dump): Same.
2890 2018-11-13  Richard Biener  <rguenther@suse.de>
2892         PR tree-optimization/87931
2893         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
2894         nested cycles we support to latch computations vectorizable_reduction
2895         handles.
2897 2018-11-13  Martin Liska  <mliska@suse.cz>
2899         PR tree-optimization/87885
2900         * cfghooks.c (account_profile_record): Rename
2901         to ...
2902         (profile_record_check_consistency): ... this.
2903         Calculate missing num_mismatched_freq_in.
2904         (profile_record_account_profile): New function
2905         that calculates time and size of a function.
2906         * cfghooks.h (struct profile_record): Remove
2907         all tuples.
2908         (struct cfg_hooks): Remove after_pass flag.
2909         (account_profile_record): Rename to ...
2910         (profile_record_check_consistency): ... this.
2911         (profile_record_account_profile): New.
2912         * cfgrtl.c (rtl_account_profile_record): Remove
2913         after_pass flag.
2914         * passes.c (check_profile_consistency): Do only
2915         checking.
2916         (account_profile): Calculate size and time of
2917         function only.
2918         (pass_manager::dump_profile_report): Reformat
2919         output.
2920         (execute_one_ipa_transform_pass): Call
2921         consistency check before clean upand call account_profile
2922         after a clean up is done.
2923         (execute_one_pass): Call check_profile_consistency and
2924         account_profile instead of using after_pass flag..
2925         * tree-cfg.c (gimple_account_profile_record): Likewise.
2927 2018-11-13  Martin Liska  <mliska@suse.cz>
2929         PR sanitizer/87930
2930         * config/i386/i386.c (ix86_option_override_internal): Error
2931         about usage -mabi=ms and -fsanitize={,kernel-}address.
2933 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2935         * config/aarch64/aarch64.c
2936         (aarch64_vectorize_preferred_vector_alignment): Change return type to
2937         poly_uint64.
2938         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
2939         alignment being a poly int.
2940         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
2941         return type to poly_uint64.
2942         * target.def (default_preferred_vector_alignment): Likewise.
2943         * targhooks.c (default_preferred_vector_alignment): Likewise.
2944         * targhooks.h (default_preferred_vector_alignment): Likewise.
2945         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
2946         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
2947         poly int.
2948         (vect_update_misalignment_for_peel): Likewise.
2949         (vect_enhance_data_refs_alignment): Likewise.
2950         (vect_find_same_alignment_drs): Likewise.
2951         (vect_duplicate_ssa_name_ptr_info): Likewise.
2952         (vect_setup_realignment): Likewise.
2953         (vect_can_force_dr_alignment_p): Change alignment parameter type to
2954         poly_uint64.
2955         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
2956         mask with a compile time variable vector alignment.
2957         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
2958         int.
2959         (vect_do_peeling): Exit early if vector alignment is not constant.
2960         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
2961         a poly int.
2962         (vectorizable_store): Likewise.
2963         (vectorizable_load): Likweise.
2964         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
2965         poly_uint64.
2966         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
2967         poly int.
2968         (vect_can_force_dr_alignment_p): Change alignment parameter type to
2969         poly_uint64.
2971 2018-11-13  Richard Biener  <rguenther@suse.de>
2973         PR tree-optimization/87962
2974         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
2975         detect outer reduction for disqualifying in-loop uses.
2977 2018-11-13  Richard Biener  <rguenther@suse.de>
2979         PR tree-optimization/87967
2980         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
2981         for constants for the scalar loop.
2983 2018-11-13  Alan Modra  <amodra@gmail.com>
2985         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
2986         integer constants when -mcmodel=medium.
2988 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
2990         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
2991         * config/arc/elf.h (PROFILE_HOOK): Define.
2992         * config/arc/linux.h (PROFILE_HOOK): Likewise.
2994 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
2996         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
2997         jump table data in the text section.
2999 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3001         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
3002         to fit the eh_return pattern.
3003         * config/arc/arc.md (eh_return): Define.
3004         (VUNSPEC_ARC_EH_RETURN): Likewise.
3005         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
3006         new implementation.
3007         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
3010 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3012         * common/config/arc/arc-common.c (arc_option_optimization_table):
3013         Millicode optimization is default on for size optimizations.
3014         * config/arc/arc-protos.h (arc_check_multi): New function.
3015         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
3016         (ENTER_LEAVE_START_REG): Define.
3017         (ENTER_LEAVE_END_REG): Likewise.
3018         (arc_override_options): Disable millicode when long calls option
3019         is on.
3020         (arc_frame_info): Change it from int to bool.
3021         (arc_compute_frame_size): Clean up.
3022         (arc_save_restore): Remove.
3023         (frame_save_reg): New function.
3024         (frame_restore_reg): Likewise.
3025         (arc_enter_leave_p): Likewise.
3026         (arc_save_callee_saves): Likewise.
3027         (arc_restore_callee_saves): Likewise.
3028         (arc_save_callee_enter): Likewise.
3029         (arc_restore_callee_leave): Likewise.
3030         (arc_save_callee_milli): Likewise.
3031         (arc_restore_callee_milli): Likewise.
3032         (arc_expand_prologue): Reimplement to emit enter/leave
3033         instructions.
3034         (arc_expand_epilogue): Likewise.
3035         (arc_check_multi): New function.
3036         * config/arc/arc.md (push_multi_fp): New pattern.
3037         (push_multi_fp_blink): Likewise.
3038         (pop_multi_fp): Likewise.
3039         (pop_multi_fp_blink): Likewise.
3040         (pop_multi_fp_ret): Likewise.
3041         (pop_multi_fp_blink_ret): Likewise.
3042         * config/arc/arc.opt (mmillicode): Update option.
3043         (mcode-density-frame): New option.
3044         * config/arc/predicates.md (push_multi_operand): New predicate.
3045         (pop_multi_operand): Likewise.
3046         * doc/invoke.texi (ARC): Update ARC options information.
3049 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3051         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
3052         * config/arc/arc.c (operands_ok_ldd_std): New function.
3053         (mem_ok_for_ldd_std): Likewise.
3054         (gen_operands_ldd_std): Likewise.
3055         * config/arc/arc.md: Add peephole2 rules for std/ldd.
3057 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
3059         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
3060         test on stack_usage_file.
3061         (lang_dependent_init): Do not open the .su file if generating LTO.
3063 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
3065         PR rtl-optimization/87918
3066         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
3067         simplify_gen_relational rather than simplify_gen_binary.
3069 2018-11-13  Richard Biener  <rguenther@suse.de>
3071         * tree-ssanames.h (set_range_info): Use value_range_base.
3072         (get_range_info): Likewise.
3073         * tree-ssanames.c (set_range_info): Likewise.
3074         (get_range_info): Likewise.
3075         * tree-vrp.c (value_range_base::union_helper): Split
3076         out common parts of value_range[_base]::union_.
3077         (value_range_base::union_): Update.
3078         (value_range::union_): Likewise.
3079         (determine_value_range_1): Use value_range_base.
3080         (determine_value_range): Likewise.
3081         * tree-vrp.h (value_range_base::union_helper): Move ...
3082         (value_range::union_helper): ... from here.
3084 2018-11-13  Alan Modra  <amodra@gmail.com>
3086         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
3087         offset for PRE_DEC.
3088         (rs6000_secondary_reload_gpr): Don't call find_replacement.
3090 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
3092         PR middle-end/59634
3093         * doc/invoke.texi (Optimize Options): Clarify that the
3094         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
3095         apply to data cache size.
3097 2018-11-13  Alan Modra  <amodra@gmail.com>
3099         * config/rs6000/predicates.md (logical_const_operand),
3100         (logical_operand): Correct comment.
3101         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
3103 2018-11-13  Alan Modra  <amodra@gmail.com>
3105         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
3107 2018-11-13  Alan Modra  <amodra@gmail.com>
3109         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
3110         description.
3111         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
3112         covered by alternative.
3113         (movcc_internal1): Ignore h for register preference.
3114         (mov<mode>_hardfloat64): Likewise.
3115         (mov<mode>_softfloat): Ignore c, l, h for register preference.
3117 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
3119         PR preprocessor/47823
3120         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
3121         documentation to...
3122         (Pragmas): ...here.
3123         * doc/extend.texi (Pragmas): Note additional pragmas documented
3124         in the CPP manual.
3126 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
3128         PR middle-end/86677
3129         PR middle-end/87528
3130         * tree-scalar-evolution.c (expression_expensive_p): Make
3131         BUILTIN POPCOUNT as expensive when backend does not define it.
3133 2018-11-12  Fredrik Noring  <noring@nocrew.org>
3135         * config.gcc: Update with-llsc defaults for MIPS r5900.
3137 2018-11-12  Martin Liska  <mliska@suse.cz>
3139         PR gcov-profile/87442
3140         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
3141         options.
3142         * doc/invoke.texi: Document them.
3143         * tree-profile.c (parse_profile_filter): New.
3144         (parse_profile_file_filtering): Likewise.
3145         (release_profile_file_filtering): Likewise.
3146         (include_source_file_for_profile): Likewise.
3147         (tree_profiling): Filter source files based on the
3148         newly added options.
3150 2018-11-12  Martin Liska  <mliska@suse.cz>
3152         PR target/87903
3153         * doc/extend.texi: Add missing values for __builtin_cpu_is and
3154         __builtin_cpu_supports for x86 target.
3156 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
3158         PR middle-end/21110
3159         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
3160         have to be Pmode.
3161         (Arithmetic): Likewise for "lo_sum".
3163 2018-11-12  Renlin Li  <renlin.li@arm.com>
3165         PR target/87815
3166         * dse.c (get_stored_val): Add check for compile-time constantness
3167         of gap.
3169 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
3171         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
3172         (ARMv8_5a): New fgroup.
3173         (armv8.5-a): New arch.
3174         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
3175         sb and predres.
3176         * config/arm/arm-tables.opt: Regenerate.
3177         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
3178         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
3179         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
3180         Add matching rules for -march=armv8.5-a and extensions.
3181         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
3182         Add sb and predres to all armv8-a except armv8.5-a.
3184 2018-11-12  Richard Biener  <rguenther@suse.de>
3186         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
3187         overload for single value.
3188         (value_range[_base]::set_nonnull): New.
3189         (value_range[_base]::set_null): Likewise.
3190         (value_range): Document bitmap copying behavior, mark
3191         copy constructor and assignment operator deleted.
3192         (value_range::move): New.
3193         (value_range::set_and_canonicalize): Default bitmap to zero.
3194         (set_value_range_to_nonnull): Remove.
3195         (set_value_range_to_null): Likewise.
3196         (set_value_range): Likewise.
3197         (set_value_range_to_value): Likewise.
3198         (extract_range_from_unary_expr): Work on value_range_base.
3199         (extract_range_from_binary_expr_1): Likewise.  Rename to...
3200         (extract_range_from_binary_expr): ... this.
3201         * tree-vrp.c (value_range::update): Clear equiv bitmap
3202         if required.
3203         (value_range::move): New, move equiv bitmap.
3204         (value_range_base::set_undefined): Avoid assignment.
3205         (value_range::set_undefined): Likewise.
3206         (value_range_base::set_varying): Likewise.
3207         (value_range::set_varying): Likewise.
3208         (set_value_range): Remove.
3209         (value_range_base::set): New overload for value.
3210         (value_range::set): Likewise.
3211         (set_value_range_to_nonnull): Remove.
3212         (value_range_base::set_nonnull): New.
3213         (value_range::set_nonnull): Likewise.
3214         (set_value_range_to_null): Remove.
3215         (value_range_base::set_null): New.
3216         (value_range::set_null): Likewise.
3217         (range_is_null): Work on value_range_base.
3218         (range_is_nonnull): Likewise.
3219         (ranges_from_anti_range): Likewise.
3220         (extract_range_into_wide_ints): Likewise.
3221         (extract_range_from_multiplicative_op): Likewise.
3222         (extract_range_from_binary_expr): Likewise.  Update for API changes.
3223         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
3224         handling.
3225         (value_range::intersect_helper): Avoid copy and assignment.
3226         (value_range::union_helper): Likewise.
3227         (determine_value_range_1): Adjust.
3228         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
3229         Avoid assignment by using move.
3230         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
3231         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3232         Likewise.
3233         * tree-ssanames.c (get_range_info): Likewise.
3234         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
3235         * vr-values.c (vr_values::get_value_range): Adjust.
3236         (vr_values::update_value_range): Likewise.
3237         (symbolic_range_based_on_p): Work on value_range_base.
3238         (vr_values::extract_range_from_binary_expr): Use value_range_base.
3239         (vr_values::extract_range_from_unary_expr): Likewise.
3240         (vr_values::extract_range_from_cond_expr): Avoid assignment.
3241         (vr_values::extract_range_from_comparison): Adjust.
3242         (vr_values::check_for_binary_op_overflow): Use value_range_base.
3243         (vr_values::extract_range_basic): Adjust.
3244         (vr_values::adjust_range_with_scev): Likewise.
3245         (vr_values::vrp_visit_assignment_or_call): Likewise.
3246         (vr_values::get_vr_for_comparison): Change API to avoid
3247         assignment and copy construction.
3248         (vr_values::compare_name_with_value): Adjust accordingly.
3249         (vr_values::compare_names): Likewise.
3250         (vr_values::extract_range_from_phi_node): Avoid assignment and
3251         bogus in-place modify of equiv bitmap.
3252         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
3253         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
3254         for extract_range_from_unary_expr API change.
3255         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
3257 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
3259         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
3261 2018-11-12  Richard Biener  <rguenther@suse.de>
3263         * tree-vrp.h (value_range_base::symbolic_p,
3264         value_range_base::constant_p, value_range_base::zero_p,
3265         value_range_base::singleton_p): Move from value_range.
3266         (value_range::dump): Add.
3267         * gimple-ssa-evrp-analyze.c
3268         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
3269         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
3270         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3271         Use set_varying.
3272         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
3273         (value_range::constant_p): Likewise.
3274         (value_range::singleton_p): Likewise.
3275         (value_range_base::dump): Add.
3276         (set_value_range_to_undefined): Remove.
3277         (set_value_range_to_varying): Likewise.
3278         (range_int_cst_p): Take value_range_base argument.
3279         (range_int_cst_singleton_p): Likewise.
3280         (value_range_constant_singleton): Likewise.
3281         (vrp_set_zero_nonzero_bits): Likewise.
3282         (extract_range_from_multiplicative_op): Use set_varying.
3283         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
3284         (extract_range_from_unary_expr): Likewise.
3285         (dump_value_range_base): Change to overload of dump_value_range.
3286         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
3287         (vrp_prop::visit_stmt): Likewise.
3288         (value_range::intersect_helper): Likewise.
3289         (value_range::union_helper): Likewise.
3290         (determine_value_range_1): Likewise.
3292 2018-11-12  Richard Biener  <rguenther@suse.de>
3294         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
3295         (set_value_range_to_null): Likewise.
3296         * vr-values.c (vr_values::extract_range_from_comparison):
3297         Clear equiv for constant singleton ranges.
3299 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
3301         * config/i386/sse.md: Combine VFIXUPIMM* patterns
3302         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3303         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3304         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
3305         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3306         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3307         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
3309 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3311         PR c/69502
3312         * doc/extend.texi (Common Type Attributes): For the align type
3313         attribute, copy language about decreasing alignment from the
3314         corresponding variable attribute.
3316 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3318         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
3319         -frounding-math.
3321 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3323         PR c++/43105
3324         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
3325         -frtti and -fno-rtti code.
3327 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3329         PR c/26366
3330         * doc/extend.texi (Other Builtins): Document probability associated
3331         with __builtin_expect.
3333 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
3335         PR target/87928
3336         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
3337         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
3338         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
3339         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
3341 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3343         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
3345 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3347         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
3349 2018-11-11  Richard Biener  <rguenther@suse.de>
3351         * tree-vrp.h (class value_range_base): New base class for
3352         value_range containing all but the m_equiv member.
3353         (dump_value_range_base): Add.
3354         (range_includes_zero_p): Work on value_range_base.
3355         * tree-vrp.c (value_range_base::set): Split out base handling
3356         from...
3357         (value_range::set): this.
3358         (value_range::set_equiv): New.
3359         (value_range_base::value_range_base): New constructors.
3360         (value_range_base::check): Split out base handling from...
3361         (value_range::check): this.
3362         (value_range::equal_p): Refactor in terms of
3363         ignore_equivs_equal_p which is now member of the base.
3364         (value_range_base::set_undefined): New.
3365         (value_range_base::set_varying): Likewise.
3366         (value_range_base::dump):Split out base handling from...
3367         (value_range::dump): this.
3368         (value_range_base::set_and_canonicalize): Split out base handling
3369         from...
3370         (value_range::set_and_canonicalize): this.
3371         (value_range_base::union_): New.
3372         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
3373         for m_vr.
3374         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
3375         instead of value_range everywhere.
3376         (ipcp_vr_lattice::print): Use dump_value_range_base.
3377         (ipcp_vr_lattice::meet_with): Adjust.
3378         (ipcp_vr_lattice::meet_with_1): Likewise.
3379         (ipa_vr_operation_and_type_effects): Likewise.
3380         (propagate_vr_across_jump_function): Likewise.
3381         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
3382         (ipa_get_value_range): Likewise.
3383         (ipa_set_jfunc_vr): Likewise.
3384         (ipa_compute_jump_functions_for_edge): Likewise.
3386 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
3388         PR middle-end/65703
3389         * doc/invoke.texi (Optimize Options): Add @opindex entries
3390         for the positive forms of -fno-xxx and -mno-xxx options
3391         that were lacking them.
3393 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
3395         * combine.c (make_more_copies): Only make an intermediate copy if the
3396         dest of a move is a pseudo.
3398 2018-11-09  Maya Rashish  <coypu@sdf.org>
3400         PR target/87221
3401         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
3402         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
3404 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
3406         PR driver/41179
3407         PR middle-end/65703
3408         * doc/invoke.texi (Optimize Options): Clarify default behavior
3409         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
3411 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3413         PR tree-optimization/87940
3414         * expr.c (string_constant): Don't strip NOPS in subexpressions.
3415         Fold PLUS_EXPR correctly.
3417 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
3419         PR target/87762
3420         * config/s390/s390.c (s390_safe_relative_long_p): New function.
3421         (annotate_constant_pool_refs): Skip insns which support
3422         relative addressing.
3423         (annotate_constant_pool_refs_1): New helper function.
3424         (find_constant_pool_ref): Skip insns which support relative
3425         addression.
3426         (find_constant_pool_ref_1): New helper function.
3427         (replace_constant_pool_ref): Skip insns which support
3428         relative addressing.
3429         (replace_constant_pool_ref_1): New helper function.
3430         (s390_mainpool_start): Adapt to the new signature.
3431         (s390_mainpool_finish): Likewise.
3432         (s390_chunkify_start): Likewise.
3433         (s390_chunkify_finish): Likewise.
3434         (pass_s390_early_mach::execute): Likewise.
3435         (s390_prologue_plus_offset): Likewise.
3436         (s390_emit_prologue): Likewise.
3437         (s390_emit_epilogue): Likewise.
3439 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
3441         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
3442         but unsupported lastprivate with conditional modifier.
3444 2018-11-09  Jeff Law  <law@redhat.com>
3446         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
3447         unused argument better.  Add gcc_unreachable to silence warning.
3449 2018-11-09  Martin Sebor  <msebor@redhat.com>
3451         PR middle-end/81824
3452         * attribs.c (has_attribute): New helper function.
3453         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
3454         * attribs.h (decls_mismatched_attributes): Declare.
3455         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
3456         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
3457         * common.opt (-Wattribute-alias): Take an argument.
3458         (-Wno-attribute-alias): New option.
3459         * doc/extend.texi (Common Function Attributes): Document copy.
3460         (Common Variable Attributes): Same.
3461         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
3462         (-Wattribute-alias): Document new option argument.
3464 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
3466         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
3467         matching criteria.  Remove unused array initializer.
3469 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
3470             Jinsong Ji  <jji@us.ibm.com>
3472         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
3473         constraints by introducing a new temporary.
3474         (_mm_cvtss_si64): Likewise.
3476 2018-11-09  Martin Liska  <mliska@suse.cz>
3478         * common.opt: Add -fipa-stack-alignment flag.
3479         * doc/invoke.texi: Document it.
3480         * final.c (rest_of_clean_state): Guard stack
3481         shrinking with flag.
3483 2018-11-09  Martin Liska  <mliska@suse.cz>
3485         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
3486         to ...
3487         (ipa_discover_variable_flags): ... this.
3488         * common.opt: Come up with new flag -fipa-reference-addressable.
3489         * doc/invoke.texi: Document it.
3490         * ipa-reference.c (propagate): Call the renamed fn.
3491         * ipa-visibility.c (whole_program_function_and_variable_visibility):
3492         Likewise.
3493         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
3494         ...
3495         (ipa_discover_variable_flags): ... this.  Discover
3496         non-addressable variables only with the newly added flag.
3497         * opts.c: Enable the newly added flag with -O1 and higher
3498         optimization level.
3500 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
3502         * json.cc (selftest::test_writing_literals): Fix comment.
3504 2018-11-09  Martin Liska  <mliska@suse.cz>
3506         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
3507         string to a stack buffer.
3508         (aarch64_parse_cpu): Likewise.
3509         (aarch64_parse_tune): Likewise.
3511 2018-11-09  Richard Biener  <rguenther@suse.de>
3513         PR tree-optimization/87953
3514         * tree-vect-loop.c (vectorizable_reduction): For analysis
3515         always pass ops[0] to vectorizable_condition.
3517 2018-11-09  Stafford Horne  <shorne@gmail.com>
3518             Richard Henderson  <rth@twiddle.net>
3519             Joel Sherrill  <joel@rtems.org>
3521         * common/config/or1k/or1k-common.c: New file.
3522         * config/or1k/*: New.
3523         * config.gcc (or1k*-*-*): New.
3524         * configure.ac (or1k*-*-*): New test for openrisc tls.
3525         * configure: Regenerated.
3526         * doc/install.texi: Document OpenRISC triplets.
3527         * doc/invoke.texi: Document OpenRISC arguments.
3528         * doc/md.texi: Document OpenRISC.
3530 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
3532         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
3533         (arm7tdmi-s): Delete CPU.
3534         (arm710t): Add aliases for arm720t and arm740t.
3535         (arm720t, arm740t): Delete CPUs.
3536         (arm920t): Add aliases for arm920, arm922t and arm940t.
3537         (arm920, arm922t, arm940t): Delete CPUs.
3538         (arm10tdmi): Add alias for arm1020t.
3539         (arm1020t): Delete CPU.
3540         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
3541         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
3542         (arm10e): Add aliases for arm1020e and arm1022e.
3543         (arm1020e, arm1022e): Delete CPU.
3544         * config/arm/arm.md (generic_sched): Remove entries that are now
3545         handled by aliases.
3546         (generic_vfp): Likewise.
3547         * config/arm/arm1020e.md: Simplify tuning selection based on alias
3548         changes.
3549         * config/arm/arm-tune.md: Regenerated.
3550         * config/arm/arm-tables.opt: Regenerated.
3552 2018-11-09  Richard Biener  <rguenther@suse.de>
3554         PR tree-optimization/87621
3555         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
3556         op with only phi inputs.
3557         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
3558         (ch_base::copy_headers): Run CSE on copied loop headers.
3559         (pass_ch_vect::process_loop_p): Simplify.
3561 2018-11-09  Alexandre Oliva <oliva@adacore.com>
3563         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
3564         for non-w64 x86_64 biarch.
3566 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
3568         PR rtl-optimization/86438
3569         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
3570         of in_b for the compare if in_b is SET_DEST.
3572         PR target/87793
3573         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
3574         non-toplevel UNSPEC.
3576 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3578         * tree-vrp.c (value_range::check): Do not access internals
3579         directly.
3580         (value_range::singleton_p): Same.
3581         (value_range::type): Same.
3582         (vrp_finalize): Use value_range API.
3584 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3586         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
3588 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3590         * vr-values.c (vr_values::get_value_range): Use value_range API
3591         instead of piecing together ranges.
3592         (vr_values::update_value_range): Same.
3594 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3596         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
3597         of performing ad-hoc calculations.
3598         * tree-ssanames.c (set_range_info): New overloaded function
3599         accepting value_range &.
3600         (get_range_info): Same.
3601         * tree-ssanames.h (set_range_info_raw): Remove.
3602         (set_range_info): New prototype.
3603         (get_range_info): Same.
3604         * tree-vrp.h (value_range::null_p): Rename to zero_p.
3605         * tree-vrp.c (value_range::null_p): Same.
3607 2018-11-09  Jan Hubicka  <jh@suse.cz>
3609         * tree.c (fld_type_variant_equal_p): Test user align flag.
3610         (flt_type_variant): Copy user align flag.
3611         (fld_incomplete_type_of): Clear it.
3613 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3615         * config/arm/neon.md (div<mode>3): New pattern.
3617 2018-11-08  Andi Kleen  <ak@linux.intel.com>
3619         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
3620         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
3621         (ix86_handle_option): Handle OPT_mptwrite.
3622         * config/i386/cpuid.h (bit_PTWRITE): Add.
3623         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
3624         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
3625         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
3626         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
3627         (ix86_option_override_internal): Handle PTA_PTWRITE.
3628         (ix86_valid_target_attribute_inner_p): Define ptwrite.
3629         (def_builtin2): Force UINT64 to be 64bit only.
3630         * config/i386/i386.h (TARGET_PTWRITE): Add.
3631         (TARGET_PTWRITE_P): Add.
3632         (PTA_PTWRITE): Add.
3633         * config/i386/i386.md: Define ptwrite.
3634         * config/i386/i386.opt: Add -mptwrite.
3635         * config/i386/immintrin.h (_ptwrite64): Add.
3636         (_ptwrite32): Add
3637         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
3638         * doc/invoke.texi: Document -mptwrite.
3640 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
3642         PR rtl-optimization/87600
3643         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
3644         * lra-constraints.c (process_alt_operands): Skip illegal hard
3645         register usage.  Prefer reloading non hard register operands.
3647 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
3649         PR other/36572
3650         * doc/invoke.texi (Optimize Options): Clarify default behavior
3651         for -fno-sched-interblock and -fno-sched-spec.
3653 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
3655         * collect2.c (linker_select):  Add USE_LLD_LD.
3656         (ld_suffixes): Add ld.lld.
3657         (main): Handle -fuse-ld=lld.
3658         * common.opt (-fuse-ld=lld): New option.
3659         * doc/invoke.texi (-fuse-ld=lld): Document.
3660         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
3662 2018-11-08  Paul Koning  <ni1d@arrl.net>
3664         * config/pdp11/constraints.md: Add "Z" series constraints for use
3665         with pre-dec and post-inc addressing.
3666         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
3667         (pdp11_expand_operands): Add int argument (word count).
3668         (pdp11_sp_frame_offset): Delete.
3669         (pdp11_cmp_length): New function.
3670         (pushpop_regeq): New function.
3671         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
3672         Add hook.
3673         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
3674         frame layout.
3675         (pdp11_initial_elimination_offset): Ditto.
3676         (pdp11_expand_operands): Add word count argument.  Bugfixes.
3677         (output_move_multiple): Change how pointer adjustment is done.
3678         (pdp11_gen_int_label): Correct format.
3679         (output_ascii): Ditto.
3680         (pdp11_asm_output_var): Add code for DEC assembler case.
3681         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
3682         value.
3683         (legitimate_const_double_p): Ditto.
3684         (pdp11_register_move_cost): Adjust for new register classes.
3685         (pdp11_regno_reg_class): Ditto.
3686         (expand_block_move): Delete.
3687         (pushpop_regeq): New function.
3688         (pdp11_legitimate_address_p): Bugfix in check for constant
3689         offset.
3690         (pdp11_sp_frame_offset): Delete.
3691         (pdp11_reg_save_size): New helper function for new frame layout.
3692         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
3693         (pdp11_expand_shift): Bugfix in check for constant shift count.
3694         (pdp11_shift_length): Ditto.
3695         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
3696         (pdp11_cmp_length): New function.
3697         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
3698         some compile options.
3699         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
3700         (CALL_USED_REGISTERS): Ditto.
3701         (ELIMINABLE_REGS): Ditto.
3702         (REGISTER_NAMES): Ditto.
3703         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
3704         constraints.
3705         (REG_CLASS_NAMES): Ditto.
3706         (REG_CLASS_CONTENTS): Ditto.  Also remove
3707         HARD_FRAME_POINTER_REGNUM.
3708         (CPU_REG_CLASS): New macro.
3709         (CLASS_MAX_NREGS): Adjust for new register classes.
3710         (FUNCTION_PROFILER): Make no-op.
3711         (may_call_alloca): Remove unused declaration.
3712         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
3713         (ASM_OUTPUT_SKIP): Fix format.
3714         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
3715         (HARD_FRAME_POINTER_REGNUM): Remove.
3716         (return): Delete.
3717         (*rts): Rename.  Remove epilogue related checks.
3718         (cmpsi, cmpdi): New insn.
3719         (cbranch<mode>4): Change to apply to SI and DI modes as well.
3720         (mov<mode>): Change constraints to enforce that push/pop
3721         destination cannot use the same register as source.
3722         (*mov<mode><cc_cc>): Ditto.
3723         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
3724         at assembly output rather than as RTL expander.
3725         (zero_extendqihi2): Bugfix in check for same registers.
3726         (adddi3_nocc): Bugfix in check for constant operand.
3727         (addsi3_nocc): Ditto.
3728         (subdi3_nocc): Ditto.
3729         (subsi3_nocc): Ditto.
3730         (negdi2_nocc): Copy input to pdp11_expand_operands.
3731         (negsi2_nocc): Ditto.
3732         (bswap2_nocc): Ditto.
3733         * config/pdp11/pdp11.opt (mlra): Fix documentation.
3734         * config/pdp11/t-pdp11: Use -Os.
3736 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
3738         * config/arm/parsecpu.awk (/alias/): New parsing rule.
3739         (/begin cpu/): Check that the cpu name hasn't been previously defined.
3740         (gen_comm_data): Print out CPU alias tables.
3741         (check_cpu): Match aliases when checking the CPU name.
3742         * config/arm/arm-protos.h (cpu_alias): New structure.
3743         (cpu_option): Add entry for aliases.
3744         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
3745         strongarm1100 and strongarm1110.
3746         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
3747         (config/arm/arm-generic.md): Remove redundant references to
3748         strongarm110, strongarm1100 and strongarm1110.
3749         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3750         Scan aliases for additional hints.
3751         (arm_parse_cpu_option_name): Also match a cpu name against the list
3752         of aliases.
3753         * config/arm/arm-tables.opt: Regenerated.
3754         * config/arm/arm-tune.md: Regenerated.
3756 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
3758         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
3759         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
3760         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
3761         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
3762         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
3763         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
3764         * gengtype.c (open_base_files): Add omp-general.h.
3765         * gimple.c (gimple_build_omp_critical):
3766         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
3767         gimple_omp_taskgroup_set_clauses.
3768         (gimple_build_omp_atomic_load): Add mo argument, call
3769         gimple_omp_atomic_set_memory_order.
3770         (gimple_build_omp_atomic_store): Likewise.
3771         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
3772         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
3773         instead of GSS_OMP.
3774         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
3775         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
3776         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
3777         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
3778         different value for GF_OMP_ATOMIC_NEED_VALUE.
3779         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
3780         comments.
3781         (struct gimple_statement_omp_single_layout): And remove here.
3782         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
3783         than gimple_statement_omp_single_layout.
3784         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
3785         GIMPLE_OMP_TEAMS.
3786         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
3787         (gimple_omp_subcode): Formatting fix.
3788         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
3789         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
3790         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
3791         gimple_omp_teams_host, gimple_omp_teams_set_host,
3792         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
3793         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
3794         gimple_omp_taskgroup_set_clauses): New inline functions.
3795         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
3796         (gimple_build_omp_atomic_store): Likewise.
3797         (gimple_omp_atomic_seq_cst_p): Remove.
3798         (gimple_omp_atomic_memory_order): New function.
3799         (gimple_omp_atomic_set_seq_cst): Remove.
3800         (gimple_omp_atomic_set_memory_order): New function.
3801         (gimple_build_omp_taskgroup): Add clauses argument.
3802         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
3803         (dump_gimple_omp_task): Print taskwait with depend clauses.
3804         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
3805         dump_omp_atomic_memory_order.
3806         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
3807         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
3808         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
3809         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
3810         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
3811         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
3812         ORT_UNTIED_TASKLOOP enumerators.
3813         (enum gimplify_defaultmap_kind): New.
3814         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
3815         target_map_pointers_as_0len_arrays members, add defaultmap.
3816         (new_omp_context): Initialize defaultmap member.
3817         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3818         (maybe_fold_stmt): Don't fold even in host teams regions.
3819         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
3820         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
3821         ctx->omp_firstprivatize_variable.
3822         (omp_add_variable): Don't add private/firstprivate for VLAs in
3823         ORT_TASKGROUP.
3824         (omp_default_clause): Print "taskloop" rather than "task" if
3825         ORT_*TASKLOOP.
3826         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3827         Handle new defaultmap clause kinds.
3828         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
3829         iterator to be lastprivate or private.  Fix up diagnostics if linear
3830         is used on collapse>1 simd iterator.
3831         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3832         (gimplify_omp_depend): New function.
3833         (gimplify_scan_omp_clauses): Add shared clause on parallel for
3834         combined parallel master taskloop{, simd} if taskloop has
3835         firstprivate, lastprivate or reduction clause.  Handle
3836         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
3837         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
3838         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
3839         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
3840         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
3841         lastprivate.
3842         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
3843         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
3844         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
3845         OMP_CLAUSE_{TASK,IN}_REDUCTION.
3846         (gimplify_omp_task): Handle taskwait with depend clauses.
3847         (gimplify_omp_for): Add shared clause on parallel for combined
3848         parallel master taskloop{, simd} if taskloop has firstprivate,
3849         lastprivate or reduction clause.  Use ORT_TASKLOOP or
3850         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
3851         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
3852         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
3853         __for_end and __for_range temporaries on OMP_PARALLEL for
3854         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
3855         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
3856         sandwiched in between two taskloops.
3857         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
3858         instead of ctx->omp_firstprivatize_variable.
3859         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
3860         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
3861         host teams, use gimplify_and_return_first etc. for body like
3862         for target or target data constructs, and at the end call
3863         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
3864         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
3865         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
3866         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
3867         gimple_omp_atomic_set_seq_cst calls.
3868         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
3869         case, handle taskgroup clauses.
3870         * lto-streamer-out.c (hash_tree): Handle
3871         OMP_CLAUSE_{TASK,IN}_REDUCTION.
3872         * Makefile.in (GTFILES): Add omp-general.h.
3873         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
3874         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
3875         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
3876         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
3877         BUILT_IN_GOMP_LOOP_DOACROSS_START,
3878         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
3879         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
3880         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
3881         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
3882         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
3883         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
3884         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
3885         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
3886         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
3887         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
3888         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
3889         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
3890         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
3891         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
3892         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
3893         * omp-expand.c (workshare_safe_to_combine_p): Return false for
3894         non-worksharing loops.
3895         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
3896         (determine_parallel_type): Don't combine parallel with worksharing
3897         which has _reductemp_ clause.
3898         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
3899         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
3900         if there is nonmonotonic modifier or if there is no modifier and no
3901         ordered clause.  For dynamic and guided schedule without monotonic
3902         and nonmonotonic modifier, default to nonmonotonic.
3903         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
3904         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
3905         task reductions.
3906         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
3907         there are any reduction clauses.
3908         (expand_taskwait_call): New function.
3909         (expand_teams_call): New function.
3910         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
3911         expand_teams_call for it.  Formatting fix.  Handle taskwait with
3912         depend clauses.
3913         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
3914         of worksharing loops with task reductions.
3915         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
3916         expansion of worksharing loops with task reductions.
3917         (expand_omp_sections): Handle expansion of sections with task
3918         reductions.
3919         (expand_omp_synch): For host teams call expand_omp_taskreg.
3920         (omp_memory_order_to_memmodel): New function.
3921         (expand_omp_atomic_load, expand_omp_atomic_store,
3922         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
3923         instead of gimple_omp_atomic_seq_cst_p.
3924         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
3925         depend clauses as a standalone directive.
3926         * omp-general.c (enum omp_requires): New variable.
3927         (omp_extract_for_data): Initialize have_reductemp member.  Allow
3928         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
3929         GT_EXPR loops depending on incr sign.  Formatting fixes.
3930         * omp-general.h (struct omp_for_data): Add have_reductemp member.
3931         (enum omp_requires): New enum.
3932         (omp_requires_mask): Declare.
3933         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
3934         Fix comment typos.
3935         * omp-low.c (struct omp_context): Add task_reductions and
3936         task_reduction_map fields.
3937         (is_host_teams_ctx): New function.
3938         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
3939         (use_pointer_for_field): Use is_global_var instead of
3940         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
3941         in outer contexts.
3942         (build_outer_var_ref): Ignore taskgroup outer contexts.
3943         (delete_omp_context): Release task_reductions and task_reduction_map.
3944         (scan_sharing_clauses): Don't add any fields for reduction clause on
3945         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
3946         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
3947         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
3948         Handle OMP_CLAUSE_NONTEMPORAL.
3949         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
3950         needed.
3951         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
3952         clauses with task modifier.
3953         (scan_omp_task): Handle taskwait with depend clauses.
3954         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
3955         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
3956         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
3957         (scan_omp_for): Fix comment formatting.
3958         (scan_omp_teams): Handle host teams constructs.
3959         (check_omp_nesting_restrictions): Allow teams with no outer
3960         OpenMP context.  Adjust diagnostics for teams strictly nested into
3961         some explicit OpenMP construct other than target.  Allow OpenMP atomics
3962         inside of simd regions.
3963         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
3964         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
3965         taskreg_nesting_level while scanning host teams construct.
3966         (task_reduction_read): New function.
3967         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
3968         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
3969         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
3970         second argument create_tmp_var if it is NULL.  Don't ignore shared
3971         clauses in is_host_teams_ctx contexts.  Handle
3972         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
3973         clauses.
3974         (lower_reduction_clauses): Ignore reduction clauses with task
3975         modifier.  Remove second argument create_tmp_var if it is NULL.
3976         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
3977         (lower_send_clauses): Ignore reduction clauses with task modifier.
3978         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
3979         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
3980         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
3981         rely that it is the last stmt in body so far.  Ignore outer taskgroup
3982         contexts.
3983         (omp_task_reductions_find_first, omp_task_reduction_iterate,
3984         lower_omp_task_reductions): New functions.
3985         (lower_omp_sections): Handle reduction clauses with taskgroup
3986         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
3987         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
3988         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
3989         modifiers.
3990         (lower_omp_taskgroup): Handle taskgroup reductions.
3991         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
3992         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
3993         (lower_depend_clauses): If there are any
3994         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
3995         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
3996         seen, assume lowering is done already and return early.  Set kind
3997         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
3998         (lower_omp_taskreg): Handle reduction clauses with task modifier on
3999         parallel construct.  Handle reduction clause on taskloop construct.
4000         Handle taskwait with depend clauses.
4001         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
4002         for host teams constructs.
4003         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
4004         nontemporal and _reductemp_ clause entries.
4005         (omp_clause_code_name): Likewise.
4006         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
4007         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
4008         * tree-core.h (enum omp_clause_code): Add
4009         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
4010         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
4011         (struct tree_base): Add omp_atomic_memory_order field into union.
4012         Remove OMP_ATOMIC_SEQ_CST comment.
4013         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
4014         and OMP_CLAUSE_DEPEND_DEPOBJ.
4015         (struct tree_omp_clause): Add subcode.defaultmap_kind.
4016         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
4017         OpenMP constructs with body and clauses operands.
4018         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
4019         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
4020         (OMP_TASKGROUP_CLAUSES): Define.
4021         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
4022         OMP_CLAUSE__LOOPTEMP_.
4023         (OMP_ATOMIC_SEQ_CST): Remove.
4024         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
4025         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
4026         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
4027         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
4028         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
4029         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
4030         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
4031         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
4032         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
4033         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
4034         Define.
4035         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
4036         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4037         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
4038         (convert_local_omp_clauses): Likewise.  Remove useless test.
4039         * tree-parloops.c (create_call_for_reduction_1): Pass
4040         OMP_MEMORY_ORDER_RELAXED as new argument to
4041         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
4042         * tree-pretty-print.c (dump_omp_iterators): New function.
4043         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
4044         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
4045         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
4046         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
4047         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
4048         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
4049         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
4050         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
4051         (dump_omp_atomic_memory_order): New function.
4052         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
4053         taskwait with depend clauses.
4054         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
4055         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
4056         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
4057         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
4058         write_ts_omp_clause_tree_pointers): Likewise.
4060 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4062         PR ipa/86395
4063         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
4064         "-missed", "-note", and "-all" sub-options.
4065         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
4066         API.
4067         (can_early_inline_edge_p): Likewise.
4068         (want_early_inline_function_p): Likewise.
4069         (want_inline_self_recursive_call_p): Likewise.
4070         (recursive_inlining): Likewise.
4071         (inline_small_functions): Likewise.
4072         (flatten_function): Likewise.
4073         (ipa_inline): Likewise.
4074         (inline_always_inline_functions): Likewise.
4075         (early_inline_small_functions): Likewise.
4076         (early_inliner): Likewise.
4077         * tree-inline.c (expand_call_inline): Likewise.
4079 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4081         * pretty-print.c (pp_format): Handle %f.
4082         (selftest::test_pp_format): Add test of %f.
4083         * pretty-print.h (pp_double): New macro.
4085 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4087         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
4088         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
4089         (dump_pretty_printer::decode_format): Implement "%C" for
4090         cgraph_node *.
4091         (selftest::test_capture_of_dump_calls): Rename "where" to
4092         "stmt_loc".  Convert test_decl to a function decl and set its
4093         location.  Add a symbol_table_test RAII instance and a
4094         cgraph_node, using it to test "%C" and dump_symtab_node.
4096 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
4098         PR middle-end/87916
4099         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
4101 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4103         * cgraph.c: Include "selftest.h".
4104         (saved_symtab): New variable.
4105         (selftest::symbol_table_test::symbol_table_test): New ctor.
4106         (selftest::symbol_table_test::~symbol_table_test): New dtor.
4107         (selftest::test_symbol_table_test): New test.
4108         (selftest::cgraph_c_tests): New.
4109         * cgraph.h (saved_symtab): New decl.
4110         (selftest::symbol_table_test): New class.
4111         * selftest-run-tests.c (selftest::run_tests): Call
4112         selftest::cgraph_c_tests.
4113         * selftest.h (selftest::cgraph_c_tests): New decl.
4115 2018-11-08  Richard Biener  <rguenther@suse.de>
4117         * tree-data-ref.h (lambda_int): New typedef.
4118         (lambda_vector_gcd): Adjust.
4119         (lambda_vector_new): Likewise.
4120         (lambda_matrix_new): Likewise.
4121         * tree-data-ref.c  (print_lambda_vector): Adjust.
4123 2018-11-08  Richard Biener  <rguenther@suse.de>
4125         PR tree-optimization/87929
4126         * tree-complex.c (expand_complex_comparison): Clean EH.
4128 2018-11-08  Martin Liska  <mliska@suse.cz>
4130         * doc/extend.texi: Reword.
4131         * predict.c (expr_expected_value_1): Likewise.
4133 2018-11-08  Richard Biener  <rguenther@suse.de>
4135         PR tree-optimization/87913
4136         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
4137         of extreme values to ordered comparisons.
4139 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4141         PR middle-end/42726
4142         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
4144 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4146         * doc/invoke.texi: Remove leading dash from @opindex entries
4147         throughout the file.
4149 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4151         PR driver/80828
4152         * doc/invoke.texi (Option Summary): Add -e and --entry.
4153         (Link Options): Likewise.
4155 2018-11-07  Nathan Sidwell  <nathan@acm.org>
4157         PR 87926
4158         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
4159         --disable-checking bootstrap.
4161 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
4163         * configure: Regenerated.
4165 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4167         PR c/87691
4168         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
4169         to the mode of the widest field iff the widest field has mode class
4170         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
4171         reference.
4173 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
4175         * common.opt: Add -Wattribute-warning.
4176         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
4177         * expr.c (expand_expr_real_1): Add new attribute to warning_at
4178         call to allow user configure behavior of "warning" attribute.
4180 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
4182         * target.def: Put @: after every vs., e.g., and i.e. where it is
4183         followed by whitespace.
4184         * doc/extend.texi: Ditto.
4185         * doc/fragments.texi: Ditto.
4186         * doc/gimple.texi: Ditto.
4187         * doc/implement-c.texi: Ditto.
4188         * doc/install.texi: Ditto.
4189         * doc/invoke.texi: Ditto.
4190         * doc/md.texi: Ditto.
4191         * doc/plugins.texi: Ditto.
4192         * doc/rtl.texi: Ditto.
4193         * doc/sourcebuild.texi: Ditto.
4194         * doc/tm.texi.in: Ditto.
4195         * doc/ux.texi: Ditto.
4196         * doc/tm.texi: Regenerate.
4198 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4200         * config/arm/arm-cpus.in (ares): New entry.
4201         * config/arm/arm-tables.opt: Regenerate.
4202         * config/arm/arm-tune.md: Likewise.
4203         * doc/invoke.texi (ARM Options): Document ares.
4205 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4207         * config/aarch64/aarch64-cores.def (ares): Define.
4208         * config/aarch64/aarch64-tune.md: Regenerate.
4209         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
4211 2018-11-07  Jan Hubicka  <jh@suse.cz>
4213         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
4214         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
4215         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
4216         TYPE_DECL.
4218 2018-11-07  Richard Biener  <rguenther@suse.de>
4220         PR tree-optimization/87914
4221         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
4222         of nested cycles.
4223         (vectorizable_reduction): Handle shifts and rotates by dispatching
4224         to vectorizable_shift.
4225         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
4226         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
4227         def cases.
4228         (vectorizable_shift): Export and handle being called as
4229         vect_nested_cycle.
4230         (vect_analyze_stmt): Call vectorizable_shift after
4231         vectorizable_reduction.
4232         * tree-vectorizer.h (vectorizable_shift): Declare.
4234 2018-11-07  Jan Hubicka  <jh@suse.cz>
4236         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
4237         than const decls in TREE_VALUE of enum.
4238         (dump_type_inheritance_graph): Improve duplicate dumping.
4239         (free_enum_values): New.
4240         (build_type_inheritance_graph): Use it.
4241         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
4242         which are not main variants or not ODR types.
4243         (verify_type_variant): Expect variants to have no TYPE_VALUES.
4245 2018-11-07  Richard Biener  <rguenther@suse.de>
4247         * ipa-inline.c (want_inline_small_function_p): Compute
4248         big_speedup_p lazily and last.
4250 2018-11-07  Jan Hubicka  <jh@suse.cz>
4252         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
4253         building incomplete variant of complete type.
4254         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
4255         variant of complete type.
4257 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4259         * config/mips/mips.c: Fix typo in documentation of
4260         mips_loongson_ext2_prefetch_cookie.
4261         (mips_option_override): fix brain twister logical.
4262         * config/mips/mips.h: Fix typo in documentation of
4263         ISA_HAS_CTZ_CTO and define pattern.
4264         * config/mips/mips.md (prefetch): Hoist EXT2 above
4265         the 2EF/EXT block.
4266         (prefetch_indexed): Hoist EXT2 above the EXT block.
4268 2018-11-07  Jan Hubicka  <jh@suse.cz>
4270         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
4271         return and parameter types of function and method types.
4272         (free_lang_data_in_cgraph): Update.
4274 2018-11-07  Martin Liska  <mliska@suse.cz>
4276         PR rtl-optimization/87868
4277         * postreload-gcse.c (eliminate_partially_redundant_load): Set
4278         threshold to max_count if we would overflow.
4279         * profile-count.h: Make max_count a public constant.
4281 2018-11-07  Martin Liska  <mliska@suse.cz>
4283         * mem-stats.h: Fix GNU coding style.
4285 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4287         * config/mips/gs264e.md: New.
4288         * config/mips/mips-cpus.def: Define gs264e.
4289         * config/mips/mips-tables.opt: Regenerate.
4290         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
4291         gs264e.
4292         (mips_issue_rate): Add support for gs264e.
4293         (mips_multipass_dfa_lookahead): Likewise.
4294         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
4295         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
4296         (MIPS_ASE_MSA_SPEC): New.
4297         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
4298         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
4299         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
4300         * config/mips/mips.md: Include gs264e.md.
4301         (processor): Add gs264e.
4302         * config/mips/mips.opt (MSA): Use Mask instead of Var.
4303         * doc/invoke.texi: Add gs264e to supported architectures.
4305 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4307         * config/mips/gs464e.md: New.
4308         * config/mips/mips-cpus.def: Define gs464e.
4309         * config/mips/mips-tables.opt: Regenerate.
4310         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
4311         gs464e.
4312         (mips_issue_rate): Add support for gs464e.
4313         (mips_multipass_dfa_lookahead): Likewise.
4314         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
4315         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
4316         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
4317         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
4318         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
4319         * config/mips/mips.md: Include gs464e.md.
4320         (processor): Add gs464e.
4321         * doc/invoke.texi: Add gs464e to supported architectures.
4323 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4325         * config/mips/loongson3a.md: Rename to ...
4326         * config/mips/gs464.md: ... here.
4327         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
4328         as an alias of gs464 processor.
4329         * config/mips/mips-tables.opt: Regenerate.
4330         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
4331         instead of PROCESSOR_LOONGSON_3A.
4332         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
4333         TUNE_LOONGSON_3A.
4334         (mips_option_override): Enable MMI and EXT for gs464.
4335         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
4336         Rename TUNE_LOONGSON_3A to TUNE_GS464.
4337         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
4338         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
4339         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
4340         TARGET_LOONGSON_3A.
4341         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
4342         (processor): Add gs464;
4343         * doc/invoke.texi: Add gs464 to supported architectures.
4345 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4347         * config/mips/mips-protos.h
4348         (mips_loongson_ext2_prefetch_cookie): New prototype.
4349         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
4350         (mips_option_override): Enable TARGET_LOONGSON_EXT when
4351         TARGET_LOONGSON_EXT2 is true.
4352         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
4353         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
4354         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
4355         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
4356         TARGET_LOONGSON_EXT2.
4357         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
4358         (define_insn "ctz<mode>2"): New insn pattern.
4359         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
4360         (define_insn "prefetch_indexed_<mode>"): Include
4361         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
4362         * config/mips/mips.opt (-mloongson-ext2): Add option.
4363         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
4365 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4367         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
4368         __mips_loongson_ext.
4369         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
4370         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
4371         -mloongson-ext.
4372         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
4373         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
4374         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
4375         instead of TARGET_LOONGSON_3A.
4376         * config/mips/mips.opt (-mloongson-ext): Add option.
4377         * gcc/doc/invoke.texi (-mloongson-ext): Document.
4379 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4381         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
4382         * config/mips/loongson.md: Move to ...
4383         * config/mips/loongson-mmi.md: here; Adjustment.
4384         * config/mips/loongson.h: Move to ...
4385         State as deprecated. Include loongson-mmiintrin.h for back
4386         compatibility and warning.
4387         * config/mips/loongson-mmiintrin.h: ... here.
4388         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
4389         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
4390         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
4391         (mips_option_override): Make sure MMI use hard float;
4392         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
4393         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
4394         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
4395         TARGET_LOONGSON_VECTORS.
4396         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
4397         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
4398         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
4399         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
4400         -mloongson-mmi.
4401         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
4402         TARGET_LOONGSON_VECTORS.
4403         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
4404         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
4405         (Loongson MMI patterns): Include loongson-mmi.md instead of
4406         loongson.md.
4407         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
4408         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
4410 2018-11-07  Richard Biener  <rguenther@suse.de>
4412         PR lto/87906
4413         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
4414         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
4416 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
4418         PR rtl-optimization/87874
4419         * lra.c (lra_substitute_pseudo): Do not create a subreg for
4420         const wide ints.
4422 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
4424         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
4425         if not in indexed or indirect form.
4426         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
4427         (bswap<mode>2_store): Ditto.
4429 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
4431         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
4432         the UNSPEC.
4434 2018-11-06  Richard Biener  <rguenther@suse.de>
4436         PR tree-optimization/86850
4437         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
4438         instead of src.m_vec.
4440 2018-11-06  Jan Hubicka  <jh@suse.cz>
4442         * tree.c (fld_simplified_type_name): Break out form ...
4443         (free_lang_data_in_type): ... here.
4444         (fld_type_variant_equal_p): Use it.
4446 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4448         * config/default-d.c: Include memmodel.h.
4450         * config/sol2-d.c: New file.
4451         * config/t-sol2 (sol2-d.o): New rule.
4452         * config.gcc <*-*-solaris2*>: Set d_target_objs,
4453         target_has_targetdm.
4455 2018-11-06  Jan Hubicka  <jh@suse.cz>
4457         * tree.c (fld_type_variant): Also copy alignment; be sure that
4458         new variant is equal.
4460 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4462         PR target/87762
4463         * config/s390/s390.md: Add relative_long attribute.
4465 2018-11-06  Jan Hubicka  <jh@suse.cz>
4467         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
4468         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
4469         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
4470         stream TYPE_NEEDS_CONSTRUCTING.
4471         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
4472         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
4474 2018-11-06  Richard Biener  <rguenther@suse.de>
4476         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
4477         dump-scope ...
4478         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
4480 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4482         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
4483         REG_CLASS_CONTENTS[GEN_REGS].
4484         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
4486 2018-11-06  Jan Hubicka  <jh@suse.cz>
4488          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
4490 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4492         PR sanitizer/80953
4493         * config/sol2.h (ASAN_CC1_SPEC): Define.
4494         (LD_WHOLE_ARCHIVE_OPTION): Define.
4495         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
4496         (ASAN_REJECT_SPEC): Provide default.
4497         (LIBASAN_EARLY_SPEC): Define.
4498         (LIBTSAN_EARLY_SPEC): Define.
4499         (LIBLSAN_EARLY_SPEC): Define.
4500         * config/i386/sol2.h (CC1_SPEC): Redefine.
4501         (ASAN_REJECT_SPEC): Define.
4503         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
4504         (TARGET_ASAN_SHADOW_OFFSET): Define.
4505         (sparc_asan_shadow_offset): New function.
4506         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
4507         (ASAN_REJECT_SPEC): Define.
4509 2018-11-06  Jan Hubicka  <jh@suse.cz>
4511         * tree.c (fld_type_variant): Copy canonical type.
4512         (fld_incomplete_type_of): Check that canonical types looks sane;
4513         copy canonical type.
4514         (verify_type): Accept when incomplete type has complete canonical type.
4516 2018-11-06  Jan Hubicka  <jh@suse.cz>
4518         * tree.c (free_lang_data): Reset overwite_assembler_name,
4519         print_xnode, print_decl, print_type and print_identifier of
4520         langhooks.
4522 2018-11-06  Richard Biener  <rguenther@suse.de>
4524         PR tree-optimization/87889
4525         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
4526         Do nothing if old and new arg are the same
4528 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
4530         PR target/87723
4531         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
4532         attributes for operands 3 and 4.
4534 2018-11-06  Richard Biener  <rguenther@suse.de>
4536         PR middle-end/18041
4537         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
4538         matching bitfield insertion.
4540 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
4542         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
4543         (reg_next_debug_use): New.
4544         (attempt_change): Propagate adjusted expression into affected
4545         debug insns.
4546         (merge_in_block): Track uses in debug insns.
4547         (pass_inc_dec::execute): Allocate and release
4548         reg_next_debug_use.
4550 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
4552         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
4553         (_mm512_fixupimm_round_pd): Update parameters and builtin.
4554         (_mm512_maskz_fixupimm_round_pd): Ditto.
4555         (_mm512_fixupimm_round_ps): Ditto.
4556         (_mm512_maskz_fixupimm_round_ps): Ditto.
4557         (_mm_fixupimm_round_sd): Ditto.
4558         (_mm_maskz_fixupimm_round_sd): Ditto.
4559         (_mm_fixupimm_round_ss): Ditto.
4560         (_mm_maskz_fixupimm_round_ss): Ditto.
4561         (_mm512_fixupimm_pd): Ditto.
4562         (_mm512_maskz_fixupimm_pd): Ditto.
4563         (_mm512_fixupimm_ps): Ditto.
4564         (_mm512_maskz_fixupimm_ps): Ditto.
4565         (_mm_fixupimm_sd): Ditto.
4566         (_mm_maskz_fixupimm_sd): Ditto.
4567         (_mm_fixupimm_ss): Ditto.
4568         (_mm_maskz_fixupimm_ss): Ditto.
4569         (_mm512_mask_fixupimm_round_pd): Update builtin.
4570         (_mm512_mask_fixupimm_round_ps): Ditto.
4571         (_mm_mask_fixupimm_round_sd): Ditto.
4572         (_mm_mask_fixupimm_round_ss): Ditto.
4573         (_mm512_mask_fixupimm_pd): Ditto.
4574         (_mm512_mask_fixupimm_ps): Ditto.
4575         (_mm_mask_fixupimm_sd): Ditto.
4576         (_mm_mask_fixupimm_ss): Ditto.
4577         * config/i386/avx512vlintrin.h:
4578         (_mm256_fixupimm_pd): Update parameters and builtin.
4579         (_mm256_maskz_fixupimm_pd): Ditto.
4580         (_mm256_fixupimm_ps): Ditto.
4581         (_mm256_maskz_fixupimm_ps): Ditto.
4582         (_mm_fixupimm_pd): Ditto.
4583         (_mm_maskz_fixupimm_pd): Ditto.
4584         (_mm_fixupimm_ps): Ditto.
4585         (_mm_maskz_fixupimm_ps): Ditto.
4586         (_mm256_mask_fixupimm_pd): Update builtin.
4587         (_mm256_mask_fixupimm_ps): Ditto.
4588         (_mm_mask_fixupimm_pd): Ditto.
4589         (_mm_mask_fixupimm_ps): Ditto.
4590         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
4591         * config/i386/i386-builtin.def: Update builtin definitions.
4592         * config/i386/i386.c: Handle new builtin types and remove useless ones.
4593         * config/i386/sse.md: Update VFIXUPIMM* patterns.
4594         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4595         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4596         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
4597         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4598         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4599         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
4600         * config/i386/subst.md:
4601         (round_saeonly_sd_mask_operand4): Add new subst_attr.
4602         (round_saeonly_sd_mask_op4): Ditto.
4603         (round_saeonly_expand_operand5): Ditto.
4604         (round_saeonly_expand): Update.
4606 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
4608         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
4610 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
4612         PR rtl-optimization/87871
4613         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
4615 2018-11-05  Paul Koning  <ni1d@arrl.net>
4617         * doc/sourcebuild.texi (target attributes): Document new "inf"
4618         effective target keyword.
4620 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
4622         * config/s390/s390.c (s390_register_move_cost): Increase costs for
4623         moves involving the CC reg.
4625 2018-11-05  Richard Biener  <rguenther@suse.de>
4627         PR tree-optimization/87873
4628         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
4629         argument.
4630         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
4631         * tree-vect-loop.c (vect_transform_loop): When splitting the
4632         loop exit also create forwarder PHIs for constants.
4633         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
4634         Handle constant to_arg, add extra checking we match up the correct
4635         PHIs.
4637 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
4639         * config/s390/s390.md: QImode and HImode for load on condition.
4641 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
4643         * config/s390/predicates.md: Fix typo.
4644         * config/s390/s390.md: Allow immediates for load on condition.
4646 2018-11-05  Martin Liska  <mliska@suse.cz>
4648         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
4649         * bitmap.h (struct bitmap_usage): Likewise.
4650         * ggc-common.c (SCALE): Remove.
4651         (LABEL): Likewise.
4652         (struct ggc_usage): Use SIZE_AMOUNT. And update
4653         compare method.
4654         * ggc-page.c (SCALE): Remove.
4655         (STAT_LABEL): Remove.
4656         (ggc_print_statistics): Use SIZE_AMOUNT.
4657         * gimple.h (SCALE): Remove.
4658         (LABEL): Likewise.
4659         * input.c (ONE_K): Remove.
4660         (ONE_M): Likewise.
4661         (SCALE): Likewise.
4662         (STAT_LABEL): Likewise.
4663         (FORMAT_AMOUNT): Likewise.
4664         (dump_line_table_statistics): Use SIZE_AMOUNT.
4665         * mem-stats.h (struct mem_usage): Likewise.
4666         * rtl.c (dump_rtx_statistics): Likewise.
4667         (rtx_alloc_counts): Change type to size_t.
4668         (rtx_alloc_sizes): Likewise.
4669         (rtx_count_cmp): New.
4670         (dump_rtx_statistics): Sort first based on counts.
4671         * tree.c (tree_nodes_cmp): New.
4672         (tree_codes_cmp): New.
4673         (dump_tree_statistics): Sort first based on counts.
4674         * system.h (ONE_K): New.
4675         (ONE_M): Likewise.
4676         (SIZE_SCALE): Likewise.
4677         (SIZE_LABEL): Likewise.
4678         (SIZE_AMOUNT): Likewise.
4679         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
4680         * tree-dfa.c (dump_dfa_stats): Likewise.
4681         * tree-phinodes.c (phinodes_print_statistics): Likewise.
4682         * tree-ssanames.c (ssanames_print_statistics): Likewise.
4683         * tree.c (dump_tree_statistics): Likewise.
4684         * vec.c (struct vec_usage): Likewise.
4685         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
4686         trigger a -Werror=format-overflow with
4687         --enable-gather-detailed-stats.
4689 2018-11-05  Martin Liska  <mliska@suse.cz>
4691         * mem-stats.h (mem_alloc_description::release_instance_overhead):
4692         Return T *.
4693         * vec.c (struct vec_usage): Register m_element_size.
4694         (vec_prefix::register_overhead): New arguments: elements and
4695         element_size.
4696         (vec_prefix::release_overhead): Subtract elements.
4697         * vec.h (struct vec_prefix): Change signature.
4698         (va_heap::reserve): Pass proper arguments.
4699         (va_heap::release): Likewise.
4701 2018-11-05  Martin Liska  <mliska@suse.cz>
4703         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
4704         style.
4705         * vec.c: Likewise.
4707 2018-11-05  Richard Biener  <rguenther@suse.de>
4709         * tree-scalar-evolution.h (final_value_replacement_loop): Update
4710         prototype.
4711         * tree-scalar-evolution.c (final_value_replacement_loop): Return
4712         whether anything was done.
4713         (scev_const_prop): Remove constant propagation part, fold
4714         remains into ...
4715         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
4716         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
4717         conditionally.
4719 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
4721         PR tree-optimization/87859
4722         * gimple-ssa-store-merging.c (struct merged_store_group): Add
4723         only_constants and first_nonmergeable_order members.
4724         (merged_store_group::merged_store_group): Initialize them.
4725         (merged_store_group::do_merge): Clear only_constants member if
4726         adding something other than INTEGER_CST store.
4727         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
4728         stores with order >= first_nonmergeable_order.  Use
4729         merged_store->only_constants instead of always recomputing it.
4730         Set merged_store->first_nonmergeable_order if we've skipped any
4731         stores.  Attempt to merge overlapping INTEGER_CST stores that
4732         we would otherwise skip.
4734         PR sanitizer/87837
4735         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
4737 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
4739         PR target/87853
4740         * config/i386/emmintrin.h (__v16qs): New to cope with option
4741         -funsigned-char.
4742         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
4743         (_mm_cmplt_epi8): Likewise.
4744         (_mm_cmpgt_epi8): Likewise.
4746 2018-11-05  Richard Biener  <rguenther@suse.de>
4748         PR rtl-optimization/87852
4749         * fwprop.c (use_killed_between): Only consider single-defs of the
4750         use whose definition statement dominates the use.
4752 2018-11-05  Martin Liska  <mliska@suse.cz>
4754         PR web/87829
4755         * doc/invoke.texi: Remove options that are
4756         not disabled with -Os.
4758 2018-11-05  Martin Liska  <mliska@suse.cz>
4760         PR c/87811
4761         * doc/extend.texi: Update constrain about the last argument
4762         of __builtin_expect_with_probability.
4764 2018-11-05  Martin Liska  <mliska@suse.cz>
4766         PR c/87811
4767         * predict.c (expr_expected_value_1): Verify
4768         that last argument is a real constants and emit
4769         error.
4771 2018-11-05  Martin Liska  <mliska@suse.cz>
4773         PR gcov-profile/77698
4774         * ipa-profile.c (ipa_profile): Adjust hotness threshold
4775         only in LTO mode.
4777 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4779         PR tree-optimization/86572
4780         * builtins.c (c_strlen): Handle negative offsets in a safe way.
4782 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4784         PR tree-optimization/87672
4785         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
4786         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
4788 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
4790         PR middle-end/58372
4791         * cfgexpand.c (pass_expand::execute): Move the call to
4792         finish_eh_generation in front of the call to expand_stack_alignment.
4794 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
4796         * common/config/i386/i386-common.c (processor_alias_table): Add
4797         znver2 entry.
4798         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
4799         (case ${target}): Add znver2.
4800         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
4801         -march=native recognize znver2 processors.
4802         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
4803         * config/i386/i386.c (m_znver2): New definition.
4804         (m_ZNVER): New definition.
4805         (m_AMD_MULTIPLE): Includes m_znver2.
4806         (processor_cost_table): Add znver2 entry.
4807         (processor_target_table): Add znver2 entry.
4808         (get_builtin_code_for_version): Set priority for
4809         PROCESSOR_ZNVER2.
4810         (processor_model): Add M_AMDFAM17H_ZNVER2.
4811         (arch_names_table): Ditto.
4812         (ix86_reassociation_width): Include znver2.
4813         * config/i386/i386.h (TARGET_znver2): New definition.
4814         (struct ix86_size_cost): Add TARGET_ZNVER2.
4815         (enum processor_type): Add PROCESSOR_ZNVER2.
4816         * config/i386/i386.md (define_attr "cpu"): Add znver2.
4817         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
4818         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
4819         (ix86_adjust_cost): Add znver2.
4820         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
4821         * gcc/doc/extend.texi: Add details about znver2.
4822         * gcc/doc/invoke.texi: Add details about znver2.
4824 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
4826         PR target/87079
4828         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
4829         pattern.
4831 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4833         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
4834         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
4836 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
4838         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
4839         shrink-wrapping when -mtrack-speculation.
4841 2018-11-02  Richard Biener  <rguenther@suse.de>
4843         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
4844         (pop_cost_one_pair): Do not free pair.
4845         (pop_best_coalesce): Likewise.
4846         (create_coalesce_list): Initialize obstack.
4847         (delete_coalesce_list): Free obstack.
4848         (find_coalesce_pair): Obstack-allocate coalesce pairs.
4849         (add_cost_one_coalesce): Likewise.
4850         (struct live_track): Remove bitmap pointer indirections.
4851         (new_live_track): Adjust.
4852         (delete_live_track): Likewise.
4853         (live_track_remove_partition): Likewise.
4854         (live_track_add_partition): Likewise.
4855         (live_track_live_p): Likewise.
4856         (live_track_process_def): Likewise.
4857         (live_track_clear_base_vars): Likewise.
4859 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4861         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
4862         * configure: Regenerate.
4863         * config.in: Regenerate.
4864         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
4865         HAVE_AS_REGISTER_PSEUDO_OP guard.
4866         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
4867         !HAVE_AS_REGISTER_PSEUDO_OP support.
4868         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
4869         guard.
4871 2018-11-02  Richard Biener  <rguenther@suse.de>
4873         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
4874         with is_type_die.
4876 2018-11-02  Richard Biener  <rguenther@suse.de>
4878         PR tree-optimization/87776
4879         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
4880         executable when iterating but running into rpo-vn-max-loop-depth
4881         and not eliding the iteration.
4883 2018-11-30  Jan Hubicka  <jh@suse.cz>
4885         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
4886         TYPE_DECL.
4888 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
4890         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
4891         prototype.
4892         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
4893         Combine with rs6000_address_for_fpconvert.
4894         (rs6000_address_for_fpconvert): Combine with
4895         rs6000_force_indexed_or_indirect_mem.
4896         (rs6000_expand_vector_init): Change function call from
4897         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
4898         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
4899         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
4900         (floatsi<mode>2_lfiwax_mem): Ditto.
4901         (floatunssi<mode>2_lfiwzx): Ditto.
4902         (floatunssi<mode>2_lfiwzx_mem): Ditto.
4903         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
4904         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
4905         (fix_trunc<mode>si2_stfiwx): Ditto.
4906         (fixuns_trunc<mode>si2_stfiwx): Ditto.
4907         (float_<mode>si2_hw): Ditto.
4908         (floatuns_<mode>si2_hw): Ditto.
4909         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
4910         (vsx_splat_<mode>): Ditto.
4912 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
4914         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
4915         at end of assembler input text.
4916         * configure: Regenerate.
4918 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
4920         PR tree-optimization/87826
4921         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
4922         negative or larger or equal to type's precision.
4924 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
4926         * opts.c (default_options_table): Do not enable
4927         OPT_fdelayed_branch at -Og.
4928         * doc/invoke.texi (-fdelayed-branch): Document it.
4930 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
4932         * optabs-libfuncs.c (build_libfunc_function_visibility):
4933         New, split out from...
4934         (build_libfunc_function): ... here.
4935         (init_one_libfunc_visibility): New, split out from ...
4936         (init_one_libfunc): ... here.
4938         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
4939         scratch register need not be early-clobber.  Document the reason
4940         why we cannot use ST<OP>.
4942 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
4944         PR bootstrap/82856
4945         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
4946         line for second argument of AC_DEFINE_UNQUOTED.
4947         * doc/install.texi (Tools/packages necessary for modifying GCC):
4948         Update to autoconf 2.69 and automake 1.15.1.
4949         * aclocal.m4, config.in, configure: Regenerate.
4951 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
4953         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
4954         initialization.
4956 2018-10-31  Martin Liska  <mliska@suse.cz>
4958         PR driver/83193
4959         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
4960         Add new argument invalid_extension.
4961         (aarch64_get_all_extension_candidates): New function.
4962         (aarch64_rewrite_selected_cpu): Add NULL to function call.
4963         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
4964         new argument.
4965         (aarch64_get_all_extension_candidates): New function.
4966         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
4967         argument invalid_extension.
4968         (aarch64_parse_cpu): Likewise.
4969         (aarch64_print_hint_for_extensions): New function.
4970         (aarch64_validate_mcpu): Provide hint about invalid extension.
4971         (aarch64_validate_march): Likewise.
4972         (aarch64_handle_attr_arch): Pass new argument.
4973         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
4974         (aarch64_handle_attr_isa_flags): Likewise.
4976 2018-10-31  Richard Biener  <rguenther@suse.de>
4978         PR middle-end/70359
4979         PR middle-end/86270
4980         * tree-outof-ssa.c (insert_backedge_copies): Restrict
4981         copy generation to useful cases.  Place the copy before
4982         the definition of the backedge value when possible.
4984 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
4986         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
4987         * config/arc/arc.c (arc_active_insn): New function.
4988         (check_store_cacheline_hazard): Likewise.
4989         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
4990         (arc_override_options): Disable delay slot scheduler for older
4991         A7.
4992         (arc_store_addr_hazard_p): New implementation, old one renamed to
4993         ...
4994         (arc_store_addr_hazard_internal_p): Renamed.
4995         (arc_reorg): Don't combine into brcc instructions which are part
4996         of hardware hazard solution.
4997         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
4998         (tune_arc700): Likewise.
4999         * config/arc/arc.opt (arc7xx): New tune value.
5000         * config/arc/arc700.md: Improve A7 scheduler.
5002 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5004         * config/arc/arc.c (arc_override_options): Remove
5005         TARGET_COMPACT_CASESI.
5006         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
5007         (CASE_VECTOR_MODE): Likewise.
5008         (CASE_VECTOR_PC_RELATIVE): Likewise.
5009         (CASE_VECTOR_SHORTEN_MODE): Likewise.
5010         (CASE_VECTOR_SHORTEN_MODE1): Delete.
5011         (ADDR_VEC_ALIGN): Update.
5012         (ASM_OUTPUT_CASE_LABEL): Undefine.
5013         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
5014         (TARGET_BI_BIH): Define.
5015         (DEFAULT_BRANCH_INDEX): Likewise.
5016         * config/arc/arc.md (casesi): Rework to accept BI/BIH
5017         instructions, remove compact_casesi use case.
5018         (casesi_compact_jump): Remove.
5019         (casesi_dispatch): New pattern.
5020         * config/arc/arc.opt: Add mbranch-index option. Deprecate
5021         compact_casesi option.
5022         * doc/invoke.texi: Document mbranch-index option.
5024 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5026         * config/arc/arc.c (arc_get_tp): Remove function.
5027         (arc_emit_call_tls_get_addr): Likewise.
5028         (arc_call_tls_get_addr): New function.
5029         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
5030         * config/arc/arc.md (tls_load_tp_soft): Remove.
5031         (tls_gd_get_addr): Likewise.
5033 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5035         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
5036         (mulsi3_600_lib): Remove pattern.
5037         (umulsi3_highpart_600_lib_le): Likewise.
5038         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
5039         (umulsidi3): Remove call to umulsidi3_600_lib.
5040         (umulsidi3_600_lib): Remove pattern.
5041         (peephole2): Remove peephole using the above deprecated patterns.
5043 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5045         PR target/87374
5046         * config/arm/arm.c (arm_option_check_internal): Disable the combined
5047         use of -mslow-flash-data and -mword-relocations.
5048         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
5049         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
5050         flag_pic.
5051         * doc/invoke.texi (-mword-relocations): Mention conflict with
5052         -mslow-flash-data.
5053         (-mslow-flash-data): Reciprocally.
5055 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
5057         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
5058         16-byte modes held in GP registers to use an even regno.
5060         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
5061         (aarch64_atomic_ldop_supported_p): Remove.
5062         (aarch64_gen_atomic_ldop): Remove.
5063         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
5064         Fully expand LSE operations here.
5065         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
5066         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
5067         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
5068         and use ATOMIC_LDOP instead; use register_operand for the input;
5069         drop the split and emit insns directly.
5070         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
5071         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
5072         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
5074         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
5075         (aarch64_gen_atomic_ldop): Don't call it.
5076         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
5077         Use aarch64_reg_or_zero.
5078         (aarch64_atomic_exchange<ALLI>): Likewise.
5079         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
5080         operand 0; use aarch64_reg_or_zero for input; merge ...
5081         (@aarch64_atomic_swp<ALLI>): ... this and remove.
5083         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
5084         (aarch64_split_compare_and_swap): Use it.
5085         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
5086         test oldval against the proper predicate.
5087         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
5088         Use nonmemory_operand for expected.
5089         (cas_short_expected_pred): New.
5090         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
5091         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
5092         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
5093         (aarch64_plushi_operand): New.
5095         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
5096         Force oldval into the rval register for TARGET_LSE; emit the compare
5097         during initial expansion so that it may be deleted if unused.
5098         (aarch64_gen_atomic_cas): Remove.
5099         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
5100         Change =&r to +r for operand 0; use match_dup for operand 2;
5101         remove is_weak and mod_f operands as unused.  Drop the split
5102         and merge with...
5103         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
5104         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
5105         (@aarch64_atomic_cas<GPI>): Similarly.
5107 2018-10-31  Richard Biener  <rguenther@suse.de>
5109         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
5110         using ABSU_EXPR.
5112 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5114         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
5115         * config/aarch64/aarch64.md: Include saphira.md
5116         * config/aarch64/saphira.md: New file for pipeline description.
5118 2018-10-30  Martin Sebor  <msebor@redhat.com>
5120         PR middle-end/87041
5121         * gimple-ssa-sprintf.c (format_directive): Use %G to include
5122         inlining context.
5123         (sprintf_dom_walker::compute_format_length):
5124         Avoid setting POSUNDER4K here.
5125         (get_destination_size): Handle null argument values.
5126         (get_user_idx_format): New function.
5127         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
5128         functions, including user-defined with attribute format printf.
5129         Use %G to include inlining context.
5130         Set POSUNDER4K here.
5132 2018-10-30  Jan Hubicka  <jh@suse.cz>
5134         * params.def (lto-partitions): Bump from 32 to 128.
5136 2018-10-30  Jan Hubicka  <jh@suse.cz>
5138         * tree.c
5139         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
5140         head in file.
5141         (free_lang_data_in_type): Forward declare.
5142         (fld_type_variant_equal_p): New function.
5143         (fld_type_variant): New function
5144         (fld_incomplete_types): New hash.
5145         (fld_incomplete_type_of): New function
5146         (fld_simplfied-type): New function.
5147         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
5148         (free_lang_data): Allocate and free fld_incomplete_type; update call
5149         of free_lang_data_in_decl.
5151 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
5153         * gcov.c (output_lines): Remove duplicate line.
5155 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5157         * config/rs6000/rs6000.md (bswapdi2): Force address into register
5158         if not in indexed or indirect form.
5159         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
5160         (bswapdi2_store): Ditto.
5161         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
5162         helper function.
5163         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
5164         Prototype for helper function.
5166 2018-10-30  Martin Sebor  <msebor@redhat.com>
5168         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
5169         (target, pragma GCC optimize, pragma GCC target): Ditto.
5171 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
5173         * doc/extend.texi: Fix prototype and description of
5174         __builtin_expect_with_probability.
5176 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
5178         * cgraph.h (clone_function_name_1): Replaced by new
5179           clone_function_name_numbered that takes name as string; for
5180           privatize_symbol_name_1 use only.
5181           (clone_function_name): Renamed to
5182           clone_function_name_numbered to be explicit about numbering.
5183           (clone_function_name): New two-argument function that does
5184           not number its output.
5185           (clone_function_name): New three-argument function that
5186           takes a number to append to its output.
5187         * cgraphclones.c (duplicate_thunk_for_node):
5188           (clone_function_name_1): Renamed.
5189           (clone_function_name_numbered): Two new functions.
5190           (clone_function_name): Improved documentation.
5191           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
5192         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
5193         * final.c (final_scan_insn_1): Use the new clone_function_name
5194           without numbering.
5195         * multiple_target.c (create_dispatcher_calls): Ditto.
5196           (create_target_clone): Ditto.
5197         * omp-expand.c (grid_expand_target_grid_body): Ditto.
5198         * omp-low.c (create_omp_child_function_name): Ditto.
5199         * omp-simd-clone.c (simd_clone_create): Ditto.
5200         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
5201           new clone_function_name without numbering.
5203 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
5205         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
5206         Assert that the allocation size is not zero.
5208 2018-10-30  Richard Biener  <rguenther@suse.de>
5210         PR tree-optimization/87800
5211         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
5212         non-induction or reduction PHIs.
5214 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
5216         * config/aarch64/falkor-tag-collision-avoidance.c
5217         (execute_tag_collision_avoidance): Call df_note_add_problem.
5219 2018-10-30  Martin Liska  <mliska@suse.cz>
5221         * doc/extend.texi: Fix typo in documentation
5222         of __builtin_expect_with_probability.
5224 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
5226         PR c++/87721
5227         * input.c (get_substring_ranges_for_loc): Detect if
5228         linemap_resolve_location gives us a NULL map, and reject
5229         this case.
5231 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
5233         * config.gcc (xstormy16-*-elf): Set tm_d_file.
5235 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
5236             Martin Sebor  <msebor@redhat.com>
5237             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
5239         * doc/ux.texi (Quoting): New subsection, adapted from material at
5240         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
5241         MartinSebor and ManuelLopezIbanez.
5242         (Fix-it hints): Note that fix-it hints shouldn't be marked for
5243         translation.
5245 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
5247         PR middle-end/87469
5248         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
5249         max value.
5251 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
5253         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
5255 2018-10-29  Paul Koning  <ni1d@arrl.net>
5257         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
5259 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
5261         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
5262         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
5263         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
5264         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
5265         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
5266         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
5267         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
5268         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
5269         Change 'vector' to '__vector'.
5270         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
5271         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
5272         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
5273         _mm_avg_pu16): Likewise.  And, whitespace corrections.
5275 2018-10-29  Richard Biener  <rguenther@suse.de>
5277         PR tree-optimization/87785
5278         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
5279         internal defs.
5281 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5283         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
5285 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5287         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
5288         ports configuration macro, defaults to "gnu".
5289         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
5290         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
5292 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5294         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
5296 2018-10-29  Richard Biener  <rguenther@suse.de>
5298         PR tree-optimization/87790
5299         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
5300         (vect_make_slp_decision): Adjust.
5301         (vect_slp_analyze_bb_1): Likewise.
5302         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
5303         edges.
5305 2018-10-29  Richard Biener  <rguenther@suse.de>
5307         PR tree-optimization/87785
5308         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
5309         and processing.
5310         (vect_build_slp_tree): Likewise.
5311         (vect_gather_slp_loads): New function.
5312         (vect_analyze_slp_instance): Gather loads separately from the
5313         SLP tree build.
5315 2018-10-29  Martin Liska  <mliska@suse.cz>
5317         * Makefile.in: Make dependency to json.o.
5318         * doc/gcov.texi: Document new JSON format, remove
5319         old intermediate format documentation.
5320         * gcov.c (struct function_info): Come up with m_name and
5321         m_demangled_name.
5322         (function_info::function_info): Initialize it.
5323         (function_info::~function_info): Release it.
5324         (main): Rename flag_intermediate_format to flag_json_format.
5325         (print_usage): Describe --json-format.
5326         (process_args): Set flag_json_format.
5327         (output_intermediate_line): Remove.
5328         (output_intermediate_json_line): Likewise.
5329         (get_gcov_intermediate_filename): Return new extension
5330         ".gcov.json.gz".
5331         (output_intermediate_file): Implement JSON emission.
5332         (output_json_intermediate_file): Implement JSON emission.
5333         (generate_results): Use ::get_name for function name.
5334         Handle JSON output file.
5335         (read_graph_file): Use ::get_name instead of cplus_demangle.
5336         (read_count_file): Likewise.
5337         (solve_flow_graph): Likewise.
5338         (add_line_counts): Likewise.
5339         (accumulate_line_counts): Use new flag_json_format.
5340         (output_function_details): Use ::get_name instead of cplus_demangle.
5341         (output_lines): Likewise.
5342         * json.cc (test_writing_literals): Add new tests.
5343         * json.h (class literal): Add new boolean constructor.
5345 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
5347         PR rtl-optimization/87701
5348         PR rtl-optimization/87780
5349         * combine.c (make_more_copies): Rewrite.
5351 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
5353         * doc/generic.texi (ABSU_EXPR): Document.
5354         * match.pd (absu(x)*absu(x) -> x*x): Handle.
5355         (absu(absu(X)) -> absu(X)): Likewise.
5356         (absu(-X) -> absu(X)): Likewise.
5357         (absu(X)  where X is nonnegative -> X): Likewise.
5359 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
5361         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
5362         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
5363         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
5364         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
5365         (s-tm-texi): Also check timestamp on d-target.def.
5366         (generated_files): Add TM_D_H and d-target-hooks-def.h.
5367         (build/genhooks.o): Also depend on D_TARGET_DEF.
5368         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
5369         variables.
5370         * config/aarch64/aarch64-d.c: New file.
5371         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
5372         Define.
5373         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
5374         prototype.
5375         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
5376         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
5377         * config/arm/arm-d.c: New file.
5378         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
5379         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
5380         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
5381         * config/arm/t-arm (arm-d.o): New rule.
5382         * config/default-d.c: New file.
5383         * config/glibc-d.c: New file.
5384         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5385         * config/i386/i386-d.c: New file.
5386         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
5387         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
5388         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
5389         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
5390         * config/i386/t-i386 (i386-d.o): New rule.
5391         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5392         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5393         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
5394         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5395         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
5396         * config/mips/mips-d.c: New file.
5397         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
5398         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
5399         * config/mips/t-mips (mips-d.o): New rule.
5400         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5401         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5402         * config/powerpcspe/powerpcspe-d.c: New file.
5403         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
5404         New prototype.
5405         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5406         Support GNU D by using 0 as the language type.
5407         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
5408         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
5409         * config/riscv/riscv-d.c: New file.
5410         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
5411         prototype.
5412         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
5413         * config/riscv/t-riscv (riscv-d.o): New rule.
5414         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5415         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5416         * config/rs6000/rs6000-d.c: New file.
5417         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
5418         prototype.
5419         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5420         Support GNU D by using 0 as the language type.
5421         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
5422         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
5423         * config/s390/s390-d.c: New file.
5424         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
5425         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
5426         * config/s390/t-s390 (s390-d.o): New rule.
5427         * config/sparc/sparc-d.c: New file.
5428         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
5429         prototype.
5430         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
5431         * config/sparc/t-sparc (sparc-d.o): New rule.
5432         * config/t-glibc (glibc-d.o): New rule.
5433         * configure: Regenerated.
5434         * configure.ac (tm_d_file): New variable.
5435         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
5436         * doc/contrib.texi (Contributors): Add self for the D frontend.
5437         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
5438         * doc/install.texi (Configuration): Mention libphobos as an option for
5439         --enable-shared.  Mention d as an option for --enable-languages.
5440         (Testing): Mention check-d as a target.
5441         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
5442         name suffixes.  Mention d as a -x option.
5443         * doc/sourcebuild.texi (Top Level): Mention libphobos.
5444         * doc/standards.texi (Standards): Add section on D language.
5445         * doc/tm.texi: Regenerated.
5446         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
5447         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
5448         * dwarf2out.c (is_dlang): New function.
5449         (gen_compile_unit_die): Use DW_LANG_D for D.
5450         (declare_in_namespace): Return module die for D, instead of adding
5451         extra declarations into the namespace.
5452         (gen_namespace_die): Generate DW_TAG_module for D.
5453         (gen_decl_die): Handle CONST_DECLSs for D.
5454         (dwarf2out_decl): Likewise.
5455         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
5456         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
5457         kinds of aggregates.
5458         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
5459         * genhooks.c: Include d/d-target.def.
5461 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
5463         PR target/85669
5464         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
5465         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
5466         (STACK_DYNAMIC_OFFSET): Likewise.
5468 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
5470         PR target/80024
5471         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
5472         error message.
5474 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
5476         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
5477         return type and other typos.
5479 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
5481         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
5482         a shorter sequence with fewer branches.
5483         (emit_final_str_compare_gpr): Ditto.
5485 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
5487         * config/rs6000/tmmintrin.h: New file.
5488         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
5490 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
5492         * config/rs6000/mmintrin.h: Enable 32bit compilation.
5493         * config/rs6000/xmmintrin.h: Likewise.
5495 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
5497         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
5499 2018-10-26  Richard Biener  <rguenther@suse.de>
5501         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
5502         and wrapper.
5503         (vect_mark_slp_stmts_relevant): Likewise.
5504         (vect_detect_hybrid_slp_stmts): Likewise.
5505         (vect_bb_slp_scalar_cost): Likewise.
5506         (vect_remove_slp_scalar_calls): Likewise.
5508 2018-10-26  Jan Hubicka  <jh@suse.cz>
5510         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
5511         (warn_types_mismatch): Fix walk of DECL_NAME.
5512         (odr_types_equivalent_p): Fix overactive assert.
5514 2018-10-26  Richard Biener  <rguenther@suse.de>
5516         PR tree-optimization/87105
5517         * tree-vectorizer.h (_slp_tree::refcnt): New member.
5518         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
5519         refcnt.
5520         (vect_create_new_slp_node): Initialize refcnt to one.
5521         (bst_traits): Move.
5522         (scalar_stmts_set_t, bst_fail): Remove.
5523         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
5524         (vect_build_slp_tree): Add bst_map argument and lookup
5525         already created SLP nodes.
5526         (vect_print_slp_tree): Handle a SLP graph, print SLP node
5527         addresses.
5528         (vect_slp_rearrange_stmts): Handle a SLP graph.
5529         (vect_analyze_slp_instance): Adjust and free SLP nodes from
5530         the CSE map.  Fix indenting.
5531         (vect_schedule_slp_instance): Add short-cut.
5533 2018-10-26  Martin Liska  <mliska@suse.cz>
5535         PR testsuite/86158
5536         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
5537         addr_expr and not with pointers.
5539 2018-10-26  Jan Hubicka  <jh@suse.cz>
5541         * tree.c (free_lang_data_in_type): Only check main variants.
5542         * ipa-devirt.c (warn_odr): Make static.
5543         (types_same_for_odr): Drop strict variant.
5544         (types_odr_comparable): Likewise.
5545         (odr_or_derived_type_p): Look for main variants.
5546         (odr_name_hasher::equal): Cleanup comment.
5547         (odr_subtypes_equivalent): Add warn and warned arguments; check main
5548         variants.
5549         (type_variants_equivalent_p): break out from ...
5550         (odr_types_equivalent): ... here; go for main variants where needed.
5551         (warn_odr): ... here; turn static.
5552         (warn_types_mismatch): Compare mangled names of main variants.
5553         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
5554         (type_with_linkage_p): Sanity check that we look at main variant.
5555         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
5556         * tree.h (types_same_for_odr): Drop strict argument.
5558 2018-10-26  Richard Biener  <rguenther@suse.de>
5560         PR tree-optimization/87746
5561         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
5562         Simplify and fix WRT strided store groups with size not
5563         equal to step in element count.
5564         (vect_analyze_group_access_1): Dump the whole group.
5566 2018-10-25  Carl Love  <cel@us.ibm.com>
5568         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
5569         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
5570         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
5571         precicion entry for each overloaded builtin.
5572         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
5573         VSCEDPUO): Rename overloaded name.
5574         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
5575         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
5576         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
5577         define_expand for xscmpexqp instruction.
5578         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
5580 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
5581             Jinsong Ji <jji@us.ibm.com>
5583         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
5584         function with vec_sl.
5585         (_mm_slli_epi32): Likewise.
5586         (_mm_slli_epi64): Likewise.
5587         (_mm_srai_epi16): Replace deprecated function with vec_sra.
5588         (_mm_srai_epi32): Likewise.
5589         (_mm_srli_epi16): Replace deprecated function with vec_sr.
5590         (_mm_srli_epi32): Likewise.
5591         (_mm_srli_epi64): Likewise.
5592         (_mm_sll_epi16): Replace deprecated function with vec_sl.
5593         (_mm_sll_epi32): Likewise.
5594         (_mm_sll_epi64): Likewise.
5595         (_mm_sra_epi16): Replace deprecated function with vec_sra.
5596         (_mm_sra_epi32): Likewise.
5597         (_mm_srl_epi16): Replace deprecated function with vec_sr.
5598         (_mm_srl_epi32): Likewise.
5599         (_mm_srl_epi64): Likewise.
5601 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
5602             Jinsong Ji <jji@us.ibm.com>
5604         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
5605         comparison operators with vec_cmp* for compatibility due to
5606         unfortunate history; clean up formatting and use types more
5607         appropriately.
5608         (_mm_sll_epi32): Likewise.
5609         (_mm_sll_epi64): Likewise.
5610         (_mm_srl_epi16): Likewise.
5611         (_mm_srl_epi32): Likewise.
5612         (_mm_srl_epi64): Likewise.
5614 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
5615             Jinsong Ji <jji@us.ibm.com>
5617         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
5618         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
5619         __vector __bool int.  Use vec_cmpgt in preference to deprecated
5620         function vec_vcmpgtfp.
5621         (_mm_max_ps): Likewise.
5623 2018-10-25  Jeff Law  <law@redhat.com>
5625         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
5626         if returning false.
5628 2018-10-25  Martin Sebor  <msebor@redhat.com>
5630         * doc/extend.texi (aligned): Expand attribute description.
5631         (Alignment): Rename section.  Discuss function arguments.
5633 2018-10-25  Jan Hubicka  <jh@suse.cz>
5635         * ipa-devirt.c (main_odr_variant): Remove.
5636         (hash_odr_name, types_same_for_odr, types_odr_comparable,
5637         odr_name_hasher::equal, odr_subtypes_equivalent_p):
5638         Drop use of main_odr_variant.
5639         (add_type_duplicate): Silence confused warnings on integer types.
5640         (get_odr_type): Always look for main variant.
5641         (register_odr_type): Simplify.
5643 2018-10-25  Richard Biener  <rguenther@suse.de>
5645         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
5646         Initialize ng to silence error with release checking bootstrap.
5648 2018-10-25  Richard Biener  <rguenther@suse.de>
5650         * tree-if-conv.c: Include tree-ssa-sccvn.h.
5651         (tree_if_conversion): Run CSE on the if-converted loop body.
5653 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5655         * config/s390/constraints.md (ZL): New constraint.
5656         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
5657         operands.
5658         * config/s390/s390.md (movdi_larl): Remove.
5659         (movdi_64): Add the LARL alternative.
5661 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5663         PR bootstrap/87747
5664         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
5665         (RTX_CODE_HWINT_P): New macro.
5666         (rtx_code_size): Use RTX_CODE_HWINT_P ().
5668 2018-10-25  Jan Hubicka  <jh@suse.cz>
5670         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
5671         is anonymous.
5673 2018-10-25  Richard Biener  <rguenther@suse.de>
5675         PR tree-optimization/87665
5676         PR tree-optimization/87745
5677         * tree-vectorizer.h (get_earlier_stmt): Remove.
5678         (get_later_stmt): Pick up UID from the original non-pattern stmt.
5680 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
5682         * options.texi (Deprecated): Move list to Var section.
5684 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
5685             Jinsong Ji <jji@us.ibm.com>
5687         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
5688         __vector long to __vector long long.
5689         (_mm_cvtpd_ps): Likewise.
5690         (_mm_cvttpd_epi32): Likewise.
5691         (_mm_cvtpi32_pd): Likewise.
5692         (_mm_unpackhi_epi64): Likewise.
5693         (_mm_unpacklo_epi64): Likewise.
5695 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
5697         PR rtl-optimization/87720
5698         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
5700 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
5702         * gimple-ssa-isolate-paths.c
5703         (find_implicit_erroneous_behavior): Do not change code if the
5704         pass is running for warnings only.
5705         (find_explicit_erroneous_behavior): Likewise.
5707 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
5709         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
5710         Define as rs6000_mangle_decl_assembler_name.
5711         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
5712         long double to IEEE long double, switch the names of the long
5713         double built-in functions to be <func>f128 instead of <func>l.
5715 2018-10-24  Martin Sebor  <msebor@redhat.com>
5717         * doc/extend.texi (nonnull): List no-argument form.  Reference
5718         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
5720 2018-10-24  Richard Biener  <rguenther@suse.de>
5722         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
5724 2018-10-24  Martin Liska  <mliska@suse.cz>
5726         PR tree-optimization/84436
5727         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
5728         Remove.
5729         (switch_conversion::contains_linear_function_p): New.
5730         (switch_conversion::build_one_array): Support linear
5731         transformation on input.
5732         * tree-switch-conversion.h (struct switch_conversion): Add
5733         contains_linear_function_p declaration.
5735 2018-10-24  Richard Biener  <rguenther@suse.de>
5737         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
5738         if its argument is CONSTANT_CLASS_P.
5740 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
5742         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
5743         it is wrong for forward declarations.
5745 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5747         * config/s390/s390.c (s390_check_qrst_address): Add the missing
5748         SYMBOL_REF_P () check.
5750 2018-10-24  Richard Biener  <rguenther@suse.de>
5752         PR tree-optimization/87105
5753         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
5754         dump classification.
5755         (vect_analyze_data_ref_accesses): Handle duplicate loads and
5756         stores by splitting the affected group after the fact.
5757         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
5758         fail the SLP build because of size constraints.
5760 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5762         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
5763         * configure: Regenerate.
5764         * config.in: Regenerate.
5765         * varasm.c (mergeable_string_section): Use readonly_data_section
5766         if linker doesn't support SHF_MERGE with alignment > 8.
5767         (mergeable_constant_section): Likewise.
5769 2018-10-24  Richard Biener  <rguenther@suse.de>
5771         PR tree-optimization/84013
5772         * tree-ssa-structalias.c (struct msdi_data): New struct for
5773         marshalling data to walk_stmt_load_store_ops.
5774         (maybe_set_dependence_info): Refactor as callback for
5775         walk_stmt_load_store_ops.
5776         (compute_dependence_clique): Set restrict info on all stmt kinds.
5778 2018-10-24  Martin Liska  <mliska@suse.cz>
5780         * cgraph.c (cgraph_node::dump):
5781         Remove reduntant dumps and make tp_first_run dump more compact.
5783 2018-10-24  Richard Biener  <rguenther@suse.de>
5785         PR tree-optimization/87665
5786         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
5787         to reflect reality.
5789 2018-10-12  Jeff Law  <law@redhat.com>
5791         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
5792         for H8/S.
5794 2018-10-23  Richard Biener  <rguenther@suse.de>
5796         * tree-vrp.c (add_assert_info): Guard dump_printf with
5797         dump_enabled_p.
5798         * gimple-ssa-evrp-analyze.c
5799         (evrp_range_analyzer::record_ranges_from_incoming_edge):
5800         Use value_range::ignore_equivs_equal_p.
5802 2018-10-23  Richard Biener  <rguenther@suse.de>
5804         PR tree-optimization/87105
5805         PR tree-optimization/87608
5806         * passes.def (pass_all_early_optimizations): Add early phi-opt
5807         after dce.
5808         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
5809         addition to debug stmts.
5810         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
5811         and abs replacement early.
5812         * tree-cfg.c (gimple_empty_block_p): Likewise.
5814 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
5816         PR target/86383
5817         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
5818         specified to configure.
5819         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
5821 2018-10-23  Richard Biener  <rguenther@suse.de>
5823         PR tree-optimization/87700
5824         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
5826 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
5828         PR target/87674
5829         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
5830         second argument from __mmask16 to __mmask8.
5831         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
5832         _mm_mask_packs_epi32): Likewise.
5833         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
5834         Likewise.
5835         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
5837 2018-10-23  Richard Biener  <rguenther@suse.de>
5839         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
5841 2018-10-23  Richard Biener  <rguenther@suse.de>
5843         PR tree-optimization/86144
5844         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
5845         over simd attribute.
5847 2018-10-23  Richard Biener  <rguenther@suse.de>
5849         PR tree-optimization/87693
5850         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
5851         the case we do not find the taken edge.
5853 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
5854             Jinsong Ji  <jji@us.ibm.com>
5856         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
5857         (_mm_store_pd): Use unaligned vector type for pointer cast.
5858         (_mm_maskmoveu_si128): Likewise.
5859         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
5860         (_mm_store_ps): Use unaligned vector type for pointer cast.
5862 2018-10-22  Paul Koning  <ni1d@arrl.net>
5864         * symtab.c (symtab_node::increase_alignment): Correct max
5865         alignment check.
5867 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
5869         PR tree-optimization/87633
5870         * match.pd: Do not generate unordered integer comparisons.
5872 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
5874         PR rtl-optimization/87600
5875         * combine.c: Add include of expr.h.
5876         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
5877         register to a pseudo.
5878         (make_more_copies): New function, add a copy to a new pseudo after
5879         the moves from hard registers into pseudos.
5880         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
5881         later.  Call make_more_copies.
5883 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
5885         * lra-constraints.c (process_alt_operands): New local array,
5886         matching_early_clobber.  Check matching_early_clobber before
5887         decrementing reject, and set matching_early_clobber after.
5889 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
5891         PR target/87598
5892         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
5893         call output_addr_const and hope for the best.
5895 2018-10-22  Richard Biener  <rguenther@suse.de>
5897         * gimple-ssa-evrp-analyze.c
5898         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
5899         smarter about what ranges to use.
5900         * tree-vrp.c (add_assert_info): Dump here.
5901         (register_edge_assert_for_2): Instead of here at multiple but
5902         not all places.
5904         * gcc.dg/tree-ssa/evrp12.c: New testcase.
5905         * gcc.dg/predict-6.c: Adjust.
5906         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
5907         * gcc.dg/tree-ssa/vrp02.c: Likewise.
5908         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
5910 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
5911         Richard Biener  <rguenther@suse.de>
5913         * bitmap.h: Update data structure documentation, including a
5914         description of bitmap views as either linked-lists or splay trees.
5915         (struct bitmap_element_def): Update comments for splay tree bitmaps.
5916         (struct bitmap_head_def): Likewise.
5917         (bitmap_list_view, bitmap_tree_view): New prototypes.
5918         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
5919         tree_form fields.
5920         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
5921         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
5922         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
5923         released bitmap element here.
5924         (bitmap_element_free): Remove.
5925         (bitmap_elt_clear_from): Work on splay tree bitmaps.
5926         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
5927         this function similar ones such that linked-list bitmap implementation
5928         functions are grouped.
5929         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
5930         and moved for grouping.
5931         (bitmap_list_insert_element_after): Renamed from
5932         bitmap_elt_insert_after, and moved for grouping.
5933         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
5934         (bitmap_tree_link_left, bitmap_tree_link_right,
5935         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
5936         bitmap_tree_link_element, bitmap_tree_unlink_element,
5937         bitmap_tree_find_element): New functions for splay-tree bitmap
5938         implementation.
5939         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
5940         Renamed and moved, see above entries.
5941         (bitmap_tree_listify_from): New function to convert part of a splay
5942         tree bitmap to a linked-list bitmap.
5943         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
5944         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
5945         (bitmap_find_bit): Remove.
5946         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
5947         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
5948         Handle splay tree bitmaps.
5949         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
5950         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
5951         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
5952         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
5953         bitmap_intersect_compl_p, bitmap_ior_and_compl,
5954         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
5955         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
5956         corresponding changes to use linked-list specific bitmap_element
5957         manipulation functions as applicable for efficiency.
5958         (bitmap_tree_to_vec): New function.
5959         (debug_bitmap_elt_file): New function split out from ...
5960         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
5961         (bitmap_print): Likewise.
5963         PR tree-optimization/63155
5964         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
5965         SSA edge worklists.
5966         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
5967         in tree-view.
5969 2018-10-22  Martin Liska  <mliska@suse.cz>
5971         PR tree-optimization/87686
5972         Revert
5973         2018-08-29  Martin Liska  <mliska@suse.cz>
5975         * tree-switch-conversion.c (switch_conversion::expand):
5976         Strenghten assumption about gswitch statements.
5978 2018-10-22  Martin Liska  <mliska@suse.cz>
5980         * ipa-icf.c (sem_item::compare_attributes): Remove.
5981         (sem_item::compare_referenced_symbol_properties): Use
5982         attribute_list_equal instead.
5983         (sem_function::equals_wpa): Likewise.
5984         * ipa-icf.h: Remove compare_attributes.
5986 2018-10-22  Richard Biener  <rguenther@suse.de>
5988         PR middle-end/87682
5989         * mem-stats.h (mem_usage::operator==): Fix pasto.
5991 2018-10-22  Richard Biener  <rguenther@suse.de>
5993         PR tree-optimization/87640
5994         * tree-vrp.c (set_value_range_with_overflow): Decompose
5995         incomplete result.
5996         (extract_range_from_binary_expr_1): Adjust.
5998 2018-10-22  Martin Jambor  <mjambor@suse.cz>
6000         * tree-eh.h (stmt_could_throw_p): Add function parameter.
6001         (stmt_can_throw_external): Likewise.
6002         (stmt_can_throw_internal): Likewise.
6003         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
6004         (lower_eh_constructs_2): Likewise.
6005         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
6006         (stmt_can_throw_external): Likewise.
6007         (stmt_can_throw_internal): Likewise.
6008         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
6009         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
6010         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
6011         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
6012         (pass_lower_eh_dispatch::execute): Pass cfun to
6013         stmt_can_throw_external.
6014         (cleanup_empty_eh): Likewise.
6015         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
6016         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
6017         stmt_can_throw_external instead of pushing it to cfun.
6018         (symbol_table::create_edge): Likewise.
6019         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
6020         stmt_can_throw_internal.
6021         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
6022         to stmt_could_throw_p.
6023         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
6024         stmt_can_throw_internal.
6025         (pass_store_merging::execute): Likewise.
6026         * gimple-ssa-strength-reduction.c
6027         (find_candidates_dom_walker::before_dom_children): Pass cfun to
6028         stmt_could_throw_p.
6029         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
6030         stmt_can_throw_internal.
6031         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
6032         to stmt_can_throw_external.
6033         (check_stmt): Pass cfun to stmt_could_throw_p.
6034         (check_stmt): Pass cfun to stmt_can_throw_external.
6035         (pass_nothrow::execute): Likewise.
6036         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
6037         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
6038         stmt_can_throw_internal.
6039         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
6040         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
6041         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
6042         * tree-complex.c (expand_complex_libcall): Pass cfun to
6043         stmt_could_throw_p and to stmt_can_throw_internal.
6044         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
6045         * tree-inline.c (copy_edges_for_bb): Likewise.
6046         (maybe_move_debug_stmts_to_successors): Likewise.
6047         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
6048         stmt_could_throw_p.
6049         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
6050         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
6051         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
6052         stmt_can_throw_internal.
6053         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
6054         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
6055         stmt_could_throw_p.
6056         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
6057         stmt_can_throw_internal.
6058         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
6059         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
6060         stmt_could_throw_p.
6061         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
6062         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
6063         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6064         (convert_mult_to_fma_1): Likewise.
6065         (convert_to_divmod): Likewise.
6066         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
6067         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
6068         * tree-ssa-propagate.c
6069         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
6070         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
6071         (maybe_optimize_range_tests): Likewise.
6072         (linearize_expr_tree): Likewise.
6073         (reassociate_bb): Likewise.
6074         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6075         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
6076         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
6077         (handle_char_store): Likewise.
6078         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
6079         stmt_can_throw_internal.
6080         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
6081         stmt_could_throw_p.
6082         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6083         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
6084         (vectorizable_simd_clone_call): Likewise.
6085         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
6086         (gimple_stringop_fixed_value): Likewise.
6088 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
6090         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
6091         literal pool references.
6092         (s390_check_qrst_address): Adapt to the new behavior of
6093         s390_loadrelative_operand_p ().
6095 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6097         PR target/72782
6098         * config/i386/sse.md (*andnot<mode>3_bcst): New.
6100 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6102         PR target/72782
6103         * config/i386/sse.md (*<code><mode>3_bcst): New.
6105 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6107         PR target/72782
6108         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
6109         V4DI, V16SI and V8DI.
6110         (*sub<mode>3<mask_name>_bcst): New.
6111         (*add<mode>3<mask_name>_bcst): Likewise.
6113 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
6114             Jinsong Ji  <jji@us.ibm.com>
6116         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
6117         __m64 with __vector unsigned long long for compatibility.
6118         (_mm_movemask_epi8): Likewise.
6119         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
6120         (_mm_cvttps_pi32): Likewise.
6121         (_mm_cvtpi32_ps): Likewise.
6122         (_mm_cvtps_pi16): Likewise.
6123         (_mm_loadh_pi): Likewise.
6124         (_mm_storeh_pi): Likewise.
6125         (_mm_movehl_ps): Likewise.
6126         (_mm_movelh_ps): Likewise.
6127         (_mm_loadl_pi): Likewise.
6128         (_mm_storel_pi): Likewise.
6129         (_mm_movemask_ps): Likewise.
6130         (_mm_shuffle_pi16): Likewise.
6132 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6134         PR target/72782
6135         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
6136         __builtin_ia32_vfnmsubpd512_mask.
6137         (_mm512_mask_fnmsub_round_pd): Likewise.
6138         (_mm512_fnmsub_pd): Likewise.
6139         (_mm512_mask_fnmsub_pd): Likewise.
6140         (_mm512_maskz_fnmsub_round_pd): Use
6141         __builtin_ia32_vfnmsubpd512_maskz.
6142         (_mm512_maskz_fnmsub_pd): Likewise.
6143         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
6144         (_mm512_mask_fnmsub_round_ps): Likewise.
6145         (_mm512_fnmsub_ps): Likewise.
6146         (_mm512_mask_fnmsub_ps): Likewise.
6147         (_mm512_maskz_fnmsub_round_ps): Use
6148         __builtin_ia32_vfnmsubps512_maskz.
6149         (_mm512_maskz_fnmsub_ps): Likewise.
6150         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
6151         __builtin_ia32_vfnmsubpd256_mask.
6152         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
6153         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
6154         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
6155         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
6156         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
6157         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
6158         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
6159         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
6160         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
6161         __builtin_ia32_vfnmsubpd.
6162         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
6163         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
6164         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
6165         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
6166         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
6167         * config/i386/i386-builtin.def: Add
6168         __builtin_ia32_vfnmsubpd256_mask,
6169         __builtin_ia32_vfnmsubpd256_maskz,
6170         __builtin_ia32_vfnmsubpd128_mask,
6171         __builtin_ia32_vfnmsubpd128_maskz,
6172         __builtin_ia32_vfnmsubps256_mask,
6173         __builtin_ia32_vfnmsubps256_maskz,
6174         __builtin_ia32_vfnmsubps128_mask,
6175         __builtin_ia32_vfnmsubps128_maskz,
6176         __builtin_ia32_vfnmsubpd512_mask,
6177         __builtin_ia32_vfnmsubpd512_maskz,
6178         __builtin_ia32_vfnmsubps512_mask,
6179         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
6180         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
6181         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
6182         __builtin_ia32_vfnmsubpd256.
6183         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
6184         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
6185         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
6186         Likewise.
6187         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
6188         Likewise.
6189         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
6190         Likewise.
6191         (fmai_vmfnmsub_<mode><round_name>): Likewise.
6193 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6195         PR target/72782
6196         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
6197         __builtin_ia32_vfnmaddpd512_mask.
6198         (_mm512_mask_fnmadd_round_pd): Likewise.
6199         (_mm512_fnmadd_pd): Likewise.
6200         (_mm512_mask_fnmadd_pd): Likewise.
6201         (_mm512_maskz_fnmadd_round_pd): Use
6202         __builtin_ia32_vfnmaddpd512_maskz.
6203         (_mm512_maskz_fnmadd_pd): Likewise.
6204         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
6205         (_mm512_mask_fnmadd_round_ps): Likewise.
6206         (_mm512_fnmadd_ps): Likewise.
6207         (_mm512_mask_fnmadd_ps): Likewise.
6208         (_mm512_maskz_fnmadd_round_ps): Use
6209         __builtin_ia32_vfnmaddps512_maskz.
6210         (_mm512_maskz_fnmadd_ps): Likewise.
6211         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
6212         __builtin_ia32_vfnmaddpd256_mask.
6213         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
6214         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
6215         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
6216         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
6217         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
6218         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
6219         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
6220         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
6221         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
6222         __builtin_ia32_vfnmaddpd.
6223         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
6224         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
6225         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
6226         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
6227         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
6228         * config/i386/i386-builtin.def: Add
6229         __builtin_ia32_vfnmaddpd256_mask,
6230         __builtin_ia32_vfnmaddpd256_maskz,
6231         __builtin_ia32_vfnmaddpd128_mask,
6232         __builtin_ia32_vfnmaddpd128_maskz,
6233         __builtin_ia32_vfnmaddps256_mask,
6234         __builtin_ia32_vfnmaddps256_maskz,
6235         __builtin_ia32_vfnmaddps128_mask,
6236         __builtin_ia32_vfnmaddps128_maskz,
6237         __builtin_ia32_vfnmaddpd512_mask,
6238         __builtin_ia32_vfnmaddpd512_maskz,
6239         __builtin_ia32_vfnmaddps512_mask,
6240         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
6241         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
6242         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
6243         __builtin_ia32_vfnmaddpd256.
6244         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
6245         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
6246         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
6247         Likewise.
6248         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
6249         Likewise.
6250         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
6251         Likewise.
6252         (fmai_vmfnmadd_<mode><round_name>): Likewise.
6254 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6256         PR target/72782
6257         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
6258         __builtin_ia32_vfmsubpd512_mask.
6259         (_mm512_mask_fmsub_round_pd): Likewise.
6260         (_mm512_fmsub_pd): Likewise.
6261         (_mm512_mask_fmsub_pd): Likewise.
6262         (_mm512_maskz_fmsub_round_pd): Use
6263         __builtin_ia32_vfmsubpd512_maskz.
6264         (_mm512_maskz_fmsub_pd): Likewise.
6265         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
6266         (_mm512_mask_fmsub_round_ps): Likewise.
6267         (_mm512_fmsub_ps): Likewise.
6268         (_mm512_mask_fmsub_ps): Likewise.
6269         (_mm512_maskz_fmsub_round_ps): Use
6270         __builtin_ia32_vfmsubps512_maskz.
6271         (_mm512_maskz_fmsub_ps): Likewise.
6272         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
6273         __builtin_ia32_vfmsubpd256_mask.
6274         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
6275         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
6276         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
6277         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
6278         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
6279         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
6280         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
6281         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
6282         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
6283         __builtin_ia32_vfmsubpd.
6284         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
6285         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
6286         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
6287         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
6288         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
6289         * config/i386/i386-builtin.def: Add
6290         __builtin_ia32_vfmsubpd256_mask,
6291         __builtin_ia32_vfmsubpd256_maskz,
6292         __builtin_ia32_vfmsubpd128_mask,
6293         __builtin_ia32_vfmsubpd128_maskz,
6294         __builtin_ia32_vfmsubps256_mask,
6295         __builtin_ia32_vfmsubps256_maskz,
6296         __builtin_ia32_vfmsubps128_mask,
6297         __builtin_ia32_vfmsubps128_maskz,
6298         __builtin_ia32_vfmsubpd512_mask,
6299         __builtin_ia32_vfmsubpd512_maskz,
6300         __builtin_ia32_vfmsubps512_mask,
6301         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
6302         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
6303         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
6304         __builtin_ia32_vfmsubpd256.
6305         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
6306         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
6307         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
6308         Likewise.
6309         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
6310         Likewise.
6311         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
6312         Likewise.
6313         (fmai_vmfmsub_<mode><round_name>): Likewise.
6315 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6317         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
6318         Remove plus.  Renamed to ...
6319         (*sub<mode>3<mask_name>_bcst): This.
6320         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
6321         (*add<mode>3<mask_name>_bcst): This.
6323 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6325         PR target/72782
6326         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
6328 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6330         PR target/87662
6331         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
6332         (_mm_or_epi32): Likewise.
6333         (_mm256_xor_epi32): Likewise.
6334         (_mm_xor_epi32): Likewise.
6335         (_mm256_or_epi64): Likewise.
6336         (_mm_or_epi64): Likewise.
6337         (_mm256_xor_epi64): Likewise.
6338         (_mm_xor_epi64): Likewise.
6340 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
6342         PR target/72782
6343         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
6345 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
6347         PR middle-end/87647
6348         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
6350 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
6352         * doc/ux.texi: Move @section directly after @node.
6354 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
6356         PR middle-end/85488
6357         PR middle-end/87649
6358         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
6359         depend closely nested inside of loop with ordered clause with
6360         a parameter.
6362 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
6364         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
6365         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
6366         * doc/ux.texi: New file.
6368 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
6370         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
6371         be the first CR field allocated.
6373 2018-10-19  Richard Biener  <rguenther@suse.de>
6375         PR target/87657
6376         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
6377         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
6379 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
6381         PR target/72782
6382         * config/i386/sse.md
6383         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
6384         (*add<mode>3<mask_name>_bcst_2): Likewise.
6386 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
6388         * config/i386/sse.md
6389         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
6390         Replace nonimmediate_operand with register_operand.
6391         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
6392         Likewise.
6393         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
6394         Likewise.
6396 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
6398         PR rtl-optimization/87596
6399         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
6400         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
6401         for instructions in FROM..TO range.
6403 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
6405         * cfgexpand.c (expand_one_var): Use specific wording in error message
6406         for non-local frame variables.
6407         * stor-layout.c (layout_decl): Do not issue a warning for them.
6409 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
6411         * haifa-sched.c (priority): Add force_recompute parameter.
6412         (apply_replacement): Call priority () with force_recompute = true.
6413         (restore_pattern): Likewise.
6415 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6417         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
6418         HOST_BITS_PER_WIDE_INT.
6419         (test_vector_ops_duplicate): Likewise.
6421 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6423         PR target/72782
6424         * config/i386/sse.md (VF_AVX512): New.
6425         (avx512bcst): Likewise.
6426         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
6427         Likewise.
6428         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
6429         Likewise.
6430         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
6431         Likewise.
6433 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
6435         * doc/invoke.texi (-dumpversion): Improve grammar.
6436         (-dumpfullversion): Make more consistent with -dumpversion.
6438 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
6440         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
6441         Set JUMP_LABEL to the jump insn.
6442         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
6443         Predict emitted jump and add label to jump insn.
6445 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
6447         PR tree-optimization/87562
6448         * input.c (get_substring_ranges_for_loc): Use
6449         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
6450         getting the linemap for the endpoint.  Verify that it's either
6451         in the same linemap as the start point's spelling location, or
6452         at least in the same file.
6454 2018-10-18  Richard Biener  <rguenther@suse.de>
6456         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
6457         feed width-specific load/store costs through ix86_vec_cost.
6458         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
6459         (k8_cost): Likewise.
6460         (bdver_cost): Likewise.
6461         (znver1_cost): Likewise.
6462         (btver1_cost): Likewise.
6463         (btver2_cost): Likewise.
6465 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6467         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
6468         to simplify subreg of vec_merge.
6470 2018-10-18  Richard Biener  <rguenther@suse.de>
6472         * config/i386/i386.c: Fix costing of vector FMA.
6474 2018-10-18  Richard Biener  <rguenther@suse.de>
6476         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
6477         and argument.
6478         (ix86_builtin_vectorization_cost): For vec_construct properly
6479         cost insertion into SSE regs.
6480         (...): Adjust calls to ix86_vec_cost.
6482 2018-10-18  Richard Biener  <rguenther@suse.de>
6484         PR middle-end/87087
6485         Revert
6486         2018-02-07  Richard Biener  <rguenther@suse.de>
6488         PR tree-optimization/84204
6489         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6490         this place.
6492 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6494         PR target/87537
6495         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
6496         of vec_duplicate.
6497         (test_vector_ops_duplicate): Add test for a scalar subreg of a
6498         VEC_MERGE of a VEC_DUPLICATE.
6500 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
6502         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
6503         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
6504         * doc/standards.texi (C Language): Document C2X.
6505         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
6506         (rl78_option_override): Handle "GNU C2X" language name.
6508 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
6510         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
6511         Document C17 as published in 2018.
6513 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
6515         PR middle-end/87623
6516         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
6517         bail out if both sides do not have the same storage order.
6519 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
6521         * bitmap.c (bitmap_head::dump): New.
6522         * bitmap.h (bitmap_head): Add dump().
6523         * gimple-ssa-evrp-analyze.c
6524         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
6525         (evrp_range_analyzer::set_ssa_range_info): Same.
6526         (evrp_range_analyzer::record_ranges_from_phis): Same.
6527         (evrp_range_analyzer::record_ranges_from_stmt): Same.
6528         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
6529         * gimple-ssa-sprintf.c (get_int_range): Same.
6530         (format_integer): Same.
6531         (sprintf_dom_walker::handle_gimple_call): Same.
6532         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
6533         (ipcp_vr_lattice::top_p): Same.
6534         (ipcp_vr_lattice::bottom_p): Same.
6535         (ipcp_vr_lattice::set_to_bottom): Same.
6536         (ipa_vr_operation_and_type_effects): Same.
6537         (propagate_vr_across_jump_function): Same.
6538         (ipcp_store_vr_results): Same.
6539         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
6540         (ipa_print_node_jump_functions_for_edge): Same.
6541         (ipa_get_value_range): Same.
6542         (ipa_compute_jump_functions_for_edge): Same.
6543         (ipa_write_jump_function): Same.
6544         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
6545         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
6546         Same.
6547         * vr-values.c (set_value_range_to_nonnegative): Same.
6548         (set_value_range_to_truthvalue): Same.
6549         (vr_values::get_value_range): Same.
6550         (vr_values::set_defs_to_varying): Same.
6551         (vr_values::update_value_range): Same.
6552         (symbolic_range_based_on_p): Same.
6553         (vr_values::op_with_boolean_value_range_p): Same.
6554         (vr_values::extract_range_for_var_from_comparison_expr): Same.
6555         (vr_values::extract_range_from_ssa_name): Same.
6556         (vr_values::extract_range_from_binary_expr): Same.
6557         (vr_values::extract_range_from_unary_expr): Same.
6558         (vr_values::extract_range_from_cond_expr): Same.
6559         (vr_values::extract_range_from_comparison): Same.
6560         (vr_values::check_for_binary_op_overflow): Same.
6561         (vr_values::extract_range_basic): Same.
6562         (vr_values::extract_range_from_assignment): Same.
6563         (compare_ranges): Same.
6564         (compare_range_with_value): Same.
6565         (vr_values::adjust_range_with_scev): Same.
6566         (vrp_valueize): Same.
6567         (vrp_valueize_1): Same.
6568         (vr_values::get_vr_for_comparison): Same.
6569         (vr_values::compare_name_with_value): Same.
6570         (vr_values::compare_names): Same.
6571         (vr_values::vrp_evaluate_conditional): Same.
6572         (find_case_label_ranges): Same.
6573         (vr_values::vrp_visit_switch_stmt): Same.
6574         (vr_values::extract_range_from_phi_node): Same.
6575         (vr_values::simplify_div_or_mod_using_ranges): Same.
6576         (vr_values::simplify_bit_ops_using_ranges): Same.
6577         (test_for_singularity): Same.
6578         (range_fits_type_p): Same.
6579         (vr_values::simplify_cond_using_ranges_1): Same.
6580         (vr_values::simplify_switch_using_ranges): Same.
6581         (vr_values::simplify_float_conversion_using_ranges): Same.
6582         (vr_values::two_valued_val_range_p): Same.
6583         (vr_values::add_equivalence): Move to value_range::equiv_add.
6584         * vr-values.h (vr_values::add_equivalence): Remove.
6585         (VR_INITIALIZER): Remove.
6586         * tree-vrp.c (value_range::set): New.
6587         (value_range::equiv_add): New.
6588         (value_range::value_range): New.
6589         (value_range::deep_copy): New.
6590         (value_range::check): New.
6591         (value_range::equal_p): New.
6592         (value_range::ignore_equivs_equal_p): New.
6593         (value_range::operator==): New.
6594         (value_range::operator!=): New.
6595         (value_range::symbolic_p): New.
6596         (value_range::numeric_p): New.
6597         (value_range::set_undefined): New.
6598         (value_range::set_varying): New.
6599         (value_range::may_contain_p): New.
6600         (value_range::equiv_clear): New.
6601         (value_range::singleton_p): New.
6602         (value_range::intersect): New.
6603         (value_range::dump): New.
6604         (value_range::set_and_canonicalize): New.
6605         (set_value_range): Adjust for value_range API.
6606         (set_value_range_to_undefined): Same.
6607         (set_value_range_to_varying): Same.
6608         (set_and_canonicalize_value_range): Same.
6609         (set_value_range_to_nonnull): Same.
6610         (set_value_range_to_null): Same.
6611         (range_is_null): Same.
6612         (range_is_nonnull): Same.
6613         (range_int_cst_p): Same.
6614         (range_int_cst_singleton_p): Same.
6615         (symbolic_range_p): Same.
6616         (range_includes_zero_p): Same.
6617         (value_range_constant_singleton): Same.
6618         (vrp_set_zero_nonzero_bits): Same.
6619         (ranges_from_anti_range): Same.
6620         (extract_range_into_wide_ints): Same.
6621         (extract_range_from_multiplicative_op): Same.
6622         (set_value_range_with_overflow): Same.
6623         (extract_range_from_binary_expr_1): Same.
6624         (extract_range_from_unary_expr): Same.
6625         (dump_value_range): Same.
6626         (debug_value_range): Same.
6627         (vrp_prop::check_array_ref): Same.
6628         (vrp_prop::check_mem_ref): Same.
6629         (vrp_prop::vrp_initialize): Same.
6630         (vrp_prop::visit_stmt): Same.
6631         (intersect_ranges): Same.
6632         (vrp_prop::visit_phi): Same.
6633         (vrp_prop::vrp_finalize): Same.
6634         (determine_value_range_1): Same.
6635         (determine_value_range): Same.
6636         (vrp_intersect_ranges_1): Rename to...
6637         (vrp_intersect_1): this.
6638         (vrp_intersect_ranges): Rename to...
6639         (value_range::intersect_helper): ...this.
6640         (vrp_meet_1): Rename to...
6641         (value_range::union_helper): ...this.
6642         (vrp_meet): Rename to...
6643         (value_range::union_): ...this.
6644         (copy_value_range): Remove.
6645         * tree-vrp.h (struct value_range): Rewrite into a proper class.
6646         (value_range::vrtype): New.
6647         (value_range::type): New.
6648         (value_range::equiv): New.
6649         (value_range::min): New.
6650         (value_range::max): New.
6651         (value_range::varying_p): New.
6652         (value_range::undefined_p): New.
6653         (value_range::null_p): New.
6654         (value_range::equiv_add): New.
6655         (copy_value_range): Remove.
6657 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
6659         * Makefile.in (SELFTEST_TARGETS): New.
6660         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
6661         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
6662         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
6663         c/Make-lang.in.
6664         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
6665         (selftest-c++-gdb, selftest-c++-valgrind): Move to
6666         cp/Make-lang.in.
6667         * configure: Regenerate.
6668         * configure.ac (selftest_languages): New.
6670 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
6672         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
6673         overflow wraps argument.
6674         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
6675         wide_int_range_multiplicative_op.
6676         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
6677         overflow wraps argument.
6678         (wide_int_range_multiplicative_op): Same.
6679         (wide_int_range_lshift): Same.
6680         (wide_int_range_div): Same.
6681         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
6682         (wide_int_range_lshift): Same.
6683         (wide_int_range_div): Same.
6685 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
6687         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
6688         use sign as argument.
6689         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
6690         wide_int_range_shift_undefined_p.
6692 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6694         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
6695         Rename to...
6696         (@despeculate_copy<ALLI_TI:mode>): ... This.
6697         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
6698         switch statement.
6700 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6702         * config.gcc: Obsolete *-*-solaris2.10*.
6703         * doc/install.texi (Specific, *-*-solaris2*): Document it.
6705 2018-10-12  Jeff Law  <law@redhat.com>
6707         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
6708         reg + sym +- const_int addressing modes.
6710 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
6712         * common.opt (fdiagnostics-minimum-margin-width=): New option.
6713         * diagnostic-show-locus.c (layout::layout): Apply the minimum
6714         margin width.
6715         (layout::start_annotation_line): Only print up to 3 of the
6716         margin character, to avoid touching the left-hand side.
6717         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
6718         minimum margin width, as set by test_diagnostic_context's ctor.
6719         (selftest::test_fixit_insert_containing_newline): Likewise.
6720         (selftest::test_fixit_insert_containing_newline_2): Likewise.
6721         (selftest::test_line_numbers_multiline_range): Clear
6722         dc.min_margin_width.
6723         * diagnostic.c (diagnostic_initialize): Initialize
6724         min_margin_width.
6725         * diagnostic.h (struct diagnostic_context): Add field
6726         "min_margin_width".
6727         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
6728         * opts.c (common_handle_option): Handle
6729         OPT_fdiagnostics_minimum_margin_width_.
6730         * selftest-diagnostic.c
6731         (selftest::test_diagnostic_context::test_diagnostic_context):
6732         Initialize min_margin_width to 6.
6733         * toplev.c (general_init): Initialize global_dc->min_margin_width.
6735 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
6737         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
6738         Fix usage of "error_at_rich_loc" in the comment.
6740 2018-10-15  Renlin Li  <renlin.li@arm.com>
6742         PR target/87563
6743         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
6744         if-conversioned loop when it contains ifn with types not
6745         supported by backend.
6746         * internal-fn.c (expand_direct_optab_fn): Add an assert.
6747         (direct_internal_fn_supported_p): New helper function.
6748         * internal-fn.h (direct_internal_fn_supported_p): Declare.
6750 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
6752         PR target/87572
6753         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
6754         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
6755         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
6757 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
6759         PR tree-optimization/87022
6760         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
6761         bits in dist vector rather than the first one.
6763 2018-10-15  Richard Biener  <rguenther@suse.de>
6765         PR middle-end/87610
6766         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
6767         (visit_loadstore): When a used restrict tag escaped verify that
6768         the points-to solution of "other" pointers do not include
6769         escaped.
6770         (compute_dependence_clique): If a used restrict tag escaped
6771         communicated that down to visit_loadstore.
6773 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
6775         * config/s390/s390.c (s390_expand_vec_init): Force vector element
6776         into reg if it isn't a general operand.
6778 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
6780         PR target/87599
6781         * config/i386/sse.md (*vec_dupv2di): Add register source to
6782         movddup.
6784 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
6786         PR target/87572
6787         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
6788         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
6789         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
6790         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
6792 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
6794         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
6795         (notice_args_size): Set it in the current trace if no insn that can
6796         throw internally has been seen yet.
6797         (connect_traces): When connecting args_size between traces, allow the
6798         incoming values not to match if there is an insn setting it before the
6799         first insn that can throw internally; in that case, force the creation
6800         of a CFI note on this latter insn.
6802 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
6804         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
6805         the base class, because using the injected-class-name was not clearly
6806         specified until DR 176.
6808 2018-10-12  Paul Koning  <ni1d@arrl.net>
6810         * config/pdp11/pdp11.md (doloop_end): New expander.
6811         (doloop_end_insn): renamed from "doloop_end".
6812         (addqi3): New pattern.
6813         (subqi3): New pattern.
6814         * config/pdp11/predicates.md (incdec_operand): New predicate.
6816 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
6818         PR middle-end/81376
6819         * real.c (format_helper::can_represent_integral_type_p): New function
6820         * real.h (format_helper::can_represent_integral_type_p): Ditto.
6821         * match.pd: New pattern.
6823 2018-10-12  Alexandre Oliva <oliva@adacore.com>
6825         * configure.ac: Introduce --enable-large-address-aware
6826         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
6827         * doc/install.texi: Document it.
6828         * configure, config.in: Rebuilt.
6829         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
6830         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
6831         (LINK_SPEC): Insert it.
6832         * config/i386/mingw-w64.h: Likewise.
6834         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
6836 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
6838         PR rtl-optimization/87600
6839         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
6841 2018-10-12  Paul Koning  <ni1d@arrl.net>
6843         * doc/md.texi (doloop_end): Document that the pattern code may
6844         need to check operand mode.
6846 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
6848         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
6849         to zero-extend between int and floating-point registers.
6850         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
6851         ldp into floating-point registers.  Add type and arch attributes.
6852         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
6853         Use f_loads for type attribute.
6855 2018-10-11  Martin Sebor  <msebor@redhat.com>
6857         * doc/extend.texi (attribute packed): Correct typos.
6859 2018-10-11  Martin Sebor  <msebor@redhat.com>
6861         * doc/extend.texi (attribute flatten): Mention interaction with
6862         noinline.
6864 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
6866         PR target/87156
6867         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6868         Set new_decl virtual flag to zero.
6870 2018-10-11  Martin Sebor  <msebor@redhat.com>
6872         PR middle-end/87593
6873         * doc/extend.texi (attribute format_arg): Discuss using multiple
6874         attributes on a single function.
6876 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6878         PR tree-optimization/86829
6879         * match.pd (sin (atan (x))): New simplification rules.
6880         (cos (atan (x))): Likewise.
6881         * real.c (build_sinatan_real): New function.
6882         * real.h (build_sinatan_real): Prototype.
6884 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
6886         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
6887         function.
6888         (fold_mergeeo_helper): New helper function.
6889         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
6890         intrinsics.  Correct some whitespace indentation issues.
6892 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
6894         PR target/87511
6895         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
6896         Use HOST_WIDE_INT_1U for shift.
6898 2018-10-11  Doug Rupp  <rupp@adacore.com>
6899             Olivier Hainque  <hainque@adacore.com>
6901         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
6902         Pass --relax to the linker for RTPs.
6903         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
6905 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
6906             Jan Hubicka  <jh@suse.cz>
6907             Martin Jambor  <mjambor@suse.cz>
6909         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
6910         the same elements are repeated rather than printing all of them.
6911         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
6912         "repeated" elements.
6913         * read-rtl-function.c (test_loading_repeat): New function.
6914         (read_rtl_function_c_tests): Call test_loading_repeat.
6915         * rtl-tests.c (test_dumping_repeat): New function.
6916         (rtl_tests_c_tests): Call test_dumping_repeat.
6918 2018-10-11  Richard Biener  <rguenther@suse.de>
6920         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
6921         bdver?_cost): Unify to ...
6922         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
6923         * config/i386/i386.c (processor_cost_table): Adjust.
6925 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
6927         PR middle-end/87574
6928         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
6929         the thunk when expanding to GIMPLE.
6931 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6933         PR bootstrap/87551
6934         * varasm.c (mergeable_string_section): Don't try to move zero-length
6935         strings to the merge section.
6937 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
6939         PR target/87573
6940         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
6942 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
6944         PR target/87550
6945         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
6946         to special_args set.
6948 2018-10-10  Richard Biener  <rguenther@suse.de>
6950         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
6951         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
6952         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
6953         and recursing and pattern terminating the recursion on SSE
6954         vector width using ix86_expand_reduc.
6955         (reduc_sminmax_scal_<mode>): Split into part reducing to half
6956         width and recursing and SSE2 vector variant doing the final
6957         reduction with ix86_expand_reduc.
6958         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
6959         with terminating the recursion at AVX level, splitting that
6960         to SSE there.
6962 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
6964         * genmatch.c (error_cb): Rename to...
6965         (diagnostic_cb): ...this, converting int params to enums.
6966         (fatal_at): Update for renaming.
6967         (warning_at): Likewise.
6968         (main): Likewise.
6969         * input.c (selftest::ebcdic_execution_charset::apply):
6970         Update for renaming of...
6971         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
6972         to...
6973         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
6974         converting level and reason to enums.
6975         (class selftest::lexer_error_sink): Rename to...
6976         (class selftest::lexer_test_options): ...this, renaming field
6977         "m_errors" to "m_diagnostics".
6978         (selftest::lexer_test_options::apply): Update for renaming of...
6979         (selftest::lexer_test_options::on_error): ...this, renaming to...
6980         (selftest::lexer_test_options::on_diagnostic): ...this
6981         converting level and reason to enums.
6982         (selftest::test_lexer_string_locations_raw_string_unterminated):
6983         Update for renamings.
6984         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
6985         "reason".
6987 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
6989         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
6990         * config/rs6000/pmmintrin.h: New file.
6992 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
6994         PR tree-optimization/86659
6995         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
6997 2018-10-09  Richard Biener  <rguenther@suse.de>
6999         PR tree-optimization/63155
7000         * tree-ssa-structalias.c: Include tree-ssa.h.
7001         (get_constraint_for_ssa_var): For undefs return nothing_id.
7002         (find_func_aliases): Cleanup PHI handling.
7004 2018-10-09  Richard Biener  <rguenther@suse.de>
7006         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
7007         replacements.
7009 2018-10-09  Martin Liska  <mliska@suse.cz>
7011         * asan.c (asan_emit_stack_protection): If a stack variable
7012         is located in a same file as current function, then emit
7013         line info into variable definition string.
7015 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7017         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
7018         information.
7020 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7022         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
7023         on the thunk.
7025 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7027         PR tree-optimization/86659
7028         * gimple-match.h (struct gimple_match_op): Add reverse field.
7029         (gimple_match_op::set_op): New overloaded method.
7030         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
7031         the REF_REVERSE_STORAGE_ORDER flag on the value.
7032         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
7033         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
7035 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
7037         PR middle-end/63155
7038         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
7039         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
7041 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
7043         PR target/87517
7044         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
7045         Defined with __builtin_ia32_vfmaddsubpd512_mask.
7047 2018-10-08  Richard Biener  <rguenther@suse.de>
7049         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
7050         cost the same as AVX128 ones.
7052 2018-10-08  Paul Koning  <ni1d@arrl.net>
7054         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
7055         (expand_block_move): New function.
7056         * config/pdp11/pdp11.c (output_block_move): Remove.
7057         (expand_block_move): New function.
7058         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
7059         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
7060         (*movmemhi1): Remove.
7062 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
7064         * config/s390/2827.md: Increase latencies for some FP instructions.
7066 2018-10-08  Richard Biener  <rguenther@suse.de>
7068         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7069         Open a dump scope.
7070         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
7071         * tree-vectorizer.h (dump_stmt_cost): Adjust.
7072         (add_stmt_cost): Dump return value of the hook.
7074 2018-10-08  Richard Biener  <rguenther@suse.de>
7076         PR tree-optimization/63155
7077         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
7078         (ssa_propagation_engine::ssa_propagate): Remove redundant
7079         bitmap bit clearing.
7081 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
7083         PR rtl-optimization/86939
7084         PR rtl-optimization/87479
7085         * ira.h (non_conflicting_reg_copy_p): New prototype.
7086         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
7087         (make_hard_regno_dead): Don't add conflicts for register
7088         ignore_reg_for_conflicts.
7089         (make_object_dead): Likewise.
7090         (non_conflicting_reg_copy_p): New function.
7091         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
7092         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
7093         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
7094         (make_hard_regno_dead): Don't add conflicts for register
7095         ignore_reg_for_conflicts.  Remove special conflict handling of
7096         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
7097         check_pic_pseudo_p and update callers.
7098         (mark_pseudo_dead): Don't add conflicts for register
7099         ignore_reg_for_conflicts.
7100         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
7102 2018-10-05  Andrew Waterman  <andrew@sifive.com>
7103             Jim Wilson  <jimw@sifive.com>
7105         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
7106         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
7107         new pattern using HONOR_SNANS that emits one extra instruction.
7109 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
7111         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
7112         patterns): Merge SI and DI patterns to a GPR pattern.
7113         (unnamed define_insn and define_split for record form of that): Merge
7114         to a single define_insn_and_split pattern.
7116 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
7118         PR c++/56856
7119         * input.c (expand_location_to_spelling_point): Add param "aspect"
7120         and use rather than hardcoding LOCATION_ASPECT_CARET.
7121         (get_substring_ranges_for_loc): Handle the case of a single token
7122         within a macro expansion.
7123         * input.h (expand_location_to_spelling_point): Add "aspect" param,
7124         defaulting to LOCATION_ASPECT_CARET.
7126 2018-10-05  Paul Koning  <ni1d@arrl.net>
7128         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
7129         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
7130         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
7131         (pdp11_guard_type): New function.
7133 2018-10-05  Paul Koning  <ni1d@arrl.net>
7135         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
7136         * config/pdp11/pdp11.opt (mfloat32): Remove.
7137         (mfloat64): Remove.
7138         * doc/invoke.texi (pdp11 -mfloat32): Remove:
7139         (pdp11 -mfloat64): Remove.
7141 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7143         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
7144         (*cmp<mode>_cc_i387): Ditto.
7145         (*cmpu<mode>_cc_i387): Ditto.
7146         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
7147         * config/i386/i386.c (ix86_expand_fp_compare): Remove
7148         "scratch" argument.
7149         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
7150         Emit x86_sahf_1 pattern.
7151         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
7152         (ix86_expand_carry_flag_compare): Ditto.
7154 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7156         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
7157         to reg_or_0_operand.  Add "C" constraint.
7158         (*cmpxf_cc_i387): Ditto.
7159         (*cmp<mode>_i387): Change operand 2 predicate
7160         to nonimm_or_0_operand.  Add "C" constraint.
7161         (*cmp<mode>_cc_i387): Ditto.
7162         (*cmp<mode>_0_i387): Remove insn pattern.
7163         (*cmp<mode>_0_cc_i387): Ditto.
7165 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7167         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
7168         * config/i386/predicates.md (nonimm_or_0_operand): Rename
7169         from vector_move_operand.  Update all uses.
7171 2018-10-05  Martin Sebor  <msebor@redhat.com>
7173         PR tree-optimization/87490
7174         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
7175         consistently.
7177 2018-10-05  Richard Biener  <rguenther@suse.de>
7179         PR tree-optimization/63155
7180         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
7181         vertical space in dumpfiles.
7182         * tree-ssa-propagate.h
7183         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
7184         * tree-ssa-propagate.c (cfg_blocks_back): New global.
7185         (ssa_edge_worklist_back): Likewise.
7186         (curr_order): Likewise.
7187         (cfg_blocks_get): Remove abstraction.
7188         (cfg_blocks_add): Likewise.
7189         (cfg_blocks_empty_p): Likewise.
7190         (add_ssa_edge): Add to current or next worklist based on
7191         RPO index.
7192         (add_control_edge): Likewise.
7193         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
7194         into ...
7195         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
7196         iteration from CFG and SSA edge worklist so we process
7197         everything in RPO order, prioritizing forward progress
7198         over iteration.
7199         (ssa_prop_init): Allocate new worklists, do not dump
7200         immediate uses.
7201         (ssa_prop_fini): Free new worklists.
7203 2018-10-05  Richard Biener  <rguenther@suse.de>
7205         * tree-core.h (tree_block::abstract_flag): Remove.
7206         (tree_block::block_num): Make full 32bits.
7207         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
7208         * tree.h (BLOCK_ABSTRACT): Remove.
7209         * dwarf2out.c (gen_lexical_block_die): Remove dead code
7210         resulting from BLOCK_ABSTRACT being always false.
7211         (gen_inlined_subroutine_die): Likewise.
7212         (gen_block_die): Likewise.
7213         * tree.c (block_ultimate_origin): Likewise.
7214         * tree-pretty-print.c (dump_block_node): Remove code dealing
7215         with BLOCK_ABSTRACT.
7216         * tree-ssa-live.c (dump_scope_block): Likewise.
7217         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
7218         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
7220 2018-10-05   Richard Biener  <rguenther@suse.de>
7222         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
7223         is asked for initialize mode to the component mode of the
7224         vector type.
7226 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
7228         PR target/87522
7229         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
7230         assembler for -mavx.
7231         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
7233 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
7235         PR target/87509
7236         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
7237         RS6000_BTM_DFP.
7238         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
7239         to be DImode.  When using mffscrn, force the operand to a register.
7241 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
7243         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
7244         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
7245         X87MODEF mode iterator.
7246         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
7247         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
7248         X87MODEF mode iterator.
7250 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
7252         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
7253         -Wno-prio-ctor-dtor.
7255 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7257         * Makefile.in (OBJS): Add opt-problem.o.
7258         * dump-context.h: Include "selftest.h.
7259         (selftest::temp_dump_context): New forward decl.
7260         (class dump_context): Make friend of class
7261         selftest::temp_dump_context.
7262         (dump_context::dump_loc_immediate): New decl.
7263         (class dump_pretty_printer): Move here from dumpfile.c.
7264         (class temp_dump_context): Move to namespace selftest.
7265         (temp_dump_context::temp_dump_context): Add param
7266         "forcibly_enable_dumping".
7267         (selftest::verify_dumped_text):
7268         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
7269         (selftest::verify_item):
7270         (ASSERT_IS_TEXT): Move here from dumpfile.c.
7271         (ASSERT_IS_TREE): Likewise.
7272         (ASSERT_IS_GIMPLE): Likewise.
7273         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
7274         to...
7275         (dump_context::dump_loc_immediate): ...this new function.
7276         (class dump_pretty_printer): Move to dump-context.h.
7277         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
7278         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
7279         (temp_dump_context::temp_dump_context): Move to "selftest"
7280         namespace.  Add param "forcibly_enable_dumping", and use it to
7281         conditionalize the use of m_pp;
7282         (selftest::verify_dumped_text): Make non-static.
7283         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
7284         (selftest::verify_item): Make non-static.
7285         (ASSERT_IS_TEXT): Move to dump-context.h.
7286         (ASSERT_IS_TREE): Likewise.
7287         (ASSERT_IS_GIMPLE): Likewise.
7288         (selftest::test_capture_of_dump_calls): Pass "true" for new
7289         param of temp_dump_context.
7290         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
7291         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
7292         TDF_COMPARE_DEBUG.
7293         * opt-problem.cc: New file.
7294         * opt-problem.h: New file.
7295         * optinfo-emit-json.cc
7296         (selftest::test_building_json_from_dump_calls): Pass "true" for
7297         new param of temp_dump_context.
7298         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
7299         (optinfo::emit_for_opt_problem): New function.
7300         (optinfo::emit): Clarity which emit_item is used.
7301         * optinfo.h (optinfo::get_dump_location): New accessor.
7302         (optinfo::emit_for_opt_problem): New decl.
7303         (optinfo::emit): Make const.
7304         * selftest-run-tests.c (selftest::run_tests): Call
7305         selftest::opt_problem_cc_tests.
7306         * selftest.h (selftest::opt_problem_cc_tests): New decl.
7307         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
7308         bool to opt_result, converting fprintf messages to
7309         opt_result::failure_at calls.  Add "stmt" param for use by the
7310         failure_at calls.
7311         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
7312         (runtime_alias_check_p): Convert return type from bool to
7313         opt_result, converting dump_printf calls to
7314         opt_result::failure_at, using the statement DDR_A for their
7315         location.
7316         (find_data_references_in_stmt): Convert return type from bool to
7317         opt_result, converting "return false" to opt_result::failure_at
7318         with a new message.
7319         * tree-data-ref.h: Include "opt-problem.h".
7320         (dr_analyze_innermost): Convert return type from bool to opt_result,
7321         and add a const gimple * param.
7322         (find_data_references_in_stmt): Convert return type from bool to
7323         opt_result.
7324         (runtime_alias_check_p): Likewise.
7325         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
7326         dr_analyze_innermost.
7327         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
7328         Convert return type from bool to opt_result, adding a message for
7329         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
7330         (vect_analyze_data_ref_dependence): Convert return type from bool
7331         to opt_result.  Change sense of return type from "false"
7332         effectively meaning "no problems" to "false" meaning a problem,
7333         so that "return false" becomes "return opt_result::success".
7334         Convert "return true" calls to opt_result::failure_at, using
7335         the location of statement A rather than vect_location.
7336         (vect_analyze_data_ref_dependences): Convert return type from bool
7337         to opt_result.
7338         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
7339         calls to opt_result::failure_at, using the stmt location rather
7340         than vect_location.
7341         (vect_verify_datarefs_alignment): Convert return type from bool
7342         to opt_result.
7343         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
7344         into multiple more-tightly-scoped copies.
7345         (vect_analyze_data_refs_alignment): Convert return type from bool
7346         to opt_result.
7347         (vect_analyze_data_ref_accesses): Likewise, converting a
7348         "return false" to a "return opt_result::failure_at", adding a
7349         new message.
7350         (vect_prune_runtime_alias_test_list): Convert return type from
7351         bool to opt_result, converting dump_printf_loc to
7352         opt_result::failure_at.  Add a %G to show the pertinent statement,
7353         and use the stmt's location rather than vect_location.
7354         (vect_find_stmt_data_reference): Convert return type from
7355         bool to opt_result, converting dump_printf_loc to
7356         opt_result::failure_at, using stmt's location.
7357         (vect_analyze_data_refs):  Convert return type from bool to
7358         opt_result.  Convert "return false" to "return
7359         opt_result::failure_at", adding messages as needed.
7360         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
7361         type from bool to opt_result.
7362         (vect_determine_vf_for_stmt): Likewise.
7363         (vect_determine_vectorization_factor): Likewise, converting
7364         dump_printf_loc to opt_result::failure_at, using location of phi
7365         rather than vect_location.
7366         (vect_analyze_loop_form_1): Convert return type from bool to
7367         opt_result, converting dump_printf_loc calls, retaining the use of
7368         vect_location.
7369         (vect_analyze_loop_form): Convert return type from loop_vec_info
7370         to opt_loop_vec_info.
7371         (vect_analyze_loop_operations): Convert return type from bool to
7372         opt_result, converting dump_printf_loc calls, using the location
7373         of phi/stmt rather than vect_location where available.  Convert
7374         various "return false" to "return opt_result::failure_at" with
7375         "unsupported phi" messages.
7376         (vect_get_datarefs_in_loop): Convert return type from bool to
7377         opt_result.  Add a message for the
7378         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
7379         (vect_analyze_loop_2): Convert return type from bool to
7380         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
7381         each "goto again;", adding new messages where needed.
7382         Add "unsupported grouped {store|load}" messages.
7383         (vect_analyze_loop): Convert return type from loop_vec_info to
7384         opt_loop_vec_info.
7385         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
7386         bool to opt_result.
7387         * tree-vect-stmts.c (process_use): Likewise, converting
7388         dump_printf_loc call and using stmt location, rather than
7389         vect_location.
7390         (vect_mark_stmts_to_be_vectorized): Likeise.
7391         (vect_analyze_stmt): Likewise, adding a %G.
7392         (vect_get_vector_types_for_stmt): Convert return type from bool to
7393         opt_result, converting dump_printf_loc calls and using stmt
7394         location, rather than vect_location.
7395         (vect_get_mask_type_for_stmt): Convert return type from tree to
7396         opt_tree, converting dump_printf_loc calls and using stmt location.
7397         * tree-vectorizer.c: Include "opt-problem.h.
7398         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
7399         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
7400         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
7401         enabled, use it to report at the top level "couldn't vectorize
7402         loop" followed by the problem.
7403         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
7404         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
7405         to opt_result.
7406         (vect_analyze_stmt): Likewise.
7407         (vect_get_vector_types_for_stmt): Likewise.
7408         (tree vect_get_mask_type_for_stmt): Likewise.
7409         (vect_analyze_data_ref_dependences): Likewise.
7410         (vect_enhance_data_refs_alignment): Likewise.
7411         (vect_analyze_data_refs_alignment): Likewise.
7412         (vect_verify_datarefs_alignment): Likewise.
7413         (vect_analyze_data_ref_accesses): Likewise.
7414         (vect_prune_runtime_alias_test_list): Likewise.
7415         (vect_find_stmt_data_reference): Likewise.
7416         (vect_analyze_data_refs): Likewise.
7417         (vect_analyze_loop): Convert return type from loop_vec_info to
7418         opt_loop_vec_info.
7419         (vect_analyze_loop_form): Likewise.
7420         (vect_analyze_slp): Convert return type from bool to opt_result.
7422 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7424         * doc/invoke.texi (-fopt-info): Document new "internals"
7425         sub-option.
7426         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
7427         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
7428         MSG_ALL_KINDS.
7429         (optinfo_verbosity_options): Add "internals".
7430         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
7431         (dump_context::apply_dump_filter_p): New member function.
7432         (dump_context::dump_loc): Use apply_dump_filter_p rather than
7433         explicitly masking the dump_kind.
7434         (dump_context::begin_scope): Increment the scope depth first.  Use
7435         apply_dump_filter_p rather than explicitly masking the dump_kind.
7436         (dump_context::emit_item): Use apply_dump_filter_p rather than
7437         explicitly masking the dump_kind.
7438         (dump_dec): Likewise.
7439         (dump_hex): Likewise.
7440         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
7441         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
7442         (opt_info_switch_p): Update handling of default
7443         MSG_OPTIMIZED_LOCATIONS to cope with default of
7444         MSG_PRIORITY_USER_FACING.
7445         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
7446         masking the dump_kind.
7447         (selftest::test_capture_of_dump_calls): Update test_dump_context
7448         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
7449         than MSG_ALL.  Generalize scope test to be run at all four
7450         combinations of with/without MSG_PRIORITY_USER_FACING and
7451         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
7452         for each of the two values.
7453         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
7454         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
7455         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
7456         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
7457         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
7458         with MSG_PRIORITY_*.
7459         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
7460         dump messages as MSG_PRIORITY_USER_FACING.
7461         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
7462         about the interaction with MSG_PRIORITY_*.
7464 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7466         * varasm.c (output_constant): Add new parameter merge_strings.
7467         Make strings properly zero terminated in merge string sections.
7468         (mergeable_string_section): Don't fail if the last char is non-zero.
7469         (assemble_variable_contents): Handle merge string sections.
7470         (assemble_variable): Likewise.
7471         (assemble_constant_contents): Likewise.
7472         (output_constant_def_contents): Likewise.
7473         (output_constructor_array_range,
7474         output_constructor_regular_field): Adjust call to output_constant.
7475         (output_object_block): Adjust call to assemble_constant_contents
7476         and assemble_variable_contents.
7478 2018-10-04  Martin Liska  <mliska@suse.cz>
7480         PR c/87483
7481         * cgraphunit.c (process_function_and_variable_attributes):
7482         Warn about a function with alias attribute and a body.
7484 2018-10-04  Martin Liska  <mliska@suse.cz>
7486         PR ipa/82625
7487         * multiple_target.c (redirect_to_specific_clone): New function.
7488         (ipa_target_clone): Use it.
7489         * tree-inline.c: Fix comment.
7491 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7493         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
7494         fields.
7495         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
7496         (gcc::dump_manager::register_pass): New member function, adapted
7497         from loop body in gcc::pass_manager::register_pass, adding a
7498         call to update_dfi_for_opt_info.
7499         (gcc::dump_manager::opt_info_enable_passes): Store the
7500         -fopt-info options into the new fields.  Move the loop
7501         bodies into...
7502         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
7503         function.
7504         * dumpfile.h (struct opt_pass): New forward decl.
7505         (gcc::dump_manager::register_pass): New decl.
7506         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
7507         (class gcc::dump_manager): Add fields "m_optgroup_flags",
7508         "m_optinfo_flags", and "m_optinfo_filename".
7509         * passes.c (gcc::pass_manager::register_pass): Move all of the
7510         dump-handling code to gcc::dump_manager::register_pass.
7512 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
7514         PR rtl-optimization/87466
7515         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
7516         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
7517         * doc/tm.texi: Regenerate.
7518         * ira-lives.c (process_bb_node_lives): Use the new target hook.
7519         * lra-lives.c (process_bb_lives): Likewise.
7520         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
7521         Define.
7523 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
7525         * params.c (add_params): Fix initialization.
7527 2018-10-04  Martin Liska  <mliska@suse.cz>
7529         PR gcov-profile/84107
7530         * tree-profile.c (init_ic_make_global_vars):
7531         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
7532         Come up with new ic_tuple* variables.  Emit
7533         __gcov_indirect_call{,_topn} variables.
7534         (gimple_gen_ic_profiler): Access the variable
7535         and emit gimple.
7536         (gimple_gen_ic_func_profiler): Access
7537         __gcov_indirect_call.callee field.
7538         (gimple_init_gcov_profiler): Use ptr_type_node.
7539         * value-prof.c (gimple_ic): Use ptr_type_node.
7541 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
7543         PR tree-optimization/85787
7544         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
7545         malloc_candidate_p into this function and add support for
7546         detecting multiple phis.
7547         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
7549 2018-10-04  Martin Liska  <mliska@suse.cz>
7551         PR ipa/87491
7552         * ipa-inline.c (inline_to_all_callers_1):
7553         Call ultimate_alias_target for node being inlined.
7555 2018-10-03  Jeff Law  <law@redhat.com>
7557         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
7558         target's wchar_t.
7559         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
7560         * tree.h (get_typenode_from_name): Prototype.
7562 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
7564         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
7565         Change operand 2 predicate to nonimmediate_operand.
7566         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
7568 2018-10-03  Martin Sebor  <msebor@redhat.com>
7569             Jeff Law  <law@redhat.com>
7571         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
7572         initialize it.
7573         (get_string_length): Detect unterminated arrays.
7574         (format_string): Same.
7575         (format_directive): Warn about unterminated arrays.
7576         (handle_gimple_call): Mark statements with no_warning as needed.
7578 2018-10-03  Jim Wilson  <jimw@sifive.com>
7580         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
7581         also define __riscv_abi_rve.  Delete trailing white space.
7583 2018-10-03  Paul Koning  <ni1d@arrl.net>
7585         Enable LRA register allocator for PDP11.
7586         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
7587         (R): Likewise.
7588         (D): Likewise.
7589         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
7590         * config/pdp11/pdp11.opt (-mlra): New option.
7591         * doc/invoke.texi (PDP-11 Options): Document -mlra.
7593 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
7595         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
7596         (*<absneg:code>extend<mode>xf2): Ditto.
7598 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
7600         PR tree-optimization/87415
7601         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
7602         precision fields.
7604 2018-10-02  Jeff Law  <law@redhat.com>
7606         * gimple-fold.c (get_range_strlen): Only set *nonstr when
7607         an unterminated string is discovered.  Bubble up range
7608         even for unterminated strings.
7609         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
7610         indicates the string was not terminated via NONSTR.
7612 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
7614         * tree-vrp.c (extract_range_from_unary_expr): Special case all
7615         pointer conversions.
7616         Do not do anything special for anti-ranges.
7618 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
7620         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
7621         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
7623 2018-10-03  Martin Liska  <mliska@suse.cz>
7625         PR gcov-profile/86109
7626         * coverage.c (coverage_begin_function): Do not
7627         mark lambdas as artificial.
7628         * tree-core.h (struct GTY): Remove tm_clone_flag
7629         and introduce new lambda_function.
7630         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
7632 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7634         PR target/87474
7635         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
7636         P8_VECTOR and VSX are enabled.
7638 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
7640         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
7641         0x3907 as CPU model number.
7643 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
7645         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
7646         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
7647         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
7648         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
7649         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
7650         * config/s390/s390.md: Likewise. Rename also the cpu attribute
7651         value from arch12 to z14.
7653 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
7655         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
7656         (isinfxf2): Ditto.
7657         (isinf<mode>2): Ditto.
7659 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
7661         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
7662         before emitting fxam.  Perform calculations in XFmode.
7664 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
7666         * match.pd (((X /[ex] A) +- B) * A): New transformation.
7668 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
7670         PR middle-end/87319
7671         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
7672         * tree.c (signed_or_unsigned_type_for): Handle complex.
7674 2018-10-02  Jeff Law  <law@redhat.com>
7676         * gimple-fold.c (get_range_strlen): Remove dead code.
7678 2018-10-02  Martin Sebor  <msebor@redhat.com>
7679             Jeff Law  <law@redhat.com>
7681         * builtins.c (unterminated_array): Add new arguments.
7682         If argument is not terminated, bubble up size and exact
7683         state to callers.
7684         (expand_builtin_strnlen): Detect, avoid expanding
7685         and diagnose unterminated arrays.
7686         (c_strlen): Fill in offset of start of unterminated strings.
7687         * builtins.h (unterminated_array): Update prototype.
7689 2018-10-02  Richard Biener  <rguenther@suse.de>
7691         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
7692         of haddv4df, first reduce to SSE width and exploit the fact
7693         that we only need element zero with the reduction result.
7694         (reduc_plus_scal_v2df): Likewise.
7696 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
7698         * dojump.h (do_jump): Delete.
7699         (do_jump_1): Likewise.
7700         (split_comparison): Move around.
7701         * dojump.c (do_jump): Make static.
7702         (do_jump_1): Likewise.
7703         (jumpifnot): Move around.
7704         (jumpifnot_1): Likewise.
7705         (jumpif): Likewise.
7706         (jumpif_1): Likewise.
7707         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
7709 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
7711         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
7712         insns in the delay slot and add_insn_after for the jump insn.
7714 2018-10-02  Richard Biener  <rguenther@suse.de>
7716         * tree-inline.c (expand_call_inline): Use the location of
7717         the callee declaration for the inline-entry marker.
7718         * final.c (notice_source_line): Remove special-casing of
7719         NOTE_INSN_INLINE_ENTRY.
7721 2018-10-01  Carl Love  <cel@us.ibm.com>
7723         PR 69431
7724         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
7725         (__builtin_mtfsb0): New.
7726         (__builtin_mtfsb1): New.
7727         ( __builtin_set_fpscr_rn): New.
7728         (__builtin_set_fpscr_drn): New.
7729         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
7730         (rs6000_expand_set_fpscr_rn_builtin): Add.
7731         (rs6000_expand_set_fpscr_drn_builtin): Add.
7732         (rs6000_expand_builtin): Add case statement entries for
7733         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
7734         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
7735         RS6000_BUILTIN_MFFSL.
7736         (rs6000_init_builtins): Add ftype initialization and def_builtin
7737         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
7738         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
7739         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
7740         rs6000_mffscdrn): Add define_insn.
7741         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
7742         * doc/extend.texi: Add documentation for the builtins.
7744 2018-10-01  Richard Biener  <rguenther@suse.de>
7746         PR tree-optimization/87465
7747         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
7748         causing branch miscounts.
7750 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7752         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
7753         aarch64_option_default_param):  New.
7754         (params.h): Include.
7755         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
7756         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
7757         stack-clash protection validation code.
7759 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7761         * params.c (validate_param): New.
7762         (add_params): Use it.
7763         (set_param_value): Refactor param validation into validate_param.
7764         (diagnostic.h): Include.
7765         * diagnostic.h (diagnostic_ready_p): New.
7767 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7769         * params.c (set_param_value):
7770         Add index of parameter being validated.
7771         * common/common-target.def (option_validate_param): New.
7772         * common/common-targhooks.h (default_option_validate_param): New.
7773         * common/common-targhooks.c (default_option_validate_param): New.
7774         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
7775         * doc/tm.texi: Regenerate.
7777 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7779         PR target/86486
7780         * config/aarch64/aarch64.c (aarch64_override_options_internal):
7781         Add validation for stack-clash parameters and set defaults.
7783 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7785         PR target/86486
7786         * configure.ac: Add stack-clash-protection-guard-size.
7787         * doc/install.texi: Document it.
7788         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
7789         * params.def: Update comment for guard-size.
7790         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
7791         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
7792         * configure: Regenerate.
7794 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7796         PR target/86486
7797         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
7798         STACK_DYNAMIC_OFFSET): New.
7799         * config/aarch64/aarch64.c (aarch64_layout_frame):
7800         Update outgoing args size.
7801         (aarch64_stack_clash_protection_alloca_probe_range,
7802         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
7804 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7806         PR target/86486
7807         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
7808         probe ranges.
7809         * target.def (stack_clash_protection_alloca_probe_range): New.
7810         (stack_clash_protection_final_dynamic_probe): Remove.
7811         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
7812         (default_stack_clash_protection_final_dynamic_probe): Remove.
7813         * targhooks.c: Likewise.
7814         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
7815         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
7816         * doc/tm.texi: Regenerate.
7818 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7820         PR target/86486
7821         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
7822         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
7823         aarch64_clamp_to_uimm12_shift): New.
7824         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
7825         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
7827 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7829         PR target/86486
7830         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
7832 2018-10-01  Jeff Law  <law@redhat.com>
7833             Richard Sandiford <richard.sandiford@linaro.org>
7834             Tamar Christina  <tamar.christina@arm.com>
7836         PR target/86486
7837         * config/aarch64/aarch64.md
7838         (probe_stack_range): Add k (SP) constraint.
7839         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
7840         STACK_CLASH_MAX_UNROLL_PAGES): New.
7841         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
7842         stack probes for stack clash.
7843         (aarch64_allocate_and_probe_stack_space): New.
7844         (aarch64_expand_prologue): Use it.
7845         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
7846         (aarch64_sub_sp): Add emit_move_imm optional param.
7848 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
7850         PR tree-optimization/87261
7851         * match.pd: Remove trailing whitespace.
7852         Add (x & y) | ~(x | y) -> ~(x ^ y),
7853         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
7855 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
7857         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
7858         constraints.
7859         (ashlsi3_insn): Update instruction constraints.
7860         (ashrsi3_insn): Likewise.
7861         (rotrsi3): Likewise.
7862         (add_shift): Likewise.
7863         * config/arc/constraints.md (Csz): New 32 bit constraint. It
7864         avoids placing in the limm field small constants which, otherwise,
7865         could end into a small instruction.
7867 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
7869         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
7870         destination register is not odd-even.
7871         (umaddsidi4_split): Likewise.
7873 2018-10-01  Richard Biener  <rguenther@suse.de>
7875         * tree-inline.c (expand_call_inline): Store origin of fn
7876         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
7877         * tree.c (block_ultimate_origin): Simplify and do some
7878         checking.
7880 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
7882         * config/i386/mmx.md (EMMS): New int iterator.
7883         (emms): New int attribute.
7884         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
7885         EMMS int iterator.  Explicitly declare clobbers.
7886         (mmx_emms): Remove expander.
7887         (mmx_femms): Ditto.
7888         * config/i386/predicates.md (emms_operation): Remove predicate.
7889         (vzeroall_pattern): New predicate.
7890         (vzeroupper_pattern): Rename from vzeroupper_operation.
7891         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
7892         vzeroupper_pattern and vzeroall_pattern predicates.
7894 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
7896         PR rtl-optimization/86939
7897         * ira-lives.c (make_hard_regno_born): Rename from this...
7898         (make_hard_regno_live): ... to this.  Remove update to conflict
7899         information.  Update function comment.
7900         (make_hard_regno_dead): Add conflict information update.  Update
7901         function comment.
7902         (make_object_born): Rename from this...
7903         (make_object_live): ... to this.  Remove update to conflict information.
7904         Update function comment.
7905         (make_object_dead):  Add conflict information update.  Update function
7906         comment.
7907         (mark_pseudo_regno_live): Call make_object_live.
7908         (mark_pseudo_regno_subword_live): Likewise.
7909         (mark_hard_reg_dead): Update function comment.
7910         (mark_hard_reg_live): Call make_hard_regno_live.
7911         (process_bb_node_lives): Likewise.
7912         * lra-lives.c (make_hard_regno_born): Rename from this...
7913         (make_hard_regno_live): ... to this.  Remove update to conflict
7914         information.  Remove now uneeded check_pic_pseudo_p argument.
7915         Update function comment.
7916         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
7917         to conflict information.  Update function comment.
7918         (mark_pseudo_live): Remove update to conflict information.  Update
7919         function comment.
7920         (mark_pseudo_dead): Add conflict information update.
7921         (mark_regno_live): Call make_hard_regno_live.
7922         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
7923         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
7925 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
7927         PR target/87370
7928         * config/i386/i386.c (construct_container): Use TImode for
7929         BLKmode values in 2 integer registers.
7931 2018-09-29  Jeff Law  <law@redhat.com>
7933         * builtins.c (unterminated_array): Pass in c_strlen_data * to
7934         c_strlen rather than just a tree *.
7935         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
7936         Update recursive calls appropriately.  If caller did not provide a
7937         suitable data pointer, create a local one.  When a non-terminated
7938         string is discovered, bubble up information about the string via the
7939         c_strlen_data object.
7940         * builtins.h (c_strlen): Update prototype.
7941         (c_strlen_data): New structure.
7942         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
7943         For a type 2 call, if c_strlen indicates a non-terminated string
7944         use the length of the non-terminated string.
7945         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
7947 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
7949         PR target/87467
7950         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
7951         __m512d type for __A argument rather than __m512.
7953 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
7955         * match.pd (simple_comparison): Don't optimize if either operand is
7956         a function pointer when target needs function pointer canonicalization.
7958 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
7960         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
7961         power5 .. power9 to remove indirection.
7962         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
7963         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
7964         ASM_CPU_476_SPEC): Delete.
7965         (ASM_CPU_SPEC): Adjust.
7966         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
7967         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
7969 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
7971         * config.in: Delete HAVE_AS_DCI.
7972         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
7973         * config/rs6000/rs6000.h: Ditto.
7974         * configure.ac: Delete HAVE_AS_DCI.
7975         * configure: Regenerate.
7977 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
7979         * config.in (HAVE_AS_LWSYNC): Delete.
7980         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
7981         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
7982         do it as a .long .
7983         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
7984         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
7985         as a .long .
7986         * configure.ac: Delete HAVE_AS_LWSYNC.
7987         * configure: Regenerate.
7989 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
7990             Pierre-Marie de Rodat  <derodat@adacore.com>
7992         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
7993         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
7994         (cgraph_node::create_thunk): Add indirect_offset parameter.
7995         (thunk_adjust): Likewise.
7996         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
7997         and initialize the corresponding field with it.
7998         (cgraph_node::dump): Dump indirect_offset field.
7999         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
8000         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
8001         (thunk_adjust): Add indirect_offset parameter and deal with it.
8002         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
8003         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
8004         or if the thunk is external or local.  Fix formatting.  Do not chain
8005         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
8006         if any, in the GIMPLE representation.
8007         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
8008         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
8009         (input_node): Read indirect_offset field.
8010         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
8011         call to thunk_adjust.
8012         * tree-nested.c (struct nesting_info): Add thunk_p field.
8013         (create_nesting_tree): Set it.
8014         (convert_all_function_calls): Copy static chain from targets to thunks.
8015         (finalize_nesting_tree_1): Return early for thunks.
8016         (unnest_nesting_tree_1): Do not finalize thunks.
8017         (gimplify_all_functions): Do not gimplify thunks.
8019 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
8021         * opt-suggestions.c (option_proposer::build_option_suggestions):
8022         Release "option_values".
8024 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
8026         * coverage.c (get_coverage_counts): Convert problem-reporting dump
8027         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
8028         * dumpfile.c (kind_as_string): New function.
8029         (dump_loc): Rather than a hardcoded prefix of "note: ", use
8030         kind_as_string to vary the prefix based on dump_kind.
8031         (selftest::test_capture_of_dump_calls): Update for above.
8033 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
8035         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
8036         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
8038 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
8040         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
8041         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
8042         INVALID_REGNUM instead of FPSR_REG.
8043         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
8044         * config/i386/i386.md: Update comment of FP compares.
8045         (fldenv): Do not clobber FPSR_REG.
8047 2018-09-28  Richard Biener  <rguenther@suse.de>
8049         * tree.h (BLOCK_ORIGIN): New.
8050         * omp-expand.c (grid_expand_target_grid_body): Assign
8051         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
8052         * tree-inline.c (remap_block): Likewise.
8053         * auto-profile.c (get_function_decl_from_block): Simplify
8054         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
8055         * langhooks.c (lhd_print_error_function): Likewise.
8056         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
8057         Likewise.
8058         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8059         * tree.c (block_nonartificial_location): Likewise.
8060         (block_ultimate_origin): Likewise.
8061         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
8062         no longer needed LTO case.
8064 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
8065             Jan Hubicka  <jh@suse.cz>
8066             Martin Jambor  <mjambor@suse.cz>
8068         * simplify-rtx.c (simplify_merge_mask): New function.
8069         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
8070         same masks are used in op1 or op2.
8071         (test_vec_merge): New function.
8072         (test_vector_ops): Call test_vec_merge.
8074 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
8076         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
8077         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
8078         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
8079         (cypress_costs): Set it.
8080         (supersparc_costs): Likewise.
8081         (hypersparc_costs): Likewise.
8082         (leon_cost): Likewise.
8083         (leon3_costs): Likewise.
8084         (sparclet_costs): Likewise.
8085         (ultrasparc_costs): Likewise.
8086         (ultrasparc_costs): Likewise.
8087         (niagara_costs): Likewise.
8088         (niagara2_costs): Likewise.
8089         (niagara3_costs): Likewise.
8090         (niagara4_costs): Likewise.
8091         (niagara7_costs): Likewise.
8092         (m8_costs): Likewise.
8093         (TARGET_CAN_FOLLOW_JUMP): Define.
8094         (pass_work_around_errata::gate): Minor tweak.
8095         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
8096         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
8097         Automaitcally clear MASK_FSMULD mask for V7 processors.
8098         (sparc_can_follow_jump): New static function.
8099         (output_ubranch): Deal with CROSSING_JUMP_P.
8100         (sparc_use_sched_lookahead): Rewrite using switch statement.
8101         (sparc_issue_rate): Reorder.
8102         (sparc_branch_cost): New function.
8104 2018-09-27  Martin Sebor  <msebor@redhat.com>
8106         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
8107         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
8108         (int_fits_type_p): Same.
8110 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
8112         * config/i386/i386.md (FPCR_REG): Remove.
8113         (UNSPEC_FLDCW): Remove.
8114         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
8115         (x86_fldcw_1): Remove insn pattern.
8116         (fnstenv): Do not clobber FPCR_REG.
8117         (fldenv): Ditto.
8118         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
8119         (CALL_USED_REGISTERS): Ditto.
8120         (REG_ALLOC_ORDER): Ditto.
8121         (REG_CLASS_CONTENTS): Ditto.
8122         (HI_REGISTER_NAMES): Ditto.
8123         (ADDITIONAL_REGISTER_NAMES): Use defines instead
8124         of numerical constants.
8125         * config/i386/i386.c (regclass_map): Remove fpsr register.
8126         (dbx_register_map): Ditto.
8127         (dbx64_register_map): Ditto.
8128         (svr4_dbx_register_map): Ditto.
8129         (print_reg): Do not handle FPCR_REG.
8131 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
8133         PR target/87149
8134         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
8135         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
8136         Delete, always treat as true.
8137         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8138         Ditto.  Simplify remaining code.
8139         * config/powerpcspe/powerpcspe.h: Ditto.
8140         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
8141         Simplify remaining code.
8142         (rs6000_expand_builtin): Ditto.
8143         * config/rs6000/rs6000.h: Ditto.
8144         * configure.ac: Ditto.
8145         * configure: Regenerate.
8147 2018-09-27  Martin Liska  <mliska@suse.cz>
8149         * coverage.c (get_coverage_counts): Revert the formatting
8150         of missing profile opt info.
8152 2018-09-27  Richard Biener  <rguenther@suse.de>
8154         PR debug/37801
8155         PR debug/87440
8156         * dwarf2out.c (set_block_origin_self): Do not mark outermost
8157         block as we do not output that.
8158         (gen_inlined_subroutine_die): Elide the originally outermost
8159         block, matching what we do for concrete instances.
8160         (decls_for_scope): Add parameter specifying whether to recurse
8161         to subblocks.
8163 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
8164             Tom de Vries  <tom@codesourcery.com>
8166         PR 82089
8168         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
8169         STORE_FLAG_VALUE == 1.
8171 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
8173         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
8174         constant definitions.
8175         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
8176         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
8177         ("speculation_barrier"): New expander definition.
8179 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
8181         PR gcov-profile/86957
8182         * common.opt: New warning option -Wmissing-profile.
8183         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
8184         * doc/invoke.texi: Document -Wmissing-profile.
8186 2018-09-26  Jim Wilson  <jimw@sifive.com>
8188         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
8189         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
8191 2018-09-26  Martin Sebor  <msebor@redhat.com>
8193         * tree.c (zerop): Change return type to bool.
8194         (integer_zerop, integer_onep, integer_each_onep): Same.
8195         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
8196         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
8197         (real_onep, real_minus_onep, chain_index): Same.
8198         (print_type_hash_statistics, type_list_equal): Same.
8199         * tree.h (zerop): Same.
8200         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
8201         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
8202         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
8203         (real_onep, real_minus_onep, chain_index): Same.
8204         (print_type_hash_statistics, type_list_equal): Same.
8206 2018-09-26  Jim Wilson  <jimw@sifive.com>
8208         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
8210 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
8212         PR target/87414
8213         * config/i386/i386.c: Include debug.h and dwarf2out.h.
8214         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
8215         call.
8217 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
8219         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
8221 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
8223         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
8224         and FP_SECOND_SSE_REGS.
8225         (REG_CLASS_NAMES): Ditto.
8226         (REG_CLASS_CONTENTS): Ditto.
8227         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
8228         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
8229         (ix86_preferred_output_reload_class): Ditto.
8230         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
8231         clobber constraint to "=&f".
8232         (fix_truncdi_i387): Ditto.
8233         (lrintxfdi2): Ditto.
8234         (fistdi2_<rounding>): Ditto.
8235         (fpremxf4_i387): Change "=u" constraint to "=f".
8236         (fprem1xf4_i387): Ditto.
8237         (sincosxf3): Ditto.
8238         (fptanxf4_i387): Ditto.
8239         (fxtractxf3_i387): Ditto.
8240         (fscalexf4_i387): Ditto.
8241         (atan2xf3): Change "u" constraint to "f".
8242         (fyl2xxf3_i387): Ditto.
8243         (fyl2xp1xf3_i387): Ditto.
8245 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
8247         PR target/87439
8248         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
8249         for removed I387_MASK_PM entity.
8252 2018-09-26  Jeff Law  <law@redhat.com>
8253         Revert
8254         2018-09-26  Alexey Neyman  <stilor@att.net>
8256         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
8257         headers are no longer pulled in by <isl/val.h>.
8259 2018-09-26  Richard Biener  <rguenther@suse.de>
8261         PR debug/87443
8262         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
8263         or concrete instance DIE to the tree.  Create abstract origin
8264         attributes also for concrete instances.
8266 2018-09-26  Alexey Neyman  <stilor@att.net>
8268         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
8269         headers are no longer pulled in by <isl/val.h>.
8271 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
8273         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
8274         Use new helper functions.
8275         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
8276         Use new helper functions.
8277         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
8278         aarch_mm_needs_release): New declarations.
8279         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
8280         aarch_mm_needs_release): New.
8282 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
8284         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
8285         (arm32_output_mi_thunk): Deal with long calls.
8287 2018-09-26  Richard Biener  <rguenther@suse.de>
8289         PR debug/87428
8290         PR debug/87362
8291         * tree-inline.c (expand_call_inline): When the location
8292         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
8293         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
8294         the inserted BLOCK to make inlined_function_outer_scope_p
8295         recognize it.
8296         * dwarf2out.c (add_call_src_coords_attributes): Do not add
8297         coords for reserved locations.
8299 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8301         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
8302         (*call_indirect_nonlocal_sysv<mode>): Ditto.
8303         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
8304         (*sibcall_nonlocal_sysv<mode>): Ditto.
8305         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
8306         (<bd>_<mode>): Ditto.
8307         (<bd>tf_<mode>): Ditto.
8309 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8311         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
8312         control string as a list of templates instead of as C code.
8313         (*altivec_movti): Ditto.
8314         * config/rs6000/darwin.md (movdf_low_di): Ditto.
8316 2018-09-25  Jim Wilson  <jimw@sifive.com>
8318         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
8319         when target symbol is weak.
8321 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8323         PR c/87387
8324         * builtins.c (unterminated_array): Simplify.
8325         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
8326         where pointer arithmetic is safe.
8328 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8330         PR target/86987
8331         * config/rs6000/altivec.md (altivec_vspltb): Use
8332         const_0_to_15_operand instead of u5bit_cint_operand.
8333         (*altivec_vspltb_internal): Ditto.
8334         (altivec_vspltb_direct): Ditto.
8335         (altivec_vsplth): Use const_0_to_7_operand instead of
8336         u5bit_cint_operand.
8337         (*altivec_vsplth_internal): Ditto.
8338         (altivec_vsplth_direct): Ditto.
8339         (altivec_vspltw): Use const_0_to_3_operand instead of
8340         u5bit_cint_operand.
8341         (*altivec_vspltw_internal): Ditto.
8342         (altivec_vspltw_direct): Ditto.
8343         (altivec_vspltsf): Ditto.
8344         (*altivec_vspltsf_internal): Ditto.
8345         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
8346         various splats with the proper size immediate.  Reorder the various
8347         cases by ascending size of immediate, and put all such together.
8349 2018-09-25  Richard Biener  <rguenther@suse.de>
8351         PR debug/83941
8352         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
8353         GC-ification.
8354         (maybe_create_die_with_external_ref): Do not create
8355         DW_TAG_imported_unit here.
8356         (add_abstract_origin_attribute): Handle external BLOCK refs.
8357         (dwarf2out_abstract_function): Simplify LTO case.
8358         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
8359         rather than using maybe_create_die_with_external_ref.
8361 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
8363         PR target/71278
8364         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
8365         (frndintxf2_mask_pm_i387): Ditto.
8366         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
8367         Enable for !flag_trapping_math.
8368         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
8369         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
8370         Change operand 1 predicate to nonimmediate_operand.
8371         (attr "i387_cw"): Remove mask_pm.
8372         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
8373         (enum ix86_entity): Remove I387_MASK_PM.
8374         * config/i386/i386.c (ix86_i387_mode_needed): Do not
8375         handle I387_MASK_PM.
8376         (ix86_mode_needed): Ditto.
8377         (ix86_mode_after): Ditto.
8378         (ix86_mode_entry): Ditto.
8379         (ix86_mode_exit): Ditto.
8380         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
8382 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
8384         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
8385         to_update_switch_stmts to vNULL instead of calling create on them
8386         immediately.
8388 2018-09-25  Richard Biener  <rguenther@suse.de>
8390         PR tree-optimization/87402
8391         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
8392         (visit_phi): Re-instantiate handling of supposed to be VARYING
8393         but non-VARYING backedge value.
8395 2018-09-25  Richard Biener  <rguenther@suse.de>
8397         PR debug/83941
8398         * dwarf2out.c (struct sym_off_pair): New.
8399         (external_die_map): New global.
8400         (lookup_decl_die): When in LTO create DIEs lazily from the
8401         external_die_map.
8402         (lookup_block_die): New function, create DIEs lazily in LTO.
8403         (equate_block_to_die): New function.
8404         (dwarf2out_die_ref_for_decl): During WPA get the association
8405         from the external DIE map.
8406         (dwarf2out_register_external_die): Record mapping into the
8407         external DIE map.
8408         (maybe_create_die_with_external_ref): New function split out from
8409         DIE generation part of old dwarf2out_register_external_die.
8410         (add_abstract_origin_attribute): Do not return the DIE.  When
8411         in LTO reference externals directly.
8412         (dwarf2out_abstract_function): When in LTO ignore calls for
8413         decls with external DIEs (already present abstract instances).
8414         (gen_call_site_die): Adjust.
8415         (add_high_low_attributes): Likewise.
8416         (gen_lexical_block_die): Likewise.
8417         (gen_inlined_subroutine_die): Likewie.
8418         (gen_block_die): Likewise.
8419         (dwarf2out_inline_entry): Likewise.
8420         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
8421         DIEs.
8423 2018-09-25  Martin Liska  <mliska@suse.cz>
8425         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
8426         integers and not by a float value.
8428 2018-09-25  Martin Liska  <mliska@suse.cz>
8430         PR fortran/87394
8431         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
8432         instead of NULL.
8433         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
8434         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
8435         * godump.c (go_format_type): Remove extra parenthesis.
8437 2018-09-25  Martin Liska  <mliska@suse.cz>
8439         * alias.c (set_dest_equal_p): Remove unused function.
8440         * config/i386/i386.c (def_builtin_pure2): Likewise.
8441         * diagnostic-show-locus.c (class layout): Remove
8442         unused field.
8443         (layout::layout): Likewise here.
8444         * dump-context.h (class temp_dump_context): Likewise.
8445         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
8446         (add_AT_loclistsptr): Likewise.
8447         (add_AT_offset): Likewise.
8448         (get_AT_hi_pc): Likewise.
8449         (is_comdat_die): Likewise.
8450         (type_is_enum): Likewise.
8451         (ceiling): Likewise.
8452         (add_AT_vms_delta): Likewise.
8453         (is_class_die): Likewise.
8454         * edit-context.c (class line_event): Remove unused field.
8455         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
8456         unused function.
8457         * ipa-cp.c (ipa_get_vr_lat): Likewise.
8458         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
8459         (ok_for_base_p_nonstrict): Likewise.
8460         * tree-chrec.c (is_not_constant_evolution): Likewise.
8461         (chrec_fold_poly_cst): Likewise.
8462         * tree-if-conv.c (has_pred_critical_p): Likewise.
8463         * tree-ssa-coalesce.c (print_exprs): Likewise.
8464         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
8465         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
8466         * tree-vrp.c (value_ranges_intersect_p): Likewise.
8467         (value_range_nonnegative_p): Likewise.
8469 2018-09-25  Martin Liska  <mliska@suse.cz>
8471         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
8472         Do not handle "GNU Pascal".
8473         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
8474         Likewise.
8475         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
8476         from documentation. Likewise.
8477         * dbxout.c (dbxout_range_type): Likewise.
8478         * doc/cpp.texi: Likewise.
8479         * doc/extend.texi: Likewise.
8480         * doc/frontends.texi: Likewise.
8481         * doc/invoke.texi: Remove Pascal entry.
8482         * tree.def (CLEANUP_POINT_EXPR): Likewise.
8483         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
8485 2018-09-25  Martin Liska  <mliska@suse.cz>
8487         PR middle-end/86078
8488         * doc/invoke.texi: Document all parameters and remove default
8489         of the parameters.
8491 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
8493         PR bootstrap/87417
8494         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
8495         contains HOST_WIDE_INTs when computing its size.
8497 2018-09-24  Jim Wilson  <jimw@sifive.com>
8499         PR target/87391
8500         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
8501         not TARGET_RVE.
8502         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
8504 2018-09-24  Andrew Pinski  <apinski@marvell.com>
8506         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
8507         access prev before checking it for NULLness in the
8508         AARCH64_FUSE_CMP_BRANCH case.
8510 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
8512         PR target/82699
8513         * config/i386/i386.c (rest_of_insert_endbranch): Set
8514         endbr_queued_at_entrance to true and don't insert ENDBR if
8515         x86_function_profiler will be called.
8516         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
8517         is true.
8518         * config/i386/i386.h (machine_function): Add
8519         endbr_queued_at_entrance.
8521 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
8523         * genattrtab.c (mk_attr_alt): Use alternative_mask.
8524         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
8525         types.
8526         (check_attr_test): Use alternative_mask.
8527         (get_attr_value): Likewise.
8528         (compute_alternative_mask): Use alternative_mask and XWINT.
8529         (make_alternative_compare): Use alternative_mask.
8530         (attr_alt_subset_p): Use XWINT.
8531         (attr_alt_subset_of_compl_p): Likewise.
8532         (attr_alt_intersection): Use alternative_mask and XWINT.
8533         (attr_alt_union): Likewise.
8534         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
8535         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
8536         (simplify_test_exp): Use alternative_mask and XWINT.
8537         (write_test_expr): Use alternative_mask and XWINT, adjust bit
8538         number calculation to support 64 bits.  Generate code that
8539         checks 64-bit masks.
8540         (main): Use alternative_mask.
8541         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
8543 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
8545         PR target/80080
8546         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
8547         RETURN+USE when returning via %r14.
8549 2018-09-24  Martin Liska  <mliska@suse.cz>
8551         * gcov.c (output_lines): Print colorization legend
8552         for both flag_use_colors and flag_use_hotness_colors.
8553         Reword the help.
8555 2018-09-24  Martin Liska  <mliska@suse.cz>
8557         * coverage.c (get_coverage_counts): Use warning_at
8558         with current_function_decl location. Use %qD in warning
8559         message.
8561 2018-09-24  Martin Liska  <mliska@suse.cz>
8563         * memory-block.h (memory_block_pool::release): Annotate with
8564         valgrind that the memory is not accessible.
8566 2018-09-24  Martin Liska  <mliska@suse.cz>
8568         PR sanitizer/85774
8569         * asan.c: Make asan_handled_variables extern.
8570         * asan.h: Likewise.
8571         * cfgexpand.c (expand_stack_vars): Make sure
8572         a representative is unpoison if another
8573         variable in the partition is handled by
8574         use-after-scope sanitization.
8576 2018-09-24  Richard Biener  <rguenther@suse.de>
8578         PR tree-optimization/63155
8579         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
8580         the worklist when the edge of the respective argument isn't
8581         executable.
8583 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
8585         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
8586         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
8587         (MASK_CLASS_P): Update for rename.
8588         (MAYBE_MASK_CLASS_P): Ditto.
8589         (REG_CLASS_NAMES): Update.
8590         (REG_CLASS_CONTENT): Update.
8591         * config/i386/i386.c (regclass_map): Update for MASK_REG
8592         and ALL_MASK_REGS rename.
8593         * config/i386/constraints.md (Yk): Update for rename.
8594         (k): Ditto.
8596 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
8598         * config/i386/i386.h (enum reg_class): Remove
8599         EVEX_SSE_REGS and MOD4_SSE_REGS.
8600         (REG_CLASS_NAMES): Update.
8601         (REG_CLASS_CONTENT): Update.
8602         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
8603         registers as ALL_SSE_REGS.
8604         (ix86_additional_allocno_class_p): Remove.
8605         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
8606         (ix86_register_priority): Lower priority of EVEX SSE registers.
8607         Use IN_RANGE macro where appropriate.
8608         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
8609         AVX-5124VNNIW checks.
8610         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
8611         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
8612         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
8613         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
8614         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
8615         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
8616         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
8617         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
8618         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
8619         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
8620         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
8621         * config/i386/constraints.md (Yh): Remove.
8623 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
8625         * config/i386/i386.c (regclass_map): Declare integer REX registers
8626         as GENERAL_REGS.
8628 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
8630         * doc/service.texi (Service): Switch the fsf.org link to https.
8632 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8634         PR target/86798
8635         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8636         Define to speculation_safe_value_not_needed.
8638 2018-09-21  Florian Weimer  <fweimer@redhat.com>
8640         PR middle-end/81035
8641         * doc/extend.texi (Common Function Attributes): Mention that
8642         noreturn suppresses tail call optimization.
8644 2018-09-21  Jeff Law  <law@redhat.com>
8646         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
8647         vr_values::cleanup_edges_and_switches.
8648         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
8649         vr_values class.
8650         (identify_jump_threads): Remove EDGE_IGNORE handling.
8651         (execute_vrp): Move handling of to_remove_edges and
8652         to_update_switch_stmts into vr_values class member functions.
8653         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
8654         (to_update_switch_stmts): Likewise.
8655         * vr-values.c: Include cfghooks.h.
8656         (vr_values::vr_values): Initialize to_remove_edges and
8657         to_update_switch_stmts.
8658         (vr_values::~vr_values): Verify to_remove_edges and
8659         to_update_switch_stmts are empty.
8660         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
8661         (vr_values::cleanup_edges_and_switches): New member function.
8662         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
8663         function.  Add new data members.
8665 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
8667         PR tree-optimization/87309
8668         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
8669         calls with pflags and alt_flags.
8670         (selftest::test_capture_of_dump_calls): Add test of interaction of
8671         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
8673 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8675         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
8677 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8679         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
8681 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8683         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
8684         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
8685         (VXWORKS_LIBS_RTP): Minor reordering.
8687 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8689         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
8690         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
8692 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8694         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
8695         (PTRDIFF_TYPE): Likewise.
8697 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8699         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
8700         triplet, similar to support for VxWorks7.
8701         * config/vxworks-dummy.h: Provide a default definition
8702         of TARGET_VXWORKS64 to 0.
8704 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8706         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
8707         * config/vxworks-dummy.h: here.
8709 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8711         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
8713 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
8714             Bo Zhou  <zbo.zhou@hisilicon.com>
8716         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
8717         * config/aarch64/aarch64-tune.md: Regenerated.
8718         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
8719         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
8720         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
8722 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
8723             Julian Brown  <julian@codesourcery.com>
8725         * builtins.c (get_builtin_sync_mem): Handle address spaces.
8727 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
8729         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
8730         if the call takes a static chain.
8732 2018-09-21  Martin Liska  <mliska@suse.cz>
8734         * auto-profile.c (autofdo_source_profile::read): Do not
8735         set sum_all.
8736         (read_profile): Do not add working sets.
8737         (read_autofdo_file): Remove sum_all.
8738         (afdo_callsite_hot_enough_for_early_inline): Remove const
8739         qualifier.
8740         * coverage.c (struct counts_entry): Remove gcov_summary.
8741         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
8742         do not support GCOV_TAG_PROGRAM_SUMMARY.
8743         (get_coverage_counts): Remove summary and expected
8744         arguments.
8745         * coverage.h (get_coverage_counts): Likewise.
8746         * doc/gcov-dump.texi: Remove -w option.
8747         * gcov-dump.c (dump_working_sets): Remove.
8748         (main): Do not support '-w' option.
8749         (print_usage): Likewise.
8750         (tag_summary): Likewise.
8751         * gcov-io.c (gcov_write_summary): Do not dump
8752         histogram.
8753         (gcov_read_summary): Likewise.
8754         (gcov_histo_index): Remove.
8755         (gcov_histogram_merge): Likewise.
8756         (compute_working_sets): Likewise.
8757         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
8758         it not obsolete.
8759         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
8760         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
8761         (GCOV_HISTOGRAM_SIZE): Remove.
8762         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
8763         (struct gcov_summary): Simplify rapidly just
8764         to runs and sum_max fields.
8765         (gcov_histo_index): Remove.
8766         (NUM_GCOV_WORKING_SETS): Likewise.
8767         (compute_working_sets): Likewise.
8768         * gcov-tool.c (print_overlap_usage_message): Remove
8769         trailing empty line.
8770         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
8771         (output_lines): Remove program related line.
8772         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
8773         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
8774         histogram.
8775         (input_profile_summary): Do not read it.
8776         (merge_profile_summaries): And do not merge it.
8777         (input_symtab): Do not call removed function.
8778         * modulo-sched.c (sms_schedule): Do not print sum_max.
8779         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
8780         removed when histogram method was invented.
8781         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
8782         mode.
8783         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
8784         GCOV coding style.
8785         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
8786         and dump selected value.
8787         * profile.c (add_working_set): Remove.
8788         (get_working_sets): Likewise.
8789         (find_working_set): Likewise.
8790         (get_exec_counts): Do not work with working sets.
8791         (read_profile_edge_counts): Do not inform as sum_max is removed.
8792         (compute_branch_probabilities): Likewise.
8793         (compute_value_histograms): Remove argument for call of
8794         get_coverage_counts.
8795         * profile.h: Do not make gcov_summary const.
8797 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
8799         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
8801 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
8803         PR tree-optimization/86990
8804         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
8805         Check that the entire merged store group is made of constants only for
8806         overlapping stores.
8808 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
8810         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
8811         (VTABLE_VERIFICATION_SPEC): Likewise.
8812         (SANITIZER_EARLY_SPEC): Likewise.
8813         (SANITIZER_SPEC): Likewise.
8814         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
8815         * doc/invoke.texi (Link Options): Document -r.
8817 2018-09-20  Richard Biener <rguenther@suse.de>
8819         PR middle-end/87054
8820         * gimplify.c (gimplify_expr): Retain alignment of
8821         addressable lvalue in dereference.
8823 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
8825         PR bootstrap/87013
8826         * configure.ac: Check for .loc is_stmt support.
8827         * configure, config.in: Rebuilt.
8828         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
8829         if not supported.
8831 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
8833         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
8834         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
8835         -misel=no.
8837 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
8839         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8840         VECTOR_OTHER.
8841         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
8842         case VECTOR_OTHER.
8844 2018-09-20  Marek Polacek  <polacek@redhat.com>
8846         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
8848 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
8850         PR tree-optimization/87288
8851         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
8852         into account when determining PEELING_FOR_NITERS.
8854 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
8856         PR tree-optimization/86877
8857         * tree-vect-loop.c (vect_analyze_loop_2): Call
8858         vect_verify_datarefs_alignment.
8860 2018-09-19  Marek Polacek  <polacek@redhat.com>
8862         * doc/invoke.texi: Document -Wclass-conversion.
8864 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
8866         * config/pa/pa.c (pa_adjust_priority): Delete.
8867         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
8869         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
8870         (atomic_storehi): Likewise.
8871         (atomic_storesi): Likewise.
8872         (atomic_loaddi): Restore compare and swap exchange loop code.
8874 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
8876         PR rtl-optimization/86902
8877         * combine.c (try_combine): When changing the CC mode used, don't change
8878         an unrelated mode in other_insn to that new CC mode.
8880 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
8882         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
8883         with %T in place of calls to dump_generic_expr.
8884         (prune_runtime_alias_test_list): Likewise.
8885         (create_runtime_alias_checks): Likewise.
8886         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
8887         (vect_analyze_data_ref_dependence): Likewise.
8888         (vect_slp_analyze_data_ref_dependence): Likewise.
8889         (vect_record_base_alignment): Likewise.  Use %G in place of call
8890         to dump_gimple_stmt.
8891         (vect_compute_data_ref_alignment): Likewise.
8892         (verify_data_ref_alignment): Likewise.
8893         (vect_find_same_alignment_drs): Likewise.
8894         (vect_analyze_group_access_1): Likewise.
8895         (vect_analyze_data_ref_accesses): Likewise.
8896         (dependence_distance_ge_vf): Likewise.
8897         (dump_lower_bound): Likewise.
8898         (vect_prune_runtime_alias_test_list): Likewise.
8899         (vect_find_stmt_data_reference): Likewise.
8900         (vect_analyze_data_refs): Likewise.
8901         (vect_create_addr_base_for_vector_ref): Likewise.
8902         (vect_create_data_ref_ptr): Likewise.
8903         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
8904         (vect_can_advance_ivs_p): Likewise.
8905         (vect_update_ivs_after_vectorizer): Likewise.
8906         (vect_gen_prolog_loop_niters): Likewise.
8907         (vect_prepare_for_masked_peels): Likewise.
8908         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8909         (vect_determine_vectorization_factor): Likewise.
8910         (vect_is_simple_iv_evolution): Likewise.
8911         (vect_analyze_scalar_cycles_1): Likewise.
8912         (vect_analyze_loop_operations): Likewise.
8913         (report_vect_op): Likewise.
8914         (vect_is_slp_reduction): Likewise.
8915         (check_reduction_path): Likewise.
8916         (vect_is_simple_reduction): Likewise.
8917         (vect_create_epilog_for_reduction): Likewise.
8918         (vect_finalize_reduction:): Likewise.
8919         (vectorizable_induction): Likewise.
8920         (vect_transform_loop_stmt): Likewise.
8921         (vect_transform_loop): Likewise.
8922         (optimize_mask_stores): Likewise.
8923         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
8924         (vect_split_statement): Likewise.
8925         (vect_recog_over_widening_pattern): Likewise.
8926         (vect_recog_average_pattern): Likewise.
8927         (vect_determine_min_output_precision_1): Likewise.
8928         (vect_determine_precisions_from_range): Likewise.
8929         (vect_determine_precisions_from_users): Likewise.
8930         (vect_mark_pattern_stmts): Likewise.
8931         (vect_pattern_recog_1): Likewise.
8932         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8933         (vect_record_max_nunits): Likewise.
8934         (vect_build_slp_tree_1): Likewise.
8935         (vect_build_slp_tree_2): Likewise.
8936         (vect_print_slp_tree): Likewise.
8937         (vect_analyze_slp_instance): Likewise.
8938         (vect_detect_hybrid_slp_stmts): Likewise.
8939         (vect_detect_hybrid_slp_1): Likewise.
8940         (vect_slp_analyze_operations): Likewise.
8941         (vect_slp_analyze_bb_1): Likewise.
8942         (vect_transform_slp_perm_load): Likewise.
8943         (vect_schedule_slp_instance): Likewise.
8944         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8945         (vect_mark_stmts_to_be_vectorized): Likewise.
8946         (vect_init_vector_1): Likewise.
8947         (vect_get_vec_def_for_operand): Likewise.
8948         (vect_finish_stmt_generation_1): Likewise.
8949         (vect_check_load_store_mask): Likewise.
8950         (vectorizable_call): Likewise.
8951         (vectorizable_conversion): Likewise.
8952         (vectorizable_operation): Likewise.
8953         (vectorizable_load): Likewise.
8954         (vect_analyze_stmt): Likewise.
8955         (vect_is_simple_use): Likewise.
8956         (vect_get_vector_types_for_stmt): Likewise.
8957         (vect_get_mask_type_for_stmt): Likewise.
8958         * tree-vectorizer.c (increase_alignment): Likewise.
8960 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
8962         * doc/rtl.texi: Adjust vec_select description.
8963         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
8964         non-constant selectors.
8966 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
8968         * config/aarch64/aarch64-protos.h
8969         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
8970         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
8971         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
8972         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
8973         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
8974         (AARCH64_FL_PROFILE): Move index so flags are ordered.
8975         (AARCH64_ISA_RCPC8_4): New flag.
8976         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
8977         to aarch64_offset_9bit_signed_unscaled_p.
8978         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
8979         and use stlur.
8980         * config/aarch64/constraints.md (Ust): New constraint.
8981         * config/aarch64/predicates.md.
8982         (aarch64_9bit_offset_memory_operand): New predicate.
8983         (aarch64_rcpc_memory_operand): New predicate.
8985 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
8987         PR rtl-optimization/87361
8988         * rtlanal.c (nonzero_bits1): Revert accidental change.
8990 2018-09-19  Richard Biener  <rguenther@suse.de>
8992         PR tree-optimization/87349
8993         PR tree-optimization/87342
8994         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
8996 2018-09-18  Marek Polacek  <polacek@redhat.com>
8998         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
8999         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
9001 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
9003         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
9005 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
9007         PR rtl-optimization/86882
9008         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
9010 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
9012         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
9013         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
9015 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
9017         PR other/87353
9018         * doc/invoke.texi (Link Options): Fix formatting and grammar.
9020 2018-09-18  Richard Biener  <rguenther@suse.de>
9022         PR middle-end/63155
9023         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
9024         (compute_samebase_partition_bases): Likewise.
9025         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
9026         (gimple_can_coalesce_p): Simplify.
9028 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
9030         Handle a library implementation of ffs calling __builtin_ffs.
9031         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
9032         (mmix_init_libfuncs): New function: make __builtin_ffs expand
9033         to __ffsdi2.
9035 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
9037         * diagnostic-show-locus.c (class layout_range): Add field
9038         "m_original_idx".
9039         (layout_range::layout_range): Add "original_idx" param and use it
9040         to initialize new field.
9041         (make_range): Use 0 for original_idx.
9042         (layout::layout): Pass in index to calls to
9043         maybe_add_location_range.
9044         (layout::maybe_add_location_range): Add param "original_idx" and
9045         pass it on to layout_range.
9046         (layout::print_any_labels): Pass on range->m_original_idx to
9047         get_text call.
9048         (gcc_rich_location::add_location_if_nearby): Use 0 for
9049         original_idx.
9050         * gcc-rich-location.h (text_range_label::get_text): Update for new
9051         param.
9052         (range_label_for_type_mismatch::get_text): Likewise.
9054 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
9056         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
9058 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
9060         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
9061         format_string_diagnostic_t.
9062         (fmtwarn_n): Likewise.
9063         * substring-locations.c
9064         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
9065         (format_warning_n_va): Convert to...
9066         (format_string_diagnostic_t::emit_warning_n_va): ...this.
9067         (format_warning_va): Convert to...
9068         (format_string_diagnostic_t::emit_warning_va): ...this.
9069         (format_warning_at_substring): Convert to...
9070         (format_string_diagnostic_t::emit_warning): ...this.
9071         (format_warning_at_substring_n): Convert to...
9072         (format_string_diagnostic_t::emit_warning_n): ...this.
9073         * substring-locations.h (class format_string_diagnostic_t): New
9074         class.
9075         (format_warning_va): Convert to
9076         format_string_diagnostic_t::emit_warning_va.
9077         (format_warning_n_va): Convert to
9078         format_string_diagnostic_t::emit_warning_n_va.
9079         (format_warning_at_substring): Convert to
9080         format_string_diagnostic_t::emit_warning.
9081         (format_warning_at_substring_n): Convert to
9082         format_string_diagnostic_t::emit_warning_n.
9084 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
9085             Bernd Schmidt <bernds_cb1@t-online.de>
9087         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
9088         SImode args.
9090 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
9092         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
9093         operand 0 predicate to nonimmediate operand.
9094         (rint<mode>2_frndint): Remove insn pattern.
9095         (rint<mode>2): Change operand 1 predicate to general_operand.
9096         Extend operand 1 to XFmode and generate rintxf2 insn.
9097         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
9098         Do not use X87MODEF mode macro.
9099         (frndintxf2_<rounding>_i387): Rename from
9100         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
9101         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
9102         to XFmode and generate significandxf3 insn.
9104 2018-09-17  Richard Biener  <rguenther@suse.de>
9106         PR tree-optimization/87328
9107         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
9108         visiting unexecutable backedges when not iterating.
9109         (do_rpo_vn): Mark all edges not executable even when not
9110         iterating.
9112 2018-09-17  Martin Jambor  <mjambor@suse.cz>
9114         PR c/63886
9115         * doc/invoke.texi (Warning Options): Likewise.
9117 2018-09-17  Richard Biener  <rguenther@suse.de>
9119         PR tree-optimization/87301
9120         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
9121         clean EH info from leftover copy assignments.
9123 2018-09-17  Martin Liska  <mliska@suse.cz>
9125         PR gcov-profile/85871
9126         * gcov.c (output_intermediate_file): Fix out of bounds
9127         access.
9129 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
9131         * config/arc/arc.c: Object attributes for core4 not reflected
9132         correctly.
9133         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
9134         core3).
9136 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
9138         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
9140 2018-09-17  Martin Liska  <mliska@suse.cz>
9142         * doc/gcov.texi: Document new option --use-hotness-colors.
9143         * gcov.c (struct source_info): Declare new field.
9144         (source_info::source_info): Set default for maximum_count.
9145         (print_usage): Add new -q option.
9146         (process_args): Process it.
9147         (accumulate_line_info): Save src->maximum_count.
9148         (output_line_beginning): Make color line number if
9149         flag_use_hotness_colors is set.
9150         (output_line_details): Pass default argument value.
9151         (output_lines): Pass src->maximum_count.
9153 2018-09-17  Martin Liska  <mliska@suse.cz>
9155         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
9156         Use processor_names table.
9157         * config/i386/i386.c (ix86_default_align): Use
9158         processor_cost_table for alignment values.
9159         (ix86_option_override_internal): Use processor_names.
9160         (ix86_function_specific_print): Likewise.
9161         * config/i386/i386.h (struct processor_costs):
9162         Add alignment values.
9163         (struct ptt): Remove and replace with const char *.
9164         * config/i386/x86-tune-costs.h (struct processor_costs):
9165         Declare default alignments for all costs.
9167 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
9169         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
9170         symbolics or VR_VARYING ranges for ABS_EXPR.
9171         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
9172         when range will wrap.
9174 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
9176         PR middle-end/86864
9177         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
9178         before and after a JUMP_TABLE_DATA.
9180 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
9182         PR middle-end/87188
9183         * dojump.c (do_compare_and_jump): Canonicalize function pointers
9184         when one operand is a function pointer.  Use POINTER_TYPE_P and
9185         FUNC_OR_METHOD_TYPE_P.
9186         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
9187         * fold-const.c (build_range_check): Likewise.
9188         * match.pd (simple_comparison): Likewise.
9190 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
9192         PR c/82967
9193         * spellcheck.c (get_edit_distance_cutoff): New function.
9194         (selftest::test_edit_distance_unit_test_oneway): Rename to...
9195         (selftest::test_get_edit_distance_one_way): ...this.
9196         (selftest::test_get_edit_distance_unit): Rename to...
9197         (selftest::test_get_edit_distance_both_ways): ...this.
9198         (selftest::test_edit_distances): Move tests to this new function,
9199         and test some more pairs of strings.  Update for above renaming.
9200         (selftest::get_old_cutoff): New function.
9201         (selftest::test_get_edit_distance_cutoff): New function.
9202         (selftest::assert_suggested_for): New function.
9203         (ASSERT_SUGGESTED_FOR): New macro.
9204         (selftest::assert_not_suggested_for): New function.
9205         (ASSERT_NOT_SUGGESTED_FOR): New macro.
9206         (selftest::test_suggestions): New function.
9207         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
9208         tests to selftest::test_edit_distances and call it.  Add calls to
9209         selftest::test_get_edit_distance_cutoff and
9210         selftest::test_suggestions.
9211         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
9212         (best_match::consider): Replace hard-coded cutoff calculation with
9213         a call to...
9214         (best_match::get_cutoff): New declaration.
9215         (best_match::get_best_meaningful_candidate): Likewise.
9217 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9219         * builtins.c (fold_builtin_strlen): Remove TODO comment.
9221 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9223         revert:
9224         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9226         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9227         terminated string literal.
9229 2018-09-14  Martin Sebor  <msebor@redhat.com>
9231         * builtins.c (unterminated_array): Handle ARRAY_REF.
9232         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
9233         * builtins.h (unterminated_array): Declare extern.
9234         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
9235         arrays.
9236         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
9237         calls.
9239 2018-09-14  Martin Sebor  <msebor@redhat.com>
9240             Jeff Law  <law@redhat.com>
9242         * builtins.c (unterminated_array): New.
9243         (expand_builtin_strcpy): Adjust.
9244         (expand_builtin_strcpy_args): Detect unterminated arrays.
9245         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
9246         unterminated arrays.
9247         * gimple-fold.h (get_maxval_strlen): Add argument.
9248         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
9250         * gimple-fold.c (get_range_strlen): Add argument.
9251         (get_maxval_strlen): Adjust.
9252         * gimple-fold.h (get_range_strlen): Add argument.
9254 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
9256         * config/i386/movdirintrin.h: Fix copyright year.
9258 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
9260         * reg-stack.c: Include regs.h.
9261         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
9262         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
9263         FIRST_STACK_REG, not DFmode.
9264         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
9265         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
9266         (change stack): Default register mode to the reg_raw_mode of
9267         FIRST_STACK_REG, not DFmode.
9268         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
9269         (*swapxf): Rename from swapxf.
9271 2018-09-14  Carl Love  <cel@us.ibm.com>
9273         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
9274         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
9276 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
9278         PR target/87224
9279         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
9280         alternatives.
9282 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
9284         PR target/85628
9285         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
9287 2018-09-14  Jason Merrill  <jason@redhat.com>
9289         Fix --enable-gather-detailed-mem-stats.
9290         * hash-table.c (hash_table_usage): Change from variable to function.
9291         * hash-table.h: Adjust.
9292         * Makefile.in: Add missing dependencies on hash-table.h.
9294 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9296         PR tree-optimization/87259
9297         PR lto/87283
9298         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
9299         execute_cse_reciprocals_1 has tried transforming.
9301 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
9303         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
9304         VR_VARYING for PLUS/MINUS_EXPR.
9306 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9308         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
9309         formatting.
9311 2018-09-14  Richard Biener  <rguenther@suse.de>
9313         PR middle-end/63155
9314         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
9315         bits for the merged partition.
9317 2018-09-13  Martin Sebor  <msebor@redhat.com>
9318             Bernd Edlinger  <bernd.edlinger@hotmail.de>
9320         * builtins.h (c_srlen): Add argument.
9321         * builtins.c (warn_string_no_nul): New function.
9322         (c_strlen): Add argument and use it.  Update recursive calls.
9323         Pass DECL argument to string_constant to get info on non
9324         terminated strings.  Update *NONSTR as needed.
9325         (fold_builtin_strlen): Add argument to calls to c_strlen.
9326         Warn for unterminated arrays.
9327         (warn_string_no_null): Add prototype.
9328         * expr.c (string_constant): Update arguments.  Update recursive
9329         calls appropriately.  Detect missing NUL terminator and outermost
9330         declaration its missing in.
9331         Improve checks for arrays with nonzero lower bound or elements
9332         that are not a single byte.  Simplify offset computation.
9333         Simplify checks for non-NUL terminated strings.
9334         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
9335         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
9337 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9339         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
9340         correctly.
9341         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
9342         argument.  Fix range checks.
9343         * fold-const.h (c_getstr): Adjust protoype.
9344         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
9345         string is constant but contains no NUL byte.
9347         * expr.c (string_constant): Adjust function comment.
9348         Remove bogus check for zero termination.
9350         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
9352         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
9353         (get_constant_size): Don't make STRING_CSTs larger than they are.
9354         (check_string_literal): New check function for STRING_CSTs.
9355         (output_constant): Use it.
9357 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
9359         PR target/86812
9360         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
9362 2018-09-13  Richard Biener  <rguenther@suse.de>
9364         PR tree-optimization/87263
9365         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
9366         (struct unwind_state): Add max_rpo field.
9367         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
9368         Compute max_rpo, the max RPO number a block can be backwards reached
9369         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
9370         separating it from the iterating mode.
9372 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
9374         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
9375         (rfs_decision): New scheduling decision.
9377 2018-09-13  Richard Biener  <rguenther@suse.de>
9379         PR bootstrap/87134
9380         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
9381         (vn_nary_op_insert_pieces_predicated): Do not write useless
9382         valid_dominated_by_p entry outside of the allocated storage.
9384 2018-09-13  Omar Sandoval  <osandov@osandov.com>
9385             Tom de Vries  <tdevries@suse.de>
9387         PR debug/86985
9388         * dwarf2out.c (is_c): New function.
9389         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
9391 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
9393         PR target/85628
9394         * config/aarch64/aarch64.md (*aarch64_bfxil):
9395         Define.
9396         * config/aarch64/constraints.md (Ulc): Define.
9397         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
9398         Define.
9399         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
9400         New function.
9402 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
9404         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
9405         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
9406         aarch64_layout_frame call.
9407         (aarch64_expand_epilogue): Likewise.
9408         (aarch64_initial_elimination_offset): Likewise.
9409         (aarch64_get_separate_components): Likewise.
9410         (aarch64_use_return_insn_p): Likewise.
9411         (aarch64_layout_frame): Remove unneeded check.
9413 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
9415         * configure.ac: Only append
9416         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
9417         gcc_config_arguments if it was never reconfigured or last reconfigure
9418         was with different arguments.
9419         * configure: Regenerated.
9421 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
9422             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9424         PR middle-end/87290
9425         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
9426         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
9428 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
9430         PR tree-optimization/87287
9431         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
9432         X % C == 0 to X % (unsigned) C == 0 optimization to ...
9433         * match.pd (X % C == 0): ... here.  New optimization.
9435 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
9437         PR middle-end/82853
9438         * expr.h (maybe_optimize_mod_cmp): Declare.
9439         * expr.c (mod_inv): New function.
9440         (maybe_optimize_mod_cmp): New function.
9441         (do_store_flag): Use it.
9442         * cfgexpand.c (expand_gimple_cond): Likewise.
9444 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
9445             Julian Brown  <julian@codesourcery.com>
9447         PR middle-end/86336
9448         * gimplify.c (gimplify_scan_omp_clauses): Set
9449         target_firstprivatize_array_bases in OpenACC parallel and kernels
9450         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
9451         OpenACC data regions.
9453 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
9455         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
9456         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
9458 2018-09-12  Richard Biener  <rguenther@suse.de>
9460         PR tree-optimization/87280
9461         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
9462         edge but unreachable target.
9463         (do_rpo_vn): For conservatively handling a PHI only mark
9464         the backedge executable but not the block reachable.
9466 2018-09-12  Richard Biener  <rguenther@suse.de>
9468         PR tree-optimization/87266
9469         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
9470         visited blocks.
9472 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
9474         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
9475         constants.
9476         ("trunc<BFP:mode><DFP_ALL:mode>2")
9477         ("trunc<DFP_ALL:mode><BFP:mode>2")
9478         ("extend<BFP:mode><DFP_ALL:mode>2")
9479         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
9480         according to the target operand type.
9482 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
9483             Andreas Krebbel  <krebbel@linux.ibm.com>
9485         PR tree-optimization/86844
9486         * gimple-ssa-store-merging.c
9487         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
9488         there are any overlapping stores in between them, make sure they are
9489         also coalesced or we give up completely.
9491 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
9493         PR middle-end/87248
9494         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
9495         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
9497 2018-09-12  Tom de Vries  <tdevries@suse.de>
9499         * common.opt (gdescribe-dies): Add option.
9500         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
9501         attribute for artifical and nameless decls.
9502         (dwarf2out_register_external_die): Add description attribute to
9503         external reference die.
9504         (add_desc_attribute): New functions.
9505         (gen_subprogram_die): Add description attribute to
9506         DW_TAG_call_site_parameter.
9507         * tree-pretty-print.c (print_generic_expr_to_str): New function.
9508         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
9509         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
9510         -gno-describe-dies.
9511         (@item -gdescribe-dies): Add.
9513 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
9515         * tree-vrp.c (vrp_shift_undefined_p): Remove.
9516         (extract_range_from_binary_expr_1: Call
9517         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
9518         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
9519         depend on sign.
9521 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
9523         * gimple-ssa-warn-alloca.c
9524         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
9525         field for ALLOCA_BOUND_*_LARGE.
9527 2018-09-11  Nathan Sidwell  <nathan@acm.org>
9529         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
9531 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
9533         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
9534         for clobbers.  Remove obsolete comment.
9536 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
9538         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
9539         mpxchk, mpxld and mpxst types.
9540         (define_attr length_immediate): Remove all processing of mpx types.
9541         (define_attr prefix_0f): Ditto.
9542         (define_attr memory): Ditto.
9544 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
9546         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
9547         (log<mode>2): Change operand 1 predicate to general_operand.
9548         Extend operand 1 to XFmode and generate logxf3 insn.
9549         (log10<mode>2): Change operand 1 predicate to general_operand.
9550         Extend operand 1 to XFmode and generate log10xf3 insn.
9551         (log2<mode>2): Change operand 1 predicate to general_operand.
9552         Extend operand 1 to XFmode and generate log2xf3 insn.
9553         (fyl2xp1_extend<mode>xf3_i387): Remove.
9554         (log1p<mode>2): Change operand 1 predicate to general_operand.
9555         Extend operand 1 to XFmode and generate log1pxf3 insn.
9556         (fxtract_extend<mode>xf3_i387): Remove.
9557         (logb<mode>2): Change operand 1 predicate to general_operand.
9558         Extend operand 1 to XFmode and generate logbxf3 insn.
9559         (ilogb<mode>2): Change operand 1 predicate to general_operand.
9560         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
9561         (significand<mode>2): Change operand 1 predicate to general_operand.
9562         Extend operand 1 to XFmode and generate significandxf3 insn.
9564 2018-09-11  Nathan Sidwell  <nathan@acm.org>
9566         * gcc.c (perror_with_name, pfatal_with_name): Delete.
9567         (load_specs): Use fatal_error.
9568         (DELETE_IF_ORDINARY, process_command): Use error.
9569         (execute, run_attempt): Use fatal_error.
9571 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
9573         * diagnostic-core.h (sorry_at): New prototype.
9574         * diagnostic.c (sorry_at): New function.
9576 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
9578         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
9579         by zero as VR_UNDEFINED.
9581 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
9583         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
9584         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
9585         (<sincos>mode2): New expander.
9586         (sincos_extend<mode>xf3_i387): Remove insn pattern.
9587         (sincos -> sin, cos splitters): Remove splitter patterns.
9588         (sincos<mode>3): Change operand 2 predicate to general_operand.
9589         Extend operand 2 to XFmode and generate sincosxf3 insn.
9590         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
9591         Change operand 3 predicate to const1_operand.
9592         (fptan_extend<mode>xf4_i387): Remove insn pattern.
9593         (tanxf2): Update operands in the call to fptanxf4_i387.
9594         (tan<mode>2): Change operand 1 predicate to general_operand.
9595         Extend operand 1 to XFmode and generate tanxf3 insn.
9596         (atan2xf3): Rename from *fpatanxf3_i387.
9597         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
9598         (atan2xf3): Remove expander.
9599         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
9600         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
9601         (atan<mode>2): Change operand 1 predicate to general_operand.
9602         Extend operand 1 to XFmode and generate atanxf3 insn.
9604 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
9606         * config/i386/i386.md (x87/SSE constant load splitter): Use
9607         memory_operand instead of nonimmediate_operand for input operand
9608         predicate.
9610 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
9612         * config/i386/i386.md (float partial SSE register stall splitter): Move
9613         splitter near its instruction pattern.
9614         (float_extend partial SSE register stall splitter): Ditto.
9615         (float_truncate partial SSE register stall splitter): Ditto.
9617 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
9619         PR target/86794
9620         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9621         to speculation_safe_value_not_needed.
9623         PR target/85666
9624         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
9625         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
9626         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
9627         leaf_function_p, instead use has_hard_reg_initial_val.
9629 2018-09-09  Nathan Sidwell  <nathan@acm.org>
9631         * gcc.h (pfatal_with_name): Don't declare here.
9632         * gcc.c (pfatal_with_name): Make static.
9634 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
9636         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
9637         earlyclobber.
9639 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
9641         PR rtl-optimization/85458
9642         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
9643         priority hook to reduce the priority of EXPR.
9645 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
9647         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
9648         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
9649         with X87_ENABLE_FLOAT.  Remove preparation code.
9650         (*float<SWI48:mode><MODEF:mode>2): Rename from
9651         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
9652         math using "enabled" attribute.
9653         (*floatdi<MODEF:mode>2_i387): Rename from
9654         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
9655         enable for 32bit targets only.
9656         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
9657         splitter.
9658         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
9659         as operand 1 predicate.  Rewrite as define_insn_and_split.
9660         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
9662 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
9664         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
9665         to fallthru to FLOAT case.
9667 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
9669         PR target/86731
9670         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
9671         around folding of vec_sl to handle out of range shift values.
9673 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
9675         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
9676         Update callers to gen_fix_trunc<mode>_i387_fisttp
9677         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
9678         nonimmediate_operand.
9679         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
9680         and corresponding splitters.
9681         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
9682         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
9683         (fix_truncdi_i387_with_temp): Remove insn pattern
9684         and corresponding splitters.
9685         (fix_trunc<mode>_i387): Change operand 0 predicate to
9686         nonimmediate_operand.
9687         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
9688         and corresponding splitters.
9689         (*fistdi2_1): Remove.
9690         (fistdi2): Ditto.
9691         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
9692         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
9693         (*fist<mode>2_1): Remove.
9694         (fist<mode>2): Ditto.
9695         (fist<mode>2_with_temp): Remove insn pattern and corresponding
9696         splitters.
9697         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
9698         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
9699         (fistdi2_<rounding>): Change operand 0 predicate to
9700         nonimmediate_operand.
9701         (fistdi2_<rounding>_with_temp): Remove insn pattern
9702         and corresponding splitters.
9703         (fist<mode>2_<rounding>): Change operand 0 predicate to
9704         nonimmediate_operand.
9705         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
9706         and corresponding splitters.
9708         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
9710 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9712         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
9713         the init value.
9715 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
9717         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
9718         early gimple folding of vec_splat().
9719         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
9720         * gimple-fold.h: Add an extern define for tree_vec_extract().
9722 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
9724         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
9725         wrappers around TREE_TYPE comparisons.
9727 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
9729         PR target/80080
9730         * config/s390/predicates.md: Add nonsym_memory_operand.
9731         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
9732         contains a SYMBOL_REF, load it into an intermediate pseudo.
9733         (s390_emit_compare_and_swap): Legitimize operand.
9734         * config/s390/s390.md: Use the new nonsym_memory_operand
9735         with UNSPECV_CAS patterns.
9737 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
9739         PR target/80080
9740         * config/s390/s390-passes.def: New file.
9741         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
9742         declaration.
9743         (make_pass_s390_early_mach): Add declaration.
9744         * config/s390/s390.c (make_pass_s390_early_mach):
9745         (s390_option_override): Remove dynamic registration.
9746         * config/s390/t-s390: Add s390-passes.def.
9748 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
9750         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9751         Remove UNSPEC_LTREL_BASE check.
9752         (annotate_constant_pool_refs): Likewise.
9753         (find_constant_pool_ref): Likewise.
9754         (find_ltrel_base): Removed.
9755         (replace_ltrel_base): Removed.
9756         (s390_mainpool_finish): Remove replace_ltrel_base call.
9757         (s390_chunkify_start): Remove pending LTREL_BASE logic.
9758         (s390_chunkify_finish): Remove replace_ltrel_base call.
9759         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
9761 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
9763         PR target/86779
9764         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9765         to speculation_safe_value_not_needed.
9767 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
9768             Bernd Schmidt  <bernds_cb1@t-online.de>
9770         * config/nvptx/nvptx-opts.h: New file.
9771         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
9772         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
9773         (ASM_SPEC): Define.
9774         (TARGET_SM35): New macro.
9775         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
9776         correct predicate.
9777         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
9778         values.
9779         (misa=): New option.
9780         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
9782 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
9784         * config/i386/i386.md (truncdfsf2): Remove expander.
9785         (truncdfsf2_with_temp): Ditto.
9786         (truncxf<mode>2): Ditto.
9787         (*truncdfsf_fast_mixed): Remove insn pattern.
9788         (*truncdfsf_fast_i387): Ditto.
9789         (*truncdfsf_mixed): Ditto.
9790         (*truncdfsf_i387): Ditto.
9791         (*truncdfsf2_i387_1): Ditto.
9792         (*truncxfsf2_mixed): Ditto.
9793         (*truncxfdf2_mixed): Ditto.
9794         (*truncxf<mode>2_i387_noop): Ditto. Update callers
9795         to call gen_truncxf<mode>2 instead.
9796         (*truncxf<mode>2_i387): Remove.
9797         (reg->reg splitters): Remove splitter pattern.
9798         (reg->mem splitters): Ditto.
9800         (truncdfsf2): New insn pattern.
9801         (truncxf<mode>2): Ditto.
9803 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9805         * tree-ssa-math-opts.c (is_mult_by): New function.
9806         (is_square_of): Use the above.
9807         (optimize_recip_sqrt): New function.
9808         (pass_cse_reciprocals::execute): Use the above.
9810 2018-09-05  Richard Biener  <rguenther@suse.de>
9812         PR bootstrap/87134
9813         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
9814         to zero-init the emplaced vec.
9816 2018-09-05  Martin Liska  <mliska@suse.cz>
9818         PR tree-optimization/87205
9819         * tree-switch-conversion.c (pass_lower_switch::execute):
9820         Group cases for switch statements.
9822 2018-09-05  Richard Biener  <rguenther@suse.de>
9824         PR tree-optimization/87217
9825         * tree-ssa-sccvn.c (vuse_valueize): New.
9826         (vn_reference_lookup_pieces): Use it.
9827         (vn_reference_lookup): Likewise.
9829 2018-09-05  Nathan Sidwell  <nathan@acm.org>
9831         PR c++/87137
9832         * stor-layout.c (place_field): Scan forwards to check last
9833         bitfield when ms_bitfield_placement is in effect.
9835 2018-09-05  Richard Biener  <rguenther@suse.de>
9837         PR bootstrap/87225
9838         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
9839         return.
9841 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9842             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
9844         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
9845         * config.gcc (extra_objs): Build it.
9846         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
9847         Likewise.
9848         * config/aarch64/aarch64-passes.def
9849         (pass_tag_collision_avoidance): New pass.
9850         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
9851         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
9852         (aarch64_classify_address): Remove static qualifier.
9853         (aarch64_address_info, aarch64_address_type): Move to...
9854         * config/aarch64/aarch64-protos.h: ... here.
9855         (make_pass_tag_collision_avoidance): New function.
9856         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
9857         New tuning flag.
9859 2018-09-05  Martin Liska  <mliska@suse.cz>
9861         * doc/gcov.texi: Update documentation of humar
9862         readable mode.
9863         * gcov.c (format_count): Print one decimal place, it provides
9864         more fine number of situations like '1G' vs. '1.4G'.
9866 2018-09-05  Martin Liska  <mliska@suse.cz>
9868         PR target/87164
9869         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
9870         * optc-gen.awk: Allow 'Var' for Deprecated options in order
9871         to generate a MASK value.
9873 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
9875         PR debug/86593
9876         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
9877         if frame pointer isn't used.
9878         (compute_frame_pointer_to_fb_displacement): Likewise.
9880 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
9882         PR target/87198
9883         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
9884         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
9885         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
9886         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
9887         and OPTION_MASK_ISA_XSAVEC_UNSET.
9889 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
9891         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
9892         XOR operations in NAND case.
9894 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
9896         * wide-int-range.cc (wide_int_range_convert): New.
9897         * wide-int-range.h (wide_int_range_convert): New.
9898         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
9899         code into wide_int_range_convert.
9900         (extract_range_into_wide_ints): Do not munge anti range constants
9901         into the entire domain.  Just return the range back.
9903 2018-09-04  Martin Liska  <mliska@suse.cz>
9905         * genmatch.c (output_line_directive): Add new argument
9906         fnargs.
9907         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
9909 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
9911         * doc/invoke.texi (Option Summary): Add whitespace.
9913         * doc/invoke.texi (Option Summary): Add -Waligned-new.
9915 2018-09-04  Richard Biener  <rguenther@suse.de>
9917         PR tree-optimization/87211
9918         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
9919         backedge value we're supposed to treat as VARYING also number
9920         the PHI to VARYING in case it got a different value-number already.
9922 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
9924         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
9925         (extract_range_from_binary_expr_1): Do not call
9926         vrp_can_optimize_bit_op.
9927         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
9928         static.
9929         (wide_int_range_get_mask_and_bounds): New.
9930         (wide_int_range_optimize_bit_op): New.
9931         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
9932         (wide_int_range_bit_and): Same.
9933         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
9934         (wide_int_range_optimize_bit_op): New.
9935         (wide_int_range_get_mask_and_bounds): New.
9937 2018-09-04  Richard Biener  <rguenther@suse.de>
9939         PR tree-optimization/87176
9940         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
9941         variable.  When value-numbering a virtual PHI node make sure
9942         to not value-number to the backedge value.
9944 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
9946         * doc/extend.texi (Long Long, Hex Floats): Document support for
9947         long long and hex floats in more recent versions of ISO C++.
9949 2018-09-03  Richard Biener  <rguenther@suse.de>
9951         PR tree-optimization/87177
9952         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
9953         cleanup.
9955 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9957         * bb-reorder.c (edge_order): Convert to C-qsort-style
9958         tri-state comparator.
9959         (reorder_basic_blocks_simple): Change std::stable_sort to
9960         gcc_stablesort.
9962 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9964         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
9965         tri-state comparator.
9966         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
9968 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9970         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
9971         (mergesort): ... here as maximum count for using netsort.
9972         (gcc_qsort): Set nlim to 3 if stable sort is requested.
9973         (gcc_stablesort): New.
9974         * system.h (gcc_stablesort): Declare.
9976 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9978         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
9979         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
9980         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
9982 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
9984         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
9985         lxsdx and stxsdx alternatives.
9986         (*mov<mode>_hardfloat64): Ditto.
9987         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
9989 2018-09-03  Richard Biener  <rguenther@suse.de>
9991         PR tree-optimization/87200
9992         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
9993         simplify result.
9995 2018-09-03  Martin Liska  <mliska@suse.cz>
9997         PR tree-optimization/87201
9998         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
9999         Fix parenthesis in an expression.
10001 2018-09-03  Richard Biener  <rguenther@suse.de>
10003         PR tree-optimization/87197
10004         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
10005         visited.  CSE the VN_INFO hashtable lookup.
10007         PR tree-optimization/87169
10008         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
10009         iterating make sure there's no extra backedges from irreducible
10010         regions feeding the header.  Mark the destination block
10011         executable.
10013 2018-09-03  Martin Liska  <mliska@suse.cz>
10015         PR driver/83193
10016         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
10017         * common/common-targhooks.c (default_get_valid_option_values):
10018         New function.
10019         * common/common-targhooks.h (default_get_valid_option_values):
10020         Likewise.
10021         * common/config/i386/i386-common.c: Move processor_target_table
10022         from i386.c.
10023         (ix86_get_valid_option_values): New function.
10024         (TARGET_GET_VALID_OPTION_VALUES): New macro.
10025         * config/i386/i386.c (struct ptt): Move to i386-common.c.
10026         (PTA_*): Move all defined masks into i386-common.c.
10027         (ix86_function_specific_restore): Use new processor_cost_table.
10028         * config/i386/i386.h (struct ptt): Moved from i386.c.
10029         (struct pta): Likewise.
10030         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
10031         * doc/tm.texi.in: Likewise.
10032         * opt-suggestions.c (option_proposer::suggest_option):
10033         Pass prefix to build_option_suggestions.
10034         (option_proposer::get_completions): Likewise.
10035         (option_proposer::build_option_suggestions): Use the new target
10036         hook.
10037         * opts.c (struct option_help_tuple): New struct.
10038         (print_filtered_help): Use the new target hook.
10040 2018-09-03  Martin Liska  <mliska@suse.cz>
10042         PR middle-end/59521
10043         * predict.c (set_even_probabilities): Add likely_edges
10044         argument and handle cases where we have precisely one
10045         likely edge.
10046         (combine_predictions_for_bb): Catch also likely_edges.
10047         (tree_predict_by_opcode): Handle gswitch statements.
10048         * tree-cfg.h (find_case_label_for_value): New declaration.
10049         (find_taken_edge_switch_expr): Likewise.
10050         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
10051         Find pivot in decision tree based on probabily, not by number of
10052         nodes.
10054 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
10056         * doc/standards.texi (Standards): Update Objective-C reference.
10058 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10060         * doc/install.texi (Prerequisites): Update link for MPC.
10062 2018-09-01  Michael Matz  <matz@suse.de>
10064         PR tree-optimization/87074
10065         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
10066         PHIs for outer-loop uses.
10068 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10070         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
10071         * doc/invoke.texi (C Dialect Options): Ditto.
10073 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10075         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
10077 2018-08-31  Richard Biener  <rguenther@suse.de>
10079         PR tree-optimization/87168
10080         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
10081         (rpo_elim::eliminate_avail): When OP was not visited it must
10082         be available.
10084 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
10086         * tree-vrp.c (copy_value_range): Convert param "from" from
10087         "value_range *" to "const value_range *".
10088         (range_is_null): Likewise for param "vr".
10089         (range_int_cst_p): Likewise.
10090         (range_int_cst_singleton_p): Likewise.
10091         (symbolic_range_p): Likewise.
10092         (value_ranges_intersect_p): Likewise for both params.
10093         (value_range_nonnegative_p): Likewise for param "vr".
10094         (value_range_constant_singleton): Likewise.
10095         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
10096         (extract_range_into_wide_ints): Likewise for param "vr".
10097         (extract_range_from_multiplicative_op): Likewise for params "vr0"
10098         and "vr1".
10099         (vrp_can_optimize_bit_op): Likewise.
10100         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
10101         "vr1_".
10102         (extract_range_from_unary_expr): Likewise.
10103         (debug_value_range): Likewise for param "vr".
10104         (value_range::dump): Add "const" qualifier.
10105         (vrp_prop::check_array_ref): Convert local "vr" from
10106         "value_range *" to "const value_range *".
10107         (vrp_prop::check_mem_ref): Likewise.
10108         (vrp_prop::visit_stmt): Likewise for local "old_vr".
10109         (vrp_intersect_ranges_1): Likewise for param "vr_1".
10110         (vrp_intersect_ranges): Likewise.
10111         (simplify_stmt_for_jump_threading): Likewise for local "vr".
10112         (vrp_prop::vrp_finalize): Likewise.
10113         * tree-vrp.h (value_range::dump): Add "const" qualifier.
10114         (vrp_intersect_ranges): Add "const" qualifier to params as above.
10115         (extract_range_from_unary_expr): Likewise.
10116         (value_range_constant_singleton): Likewise.
10117         (symbolic_range_p): Likewise.
10118         (copy_value_range): Likewise.
10119         (extract_range_from_binary_expr_1): Likewise.
10120         (range_int_cst_p): Likewise.
10121         (vrp_set_zero_nonzero_bits): Likewise.
10122         (range_int_cst_singleton_p): Likewise.
10124 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
10126         * config/aarch64/arm_neon.h (vabsd_s64): New.
10127         (vnegd_s64): Likewise.
10129 2018-08-31  Martin Jambor  <mjambor@suse.cz>
10131         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
10133 2018-08-31  Martin Liska  <mliska@suse.cz>
10135         * ipa-icf.c (sem_item::add_type): Use
10136         sem_item::m_type_hash_cache.
10137         * ipa-icf.h: Move the cache from sem_item_optimizer
10138         to sem_item.
10140 2018-08-31  Nathan Sidwell  <nathan@acm.org>
10142         * doc/extend.texi (Backwards Compatibility): Remove implicit
10143         extern C leeway of () being (...).
10145 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10147         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
10149 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
10151         PR target/86684
10152         PR target/87149
10153         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
10155 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
10157         PR middle-end/87138
10158         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
10159         gen_int_mode.  Formatting fixes.
10161 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
10163         * target.def (custom_function_descriptors): Improve documentation.
10164         * doc/tm.texi.in (Trampolines): Expand discussion of function
10165         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
10166         beginning of the section.
10167         * doc/tm.texi: Regenerated.
10169 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
10171         * cfg.h (class auto_edge_flag): Spell out the template-id of the
10172         base class in the initializer list.  This is a workaround for
10173         building with older GCC.
10174         (class auto_bb_flag): Likewise.
10176 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
10178         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
10179         (altivec_vcmpequ<VI_char>_p): Remove star.
10180         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
10181         vector load modes.
10182         (expand_strncmp_vec_sequence): New function.
10183         (emit_final_str_compare_vec): New function.
10184         (expand_strn_compare): Add support for vector strncmp.
10185         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
10186         length specification to bytes.
10187         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
10188         (vcmpnezb_p): New pattern.
10189         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
10190         for option -mstring-compare-inline-limit.
10192 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
10194         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
10195         (PTA_SKYLAKE): Add PTA_AES.
10196         (PTA_GOLDMONT): Likewise.
10198 2018-08-29  Jan Hubicka  <jh@suse.cz>
10200         PR lto/86517
10201         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
10202         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
10204 2018-08-29  Jan Hubicka  <jh@suse.cz>
10206         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
10207         TYPE_STUB_DECL.
10208         (hash_tree): Do not visit TYPE_STUB_DECL.
10209         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
10210         stream TYPE_STUB_DECL.
10211         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
10212         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
10213         after free_lang_data.
10214         (type_in_anonymous_namespace_p): Likewise.
10216 2018-08-29  Jan Hubicka  <jh@suse.cz>
10218         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
10219         comment that it has to be even number.
10220         (class sreal): Change m_sig type to int32_t.
10221         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
10222         int64_t for temporary calculations.
10223         (sreal_verify_basics): Drop one bit from minimum and maximum.
10225 2018-08-30  Richard Biener  <rguenther@suse.de>
10227         PR tree-optimization/87147
10228         * tree-ssa-sccvn.c (SSA_VISITED): New function.
10229         (visit_phi): When the degenerate result is from the backedge and
10230         we didn't visit its definition yet drop to VARYING.
10231         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
10233 2018-08-29  Jan Hubicka  <jh@suse.cz>
10235         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
10236         DECL_VINDEX.
10237         (hash_tree): Likewise.
10239 2018-08-29  Jan Hubicka  <jh@suse.cz>
10241         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
10242         and TYPE_NEXT_REF_TO.
10244 2018-08-29  Jan Hubicka  <jh@suse.cz>
10246         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
10247         comment that it has to be even number.
10248         (class sreal): Change m_sig type to int32_t.
10249         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
10250         int64_t for temporary calculations.
10251         (sreal_verify_basics): Drop one bit from minimum and maximum.
10253 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
10255         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
10257 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
10259         PR middle-end/86995
10260         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
10261         if to_add is negative.
10263 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10265         PR middle-end/87053
10266         * builtins.c (c_strlen): Improve range checks.
10268 2018-08-29  Martin Sebor  <msebor@redhat.com>
10269             Jeff Law  <law@redhat.com>
10271         PR tree-optimization/86714
10272         PR tree-optimization/86711
10273         * builtins.c (c_strlen): Add arguments to call to string_constant.
10274         * expr.c (string_constant): Add argument.  Detect missing nul
10275         terminator and outermost declaration it's missing in.
10276         * expr.h (string_constant): Add argument.
10277         * fold-const.c (read_from_constant_string): Add arguments to call to
10278         string_constant.
10279         (c_getstr): Likewise.
10280         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
10281         to string_constant.
10282         * tree-ssa-strlen.c (get_stridx): Likewise.
10284 2018-08-29  Jan Hubicka  <jh@suse.cz>
10286         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
10287         Do not stream DECL_VINDEX.
10288         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
10289         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
10290         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
10292 2018-08-29  Richard Biener  <rguenther@suse.de>
10294         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
10295         virtual operands that are not default defs to honor region
10296         boundaries.
10297         (rpo_vn_valueize): Remove ineffective code here.
10299 2018-08-29  Richard Biener  <rguenther@suse.de>
10301         PR tree-optimization/87132
10302         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
10303         when skipping defs reachable over backedges.
10305 2018-08-29  Richard Biener  <rguenther@suse.de>
10307         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
10308         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
10309         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
10310         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
10311         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
10312         (vn_reference_lookup_3): Remove use of const_parms.
10313         (free_rpo_vn): Do not free const_parms.
10314         (do_rpo_vn): Do not call init_const_parms.
10315         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
10316         SSA_NAME_POINTS_TO_READONLY_MEMORY.
10317         (call_may_clobber_ref_p_1): Likewise.
10319 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
10321         PR other/86726
10322         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
10323         (-O): Ditto.
10324         (-ftree-scev-cprop): Document.
10326 2018-08-29  Jan Hubicka  <jh@suse.cz>
10328         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
10329         parameters.
10330         (sreal constructor): Update.
10331         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
10332         sreal:operator/): Update.
10334 2018-08-29  Martin Liska  <mliska@suse.cz>
10336         * tree-switch-conversion.c (switch_conversion::expand):
10337         Strenghten assumption about gswitch statements.
10339 2018-08-29  Richard Biener  <rguenther@suse.de>
10341         PR tree-optimization/87117
10342         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
10343         re-value-number released SSA VDEFs.
10345 2018-08-29  Richard Biener  <rguenther@suse.de>
10347         PR tree-optimization/87126
10348         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
10350 2018-08-28  Jim Wilson  <jimw@sifive.com>
10352         * config/riscv/pic.md: Rewrite.
10353         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
10354         invalid address.
10355         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
10356         (SOFTF, default_load, softload, softstore): New.
10358 2018-08-28  Jeff Law  <law@redhat.com>
10360         * fold-const.c (fold_binary_loc): Remove recently added assert.
10362 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
10364         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
10365         to OP parmeter of generated function.
10367 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
10369         PR tree-optimization/87009
10370         * match.pd: Add boolean optimizations.
10372 2018-08-28  Martin Sebor  <msebor@redhat.com>
10374         PR middle-end/86631
10375         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
10376         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
10377         (pass_walloca::gate): Use it.
10378         (alloca_call_type): Same.
10379         (pass_walloca::execute): Same.
10380         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
10382 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
10384         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
10385         GCC_VERSION for usage of "__gcc_dump_printf__" format from
10386         >= 3005 to >= 9000.
10388 2018-08-28  Richard Biener  <rguenther@suse.de>
10390         PR tree-optimization/87124
10391         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
10392         constants before looking up avail.
10394 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
10396         PR middle-end/87099
10397         * calls.c (maybe_warn_nonstring_arg): Punt early if
10398         warn_stringop_overflow is zero.  Don't call get_range_strlen
10399         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
10400         Swap comparison operands to have constants on rhs.  Only use
10401         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
10402         increment lenrng[0].
10404 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
10406         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
10407         use of tree_to_shwi.  Remove duplicated test for the size being
10408         a whole number of bytes.
10410 2018-08-28  Richard Biener  <rguenther@suse.de>
10412         PR tree-optimization/87117
10413         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
10414         Handle removed stmt without LHS (GIMPLE_NOP).
10416 2018-08-28  Richard Biener  <rguenther@suse.de>
10418         PR tree-optimization/87117
10419         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
10420         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
10422 2018-08-28  Richard Biener  <rguenther@suse.de>
10424         PR tree-optimization/87117
10425         * tree-ssa-pre.c (compute_avail): Do not make expressions
10426         with predicated values available.
10427         (get_expr_value_id): Assert we do not run into predicated value
10428         expressions.
10430 2018-08-28  Richard Biener  <rguenther@suse.de>
10432         PR tree-optimization/87117
10433         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
10434         get virtual operands.
10435         (get_expr_operands): Handle STRING_CST like other decls.
10437 2018-08-28  Martin Liska  <mliska@suse.cz>
10439         * tree.h: Update documentation of fndecl_built_in_p
10440         functions.
10443 2018-08-27  Jeff Law  <law@redhat.com>
10444         PR tree-optimization/87110
10445         * tree-ssa-dse.c (compute_trims): Handle non-constant
10446         TYPE_SIZE_UNIT.
10448 2018-08-27  Martin Sebor  <msebor@redhat.com>
10450         PR tree-optimization/86914
10451         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
10453 2018-08-27  Martin Sebor  <msebor@redhat.com>
10455         PR tree-optimization/87112
10456         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
10457         the type of the bound argument.
10459 2018-08-27  Jeff Law  <law@redhat.com>
10461         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
10462         type does not have a TYPE_SIZE_UNIT.
10464 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
10466         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
10467         with include of backend.h.
10469 2018-08-27  Richard Biener  <rguenther@suse.de>
10471         PR tree-optimization/86927
10472         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
10473         use const cond reduction code.
10475 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
10477         PR tree-optimization/85758
10478         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
10480 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
10482         PR c++/87091
10483         * diagnostic-show-locus.c (class layout_range): Update for
10484         conversion of show_caret_p to a tri-state.
10485         (layout_range::layout_range): Likewise.
10486         (make_range): Likewise.
10487         (layout::maybe_add_location_range): Likewise.
10488         (layout::should_print_annotation_line_p): Don't show annotation
10489         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
10490         (layout::get_state_at_point): Update for conversion of
10491         show_caret_p to a tri-state.  Bail out early for
10492         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
10493         underlining or source colorization.
10494         (gcc_rich_location::add_location_if_nearby): Update for conversion
10495         of show_caret_p to a tri-state.
10496         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
10497         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
10498         Likewise.
10499         (selftest::test_one_liner_labels): Likewise.
10500         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
10501         conversion of show_caret_p to a tri-state.
10502         * pretty-print.c (text_info::set_location): Likewise.
10503         * pretty-print.h (text_info::set_location): Likewise.
10504         * substring-locations.c (format_warning_n_va): Likewise.
10505         * tree-diagnostic.c (default_tree_printer): Likewise.
10506         * tree-pretty-print.c (newline_and_indent): Likewise.
10508 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
10510         PR c++/87091
10511         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
10512         line above for line-insertion fix-it hints.
10513         (selftest::test_fixit_insert_containing_newline): Update the
10514         expected results, and add a test with line-numbering enabled.
10516 2018-08-27  Martin Liska  <mliska@suse.cz>
10518         PR sanitizer/86962
10519         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
10520         params with DECL_HAS_VALUE_EXPR_P.
10522 2018-08-27  Martin Liska  <mliska@suse.cz>
10524         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
10525         selected expansion strategy.
10527 2018-08-27  Martin Liska  <mliska@suse.cz>
10529         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
10530         * builtins.c (is_builtin_fn): Likewise.
10531         * attribs.c (diag_attr_exclusions): Use new function
10532         fndecl_built_in_p and remove check for FUNCTION_DECL if
10533         possible.
10534         (builtin_mathfn_code): Likewise.
10535         (fold_builtin_expect): Likewise.
10536         (fold_call_expr): Likewise.
10537         (fold_builtin_call_array): Likewise.
10538         (fold_call_stmt): Likewise.
10539         (set_builtin_user_assembler_name): Likewise.
10540         (is_simple_builtin): Likewise.
10541         * calls.c (gimple_alloca_call_p): Likewise.
10542         (maybe_warn_nonstring_arg): Likewise.
10543         * cfgexpand.c (expand_call_stmt): Likewise.
10544         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
10545         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
10546         (cgraph_node::verify_node): Likewise.
10547         * cgraphclones.c (build_function_decl_skip_args): Likewise.
10548         (cgraph_node::create_clone): Likewise.
10549         * config/arm/arm.c (arm_insert_attributes): Likewise.
10550         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
10551         * dse.c (scan_insn): Likewise.
10552         * expr.c (expand_expr_real_1): Likewise.
10553         * fold-const.c (operand_equal_p): Likewise.
10554         (fold_binary_loc): Likewise.
10555         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
10556         * gimple-low.c (lower_stmt): Likewise.
10557         * gimple-pretty-print.c (dump_gimple_call): Likewise.
10558         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10559         Likewise.
10560         * gimple.c (gimple_build_call_from_tree): Likewise.
10561         (gimple_call_builtin_p): Likewise.
10562         (gimple_call_combined_fn): Likewise.
10563         * gimplify.c (gimplify_call_expr): Likewise.
10564         (gimple_boolify): Likewise.
10565         (gimplify_modify_expr): Likewise.
10566         (gimplify_addr_expr): Likewise.
10567         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
10568         * ipa-cp.c (determine_versionability): Likewise.
10569         * ipa-fnsummary.c (compute_fn_summary): Likewise.
10570         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
10571         * ipa-split.c (visit_bb): Likewise.
10572         (split_function): Likewise.
10573         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10574         * lto-cgraph.c (input_node): Likewise.
10575         * lto-streamer-out.c (write_symbol): Likewise.
10576         * omp-low.c (setjmp_or_longjmp_p): Likewise.
10577         (lower_omp_1): Likewise.
10578         * predict.c (strip_predict_hints): Likewise.
10579         * print-tree.c (print_node): Likewise.
10580         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
10581         * trans-mem.c (is_tm_irrevocable): Likewise.
10582         (is_tm_load): Likewise.
10583         (is_tm_simple_load): Likewise.
10584         (is_tm_store): Likewise.
10585         (is_tm_simple_store): Likewise.
10586         (is_tm_abort): Likewise.
10587         (tm_region_init_1): Likewise.
10588         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
10589         * tree-cfg.c (verify_gimple_call): Likewise.
10590         (move_stmt_r): Likewise.
10591         (stmt_can_terminate_bb_p): Likewise.
10592         * tree-eh.c (lower_eh_constructs_2): Likewise.
10593         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
10594         * tree-inline.c (remap_gimple_stmt): Likewise.
10595         (copy_bb): Likewise.
10596         (estimate_num_insns): Likewise.
10597         (fold_marked_statements): Likewise.
10598         * tree-sra.c (scan_function): Likewise.
10599         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
10600         (optimize_stack_restore): Likewise.
10601         (pass_fold_builtins::execute): Likewise.
10602         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
10603         (mark_all_reaching_defs_necessary_1): Likewise.
10604         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
10605         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
10606         (pass_forwprop::execute): Likewise.
10607         * tree-ssa-loop-im.c (stmt_cost): Likewise.
10608         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
10609         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
10610         * tree-ssa-strlen.c (get_string_length): Likewise.
10611         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
10612         (find_func_aliases_for_call): Likewise.
10613         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
10614         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
10615         * tree-tailcall.c (find_tail_calls): Likewise.
10616         * tree.c (need_assembler_name_p): Likewise.
10617         (free_lang_data_in_decl): Likewise.
10618         (get_call_combined_fn): Likewise.
10619         * ubsan.c (is_ubsan_builtin_p): Likewise.
10620         * varasm.c (incorporeal_function_p): Likewise.
10621         * tree.h (DECL_BUILT_IN): Remove and replace with
10622         fndecl_built_in_p.
10623         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
10624         (fndecl_built_in_p): New.
10626 2018-08-27  Martin Liska  <mliska@suse.cz>
10628         PR tree-optimization/86847
10629         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
10630         Dump also subtree probability.
10631         (switch_decision_tree::do_jump_if_equal): New function.
10632         (switch_decision_tree::emit_case_nodes): Handle special
10633         situations in balanced tree that can be emitted much simpler.
10634         Fix calculation of probabilities that happen in tree expansion.
10635         * tree-switch-conversion.h (struct cluster): Add
10636         is_single_value_p.
10637         (struct simple_cluster): Likewise.
10638         (struct case_tree_node): Add new function has_child.
10639         (do_jump_if_equal): New.
10641 2018-08-27  Martin Liska  <mliska@suse.cz>
10643         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
10644         Add new argument to bit_test_cluster constructor.
10645         (bit_test_cluster::emit): Set bits really number of values
10646         handlel by a test.
10647         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
10648         probability argument.
10649         * tree-switch-conversion.h (struct bit_test_cluster):
10650         Add m_handles_entire_switch.
10652 2018-08-27  Martin Liska  <mliska@suse.cz>
10654         PR tree-optimization/86702
10655         * tree-switch-conversion.c (jump_table_cluster::emit):
10656         Make probabilities even for values in jump table
10657         according to number of cases handled.
10658         (switch_decision_tree::compute_cases_per_edge): Pass
10659         argument to reset_out_edges_aux function.
10660         (switch_decision_tree::analyze_switch_statement): Likewise.
10661         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
10662         Make it static.
10664 2018-08-27  Martin Liska  <mliska@suse.cz>
10666         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
10667         cfun argument explicitly.
10668         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
10669         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
10670         function gimple_switch_default_bb.
10671         (convert_switch_statements):
10672         (expand_builtins):
10673         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
10674         * stmt.c (label_to_block_fn): Use label_to_block and pass
10675         cfun argument explicitly and use gimple_switch_label_bb.
10676         (expand_case): Likewise.
10677         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
10678         cfun argument explicitly. Likewise.
10679         (make_edges_bb): Likewise.
10680         (make_cond_expr_edges): Likewise.
10681         (get_cases_for_edge): Likewise.
10682         (make_gimple_switch_edges): Likewise.
10683         (label_to_block_fn): Likewise.
10684         (label_to_block): Likewise.
10685         (make_goto_expr_edges): Likewise.
10686         (make_gimple_asm_edges): Likewise.
10687         (main_block_label): Likewise.
10688         (group_case_labels_stmt): Likewise.
10689         (find_taken_edge_computed_goto): Likewise.
10690         (find_taken_edge_switch_expr): Likewise.
10691         (gimple_verify_flow_info): Likewise.
10692         (gimple_redirect_edge_and_branch): Likewise.
10693         (gimple_switch_label_bb): New function.
10694         (gimple_switch_default_bb): Likewise.
10695         (gimple_switch_edge): Likewise.
10696         (gimple_switch_default_edge): Likewise.
10697         * tree-cfg.h (label_to_block_fn): Remove and replace ...
10698         (label_to_block): ... with this.
10699         (gimple_switch_label_bb): New.
10700         (gimple_switch_default_bb): Likewise.
10701         (gimple_switch_edge): Likewise.
10702         (gimple_switch_default_edge): Likewise.
10703         * tree-cfgcleanup.c (convert_single_case_switch): Use
10704         new gimple functions and pass new argument to label_to_block.
10705         (cleanup_control_flow_bb):
10706         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
10707         cfun argument explicitly.
10708         (make_eh_edges): Likewise.
10709         (redirect_eh_dispatch_edge): Likewise.
10710         (lower_resx): Likewise.
10711         (lower_eh_dispatch): Likewise.
10712         (maybe_remove_unreachable_handlers): Likewise.
10713         (unsplit_eh): Likewise.
10714         (cleanup_empty_eh): Likewise.
10715         (verify_eh_edges): Likewise.
10716         (verify_eh_dispatch_edge): Likewise.
10717         * tree-ssa-dom.c (record_edge_info): Likewise.
10718         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
10719         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
10720         (thread_through_normal_block): Likewise.
10721         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
10722         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
10723         * tree-switch-conversion.c (switch_conversion::collect): Use new
10724         gimple functions.
10725         (switch_conversion::check_final_bb): Likewise.
10726         (switch_conversion::gather_default_values): Pass new argument
10727         to label_to_block.
10728         (switch_conversion::build_constructors): Likewise.
10729         (switch_decision_tree::compute_cases_per_edge): Use new
10730         gimple_switch_edge function.
10731         (switch_decision_tree::analyze_switch_statement): Pass new argument
10732         to label_to_block.
10733         (switch_decision_tree::try_switch_expansion): Use
10734         gimple_switch_default_edge.
10735         * tree-vrp.c (find_switch_asserts): Pass new argument
10736         to label_to_block.
10737         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
10738         (vr_values::simplify_switch_using_ranges): Likewise.
10740 2018-08-27  Richard Biener  <rguenther@suse.de>
10742         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
10743         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
10745         * tree-ssa-sccvn.h (struct vn_pval): New structure.
10746         (struct vn_nary_op_s): Add unwind_to member.  Add
10747         predicated_values flag and put result into a union together
10748         with a linked list of vn_pval.
10749         (struct vn_ssa_aux): Add name member to make maintaining
10750         a map of SSA name to vn_ssa_aux possible.  Remove no longer
10751         needed info, dfsnum, low, visited, on_sccstack, use_processed
10752         and range_info_anti_range_p members.
10753         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
10754         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
10755         New functions.
10756         (vn_valueize): New global.
10757         (vn_context_bb): Likewise.
10758         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
10759         VN_INFO_PTR_INFO): Remove.
10760         * tree-ssa-sccvn.c: ... (rewrite)
10761         (pass_fre::execute): For -O2+ initialize loops and run
10762         RPO VN in optimistic mode (iterating).  For -O1 and -Og
10763         run RPO VN in non-optimistic mode.
10764         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
10765         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
10766         * doc/invoke.texi (sccvn-max-scc-size): Remove.
10767         (rpo-vn-max-loop-depth): Document.
10768         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
10769         when valuezing the VUSE signals we walked out of the region.
10770         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
10771         (phi_translate): Set VN context block to use for availability
10772         lookup.
10773         (compute_avail): Likewise.
10774         (pre_valueize): New function.
10775         (pass_pre::execute): Adjust to the RPO VN API.
10777         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
10778         (propagate_constants_for_unrolling): Remove.
10779         (tree_unroll_loops_completely): Perform value-numbering
10780         on the unrolled bodies loop parent.
10782 2018-08-27  Richard Biener  <rguenther@suse.de>
10784         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
10785         for partial antic compute.
10787 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
10789         PR rtl-optimization/87065
10790         * combine.c (simplify_if_then_else): Formatting fix.
10791         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
10792         check.
10793         (known_cond): Don't return const_true_rtx for vector modes.  Use
10794         CONST0_RTX instead of const0_rtx.  Formatting fixes.
10796 2018-08-27  Martin Liska  <mliska@suse.cz>
10798         PR gcov-profile/87069
10799         * gcov.c (process_file): Record files already processed
10800         and warn about a file being processed multiple times.
10802 2018-08-27  Martin Liska  <mliska@suse.cz>
10804         PR driver/83193
10805         * config/aarch64/aarch64.c (aarch64_override_options_internal):
10806         Set default values for x_aarch64_*_string strings.
10807         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
10808         prefix.  For -mabi do not print '=ABI' in help and use
10809         <option_value> format for -msve-vector-bits and -moverride
10810         options.
10812 2018-08-26  Jeff Law  <law@redhat.com>
10814         * config/mips/frame-header-opt.c: Include "backend.h" rather than
10815         "cfg.h"
10817 2018-08-26  Marek Polacek  <polacek@redhat.com>
10819         PR c++/87029, Implement -Wredundant-move.
10820         * doc/invoke.texi: Document -Wredundant-move.
10822 2018-08-25  Martin Sebor  <msebor@redhat.com>
10824         PR tree-optimization/87059
10825         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
10826         to the same type as the other.
10827         * fold-const.c (fold_binary_loc): Assert expectation.
10829 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
10831         * config/darwin.c (machopic_legitimize_pic_address): Clean up
10832         extraneous parentheses, dead code section and formatting.
10834 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
10836         PR c++/87091
10837         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
10838         wide enough for jumps in the line-numbering to be visible.
10839         (layout::print_gap_in_line_numbering): New member function.
10840         (layout::calculate_line_spans): When using line numbering, merge
10841         line spans that are only 1 line apart.
10842         (diagnostic_show_locus): When printing line numbers, show gaps in
10843         line numbering directly, rather than printing headers.
10844         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
10845         line-numbering with multiple line spans.
10846         (selftest::test_fixit_insert_containing_newline_2): Add test of
10847         line-numbering, in which the spans are close enough to be merged.
10849 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
10851         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
10852         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
10853         * tree-vrp.c (range_is_nonnull): Remove.
10854         (range_includes_zero_p): Accept value_range instead of min/max.
10855         (extract_range_from_binary_expr_1): Do not early bail on
10856         POINTER_PLUS_EXPR.
10857         Use range_includes_zero_p instead of range_is_nonnull.
10858         (extract_range_from_unary_expr): Use range_includes_zero_p instead
10859         of range_is_nonnull.
10860         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
10861         special case VR_ANTI_RANGE.
10862         (vrp_finalize): Same.
10863         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
10864         instead of min/max.
10865         (range_is_nonnull): Remove.
10866         * vr-values.c (vrp_stmt_computes_nonzero): Use
10867         range_includes_zero_p instead of range_is_nonnull.
10868         (extract_range_basic): Pass value_range to range_includes_zero_p
10869         instead of range_is_nonnull.
10871 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
10873         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
10874         * emit-rtl.h (rtl_data): Remove return_bnd.
10875         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
10876         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
10877         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
10878         (POINTER_BOUNDS_MODE): Remove definition.
10879         (make_pointer_bounds_mode): Remove.
10880         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
10881         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
10882         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
10883         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
10884         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
10885         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
10886         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
10888         * config/i386/i386-modes.def (BND32, BND64): Remove.
10889         * config/i386/i386.c (dbx_register_map): Remove bound registers.
10890         (dbx64_register_map): Ditto.
10891         (svr4_dbx_register_map): Ditto.
10892         (indirect_thunk_bnd_needed): Remove.
10893         (indirect_thunks_bnd_used): Ditto.
10894         (indirect_return_bnd_needed): Ditto.
10895         (indirect_return_via_cx_bnd): Ditto.
10896         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
10897         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
10898         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
10899         (output_indirect_thunk_function): Remove handling of
10900         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
10901         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
10902         (ix86_save_reg): Remove handling of crtl->return_bnd.
10903         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
10904         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
10905         and UNSPEC_BNDLX_ADDR.
10906         (ix86_output_indirect_branch_via_reg): Remove handling of
10907         indirect_thunk_prefix_bnd.
10908         (ix86_output_indirect_branch_via_push): Ditto.
10909         (ix86_output_function_return): Ditto.
10910         (ix86_output_indirect_function_return): Ditto.
10911         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
10912         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
10913         (CALL_USED_REGISTERS): Ditto.
10914         (REG_ALLOC_ORDER): Update for removal of bound registers.
10915         (HI_REGISTER_NAMES): Ditto.
10916         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
10917         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
10918         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
10919         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
10920         (FIRST_PSEUDO_REG): Update.
10921         (BND): Remove mode iterator.
10922         * config/i386/predicates.md (bnd_mem_operator): Remove.
10924 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
10926         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
10927         vectors.
10929 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
10931         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
10932         the case in which the permute needs only a single element and
10933         repeats for every vector of the result.  Extend that case to
10934         handle variable-length vectors.
10935         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10937 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
10939         PR debug/79342
10940         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
10941         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
10943 2018-08-24  Richard Biener  <rguenther@suse.de>
10945         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
10946         bb_flags_allocated members.
10947         (auto_flag): New RAII class for allocating flags.
10948         (auto_edge_flag): New RAII class for allocating edge flags.
10949         (auto_bb_flag): New RAII class for allocating bb flags.
10950         * cfgloop.c (verify_loop_structure): Allocate temporary edge
10951         flag dynamically.
10952         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
10953         in favor of temporarily allocated BB flag.
10954         * hsa-brig.c: Re-order includes.
10955         * hsa-dump.c: Likewise.
10956         * hsa-regalloc.c: Likewise.
10957         * print-rtl.c: Likewise.
10958         * profile-count.c: Likewise.
10960 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
10962         PR target/86989
10963         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
10964         the TOC register.
10966 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
10968         PR 87073/bootstrap
10969         * wide-int-range.cc (wide_int_range_div): Do not ignore result
10970         from wide_int_range_multiplicative_op.
10972 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10974         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
10975         "permutaion".
10977 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
10979         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
10980         to 'expanded'.
10982 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
10984         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
10985         full GENERIC expression used for replacement.
10987 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
10989         * tree-vrp.c (abs_extent_range): Remove.
10990         (extract_range_into_wide_ints): Pass wide ints by reference.
10991         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
10992         Pass wide ints by reference in all calls to
10993         extract_range_into_wide_ints.
10994         * wide-int-range.cc (wide_int_range_div): New.
10995         * wide-int-range.h (wide_int_range_div): New.
10996         (wide_int_range_includes_zero_p): New.
10997         (wide_int_range_zero_p): New.
10999 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
11001         * config/aarch64/aarch64.md (arches): New enum.
11002         (arch): New enum attr.
11003         (arch_enabled): New attr.
11004         (enabled): Now uses arch_enabled only.
11005         (simd, sve, fp16): Removed attribute.
11006         (fp): Attr now defined in terms of 'arch'.
11007         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
11008         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
11009         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
11010         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
11011         attributes into 'arch'.
11012         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
11013         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
11014         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
11015         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
11016         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
11017         'simd' attribute into 'arch'.
11018         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
11019         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
11020         Convert use of 'fp' attribute to 'arch'.
11021         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
11022         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
11023         into 'arch'.
11024         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
11025         (different modes) Merge 'fp' and 'simd' into 'arch'.
11026         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
11027         'simd' into 'arch'.
11029 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
11031         PR rtl-optimization/87026
11032         * expmed.c (canonicalize_comparison): If we can no longer create
11033         pseudoregisters, don't.
11035 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
11037         PR target/86951
11038         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
11039         prototype.
11040         * config/arm/arm.c (speculation_barrier_libfunc): New static
11041         variable.
11042         (arm_init_libfuncs): Initialize it.
11043         (arm_emit_speculation_barrier): New function.
11044         * config/arm/arm.md (speculation_barrier): Call
11045         arm_emit_speculation_barrier for architectures that do not have
11046         DSB or ISB.
11047         (speculation_barrier_insn): Only match on Armv7 or later.
11049 2018-08-23  Richard Biener  <rguenther@suse.de>
11051         PR middle-end/87024
11052         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
11053         calls.
11055 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
11057         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
11058         of single-vector TBLs.
11059         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
11060         one input is given.
11062 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
11064         PR target/85910
11065         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
11066         aarch64_evpc_tbl guard.
11068 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11070         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
11071         behaviour.
11073 2018-08-22  Martin Sebor  <msebor@redhat.com>
11075         PR middle-end/87052
11076         * tree-pretty-print.c (pretty_print_string): Add argument.
11077         (dump_generic_node): Call to pretty_print_string with string size.
11079 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
11081         PR rtl-optimization/86771
11082         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
11083         of two SETs into those two SETs, one to be placed at i2, if that SETs
11084         destination is modified between i2 and i3.
11086 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11088         PR tree-optimization/86725
11089         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
11090         function.
11091         (vect_analyze_scalar_cycles_1): Check it.
11093 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11095         PR tree-optimization/86725
11096         * tree-vect-loop.c (vect_is_simple_reduction): When treating
11097         an outer loop phi as a double reduction, make sure that the
11098         single user of the phi result is an inner loop phi.
11100 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11102         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
11103         grouped stores with gaps to a strided group.
11105 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11107         * tree-vect-stmts.c (get_group_load_store_type)
11108         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
11109         first statement in a group.
11111 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11113         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
11114         the sequence used in gcc/gcc.c.
11116 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11118         PR other/704
11119         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
11120         building it.
11122 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11124         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
11125         Darwin10-specific unwinder-shim.
11126         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
11127         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
11128         New to cater for Darwin10 Rosetta.
11130 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11132         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
11133         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
11135 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11137         PR bootstrap/81033
11138         PR target/81733
11139         PR target/52795
11140         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
11141         (dwarf2out_switch_text_section): Generate a local label for the
11142         second function sub-section and apply it as the second FDE start
11143         label.
11144         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
11145         second sub-section start.
11147 2018-08-22  Richard Biener  <rguenther@suse.de>
11149         PR tree-optimization/86988
11150         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
11152 2018-08-22  Richard Biener  <rguenther@suse.de>
11154         PR tree-optimization/86945
11155         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
11157 2018-08-22  Alexandre Oliva <oliva@adacore.com>
11159         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
11160         a comment about how uses of r2 for .sdata2 come about.
11162 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
11164         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
11166 2018-08-21  Marek Polacek  <polacek@redhat.com>
11168         PR c++/86981, Implement -Wpessimizing-move.
11169         * doc/invoke.texi: Document -Wpessimizing-move.
11171 2018-08-21  Jan Hubicka  <jh@suse.cz>
11173         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
11174         * tree.h (is_redundant_typedef): Remove.
11175         * dwarf2out.c (is_redundant_typedef): Turn into static function.
11177 2018-08-21  Jan Hubicka  <jh@suse.cz>
11179         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
11180         when possible.
11182 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
11184         * expmed.c (extract_low_bits): Reject invalid subregs early.
11186 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11188         PR middle-end/86121
11189         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
11190         behaviour.
11192 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11194         * config/vxworks.h: Guard vxworks_asm_out_constructor and
11195         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
11196         * config/vxworks.c: Likewise.
11198 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11200         * config/vxworks.c: Set targetm.have_ctors_dtors
11201         if HAVE_INITFINI_ARRAY_SUPPORT.
11202         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
11203         if HAVE_INITFINI_ARRAY_SUPPORT.
11205 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11207         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
11208         default search path for VxWorks < 7.
11210 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11212         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
11213         (get_format_string): Refer to c_getstr.
11215 2018-08-21  Tom de Vries  <tdevries@suse.de>
11217         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
11218         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
11219         (debuginfo_early_stop): Declare.
11220         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
11221         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
11222         (debuginfo_early_stop): New function.
11223         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
11224         and debuginfo_early_stop.
11225         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
11226         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
11227         (general_init): Call debuginfo_early_init.
11228         (finalize): Call debuginfo_fini.
11229         (do_compile): Call debuginfo_init.
11230         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
11231         -fdump-early-debug.
11232         (@item -fdump-debug, @item -fdump-earlydebug): Add.
11234 2018-08-21  Tom de Vries  <tdevries@suse.de>
11236         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
11237         flag_dump_noaddr and flag_dump_unnumbered.
11239 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
11241         * wide-int-range.cc (wide_int_range_abs): New.
11242         (wide_int_range_order_set): Rename from wide_int_range_min_max.
11243         * wide-int-range.h (wide_int_range_abs): New.
11244         (wide_int_range_min_max): New.
11245         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
11246         case to call wide_int_range_abs.
11247         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
11248         (extract_range_from_abs_expr): Delete.
11250 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
11252         PR target/87033
11253         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
11254         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
11255         for indexed loads.
11257 2018-08-20  Nathan Sidwell  <nathan@acm.org>
11258             Jeff Law <law@redhat.com>
11260         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
11261         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
11263 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
11265         PR other/84889
11266         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
11267         (decl_attributes): Likewise.
11268         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
11269         instance.
11270         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11271         * diagnostic-core.h (class auto_diagnostic_group): New class.
11272         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
11273         (diagnostic_report_diagnostic): Handle the first diagnostics within
11274         a group.
11275         (emit_diagnostic): Add auto_diagnostic_group instance.
11276         (inform): Likewise.
11277         (inform_n): Likewise.
11278         (warning): Likewise.
11279         (warning_at): Likewise.
11280         (warning_n): Likewise.
11281         (pedwarn): Likewise.
11282         (permerror): Likewise.
11283         (error): Likewise.
11284         (error_n): Likewise.
11285         (error_at): Likewise.
11286         (sorry): Likewise.
11287         (fatal_error): Likewise.
11288         (internal_error): Likewise.
11289         (internal_error_no_backtrace): Likewise.
11290         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
11291         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
11292         * diagnostic.h (struct diagnostic_context): Add fields
11293         "diagnostic_group_nesting_depth",
11294         "diagnostic_group_emission_count", "begin_group_cb",
11295         "end_group_cb".
11296         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
11297         Add auto_diagnostic_group instance(s).
11298         (find_explicit_erroneous_behavior): Likewise.
11299         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
11300         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
11301         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
11302         (gimplify_va_arg_expr): Likewise.
11303         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
11304         (HSA_SORRY_AT): Likewise.
11305         * ipa-devirt.c (compare_virtual_tables): Likewise.
11306         (warn_odr): Likewise.
11307         * multiple_target.c (expand_target_clones): Likewise.
11308         * opts-common.c (cmdline_handle_error): Likewise.
11309         * reginfo.c (globalize_reg): Likewise.
11310         * substring-locations.c (format_warning_n_va): Likewise.
11311         * tree-inline.c (expand_call_inline): Likewise.
11312         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
11313         * tree-ssa-loop-niter.c
11314         (do_warn_aggressive_loop_optimizations): Likewise.
11315         * tree-ssa-uninit.c (warn_uninit): Likewise.
11316         * tree.c (warn_deprecated_use): Likewise.
11318 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
11320         PR target/87014
11321         * config/i386/i386.md (eh_return): Always update EH return
11322         address in word_mode.
11324 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
11326         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
11327         TARGET_SPLIT_COMPLEX_ARG is defined.
11329 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11331         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
11333 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11335         PR target/86984
11336         * expr.c (expand_assignment): Assert that bitpos is positive.
11337         (store_field): Likewise
11338         (expand_expr_real_1): Make sure that bitpos is positive.
11339         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
11340         integer overflow.
11342 2018-08-20  Nathan Sidwell  <nathan@acm.org>
11344         * Makefile.in (CPP_ID_DATA_H): Delete.
11345         (CPP_INTERNAL_H): Don't add it.
11346         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
11347         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
11349 2018-08-20  Richard Biener  <rguenther@suse.de>
11351         PR tree-optimization/78655
11352         * tree-vrp.c (extract_range_from_binary_expr_1): Make
11353         pointer + offset nonnull if either operand is nonnull work.
11355 2018-08-20  Tom de Vries  <tdevries@suse.de>
11357         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
11358         unless the referenced die describes the added property using
11359         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
11360         Otherwise, add a DW_AT_location to the referenced die.
11362 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
11364         PR target/86994
11365         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
11366         register_operand when calling ix86_set_reg_reg_cost.
11367         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
11368         Set *total to 0 for operands that satisfy x86_64_immediate_operand
11369         predicate and to 1 otherwise.
11371 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
11373         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
11374         emit a diagnostic that it is not supported and reset the option.
11375         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
11376         supported and consume the option.  (ASM_FINAL_SPEC): New.
11378 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
11380         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
11381         a sentence.
11383 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
11385         C-SKY port: Documentation
11387         * doc/extend.texi (C-SKY Function Attributes): New section.
11388         * doc/invoke.texi (Option Summary): Add C-SKY options.
11389         (C-SKY Options): New section.
11390         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
11392 2018-08-17  Jojo  <jijie_rong@c-sky.com>
11393             Huibin Wang  <huibin_wang@c-sky.com>
11394             Sandra Loosemore  <sandra@codesourcery.com>
11395             Chung-Lin Tang  <cltang@codesourcery.com>
11397         C-SKY port: Backend implementation
11399         * config/csky/*: New.
11400         * common/config/csky/*: New.
11402 2018-08-17  Jojo  <jijie_rong@c-sky.com>
11403             Huibin Wang  <huibin_wang@c-sky.com>
11404             Sandra Loosemore  <sandra@codesourcery.com>
11405             Chung-Lin Tang  <cltang@codesourcery.com>
11406             Andrew Jenner  <andrew@codesourcery.com>
11408         C-SKY port: Configury
11410         * config.gcc (csky-*-*): New.
11411         * configure.ac: Add csky to targets for dwarf2 debug_line support.
11412         * configure: Regenerated.
11414 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
11416         * dump-context.h: Include "dumpfile.h".
11417         (dump_context::dump_printf_va): Convert final param from va_list
11418         to va_list *.  Convert from ATTRIBUTE_PRINTF to
11419         ATTRIBUTE_GCC_DUMP_PRINTF.
11420         (dump_context::dump_printf_loc_va): Likewise.
11421         * dumpfile.c: Include "stringpool.h".
11422         (make_item_for_dump_printf_va): Delete.
11423         (make_item_for_dump_printf): Delete.
11424         (class dump_pretty_printer): New class.
11425         (dump_pretty_printer::dump_pretty_printer): New ctor.
11426         (dump_pretty_printer::emit_items): New member function.
11427         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
11428         function.
11429         (dump_pretty_printer::emit_item): New member function.
11430         (dump_pretty_printer::stash_item): New member function.
11431         (dump_pretty_printer::format_decoder_cb): New member function.
11432         (dump_pretty_printer::decode_format): New member function.
11433         (dump_context::dump_printf_va): Reimplement in terms of
11434         dump_pretty_printer.
11435         (dump_context::dump_printf_loc_va): Convert final param from va_list
11436         to va_list *.
11437         (dump_context::begin_scope): Reimplement call to
11438         make_item_for_dump_printf.
11439         (dump_printf): Update for change to dump_printf_va.
11440         (dump_printf_loc): Likewise.
11441         (selftest::test_capture_of_dump_calls): Convert "stmt" from
11442         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
11443         with %T, %E, and %G.
11444         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
11445         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
11446         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
11447         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
11448         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
11449         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
11450         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
11451         within a dump_printf_loc call to "%wu".
11452         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
11453         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
11454         missing space after "=".
11455         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
11456         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
11457         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
11458         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
11459         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
11460         duplicate "vectorized" from message.
11462 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
11464         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
11465         polyNxK_t element's TYPE_STRING_FLAG.
11467 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
11469         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
11470         (they were unnamed before).  Fix comments.
11472 2018-08-17  Nathan Sidwell  <nathan@acm.org>
11474         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
11476 2018-08-17  Richard Biener  <rguenther@suse.de>
11478         PR tree-optimization/86841
11479         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
11481 2018-08-17  Martin Liska  <mliska@suse.cz>
11483         * common.opt: Remove Warn, Init and Report for options with
11484         Ignore/Deprecated flag. Warning is done automatically for
11485         Deprecated flags.
11486         * config/i386/i386.opt: Likewise.
11487         * config/ia64/ia64.opt: Likewise.
11488         * config/rs6000/rs6000.opt: Likewise.
11489         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
11490         Remove usage of flag_check_pointer_bounds.
11491         * lto-wrapper.c (merge_and_complain): Do not handle
11492         OPT_fcheck_pointer_bounds.
11493         (append_compiler_options): Likewise.
11494         * opt-functions.awk: Do not handle Deprecated.
11495         * optc-gen.awk: Check that Var, Report and Init are not
11496         used for an option with Ignore/Deprecated flag.
11497         * opts-common.c (decode_cmdline_option): Do not report
11498         CL_ERR_DEPRECATED.
11499         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
11500         options.
11501         * opts.h (struct cl_option): Remove cl_deprecated flag.
11502         (CL_ERR_DEPRECATED): Remove error enum value.
11504 2018-08-17  Richard Biener  <rguenther@suse.de>
11506         PR middle-end/86505
11507         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
11508         across a va-arg-pack using call adjust its return value accordingly.
11510 2018-08-16  Martin Sebor  <msebor@redhat.com>
11512         PR tree-optimization/86853
11513         * gimple-ssa-sprintf.c (struct format_result): Rename member.
11514         (struct fmtresult): Add member and initialize it in ctors.
11515         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
11516         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
11517         (format_directive): Set POSUNDER4K when MAYFAIL is set.
11518         (parse_directive): Handle %C same as %c.
11519         (sprintf_dom_walker::compute_format_length): Adjust.
11520         (is_call_safe): Adjust.
11522 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11524         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
11525         for determining how to count the elements.
11526         * builtins.h (c_strlen): Adjust prototype.
11527         * expr.c (string_constant): Add new parameter mem_size.
11528         Set *mem_size appropriately.
11529         * expr.h (string_constant): Adjust protoype.
11530         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
11531         * gimple-fold.h (get_range_strlen): Adjust prototype.
11532         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
11533         (format_string): Call get_string_length with eltsize.
11535 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
11537         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
11538         to emit the span, rather than setting it as the prefix.
11540 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
11542         * diagnostic-show-locus.c (layout::start_annotation_line): Add
11543         "margin_char" parameter, defaulting to space.  Use it in place
11544         of pp_space for the initial part of the margin.
11545         (layout::print_leading_fixits): Use '+' when filling the margin
11546         of line-insertion fix-it hints.
11548 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11550         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
11551         Delete.
11553 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11555         * config/rs6000/altivec.md: Don't set length attribute to the default
11556         value.
11557         * config/rs6000/darwin.md: Ditto.
11558         * config/rs6000/dfp.md: Ditto.
11559         * config/rs6000/htm.md: Ditto.
11560         * config/rs6000/rs6000.md: Ditto.
11561         * config/rs6000/sync.md: Ditto.
11562         * config/rs6000/vsx.md: Ditto.
11564 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11566         * config/rs6000/altivec.md: Don't set length attribute to the default
11567         value, for branch instructions.
11568         * config/rs6000/darwin.md: Ditto.
11569         * config/rs6000/rs6000.md: Ditto.
11571 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11573         * config/rs6000/rs6000.md (length): Always define as const_int 4.
11574         (unnamed conditional branch define_insn): Set length to 4 or 8
11575         depending on offset.
11576         (<bd>_<mode>): Similar, for alternative 0.
11577         (<bd>tf_<mode>): Ditto.
11579 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
11581         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
11583 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
11585         * doc/rtl.texi: Replace old RTX class names with new names.
11588 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
11590         * expmed.h (canonicalize_comparison): New declaration.
11591         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
11592         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
11593         * optabs.c (prepare_cmp_insn): Likewise.
11594         * rtl.h (unsigned_condition_p): New function which checks if a
11595         comparison operator is unsigned.
11597 2018-08-16  Nathan Sidwell  <nathan@acm.org>
11599         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
11600         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
11602 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
11604         PR target/84711
11605         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
11606         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
11607         (mov<mov>): ..this and enable unconditionally.
11609 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
11611         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
11613 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
11615         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
11616         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
11617         "Common" with "Target".
11619 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
11621         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
11622         * doc/invoke.texi (mmitigate-rop): Remove.
11623         * config/i386/i386.c: Do not include "regrename.h".
11624         (ix86_rop_should_change_byte_p, reg_encoded_number)
11625         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
11626         Remove.
11627         (ix86_reorg): Remove call to ix86_mitigate_rop.
11628         * config/i386/i386.md (attr "modrm_class"): Remove.
11629         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
11630         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
11631         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
11633 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
11635         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
11636         allow folding of mergeh() and mergel() for the float and double types.
11637         (fold_mergehl_helper): Rework to handle building a permute tree
11638         for float vectors.
11640 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
11642         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
11643         for TARGET_SSE.
11645 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
11647         * common.opt (fdiagnostics-show-labels): New option.
11648         * diagnostic-show-locus.c (class layout_range): Add field
11649         "m_label".
11650         (class layout): Add field "m_show_labels_p".
11651         (layout_range::layout_range): Add param "label" and use it to
11652         initialize m_label.
11653         (make_range): Pass in NULL for new "label" param of layout_range's
11654         ctor.
11655         (layout::layout): Initialize m_show_labels_p.
11656         (layout::maybe_add_location_range): Pass in loc_range->m_label
11657         when constructing layout_range instances.
11658         (struct line_label): New struct.
11659         (layout::print_any_labels): New member function.
11660         (layout::print_line): Call it if label-printing is enabled.
11661         (selftest::test_one_liner_labels): New test.
11662         (selftest::test_diagnostic_show_locus_one_liner): Call it.
11663         * diagnostic.c (diagnostic_initialize): Initialize
11664         context->show_labels_p.
11665         * diagnostic.h (struct diagnostic_context): Add field
11666         "show_labels_p".
11667         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
11668         -fno-diagnostics-show-labels.
11669         * dwarf2out.c (gen_producer_string): Add
11670         OPT_fdiagnostics_show_labels to the ignored options.
11671         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
11672         param.
11673         (gcc_rich_location::maybe_add_expr): Likewise.
11674         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
11675         label" param, defaulting to NULL.
11676         (gcc_rich_location::add_expr): Add "label" param.
11677         (gcc_rich_location::maybe_add_expr): Likewise.
11678         (class text_range_label): New class.
11679         (class range_label_for_type_mismatch): New class.
11680         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
11681         of format_warning_va.
11682         (fmtwarn_n): Likewise for new params of format_warning_n_va.
11683         * lto-wrapper.c (merge_and_complain): Add
11684         OPT_fdiagnostics_show_labels to the "pick one setting" options.
11685         (append_compiler_options): Likewise to the dropped options.
11686         (append_diag_options): Likewise to the passed-on options.
11687         * opts.c (common_handle_option): Handle the new option.
11688         * selftest-diagnostic.c
11689         (test_diagnostic_context::test_diagnostic_context): Enable
11690         show_labels_p.
11691         * substring-locations.c: Include "gcc-rich-location.h".
11692         (format_warning_n_va): Add "fmt_label" and "param_label" params
11693         and use them as appropriate.
11694         (format_warning_va): Add "fmt_label" and "param_label" params,
11695         passing them on to format_warning_n_va.
11696         (format_warning_at_substring): Likewise.
11697         (format_warning_at_substring_n): Likewise.
11698         * substring-locations.h (format_warning_va): Add "fmt_label" and
11699         "param_label" params.
11700         (format_warning_n_va): Likewise.
11701         (format_warning_at_substring): Likewise.
11702         (format_warning_at_substring_n): Likewise.
11703         * toplev.c (general_init): Initialize global_dc->show_labels_p.
11705 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
11707         PR testsuite/86519
11708         * builtins.c (expand_builtin_memcmp): Do not expand the call
11709         when overflow is detected.
11711 2018-08-15  Martin Sebor  <msebor@redhat.com>
11713         PR tree-optimization/71625
11714         * config/aarch64/aarch64-builtins.c
11715         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
11717 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
11719         * config/s390/s390.c (s390_reorg): Remove loop.
11721 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
11723         * config/darwin.c
11724          (darwin_function_switched_text_sections): Delete.
11725         * gcc/config/darwin.h
11726          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
11728 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
11730         PR target/81685
11731         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
11732         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
11733         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
11735 2018-08-15  Martin Liska  <mliska@suse.cz>
11737         PR tree-optimization/86925
11738         * predict.c (expr_expected_value_1): When taking
11739         later predictor, assign also probability.
11740         Use fold_build2_initializer_loc in order to fold
11741         the expression in -frounding-math.
11743 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
11745         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
11746         patterns.
11747         (expand_vec_perm_1): Try the new method.
11749 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
11751         PR target/86547
11752         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
11753         Check whether lra_live_max_point is 0 before dividing.
11755 2018-08-14  Martin Sebor  <msebor@redhat.com>
11757         PR tree-optimization/86650
11758         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
11759         (vrp_prop::check_mem_ref): Same.
11761 2018-08-13  Liu Hao <lh_mouse@126.com>
11763         * pretty-print.c (eat_esc_sequence): Swap the foreground and
11764         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
11765         and clear it thereafter, as it only works for DBCS.
11767 2018-08-13  Liu Hao <lh_mouse@126.com>
11769         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
11770         handle returned by _get_osf_handle().
11772 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
11774         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
11775         for folding vec_perm.
11777 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
11779         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
11780         Add support for gimple-folding of vec_pack() and vec_unpack()
11781         intrinsics.
11783 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
11785         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
11786         vec_xst variants to the list.
11787         (rs6000_gimple_fold_builtin): Add support for folding unaligned
11788         vector loads and stores.
11790 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
11792         * config.gcc (rs6000-ibm-aix4.x): Delete.
11793         (rs6000-ibm-aix5.1): Delete.
11794         (rs6000-ibm-aix5.2): Delete.
11795         (rs6000-ibm-aix5.3): Delete.
11796         * config/rs6000/aix43.h: Delete.
11797         * config/rs6000/aix51.h: Delete.
11798         * config/rs6000/aix52.h: Delete.
11799         * config/rs6000/t-aix43: Delete.
11801 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
11803         * config/s390/s390.c (s390_decompose_constant_pool_ref):
11804         New function.
11805         (s390_decompose_address): Factor out constant pool ref
11806         decomposition.
11808 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11810         * config/nds32/nds32-predicates.c
11811         (nds32_can_use_bclr_p): Change return type as bool.
11812         (nds32_can_use_bset_p): Ditto.
11813         (nds32_can_use_btgl_p): Ditto.
11814         (nds32_can_use_bitci_p): Ditto.
11815         * config/nds32/nds32-protos.h
11816         (nds32_can_use_bclr_p): Change declaration.
11817         (nds32_can_use_bset_p): Ditto.
11818         (nds32_can_use_btgl_p): Ditto.
11819         (nds32_can_use_bitci_p): Ditto.
11821 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11823         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
11824         Support -msched-prolog-epilog option.
11825         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
11827 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11829         * common/config/nds32/nds32-common.c
11830         (nds32_option_optimization_table): Enalbe -malways-align.
11832 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11834         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
11835         extra_headers.
11836         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
11837         OPT_misr_secure_ case.
11838         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
11839         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
11840         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
11841         secure attribute.
11842         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
11843         (nds32_isr_info): New field security_level.
11844         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
11845         * config/nds32/nds32.md (return_internal): Consider critical attribute.
11846         * config/nds32/nds32.opt (misr-secure): New option.
11847         * config/nds32/nds32_init.inc: New file.
11848         * config/nds32/nds32_isr.h: New file.
11850 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
11852         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
11853         Update comment for atomic instructions.
11854         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
11855         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
11856         Remove.
11857         (atomic_loaddi): Revise fence expansion to only emit fence prior to
11858         load for __ATOMIC_SEQ_CST model.
11859         (atomic_loaddi_1): Remove float register target.
11860         (atomic_storedi): Handle CONST_INT values.
11861         (atomic_storedi_1): Remove float register source.  Add special case
11862         for zero value.
11863         (memory_barrier): New expander and insn.
11865 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
11867         PR tree-optimization/86835
11868         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
11869         new_stmt after def_gsi, make sure to insert new_square_stmt after
11870         that stmt, not 2 stmts before it.
11872 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
11874         PR target/82418
11875         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
11876         instead of SWI48.
11878 2018-08-10  Martin Liska  <mliska@suse.cz>
11880         PR target/83610
11881         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
11882         function type.
11883         * builtins.c (expand_builtin_expect_with_probability):
11884         New function.
11885         (expand_builtin_expect_with_probability): New function.
11886         (build_builtin_expect_predicate): Add new argumnet probability
11887         for BUILT_IN_EXPECT_WITH_PROBABILITY.
11888         (fold_builtin_expect):
11889         (fold_builtin_2):
11890         (fold_builtin_3):
11891         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
11892         * builtins.h (fold_builtin_expect): Set new argument.
11893         * doc/extend.texi: Document __builtin_expect_with_probability.
11894         * doc/invoke.texi: Likewise.
11895         * gimple-fold.c (gimple_fold_call): Pass new argument.
11896         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
11897         also BUILT_IN_EXPECT_WITH_PROBABILITY.
11898         * predict.c (get_predictor_value): New function.
11899         (expr_expected_value): Add new argument probability. Assume
11900         that predictor and probability are always non-null.
11901         (expr_expected_value_1): Likewise.  For __builtin_expect and
11902         __builtin_expect_with_probability set probability.  Handle
11903         combination in binary expressions.
11904         (tree_predict_by_opcode): Simplify code by simply calling
11905         get_predictor_value.
11906         (pass_strip_predict_hints::execute): Add handling of
11907         BUILT_IN_EXPECT_WITH_PROBABILITY.
11908         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
11909         new predictor.
11910         * tree.h (DECL_BUILT_IN_P): New function.
11912 2018-08-10  Martin Liska  <mliska@suse.cz>
11914         PR tree-optimization/85799
11915         * passes.def: Add argument for pass_strip_predict_hints.
11916         * predict.c (class pass_strip_predict_hints): Add new argument
11917         early_p.
11918         (strip_predictor_early): New function.
11919         (pass_strip_predict_hints::execute): Call the function to
11920         strip predictors.
11921         (strip_predict_hints): New function.
11922         * predict.def: Fix comment.
11924 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
11926         * Makefile.in: Clarify which tm.texi to copy over to assert the
11927         right to grant a GFDL license for all.
11929 2018-08-09  Jeff Law  <law@redhat.com>
11931         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
11932         unused variable.
11934 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
11936         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
11937         prototype.
11939 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
11941         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
11942         reductions for variable-length vectors.
11944 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
11946         PR other/84889
11947         * common.opt (fdiagnostics-show-line-numbers): New option.
11948         * diagnostic-show-locus.c (class layout): Add fields
11949         "m_show_line_numbers_p" and "m_linenum_width";
11950         (num_digits): New function.
11951         (test_num_digits): New function.
11952         (layout::layout): Initialize new fields.  Update m_x_offset
11953         logic to handle any left margin.
11954         (layout::print_source_line): Print line number when requested.
11955         (layout::start_annotation_line): New member function.
11956         (layout::print_annotation_line): Call it.
11957         (layout::print_leading_fixits): Likewise.
11958         (layout::print_trailing_fixits): Likewise.  Update calls to
11959         move_to_column for new parameter.
11960         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
11961         it to potentially call start_annotation_line.
11962         (layout::show_ruler): Call start_annotation_line.
11963         (selftest::test_line_numbers_multiline_range): New selftest.
11964         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
11965         and selftest::test_line_numbers_multiline_range.
11966         * diagnostic.c (diagnostic_initialize): Initialize
11967         show_line_numbers_p.
11968         * diagnostic.h (struct diagnostic_context): Add field
11969         "show_line_numbers_p".
11970         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
11971         -fno-diagnostics-show-line-numbers.
11972         * dwarf2out.c (gen_producer_string): Add
11973         OPT_fdiagnostics_show_line_numbers to the ignored options.
11974         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
11975         one setting" options.
11976         (append_compiler_options): Likewise to the dropped options.
11977         (append_diag_options): Likewise to the passed-on options.
11978         * opts.c (common_handle_option): Handle the new option.
11979         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
11981 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11983         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
11984         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
11985         third argument of type "const signed char" to descriptions of
11986         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
11987         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
11988         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
11989         __builtin_bcdsub_ov functions.
11991 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
11993         PR tree-optimization/86858
11994         * tree-vect-loop.c (vect_is_simple_reduction): Restore
11995         flow_bb_inside_loop_p calls.
11997 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
11999         PR tree-optimization/86871
12000         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
12001         instead of gimple_assign_lhs.
12003 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
12005         PR target/86887
12006         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
12007         register constraint to operand 0.
12008         (add<mode>3_carryinC): Likewise.
12009         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
12011 2018-08-09  Martin Liska  <mliska@suse.cz>
12013         PR c/86895
12014         * common.opt: Remove extra line.
12016 2018-08-09  Martin Liska  <mliska@suse.cz>
12018         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
12019         at the end of a line, make first letter capital and end up
12020         a sentence with a dot.
12021         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
12022         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
12023         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
12024         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
12025         (PARAM_MAX_ISL_OPERATIONS): Likewise.
12026         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
12027         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
12028         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
12029         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
12030         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
12031         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
12032         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
12033         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
12034         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
12035         (PARAM_TREE_REASSOC_WIDTH): Likewise.
12036         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
12037         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
12038         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
12040 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
12042         PR target/84332
12043         * config/s390/s390.c (s390_option_override_internal): Reduce the
12044         stack-clash-protection-probe-interval param if it would be too big
12045         for z900.
12047 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
12049         PR target/46179
12050         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
12051         * config/m68k/m68k.c (handle_move_double): Don't call
12052         m68k_final_prescan_insn.
12053         (m68k_adjust_decorated_operand): Renamed from
12054         m68k_final_prescan_insn, remove first and third operand and
12055         simplify.
12056         (print_operand): Call it.
12057         (print_operand_address): Call it.
12059 2018-08-08  Nathan Sidwell  <nathan@acm.org>
12061         * diagnostic.c (diagnostic_report_current_module): Use
12062         linemap_included_from & linemap_included_from_linemap.
12064 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
12066         * config/aarch64/aarch64-cores.def: Add phecda core.
12067         * config/aarch64/aarch64-tune.md: Regenerate.
12068         * doc/invoke.texi: Add phecda core.
12070 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
12072         PR target/85295
12073         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
12074         definitions.
12075         * config/s390/s390.md ("movti"): Add more alternatives for
12076         constant to GPR copies.
12078 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
12080         * config/s390/s390.c: Fix whitespace damage throughout the file.
12081         * config/s390/s390.h: Likewise.
12082         * config/s390/tpf.h: Likewise.
12084 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
12086         * config/s390/s390.c (s390_loadrelative_operand_p):
12087         Remove TARGET_CPU_ZARCH usages.
12088         (s390_rtx_costs): Likewise.
12089         (s390_legitimate_constant_p): Likewise.
12090         (s390_cannot_force_const_mem): Likewise.
12091         (legitimate_reload_constant_p): Likewise.
12092         (s390_preferred_reload_class): Likewise.
12093         (legitimize_pic_address): Likewise.
12094         (legitimize_tls_address): Likewise.
12095         (s390_split_branches): Removed.
12096         (s390_add_execute): Removed.
12097         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
12098         (s390_mainpool_start): Likewise.
12099         (s390_mainpool_finish): Likewise.
12100         (s390_mainpool_cancel): Removed.
12101         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
12102         (s390_chunkify_cancel): Likewise.
12103         (s390_return_addr_rtx): Likewise.
12104         (s390_register_info): Remove split_branches_pending_p uages.
12105         (s390_optimize_register_info): Likewise.
12106         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
12107         split_branches_pending_p usages.
12108         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
12109         (s390_load_got): Likewise.
12110         (s390_expand_split_stack_prologue): Likewise.
12111         (output_asm_nops): Likewise.
12112         (s390_function_profiler): Likewise.
12113         (s390_emit_call): Likewise.
12114         (s390_conditional_register_usage): Likewise.
12115         (s390_optimize_prologue): Likewise.
12116         (s390_reorg): Remove TARGET_CPU_ZARCH and
12117         split_branches_pending_p usages.
12118         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
12119         usages.
12120         (s390_output_indirect_thunk_function): Likewise.
12121         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
12122         (TARGET_CPU_ZARCH_P): Removed.
12123         (struct machine_function): Remove split_branches_pending_p.
12124         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
12126 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
12128         * common/config/s390/s390-common.c (processor_flags_table):
12129         Remove flags.
12130         * config.gcc: Remove with_arch/with_tune support.
12131         * config/s390/2064.md: Remove cpu attribute comparisons.
12132         * config/s390/driver-native.c (s390_host_detect_local_cpu):
12133         Remove MTN.
12134         * config/s390/linux.h (ASM_SPEC):
12135         Remove -march support.
12136         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
12137         Use a table to get an arch level.
12138         * config/s390/s390-opts.h (enum processor_type):
12139         Remove enum values.
12140         * config/s390/s390.c
12141         (processor_table): Remove entries, add arch_level values.
12142         (s390_issue_rate): Remove cases.
12143         (s390_option_override): Adjust
12144         s390_option_override_internal() call.
12145         (s390_option_override_internal): Remove deprecation warning.
12146         (s390_valid_target_attribute_tree): Adjust
12147         s390_option_override_internal() call.
12148         * config/s390/s390.h (struct s390_processor):
12149         Share with s390-c.c, add arch_level field.
12150         * config/s390/s390.md:
12151         Remove occurrences in cpu attribute.
12152         * config/s390/s390.opt: Remove -march/-mtune support.
12153         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
12154         * doc/invoke.texi: Remove deprecation warning.
12156 2018-08-08  Luis Machado  <luis.machado@linaro.org>
12158         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
12159         global.
12160         (qdf24xx_tunings): Set vector cost structure to
12161         qdf24xx_vector_cost.
12163         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
12164         <register_sextend>: Set to 3.
12166 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
12168         PR target/86838
12169         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
12170         * config/aarch64/aarch64-simd.md
12171         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
12172         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
12173         (aarch64_frecpx<mode>): ...this new pattern.
12174         * config/aarch64/aarch64-simd-builtins.def: Remove comment
12175         about aarch64_frecp<FRECP:frecp_suffix><mode>.
12177 2018-08-07  Martin Liska  <mliska@suse.cz>
12179         PR middle-end/83023
12180         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
12181         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
12182         * predict.def (PRED_MALLOC_NONNULL): New predictor.
12183         * doc/extend.texi: Document that malloc attribute adds
12184         hit to compiler.
12186 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
12188         PR target/86785
12189         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12190         Define to speculation_safe_value_not_needed.
12192 2018-08-06  Jeff Law  <law@redhat.com>
12194         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
12195         the vr_values instance to cprop_into_stmt.
12196         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
12197         (cprop_operand): Also query EVRP to determine if OP is a constant.
12199 2018-08-06  Nathan Sidwell  <nathan@acm.org>
12201         * diagnostic.c (diagnostic_report_current_module): Reroll
12202         included-at loop.  Translate text.
12204 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
12206         * function-tests.c (selftest::test_expansion_to_rtl): Call
12207         free_after_compilation.
12209 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12211         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
12213 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
12215         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
12216         loops with memory block operations from getting unrolled.
12218 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
12220         PR target/86807
12221         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12222         Define to speculation_safe_value_not_needed.
12224 2018-08-06  Jeff Law  <law@redhat.com>
12226         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
12227         assert.
12229 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12231         PR target/86662
12232         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
12233         with all enabled __intN types.
12235         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
12237 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12239         * alias.c (record_set): Check for clobber high.
12240         * cfgexpand.c (expand_gimple_stmt): Likewise.
12241         * combine-stack-adj.c (single_set_for_csa): Likewise.
12242         * combine.c (find_single_use_1): Likewise.
12243         (set_nonzero_bits_and_sign_copies): Likewise.
12244         (get_combine_src_dest): Likewise.
12245         (is_parallel_of_n_reg_sets): Likewise.
12246         (try_combine): Likewise.
12247         (record_dead_and_set_regs_1): Likewise.
12248         (reg_dead_at_p_1): Likewise.
12249         (reg_dead_at_p): Likewise.
12250         * dce.c (deletable_insn_p): Likewise.
12251         (mark_nonreg_stores_1): Likewise.
12252         (mark_nonreg_stores_2): Likewise.
12253         * df-scan.c (df_find_hard_reg_defs): Likewise.
12254         (df_uses_record): Likewise.
12255         (df_get_call_refs): Likewise.
12256         * dwarf2out.c (mem_loc_descriptor): Likewise.
12257         * haifa-sched.c (haifa_classify_rtx): Likewise.
12258         * ira-build.c (create_insn_allocnos): Likewise.
12259         * ira-costs.c (scan_one_insn): Likewise.
12260         * ira.c (equiv_init_movable_p): Likewise.
12261         (rtx_moveable_p): Likewise.
12262         (interesting_dest_for_shprep): Likewise.
12263         * jump.c (mark_jump_label_1): Likewise.
12264         * postreload-gcse.c (record_opr_changes): Likewise.
12265         * postreload.c (reload_cse_simplify): Likewise.
12266         (struct reg_use): Add source expr.
12267         (reload_combine): Check for clobber high.
12268         (reload_combine_note_use): Likewise.
12269         (reload_cse_move2add): Likewise.
12270         (move2add_note_store): Likewise.
12271         * print-rtl.c (print_pattern): Likewise.
12272         * recog.c (decode_asm_operands): Likewise.
12273         (store_data_bypass_p): Likewise.
12274         (if_test_bypass_p): Likewise.
12275         * regcprop.c (kill_clobbered_value): Likewise.
12276         (kill_set_value): Likewise.
12277         * reginfo.c (reg_scan_mark_refs): Likewise.
12278         * reload1.c (maybe_fix_stack_asms): Likewise.
12279         (eliminate_regs_1): Likewise.
12280         (elimination_effects): Likewise.
12281         (mark_not_eliminable): Likewise.
12282         (scan_paradoxical_subregs): Likewise.
12283         (forget_old_reloads_1): Likewise.
12284         * reorg.c (find_end_label): Likewise.
12285         (try_merge_delay_insns): Likewise.
12286         (redundant_insn): Likewise.
12287         (own_thread_p): Likewise.
12288         (fill_simple_delay_slots): Likewise.
12289         (fill_slots_from_thread): Likewise.
12290         (dbr_schedule): Likewise.
12291         * resource.c (update_live_status): Likewise.
12292         (mark_referenced_resources): Likewise.
12293         (mark_set_resources): Likewise.
12294         * rtl.c (copy_rtx): Likewise.
12295         * rtlanal.c (reg_referenced_p): Likewise.
12296         (single_set_2): Likewise.
12297         (noop_move_p): Likewise.
12298         (note_stores): Likewise.
12299         * sched-deps.c (sched_analyze_reg): Likewise.
12300         (sched_analyze_insn): Likewise.
12302 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12304         * cse.c (invalidate_reg): New function extracted from...
12305         (invalidate): ...here.
12306         (canonicalize_insn): Check for clobber high.
12307         (invalidate_from_clobbers): invalidate clobber highs.
12308         (invalidate_from_sets_and_clobbers): Likewise.
12309         (count_reg_usage): Check for clobber high.
12310         (insn_live_p): Likewise.
12311         * cselib.c (cselib_expand_value_rtx_1):Likewise.
12312         (cselib_invalidate_regno): Check for clobber in setter.
12313         (cselib_invalidate_rtx): Pass through setter.
12314         (cselib_invalidate_rtx_note_stores):
12315         (cselib_process_insn): Check for clobber high.
12316         * cselib.h (cselib_invalidate_rtx): Add operand.
12318 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12320         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
12321         (mark_not_eliminable): Likewise.
12322         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
12323         * lra-lives.c (process_bb_lives): Check for clobber high.
12324         * lra.c (new_insn_reg): Remember clobber highs.
12325         (collect_non_operand_hard_regs): Check for clobber high.
12326         (lra_set_insn_recog_data): Likewise.
12327         (add_regs_to_insn_regno_info): Likewise.
12328         (lra_update_insn_regno_info): Likewise.
12330 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12332         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
12333         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
12335 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12337         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
12338         (copy_insn_1): Likewise.
12339         (gen_hard_reg_clobber_high): New gen function.
12340         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
12341         * genemit.c (gen_exp): Likewise.
12342         (gen_emit_seq): Pass through info.
12343         (gen_insn): Check for CLOBBER_HIGH.
12344         (gen_expand): Pass through info.
12345         (gen_split): Likewise.
12346         (output_add_clobbers): Likewise.
12347         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
12348         (remove_clobbers): Likewise.
12349         * rtl.h (gen_hard_reg_clobber_high): New declaration.
12351 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12353         * doc/rtl.texi (clobber_high): Add.
12354         (parallel): Add in clobber high
12355         * rtl.c (rtl_check_failed_code3): Add function.
12356         * rtl.def (CLOBBER_HIGH): Add expression.
12357         * rtl.h (RTL_CHECKC3): Add macro.
12358         (rtl_check_failed_code3): Add declaration.
12359         (XC3EXP): Add macro.
12361 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
12363         PR target/86386
12364         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
12365         cfun->machine->max_used_stack_alignment if needed.
12367 2018-08-04  Martin Sebor  <msebor@redhat.com>
12369         PR tree-optimization/86571
12370         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
12371         NaN output to 4.
12373 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
12375         PR target/86799
12376         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12377         Define.
12379 2018-08-03  Jeff Law  <law@redhat.com>
12381         PR target/86795
12382         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12383         Define to speculation_safe_value_not_needed.
12385 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
12387         * doc/gcov.texi (-x): Remove duplicate "to".
12388         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
12389         (-Wif-not-aligned): Remove duplicate "is".
12390         (-flto): Remove duplicate "the".
12391         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
12392         duplicate "v5.00.b".
12393         (MSP430 Options): Remove duplicate "and" from the description
12394         of "-mgprel-sec=regexp".
12395         (x86 Options): Remove duplicate copies of "vmldLog102" and
12396         vmlsLog104 from description of "-mveclibabi=type".
12398 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
12400         * internal-fn.h (first_commutative_argument): Declare.
12401         * internal-fn.c (first_commutative_argument): New function.
12402         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
12403         restrictions for pattern statements.  Use first_commutative_argument
12404         to look for commutative operands in calls to internal functions.
12406 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
12408         * Makefile.in (wide-int-range.o): New.
12409         * tree-vrp.c: Move all the wide_int_* functions to...
12410         * wide-int-range.cc: ...here.
12411         * tree-vrp.h: Move all the wide_int_* prototypes to...
12412         * wide-int-range.h: ...here.
12414 2018-08-03  Tom de Vries  <tdevries@suse.de>
12416         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
12417         UI_NONE.
12418         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
12419         * except.c (output_function_exception_table): Do early exit if
12420         targetm_common.except_unwind_info (&global_options) == UI_NONE.
12422 2018-08-03  Martin Liska  <mliska@suse.cz>
12424         * predict.c (dump_prediction): Change to 2 digits
12425         in fraction part.
12427 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
12429         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
12430         neon_dup_q to...
12431         (falkor_am_1_gtov_gtov): ... a new insn reservation.
12433 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
12435         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
12436         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
12437         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
12438         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
12439         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
12441 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
12443         * diagnostic-show-locus.c (diagnostic_show_locus): Use
12444         pp_take_prefix when saving the existing prefix.
12445         * diagnostic.c (diagnostic_append_note): Likewise.
12446         * langhooks.c (lhd_print_error_function): Likewise.
12447         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
12448         param's type.  Free the existing prefix.
12449         (pp_take_prefix): New function.
12450         (pretty_printer::pretty_printer): Drop the prefix parameter.
12451         Rename the length parameter to match the comment.
12452         (pretty_printer::~pretty_printer): Free the prefix.
12453         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
12454         parameter.
12455         (struct pretty_printer): Drop the "const" from "prefix" field's
12456         type and clarify memory management.
12457         (pp_set_prefix): Drop the "const" from the 2nd param.
12458         (pp_take_prefix): New decl.
12460 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
12462         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
12463         for word_mode_ok here instead of passing as argument.
12464         (expand_block_compare): Change select_block_compare_mode() call.
12465         (expand_strncmp_gpr_sequence): New function.
12466         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
12468 2018-08-02  Jeff Law  <law@redhat.com>
12470         PR target/86790
12471         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12472         Define to speculation_safe_value_not_needed.
12474         PR target/86784
12475         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12476         Define to speculation_safe_value_not_needed.
12478 2018-08-02  Tom de Vries  <tdevries@suse.de>
12480         PR target/86660
12481         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
12482         function.  Return UI_TARGET unconditionally.
12483         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
12484         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
12486 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
12488         * genemit.c (print_overload_test): Fix typo.
12490 2018-08-02  Richard Biener  <rguenther@suse.de>
12492         PR tree-optimization/86816
12493         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
12494         which checks for value availability before querying it.
12495         (gvn_uses_equal): Use it.
12496         (same_succ_hash): Likewise.
12497         (gimple_equal_p): Likewise.
12499 2018-08-02  Nick Clifton  <nickc@redhat.com>
12501         PR target/86813
12502         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12503         Define to speculation_safe_value_not_needed.
12505         PR target/86810
12506         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12507         Define to speculation_safe_value_not_needed.
12509         PR target/86810
12510         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12511         Define to speculation_safe_value_not_needed.
12513         PR target/86803
12514         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12515         Define to speculation_safe_value_not_needed.
12517         PR target/86797
12518         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12519         Define to speculation_safe_value_not_needed.
12521         PR target/86791
12522         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12523         Define to speculation_safe_value_not_needed.
12525         PR target/86789
12526         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12527         Define to speculation_safe_value_not_needed.
12529         PR target/86787
12530         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12531         Define to speculation_safe_value_not_needed.
12533         PR target/86782
12534         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
12535         speculation_safe_value_not_needed.
12537         PR target/86781
12538         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
12539         to speculation_safe_value_not_needed.
12541 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
12543         * doc/md.texi: Expand the documentation of instruction names
12544         to mention port-local uses.  Document '@' in pattern names.
12545         * read-md.h (overloaded_instance, overloaded_name): New structs.
12546         (mapping): Declare.
12547         (md_reader::handle_overloaded_name): New member function.
12548         (md_reader::get_overloads): Likewise.
12549         (md_reader::m_first_overload): New member variable.
12550         (md_reader::m_next_overload_ptr): Likewise.
12551         (md_reader::m_overloads_htab): Likewise.
12552         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
12553         m_next_overload_ptr and m_overloads_htab.
12554         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
12555         (get_mode_token, get_code_token, get_int_token): New functions.
12556         (map_attr_string): Add an optional argument that passes back
12557         the associated iterator.
12558         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
12559         (md_reader::handle_overloaded_name, add_overload_instance): New
12560         functions.
12561         (apply_iterators): Handle '@' names.  Report an error if '@'
12562         is used without iterators.
12563         (initialize_iterators): Initialize the new iterator_group fields.
12564         * genopinit.c (handle_overloaded_code_for)
12565         (handle_overloaded_gen): New functions.
12566         (main): Use them to print declarations of maybe_code_for_* and
12567         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
12568         * genemit.c (print_overload_arguments, print_overload_test)
12569         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
12570         (main): Use it to print definitions of maybe_code_for_* and
12571         maybe_gen_* functions.
12572         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
12573         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
12574         instead of explicit mode checks.
12575         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
12576         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
12577         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
12578         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
12579         (aarch64_expand_compare_and_swap): Likewise
12580         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
12581         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
12582         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
12583         (aarch64_constant_pool_reload_icode): Delete.
12584         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
12585         instead of aarch64_constant_pool_reload_icode.  Use
12586         code_for_aarch64_reload_mov instead of explicit mode checks.
12587         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
12588         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
12589         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
12590         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
12591         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
12592         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
12593         (aarch64_atomic_load_op_code): Delete.
12594         (aarch64_emit_atomic_load_op): Likewise.
12595         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
12596         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
12597         instead of aarch64_emit_atomic_load_op.
12598         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
12599         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
12600         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
12601         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
12602         character before the pattern name.
12603         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
12604         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
12605         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
12606         (aarch64_frecps<mode>): Likewise.
12607         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
12608         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
12609         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
12610         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
12611         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
12613 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
12615         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
12616         Allow HFmode constants if TARGET_FP_F16INST.
12618 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
12620         PR target/86014
12621         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
12622         No longer check last store for clobber of address register.
12624 2018-08-02  Martin Liska  <mliska@suse.cz>
12626         PR gcov-profile/86817
12627         * gcov.c (process_all_functions): New function.
12628         (main): Call it.
12629         (process_file): Move functions processing to
12630         process_all_functions.
12632 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
12634         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
12635         "const" to the "gimple *" and "rtx_insn *" parameters.
12636         * dumpfile.h (dump_user_location_t::dump_user_location_t):
12637         Likewise.
12638         (dump_location_t::dump_location_t): Likewise.
12640 2018-08-01  Martin Sebor  <msebor@redhat.com>
12642         PR tree-optimization/86650
12643         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
12644         rather than a "gcall *".  Directly pass the data of interest
12645         to percent_K_format, rather than building a temporary CALL_EXPR
12646         to hold it.
12647         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
12648         (gimple_fold_builtin_strncat): Adjust.
12649         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
12650         gcall* argument with gimple*.
12651         * gimple-ssa-warn-restrict.c (check_call): Same.
12652         (wrestrict_dom_walker::before_dom_children): Same.
12653         (builtin_access::builtin_access): Same.
12654         (check_bounds_or_overlap): Same
12655         (maybe_diag_overlap): Same.
12656         (maybe_diag_offset_bounds): Same.
12657         * tree-diagnostic.c (default_tree_printer): Move usage of
12658         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
12659         to this callsite.
12660         * tree-pretty-print.c (percent_K_format): Add argument.
12661         * tree-pretty-print.h: Add argument.
12662         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
12663         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
12664         (maybe_diag_stxncpy_trunc): Same.
12665         (handle_builtin_stxncpy): Same.
12666         (handle_builtin_strcat): Same.
12668 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12670         * match.pd: Optimise pointer range checks.
12672 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12674         PR tree-optimization/86758
12675         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
12676         to remove pattern statements.
12678 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12680         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
12681         result of dfs_enumerate_from when constructing stmt_vec_infos,
12682         instead of additionally calling get_loop_body.
12684 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12686         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
12687         parameter.
12688         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
12689         When creating an iv, assert that the step is not known to be zero.
12690         (vect_setup_realignment): Update call accordingly.
12691         * tree-vect-stmts.c (vectorizable_store): Likewise.
12692         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
12694 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12696         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
12697         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
12698         (vectorizable_reduction): Likewise.
12699         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12700         (vect_detect_hybrid_slp_stmts): Likewise.
12701         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12703 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
12705         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
12706         (wide_int_set_zero_nonzero_bits): ...this.
12707         (zero_nonzero_bits_from_vr): Rename to...
12708         (vrp_set_zero_nonzero_bits): ...this.
12709         (extract_range_from_multiplicative_op_1): Abstract wide int
12710         code...
12711         (wide_int_range_multiplicative_op): ...here.
12712         (extract_range_from_binary_expr_1): Extract wide int binary
12713         operations into their own functions.
12714         (wide_int_range_lshift): New.
12715         (wide_int_range_can_optimize_bit_op): New.
12716         (wide_int_range_shift_undefined_p): New.
12717         (wide_int_range_bit_xor): New.
12718         (wide_int_range_bit_ior): New.
12719         (wide_int_range_bit_and): New.
12720         (wide_int_range_trunc_mod): New.
12721         (extract_range_into_wide_ints): New.
12722         (vrp_shift_undefined_p): New.
12723         (extract_range_from_multiplicative_op): New.
12724         (vrp_can_optimize_bit_op): New.
12725         * tree-vrp.h (value_range::dump): New.
12726         (wide_int_range_multiplicative_op): New.
12727         (wide_int_range_lshift):New.
12728         (wide_int_range_shift_undefined_p): New.
12729         (wide_int_range_bit_xor): New.
12730         (wide_int_range_bit_ior): New.
12731         (wide_int_range_bit_and): New.
12732         (wide_int_range_trunc_mod): New.
12733         (zero_nonzero_bits_from_bounds): Rename to...
12734         (wide_int_set_zero_nonzero_bits): ...this.
12735         (zero_nonzero_bits_from_vr): Rename to...
12736         (vrp_set_zero_nonzero_bits): ...this.
12737         (range_easy_mask_min_max): Rename to...
12738         (wide_int_range_can_optimize_bit_op): this.
12739         * vr-values.c (simplify_bit_ops_using_ranges): Rename
12740         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
12742 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12744         * tree-vectorizer.h (vect_orig_stmt): New function.
12745         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
12746         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
12747         (vect_create_epilog_for_reduction): Likewise.
12748         (vectorizable_live_operation): Likewise.
12749         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
12750         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
12751         * tree-vect-stmts.c (vectorizable_call): Likewise.
12752         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
12754 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12756         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
12757         argument.
12758         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
12759         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
12760         (vect_transform_loop): Likewise.
12761         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
12763 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12765         * tree-vectorizer.h (vect_schedule_slp): Return void.
12766         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
12767         (vect_schedule_slp): Likewise.
12769 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12771         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
12772         argument.
12773         (vect_transform_loop): Update calls accordingly.  Schedule SLP
12774         instances before the main loop, if any exist.
12776 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12778         PR tree-optimization/86749
12779         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
12780         If the lhs is used in a COND_EXPR, check that it is being used
12781         as the "then" or "else" value.
12783 2018-08-01  Tom de Vries  <tdevries@suse.de>
12785         PR target/86800
12786         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
12787         speculation_safe_value_not_needed.
12789 2018-08-01  Richard Biener  <rguenther@suse.de>
12791         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
12792         as base and offset.
12794 2018-08-01  Martin Liska  <mliska@suse.cz>
12796         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
12797         format how successful transformation is dumped.
12798         (gimple_mod_pow2_value_transform): Likewise.
12799         (gimple_mod_subtract_transform): Likewise.
12800         (gimple_stringops_transform): Likewise.
12802 2018-08-01  Martin Liska  <mliska@suse.cz>
12804         PR value-prof/35543
12805         * value-prof.c (interesting_stringop_to_profile_p):
12806         Simplify the code and add BUILT_IN_MEMMOVE.
12807         (gimple_stringops_transform): Likewise.
12809 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
12811         * config/aarch64/aarch64-simd.md
12812         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
12813         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
12814         use GPI iterator instead of SI mode.
12816 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
12818         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
12819         rs6000_speculation_barrier.
12820         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
12821         new barrier pattern name.
12823 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
12825         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
12826         (speculation_barrier): New insn.
12828 2018-08-01  Richard Biener  <rguenther@suse.de>
12830         PR bootstrap/86724
12831         * graphite.h: Include isl/id.h and isl/space.h to allow build
12832         with ISL 0.20.
12834 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
12836         PR target/86651
12837         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12838         mode for COFF targets.
12839         * defaults.h (TARGET_COFF): Define.
12840         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
12841         TARGET_COFF): Define.
12842         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
12843         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
12844         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
12846 2018-07-31  Alexandre Oliva <oliva@adacore.com>
12848         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
12849         * gimple-streamer-out.c (output_bb): Save it.
12850         * lto-streamer-in.c (input_struct_function_base): Restore
12851         instance discriminator if available.  Create map on demand.
12852         * lto-streamer-out.c (output_struct_function_base): Save it if
12853         available.
12854         * final.c (decl_to_instance_map): Document LTO strategy.
12856 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
12857             Olivier Hainque  <hainque@adacore.com>
12859         * debug.h (decl_to_instance_map_t): New type.
12860         (decl_to_instance_map): Declare.
12861         (maybe_create_decl_to_instance_map): New inline function.
12862         * final.c (bb_discriminator, last_bb_discriminator): New statics,
12863         to track basic block discriminators.
12864         (final_start_function_1): Initialize them.
12865         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
12866         bb_discriminator.
12867         (decl_to_instance_map): New variable.
12868         (map_decl_to_instance, maybe_set_discriminator): New functions.
12869         (notice_source_line): Set discriminator.
12871 2018-07-31  Ian Lance Taylor  <iant@golang.org>
12873         * targhooks.c (default_have_speculation_safe_value): Add
12874         ATTRIBUTE_UNUSED.
12876 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
12878         * dump-context.h: Include "pretty-print.h".
12879         (dump_context::refresh_dumps_are_enabled): New decl.
12880         (dump_context::emit_item): New decl.
12881         (class dump_context): Add fields "m_test_pp" and
12882         "m_test_pp_flags".
12883         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
12884         (temp_dump_context::get_dumped_text): New decl.
12885         (class temp_dump_context): Add field "m_pp".
12886         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
12887         (dump_context::refresh_dumps_are_enabled): ...and add a test for
12888         m_test_pp.
12889         (set_dump_file): Update for above change.
12890         (set_alt_dump_file): Likewise.
12891         (dump_loc): New overload, taking a pretty_printer *.
12892         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
12893         to any test pretty-printer.
12894         (make_item_for_dump_gimple_stmt): New function, adapted from
12895         optinfo::add_gimple_stmt.
12896         (dump_context::dump_gimple_stmt): Call it, and use the result,
12897         eliminating the direct usage of dump_file and alt_dump_file in
12898         favor of indirectly using them via emit_item.
12899         (make_item_for_dump_gimple_expr): New function, adapted from
12900         optinfo::add_gimple_expr.
12901         (dump_context::dump_gimple_expr): Call it, and use the result,
12902         eliminating the direct usage of dump_file and alt_dump_file in
12903         favor of indirectly using them via emit_item.
12904         (make_item_for_dump_generic_expr): New function, adapted from
12905         optinfo::add_tree.
12906         (dump_context::dump_generic_expr): Call it, and use the result,
12907         eliminating the direct usage of dump_file and alt_dump_file in
12908         favor of indirectly using them via emit_item.
12909         (make_item_for_dump_printf_va): New function, adapted from
12910         optinfo::add_printf_va.
12911         (make_item_for_dump_printf): New function.
12912         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
12913         and use the result, eliminating the direct usage of dump_file and
12914         alt_dump_file in favor of indirectly using them via emit_item.
12915         (make_item_for_dump_dec): New function.
12916         (dump_context::dump_dec): Call it, and use the result,
12917         eliminating the direct usage of dump_file and alt_dump_file in
12918         favor of indirectly using them via emit_item.
12919         (make_item_for_dump_symtab_node): New function, adapted from
12920         optinfo::add_symtab_node.
12921         (dump_context::dump_symtab_node): Call it, and use the result,
12922         eliminating the direct usage of dump_file and alt_dump_file in
12923         favor of indirectly using them via emit_item.
12924         (dump_context::begin_scope): Reimplement, avoiding direct usage
12925         of dump_file and alt_dump_file in favor of indirectly using them
12926         via emit_item.
12927         (dump_context::emit_item): New member function.
12928         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
12929         Set up test pretty-printer on the underlying context.  Call
12930         refresh_dumps_are_enabled.
12931         (temp_dump_context::~temp_dump_context): Call
12932         refresh_dumps_are_enabled.
12933         (temp_dump_context::get_dumped_text): New member function.
12934         (selftest::verify_dumped_text): New function.
12935         (ASSERT_DUMPED_TEXT_EQ): New macro.
12936         (selftest::test_capture_of_dump_calls): Run all tests twice, with
12937         and then without optinfo enabled.  Add uses of
12938         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
12939         * dumpfile.h: Update comment for the dump_* API.
12940         * optinfo-emit-json.cc
12941         (selftest::test_building_json_from_dump_calls): Update for new
12942         param for temp_dump_context ctor.
12943         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
12944         and "m_owned" field.
12945         (optinfo_item::~optinfo_item): Likewise.
12946         (optinfo::add_item): New member function.
12947         (optinfo::emit): Update comment.
12948         (optinfo::add_string): Delete.
12949         (optinfo::add_printf): Delete.
12950         (optinfo::add_printf_va): Delete.
12951         (optinfo::add_gimple_stmt): Delete.
12952         (optinfo::add_gimple_expr): Delete.
12953         (optinfo::add_tree): Delete.
12954         (optinfo::add_symtab_node): Delete.
12955         (optinfo::add_dec): Delete.
12956         * optinfo.h (class dump_context): New forward decl.
12957         (optinfo::add_item): New decl.
12958         (optinfo::add_string): Delete.
12959         (optinfo::add_printf): Delete.
12960         (optinfo::add_printf_va): Delete.
12961         (optinfo::add_gimple_stmt): Delete.
12962         (optinfo::add_gimple_expr): Delete.
12963         (optinfo::add_tree): Delete.
12964         (optinfo::add_symtab_node): Delete.
12965         (optinfo::add_dec): Delete.
12966         (optinfo::add_poly_int): Delete.
12967         (optinfo_item::optinfo_item): Remove "owned" param.
12968         (class optinfo_item): Remove field "m_owned".
12970 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12972         PR middle-end/86705
12973         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
12974         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
12975         requested variable alignment.
12976         (expand_one_ssa_partition): Likewise.
12977         (expand_one_var): Likewise.
12979 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12981         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
12982         to speculation_safe_value_not_needed.
12984 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12986         * targhooks.h (speculation_safe_value_not_needed): New prototype.
12987         * targhooks.c (speculation_safe_value_not_needed): New function.
12988         * target.def (have_speculation_safe_value): Update documentation.
12989         * doc/tm.texi: Regenerated.
12991 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12993         * config/aarch64/iterators.md (ALLI_TI): New iterator.
12994         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
12995         expand.
12996         (despeculate_copy<ALLI:mode>_insn): New insn.
12997         (despeculate_copyti_insn): New insn.
12998         (despeculate_simple<ALLI:mode>): New insn
12999         (despeculate_simpleti): New insn.
13000         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
13001         function.
13002         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
13003         aarch64_speculation_safe_value.
13004         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
13006 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13008         * config/aarch64/aarch64-speculation.cc: New file.
13009         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
13010         before pass_reorder_blocks.
13011         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
13012         prototype.
13013         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
13014         X14 and X15 when tracking speculation.
13015         * config/aarch64/aarch64.md (register name constants): Add
13016         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
13017         (unspec): Add UNSPEC_SPECULATION_TRACKER.
13018         (speculation_barrier): New insn attribute.
13019         (cmp<mode>): Allow SP in comparisons.
13020         (speculation_tracker): New insn.
13021         (speculation_barrier): Add speculation_barrier attribute.
13022         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
13023         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
13024         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
13026 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13028         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
13029         aarch64_track_speculation is true.
13030         (tb<optab><mode>1): Likewise.
13031         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
13032         generate CB[N]Z when tracking speculation.
13033         (aarch64_split_compare_and_swap): Likewise.
13034         (aarch64_split_atomic_op): Likewise.
13036 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13038         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
13040 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13042         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
13043         (speculation_barrier): New insn.
13045 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13047         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
13048         * config/arm/arm.md (speculation_barrier): New expand.
13049         (speculation_barrier_insn): New pattern.
13051 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13053         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
13054         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
13055         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
13056         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
13057         list.
13058         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
13059         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
13060         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
13061         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
13062         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
13063         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
13064         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
13065         * builtins.c (expand_speculation_safe_value): New function.
13066         (expand_builtin): Call it.
13067         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
13068         * doc/extend.texi: Document __builtin_speculation_safe_value.
13069         * doc/md.texi: Document "speculation_barrier" pattern.
13070         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
13071         TARGET_HAVE_SPECULATION_SAFE_VALUE.
13072         * doc/tm.texi: Regenerated.
13073         * target.def (have_speculation_safe_value, speculation_safe_value): New
13074         hooks.
13075         * targhooks.c (default_have_speculation_safe_value): New function.
13076         (default_speculation_safe_value): New function.
13077         * targhooks.h (default_have_speculation_safe_value): Add prototype.
13078         (default_speculation_safe_value): Add prototype.
13080 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
13082         * dump-context.h (dump_context::dump_loc): New decl.
13083         * dumpfile.c (dump_context::dump_loc): New member function.
13084         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
13085         and dump_gimple_stmt.
13086         (dump_context::dump_gimple_expr_loc): Likewise, using
13087         dump_gimple_expr.
13088         (dump_context::dump_generic_expr_loc): Likewise, using
13089         dump_generic_expr.
13090         (dump_context::dump_printf_loc_va): Likewise, using
13091         dump_printf_va.
13092         (dump_context::begin_scope): Explicitly using the global function
13093         "dump_loc", rather than the member function.
13095 2018-07-31  Martin Sebor  <msebor@redhat.com>
13097         PR tree-optimization/86741
13098         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
13100 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
13102         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
13103         depend on whether prefetch instructions will be emitted or not.
13104         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
13105         will be emitted or not.
13106         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
13107         (TARGET_SETMEM_PFD): New macros.
13109 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13111         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
13112         (NULL_STMT_VEC_INFO): Delete.
13113         (stmt_vec_info::operator*): Likewise.
13114         (stmt_vec_info::operator gimple *): Likewise.
13115         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
13116         of NULL_STMT_VEC_INFO.
13117         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13118         (vect_reassociating_reduction_p): Likewise.
13119         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
13120         (vectorizable_store): Likewise.
13121         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
13122         (vec_info::free_stmt_vec_infos): Likewise.
13124 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13126         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
13127         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
13129 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13131         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
13132         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
13133         (vec_info::free_stmt_vec_info): New private member functions.
13134         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
13135         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
13136         * tree-parloops.c (gather_scalar_reductions): Remove calls to
13137         set_stmt_vec_info_vec and free_stmt_vec_infos.
13138         * tree-vect-loop.c (_loop_vec_info): Remove call to
13139         set_stmt_vec_info_vec.
13140         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
13141         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
13142         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
13143         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
13144         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
13145         assignments in {vec_info::,}new_stmt_vec_info that are redundant
13146         with the clearing in the xcalloc.
13147         (stmt_vec_info_vec): Delete.
13148         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
13149         (vectorize_loops): Likewise.
13150         (vec_info::~vec_info): Remove argument from call to
13151         free_stmt_vec_infos.
13152         (vec_info::add_stmt): Remove vinfo argument from call to
13153         new_stmt_vec_info.
13155 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13157         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
13158         rather than a gimple stmt.
13159         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
13160         information for pattern statements when passed the original
13161         statement; instead wait to be passed the pattern statement itself.
13162         Don't call set_vinfo_for_stmt here.
13163         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
13164         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
13165         stmt_vec_infos here.
13166         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
13167         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
13168         stmt_vec_infos entry.
13170 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13172         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
13173         * tree-vectorizer.c (vec_info::replace_stmt): New function.
13174         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
13175         * tree-vect-stmts.c (vectorizable_call): Likewise.
13176         (vectorizable_simd_clone_call): Likewise.
13178 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13180         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
13181         * tree-vectorizer.c (vec_info::remove_stmt): New function.
13182         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
13183         * tree-vect-loop.c (vect_transform_loop): Likewise.
13184         * tree-vect-slp.c (vect_schedule_slp): Likewise.
13185         * tree-vect-stmts.c (vect_remove_stores): Likewise.
13187 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13189         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
13190         (vect_dr_stmt): Delete.
13191         * tree-vectorizer.c (vec_info::lookup_dr): New function.
13192         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
13193         of DR_VECT_AUX.
13194         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13195         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
13196         (vect_verify_datarefs_alignment, vect_peeling_supportable)
13197         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
13198         (vect_analyze_data_refs): Likewise.
13199         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
13200         argument.
13201         (vect_find_same_alignment_drs): Likewise.
13202         (vect_slp_analyze_node_dependences): Update calls accordingly.
13203         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
13204         instead of DR_VECT_AUX.
13205         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
13206         of a vector data references.  Use vec_info::lookup_dr instead of
13207         DR_VECT_AUX.
13208         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
13209         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
13210         instead of DR_VECT_AUX.
13212 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13214         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
13215         dr_vec_info.
13216         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
13217         accordingly.
13218         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13219         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13220         (vect_gen_prolog_loop_niters): Likewise.
13222 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13224         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
13225         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
13226         (vect_known_alignment_in_bytes, vect_dr_behavior)
13227         (vect_get_scalar_dr_size): Take references as dr_vec_infos
13228         instead of data_references.  Update calls to other routines for
13229         which the same change has been made.
13230         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
13231         dr_vec_infos instead of stmt_vec_infos.
13232         (vect_analyze_data_ref_dependence): Update call accordingly.
13233         (vect_slp_analyze_data_ref_dependence)
13234         (vect_record_base_alignments): Use DR_VECT_AUX.
13235         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
13236         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
13237         (vector_alignment_reachable_p, vect_get_data_access_cost)
13238         (vect_peeling_supportable, vect_analyze_group_access_1)
13239         (vect_analyze_group_access, vect_analyze_data_ref_access)
13240         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
13241         (vect_compile_time_alias, vect_small_gap_p)
13242         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
13243         (vect_supportable_dr_alignment): Take references as dr_vec_infos
13244         instead of data_references.  Update calls to other routines for
13245         which the same change has been made.
13246         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
13247         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
13248         (vect_slp_analyze_and_verify_node_alignment)
13249         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
13250         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
13251         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
13252         above changes.
13253         (_vect_peel_info::dr): Replace with...
13254         (_vect_peel_info::dr_info): ...this new field.
13255         (vect_peeling_hash_get_most_frequent)
13256         (vect_peeling_hash_choose_best_peeling): Update accordingly.
13257         (vect_peeling_hash_get_lowest_cost):
13258         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
13259         routines for which the same change has been made.
13260         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
13261         data_reference.
13262         * tree-vect-loop-manip.c (get_misalign_in_elems)
13263         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
13264         above changes.
13265         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13266         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
13267         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
13268         (get_group_load_store_type, get_negative_load_store_type)
13269         (vect_get_data_ptr_increment, vectorizable_store)
13270         (vectorizable_load): Likewise.
13271         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
13272         Update calls to other routines for which the same change has been made.
13274 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13276         * tree-vectorizer.h (vec_info::move_dr): New member function.
13277         (dataref_aux): Rename to...
13278         (dr_vec_info): ...this and add "dr" and "stmt" fields.
13279         (_stmt_vec_info::dr_aux): Update accordingly.
13280         (_stmt_vec_info::data_ref_info): Delete.
13281         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
13282         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
13283         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
13284         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
13285         of data_ref.
13286         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
13287         (STMT_VINFO_DR_INFO): New macro.
13288         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
13289         (set_dr_misalignment): Update after rename of dataref_aux.
13290         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
13291         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
13292         initialization of STMT_VINFO_DATA_REF.
13293         * tree-vectorizer.c (vec_info::move_dr): New function.
13294         * tree-vect-patterns.c (vect_recog_bool_pattern)
13295         (vect_recog_mask_conversion_pattern)
13296         (vect_recog_gather_scatter_pattern): Use it.
13297         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
13298         the "dr" and "stmt" fields of dr_vec_info instead of
13299         STMT_VINFO_DATA_REF.
13301 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13303         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
13304         (is_pattern_stmt_p): Use it.
13305         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
13306         on pattern statements.
13308 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13310         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
13311         original stmt as a stmt_vec_info rather than a gimple stmt.
13312         (vect_pattern_recog_1): Take the statement directly as a
13313         stmt_vec_info, rather than via a gimple_stmt_iterator.
13314         Update call to vect_mark_pattern_stmts.
13315         (vect_pattern_recog): Update calls accordingly.
13317 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13319         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
13320         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
13321         a vect_def_type for the first argument.
13322         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
13323         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
13324         operand if it isn't defined by a vectorized statement.
13325         (vect_build_gather_load_calls): Remove the mask_dt argument and
13326         update calls to vect_get_vec_def_for_stmt_copy.
13327         (vectorizable_bswap): Likewise the dt argument.
13328         (vectorizable_call): Update calls to vectorizable_bswap and
13329         vect_get_vec_def_for_stmt_copy.
13330         (vectorizable_simd_clone_call, vectorizable_assignment)
13331         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
13332         (vectorizable_comparison): Update calls to
13333         vect_get_vec_def_for_stmt_copy.
13334         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
13335         vect_is_simple_use.
13336         (vect_get_loop_based_defs): Remove dt argument and update call
13337         to vect_get_vec_def_for_stmt_copy.
13338         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
13339         and vect_get_vec_def_for_stmt_copy.
13340         (vectorizable_load): Update calls to vect_build_gather_load_calls
13341         and vect_get_vec_def_for_stmt_copy.
13342         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13343         (vectorizable_reduction, vectorizable_live_operation): Update calls
13344         to vect_get_vec_def_for_stmt_copy.
13346 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13348         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
13349         and gimple stmt arguments with a stmt_vec_info.
13350         (vect_record_base_alignments): Update calls accordingly.
13351         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
13352         and gimple stmt arguments with a stmt_vec_info.
13353         (vect_build_slp_tree_1): Remove vinfo argument and update call
13354         to vect_record_max_nunits.
13355         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
13356         and vect_record_max_nunits.
13358 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13360         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
13361         file and take a stmt_vec_info instead of a gimple stmt.
13362         (supportable_widening_operation, vect_finish_replace_stmt)
13363         (vect_finish_stmt_generation, vect_get_store_rhs)
13364         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
13365         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
13366         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
13367         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
13368         (vect_create_data_ref_ptr, bump_vector_ptr)
13369         (vect_permute_store_chain, vect_setup_realignment)
13370         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
13371         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
13372         (vectorizable_reduction, vectorizable_induction)
13373         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
13374         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
13375         than gimple stmts as arguments.
13376         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
13377         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
13378         (can_group_stmts_p, vect_check_gather_scatter)
13379         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
13380         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
13381         (vect_permute_load_chain, vect_shift_permute_load_chain)
13382         (vect_transform_grouped_load)
13383         (vect_record_grouped_load_vectors): Likewise.
13384         * tree-vect-loop.c (vect_fixup_reduc_chain)
13385         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
13386         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
13387         (vectorizable_reduction, vectorizable_induction)
13388         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
13389         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
13390         (vect_get_load_store_mask): Likewise.
13391         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
13392         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
13393         * tree-vect-stmts.c (vect_mark_relevant)
13394         (is_simple_and_all_uses_invariant)
13395         (exist_non_indexing_operands_for_use_p, process_use)
13396         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
13397         (vect_get_vec_def_for_operand, vect_get_vec_defs)
13398         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
13399         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
13400         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
13401         (get_negative_load_store_type, get_load_store_type)
13402         (vect_check_load_store_mask, vect_check_store_rhs)
13403         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
13404         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
13405         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
13406         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13407         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
13408         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
13409         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
13410         (supportable_widening_operation): Likewise.
13412 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13414         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
13415         a stmt_vec_info instead of a gcall.
13416         (vect_check_gather_scatter): Update call accordingly.
13417         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
13418         of a gphi.
13419         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
13420         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
13421         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
13422         instead of a gimple stmt.
13423         (vect_transform_loop): Update calls accordingly.
13424         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
13425         stmt_vec_infos instead of gimple stmts.
13426         (vect_analyze_slp_instance): Update use accordingly.
13427         * tree-vect-stmts.c (read_vector_array, write_vector_array)
13428         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
13429         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
13430         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
13431         (vect_gen_widened_results_half, vect_get_loop_based_defs)
13432         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
13433         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
13434         down to subroutines.
13436 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13438         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
13439         of the worklist from a vector of gimple stmts to a vector of
13440         stmt_vec_infos.
13441         * tree-vect-stmts.c (vect_mark_relevant, process_use)
13442         (vect_mark_stmts_to_be_vectorized): Likewise
13444 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13446         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
13447         statement before passing it to vect_analyze_stmt.
13448         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
13449         the chain of phi vector definitions.  Track the exit phi via its
13450         stmt_vec_info.
13451         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
13452         STMT_VINFO_REDUC_DEF.
13453         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
13454         stmt_vec_infos to handle the statement chains.
13455         (vect_get_slp_defs): Record the first statement in the node
13456         using a stmt_vec_info.
13457         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
13458         statements here and pass their stmt_vec_info down to subroutines.
13459         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
13460         down to vect_finish_stmt_generation.
13461         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
13462         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
13463         stmt_vec_infos to subroutines.
13464         (vect_remove_stores): Use stmt_vec_infos to handle the statement
13465         chains.
13467 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13469         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
13470         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
13471         (vect_permute_store_chain, vect_setup_realignment)
13472         (vect_permute_load_chain, vect_shift_permute_load_chain)
13473         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
13474         stmts internally, and when passing values to other vectorizer routines.
13475         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
13476         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
13477         (vect_analyze_loop_operations, get_initial_def_for_reduction)
13478         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
13479         (vectorizable_reduction, vectorizable_induction)
13480         (vectorizable_live_operation, vect_transform_loop_stmt)
13481         (vect_transform_loop): Likewise.
13482         * tree-vect-patterns.c (vect_reassociating_reduction_p)
13483         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
13484         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
13485         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13486         (vect_slp_analyze_node_operations_1): Likewise.
13487         * tree-vect-stmts.c (vect_mark_relevant, process_use)
13488         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
13489         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
13490         (vect_finish_stmt_generation_1, get_group_load_store_type)
13491         (get_load_store_type, vect_build_gather_load_calls)
13492         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
13493         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
13494         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13495         (vectorizable_store, vectorizable_load, vectorizable_condition)
13496         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
13497         (supportable_widening_operation): Likewise.
13498         (vect_get_vector_types_for_stmt): Likewise.
13499         * tree-vectorizer.h (vect_dr_behavior): Likewise.
13501 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13503         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13504         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
13505         (vect_permute_store_chain, vect_permute_load_chain)
13506         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
13507         repeated stmt_vec_info lookups.
13508         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
13509         (vect_update_ivs_after_vectorizer): Likewise.
13510         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
13511         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
13512         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
13513         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13514         (vect_bb_slp_scalar_cost): Likewise.
13515         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
13517 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13519         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
13520         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
13521         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
13522         to get gassigns and gcalls, rather than operating on generc gimple
13523         stmts.
13524         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
13525         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
13526         (vectorizable_load, vect_analyze_stmt): Likewise.
13527         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
13529 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13531         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
13532         return stmt_vec_infos rather than gimple stmts.  Do not accept
13533         null arguments.
13534         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
13535         of a gimple stmt.
13536         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
13537         Update use of get_later_stmt.
13538         (vect_get_constant_vectors): Update call accordingly.
13539         (vect_schedule_slp_instance): Likewise
13540         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
13541         (vect_slp_analyze_instance_dependence): Likewise.
13542         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
13544 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13546         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
13547         (stmt_info_for_cost::stmt_info): ...this new field.
13548         (add_stmt_costs): Update accordingly.
13549         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
13550         (vect_get_known_peeling_cost): Likewise.
13551         (vect_estimate_min_profitable_iters): Likewise.
13552         * tree-vect-stmts.c (record_stmt_cost): Likewise.
13554 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13556         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
13557         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
13558         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
13559         accordingly.
13560         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
13562 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13564         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
13565         a gimple stmt to a stmt_vec_info.
13566         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
13568 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13570         * tree-vectorizer.h (vec_info::grouped_stores): Change from
13571         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
13572         (_loop_vec_info::reduction_chains): Likewise.
13573         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
13574         accordingly.
13575         * tree-vect-slp.c (vect_analyze_slp): Likewise.
13577 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13579         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
13580         a gimple stmt to a stmt_vec_info.
13581         (_stmt_vec_info::next_element): Likewise.
13582         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
13583         (vect_slp_analyze_and_verify_node_alignment)
13584         (vect_analyze_group_access_1, vect_analyze_group_access)
13585         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
13586         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
13587         (vect_supportable_dr_alignment): Update accordingly.
13588         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
13589         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
13590         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
13591         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13592         * tree-vect-slp.c (vect_build_slp_tree_1)
13593         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
13594         (vect_split_slp_store_group, vect_analyze_slp_instance)
13595         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
13596         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
13597         (get_group_load_store_type, get_load_store_type)
13598         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
13599         (vect_transform_stmt, vect_remove_stores): Likewise.
13601 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13603         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
13604         than a gimple stmt.
13605         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13606         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
13607         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
13608         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
13609         (vector_alignment_reachable_p, vect_get_data_access_cost)
13610         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
13611         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
13612         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
13613         (vect_analyze_group_access_1, vect_analyze_group_access)
13614         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
13615         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
13616         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
13617         result of vect_dr_stmt and use the stmt_vec_info instead of
13618         the associated gimple stmt.
13619         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13620         (vect_gen_prolog_loop_niters): Likewise.
13621         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13623 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13625         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
13626         to a vec<stmt_vec_info>.
13627         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
13628         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
13629         vec<stmt_vec_info>.
13630         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
13631         to a vec<stmt_vec_info>.
13632         (bst_traits::value_type, bst_traits::value_type): Likewise.
13633         (bst_traits::hash): Update accordingly.
13634         (vect_get_and_check_slp_defs): Change the stmts parameter from
13635         a vec<gimple *> to a vec<stmt_vec_info>.
13636         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
13637         (vect_build_slp_tree): Likewise.
13638         (vect_build_slp_tree_2): Likewise.  Update uses of
13639         SLP_TREE_SCALAR_STMTS.
13640         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
13641         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
13642         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
13643         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
13644         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
13645         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
13646         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
13647         (vect_get_constant_vectors, vect_get_slp_defs)
13648         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
13649         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
13650         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
13651         instead of gimple stmts.
13652         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
13653         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
13654         (vect_slp_analyze_instance_dependence): Update uses of
13655         SLP_TREE_SCALAR_STMTS.
13656         (vect_slp_analyze_and_verify_node_alignment): Likewise.
13657         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
13658         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
13659         (get_initial_defs_for_reduction): Likewise.
13660         (vect_create_epilog_for_reduction): Likewise.
13661         (vectorize_fold_left_reduction): Likewise.
13662         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
13663         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
13664         (can_vectorize_live_stmts): Likewise.
13666 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13668         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
13669         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
13670         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
13671         than gimple stmts.
13672         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
13673         of a gimple stmt.
13674         (gather_scalar_reductions): Update after above interface changes.
13675         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
13676         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
13677         than gimple stmts.
13678         (vect_force_simple_reduction): Likewise.
13679         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
13680         LOOP_VINFO_REDUCTIONS.
13681         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13683 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13685         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
13686         a gimple stmt to a stmt_vec_info.
13687         * tree-vect-loop.c (vect_active_double_reduction_p)
13688         (vect_force_simple_reduction, vectorizable_reduction): Update
13689         accordingly.
13691 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13693         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
13694         vec<gimple *> to a vec<stmt_vec_info>.
13695         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
13696         the reduction_phis argument from a vec<gimple *> to a
13697         vec<stmt_vec_info>.
13698         (vectorizable_reduction): Likewise the phis local variable that
13699         is passed to vect_create_epilog_for_reduction.  Update for new type
13700         of SLP_TREE_VEC_STMTS.
13701         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
13702         (vectorizable_live_operation): Likewise.
13703         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
13704         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
13706 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13708         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
13709         a gimple stmt to a stmt_vec_info.
13710         (vectorizable_condition, vectorizable_live_operation)
13711         (vectorizable_reduction, vectorizable_induction): Pass back the
13712         vectorized statement as a stmt_vec_info.
13713         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
13714         use of STMT_VINFO_VEC_STMT.
13715         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
13716         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
13717         as stmt_vec_infos rather than gimple stmts.
13718         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
13719         to a stmt_vec_info.
13720         (vectorizable_live_operation): Likewise.
13721         (vectorizable_reduction, vectorizable_induction): Likewise,
13722         updating use of STMT_VINFO_VEC_STMT.
13723         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
13724         of STMT_VINFO_VEC_STMT.
13725         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
13726         (vectorizable_simd_clone_call, vectorizable_conversion)
13727         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13728         (vectorizable_store, vectorizable_load, vectorizable_condition)
13729         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
13730         from a gimple stmt to a stmt_vec_info.
13731         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
13732         pointer to a stmt_vec_info to the vectorizable_* routines.
13734 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13736         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
13737         a gimple stmt to a stmt_vec_info.
13738         (is_pattern_stmt_p): Update accordingly.
13739         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
13740         (vect_record_grouped_load_vectors): Likewise.
13741         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
13742         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
13743         (vect_model_reduction_cost): Likewise.
13744         (vect_create_epilog_for_reduction): Likewise.
13745         (vectorizable_reduction, vectorizable_induction): Likewise.
13746         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13747         Return the stmt_vec_info for the pattern statement.
13748         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
13749         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
13750         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
13751         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
13752         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
13753         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
13754         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
13755         (free_stmt_vec_info, vect_is_simple_use): Likewise.
13757 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13759         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
13760         (vect_finish_stmt_generation): Likewise.
13761         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
13762         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
13763         (vect_build_gather_load_calls): Use the return value of the above
13764         functions instead of a separate call to vinfo_for_stmt.  Use narrow
13765         scopes for the input gimple stmt and wider scopes for the associated
13766         stmt_vec_info.  Use vec_info::lookup_def when setting these
13767         stmt_vec_infos from an SSA_NAME definition.
13768         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
13769         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
13770         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13771         (vectorizable_store, vectorizable_load, vectorizable_condition)
13772         (vectorizable_comparison): Likewise.
13773         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
13774         (vectorizable_reduction): Likewise.
13776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13778         * tree-vectorizer.h (vect_is_simple_use): Add an optional
13779         stmt_vec_info * parameter before the optional gimple **.
13780         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13781         (process_use, vect_get_vec_def_for_operand_1): Update callers.
13782         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
13783         * tree-vect-loop.c (vectorizable_reduction): Likewise.
13784         (vectorizable_live_operation): Likewise.
13785         * tree-vect-patterns.c (type_conversion_p): Likewise.
13786         (vect_look_through_possible_promotion): Likewise.
13787         (vect_recog_rotate_pattern): Likewise.
13788         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13790 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13792         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
13793         a typedef to a wrapper class.
13794         (NULL_STMT_VEC_INFO): New macro.
13795         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
13796         (stmt_vec_info::operator*): New function.
13797         (stmt_vec_info::operator gimple *): Likewise.
13798         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
13799         (add_stmt_costs): Likewise.
13800         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
13801         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
13802         (vect_get_known_peeling_cost): Likewise.
13803         (vect_estimate_min_profitable_iters): Likewise.
13804         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13805         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
13806         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
13807         (vectorizable_store, free_stmt_vec_infos): Likewise.
13808         (new_stmt_vec_info): Change return type of xcalloc to
13809         _stmt_vec_info *.
13811 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13813         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
13814         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
13815         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
13816         a single_imm_use-based sequence.
13817         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
13819 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13821         * tree-vectorizer.h (vec_info::lookup_def): Declare.
13822         * tree-vectorizer.c (vec_info::lookup_def): New function.
13823         * tree-vect-patterns.c (vect_get_internal_def): Use it.
13824         (vect_widened_op_tree): Likewise.
13825         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13826         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
13827         (vectorizable_reduction): Likewise.
13828         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
13829         of a gimple *.
13830         (vect_is_slp_reduction): Update calls accordingly.  Use
13831         vec_info::lookup_def.
13832         (vect_is_simple_reduction): Likewise
13833         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
13835 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13837         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
13838         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
13839         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
13840         of vinfo_for_stmt.
13841         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
13842         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
13843         (vect_update_vf_for_slp, vect_analyze_loop_operations)
13844         (vect_is_slp_reduction, vectorizable_induction)
13845         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
13846         * tree-vect-patterns.c (vect_init_pattern_stmt):
13847         (vect_determine_min_output_precision_1, vect_determine_precisions)
13848         (vect_pattern_recog): Likewise.
13849         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
13850         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
13851         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
13852         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
13853         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
13854         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
13855         info field from a loop to a loop_vec_info.
13857 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13859         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
13860         (vec_info::add_stmt): Declare.
13861         * tree-vectorizer.c (vec_info::add_stmt): New function.
13862         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
13863         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
13864         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
13865         (vectorizable_induction): Likewise.
13866         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
13867         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
13868         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
13869         (vectorizable_load): Likewise.
13870         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13871         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
13872         (vect_recog_gather_scatter_pattern): Likewise.
13873         (append_pattern_def_seq): Likewise.  Remove a check that is
13874         performed by add_stmt itself.
13876 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13878         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
13879         which make_ssa_name was called with new_stmt before new_stmt
13880         had been created.
13882 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13884         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
13885         split out from...
13886         (vect_is_slp_reduction): ...here...
13887         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
13888         that are already known to be false.
13890 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13892         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
13893         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
13894         STMT_VINFO_NUM_SLP_USES when it's true.
13895         (vect_free_slp_instance): Add a final_p parameter and pass it to
13896         vect_free_slp_tree.
13897         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
13898         (vect_analyze_slp_instance): Likewise.
13899         (vect_slp_analyze_operations): Likewise.
13900         (vect_slp_analyze_bb_1): Likewise.
13901         * tree-vectorizer.c (vec_info): Likewise.
13902         * tree-vect-loop.c (vect_transform_loop): Likewise.
13904 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13906         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13907         function is not called for second and subsequent members of
13908         a reduction group.
13910 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13912         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
13913         cases for nested loops from here to ...
13914         (vect_create_epilog_for_reduction): ...here.  Only call
13915         vect_is_simple_use for inner-loop reductions.
13917 2018-07-31  Martin Liska  <mliska@suse.cz>
13919         PR gcov-profile/85338
13920         PR gcov-profile/85350
13921         PR gcov-profile/85372
13922         * profile.c (struct location_triplet): New.
13923         (struct location_triplet_hash): Likewise.
13924         (output_location): Do not output a BB that
13925         is already recorded for a line.
13926         (branch_prob): Use streamed_locations.
13928 2018-07-31  Martin Liska  <mliska@suse.cz>
13930         PR gcov-profile/85370
13931         * coverage.c (coverage_begin_function): Do not mark target
13932         clones as artificial functions.
13934 2018-07-31  Martin Liska  <mliska@suse.cz>
13936         PR gcov-profile/83813
13937         PR gcov-profile/84758
13938         PR gcov-profile/85217
13939         PR gcov-profile/85332
13940         * profile.c (branch_prob): Do not record GOTO expressions
13941         for GIMPLE statements which locations are already streamed.
13943 2018-07-31  Olivier Hainque  <hainque@adacore.com>
13945         * gcc.c (handle_spec_function): Accept a soft_matched_part
13946         argument, as do_spec_1.  Pass it down to ...
13947         (eval_spec_function): Accept a soft_matched_part argument,
13948         and pass it down to ...
13949         (do_spec_2): Accept a soft_matched_part argument, and pass
13950         it down to do_spec_1.
13951         (do_spec_1): Pass soft_matched_part to handle_spec_function.
13952         (handle_braces): Update call to handle_spec_function.
13953         (driver::set_up_specs): Update calls to do_spec_2.
13954         (compare_debug_dump_opt_spec_function): Likewise.
13955         (compare_debug_self_opt_spec_function): Likewise.
13957 2018-07-31  Olivier Hainque  <hainque@adacore.com>
13959         * common.opt (nolibc): New option.
13960         * doc/invoke.texi (Link Options): Document it.
13961         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
13962         * config/alpha/linux.h: Likewise.
13963         * config/arc/elf.h: Likewise.
13964         * config/arm/uclinux-elf.h: Likewise.
13965         * config/arm/unknown-elf.h: Likewise.
13966         * config/avr/avrlibc.h: Likewise.
13967         * config/bfin/bfin.h: Likewise.
13968         * config/bfin/linux.h: Likewise.
13969         * config/bfin/uclinux.h: Likewise.
13970         * config/darwin.h: Likewise.
13971         * config/darwin10.h: Likewise.
13972         * config/darwin12.h: Likewise.
13973         * config/gnu-user.h: Likewise.
13974         * config/lm32/uclinux-elf.h: Likewise.
13975         * config/pa/pa-hpux11.h: Likewise.
13976         * config/pa/pa64-hpux.h: Likewise.
13977         * config/sparc/sparc.h: Likewise.
13979 2018-07-31  Olivier Hainque  <hainque@adacore.com>
13981         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
13982         undefined variables.
13984 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
13986         PR target/86640
13987         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
13988         instead of GEN_INT.
13990 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13992         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
13993         terminated string literal.
13995 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
13997         PR rtl-optimization/85160
13998         * combine.c (is_just_move): New function.
13999         (try_combine): Allow combining two instructions into two if neither of
14000         the original instructions was a move.
14002 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
14004         PR target/86673
14005         * doc/extend.texi (Global Register Variables): Discourage use of type
14006         qualifiers.
14007         (Local Register Variables): Likewise.
14009 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
14011         PR tree-optimization/86506
14012         * hwint.h (ceil_log2): Resync with hwint.c implementation.
14014 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
14016         PR target/86547
14017         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
14018         hard_regno, make sure no insn between `from` and `to` clobbers it.
14020 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
14021             Tom de Vries  <tdevries@suse.de>
14023         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
14024         (PTX_DEFAULT_RUNTIME_DIM): ... this.
14025         (nvptx_goacc_validate_dims): Set default worker and gang dims to
14026         PTX_DEFAULT_RUNTIME_DIM.
14027         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
14029 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
14031         * config/pa/pa.c (pa_output_addr_vec): Align address table.
14032         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
14033         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
14035 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
14037         * config/rs6000/constraints.md (wG constraint): Delete, no longer
14038         used.
14039         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
14040         predicate to reflect toc fusion has been deleted.
14041         (toc_fusion_mem_raw): Delete, no longer used.
14042         (toc_fusion_mem_wrapped): Likewise.
14043         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
14044         fusion mask bit.
14045         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
14046         Delete, no longer used.
14047         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
14048         meant to be used for toc fusion.
14049         (rs6000_debug_print_mode): Delete toc fusion debugging.
14050         (rs6000_debug_reg_global): Likewise.
14051         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
14052         fusion and secondary reload support that were never used.
14053         (rs6000_option_override_internal): Delete TOC fusion, that was only
14054         partially defined, and it did not work unless you also used the
14055         -mcmodel= switch.
14056         (rs6000_legitimate_address_p): Delete TOC fusion support.
14057         (rs6000_opt_masks): Likewise.
14058         (fusion_wrap_memory_address): Delete function, no longer used.
14059         (fusion_split_address); Delete TOC fusion support.
14060         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
14061         longer used with toc fusion being deleted.
14062         (TARGET_TOC_FUSION_FP): Likewise.
14063         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
14064         UNSPEC.
14065         (toc fusion spliter): Delete TOC fusion support.
14066         (toc_fusionload_<mode>): Likewise.
14067         (toc_fusionload_di): Likewise.
14068         (fusion_gpr_load_<mode>): Delete generator function, this insn no
14069         longer needs to be named.  Rename predicate to delete TOC fusion.
14070         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
14071         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
14072         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
14073         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
14074         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
14076 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14078         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
14079         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
14080         __int128 in built-in function prototypes.
14081         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
14082         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
14084 2018-07-27  Martin Sebor  <msebor@redhat.com>
14086         PR tree-optimization/86696
14087         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
14088         types, including enums.
14089         (handle_char_store): Be prepared for the above function to fail.
14091 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
14093         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
14094         when optimization level is lower than 2 or optimize for size.
14096 2018-07-26  Martin Sebor  <msebor@redhat.com>
14098         PR tree-optimization/86043
14099         PR tree-optimization/86042
14100         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
14101         (get_string_cst_length): Rename...
14102         (get_min_string_length): ...to this.  Add argument.
14103         (handle_char_store): Extend to handle multi-character stores by
14104         MEM_REF.
14105         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
14106         * tree.h (initializer_zerop): Add argument.
14108 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
14110         PR middle-end/86660
14111         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
14112         declare target to variables if they have always,{to,from,tofrom} map
14113         kinds.
14115 2018-07-26  Martin Liska  <mliska@suse.cz>
14117         PR lto/86548
14118         * lto-wrapper.c: Add linker_output as prefix
14119         for ltrans_output_file.
14121 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
14123         PR rtl-optimization/85805
14124         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
14125         value for hard registers if that was written in the same mode.
14127 2018-07-26  Martin Liska  <mliska@suse.cz>
14129         PR gcov-profile/86536
14130         * gcov.c (format_gcov): Use printf format %.*f directly
14131         and do not handle special values.
14133 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14135         * common/config/arc/arc-common.c (arc_option_optimization_table):
14136         Update default optimizations for size.
14138 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14140         * config/arc/arc.md (movsf_insn): Add short instruction selection.
14141         * config/arc/constraints.md (CfZ): New constraint.
14142         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
14143         (subsf3_fpu): Likewise.
14144         (cmpsf_fpu): Likewise.
14145         (cmpsf_fpu_uneq): Likewise.
14147 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14149         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
14150         accesses as well.
14151         (arc_is_uncached_mem_p): uncached applies to both the variable and
14152         the pointer.
14154 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14156         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
14157         register names.
14159 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
14161         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
14162         field "m_scopes" from vec to auto_vec.
14164 2018-07-25  Martin Liska  <mliska@suse.cz>
14166         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
14167         return type.
14169 2018-07-25  Richard Biener  <rguenther@suse.de>
14171         PR debug/86654
14172         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
14173         special wrt context_die late.
14174         (gen_subprogram_die): Re-use DIEs in local scope.
14176 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
14178         PR tree-optimization/86644
14179         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
14181 2018-07-25  Martin Liska  <mliska@suse.cz>
14183         PR middle-end/86645
14184         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
14185         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
14187 2018-07-25  Martin Liska  <mliska@suse.cz>
14189         PR sanitizer/79635
14190         * params.def: Explain ASan abbreviation and provide
14191         a documentation link.
14193 2018-07-24  Martin Sebor  <msebor@redhat.com>
14195         PR tree-optimization/86622
14196         PR tree-optimization/86532
14197         * builtins.h (string_length): Declare.
14198         * builtins.c (c_strlen): Correct handling of non-constant offsets.
14199         (check_access): Be prepared for non-constant length ranges.
14200         (string_length): Make extern.
14201         * expr.c (string_constant): Only handle the minor non-constant
14202         array index.  Use string_constant to compute the length of
14203         a generic string constant.
14205 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
14207         PR tree-optimization/86618
14208         * tree-vect-stmts.c (vectorizable_call): Don't take the address
14209         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
14211 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
14213         PR tree-optimization/86636
14214         * json.cc (json::object::set): Fix comment.  Add assertions.
14215         (json::array::append): Move here from json.h.  Add comment and an
14216         assertion.
14217         (json::string::string): Likewise.
14218         * json.h (json::array::append): Move to json.cc.
14219         (json::string::string): Likewise.
14220         * optinfo-emit-json.cc
14221         (optrecord_json_writer::impl_location_to_json): Assert that we
14222         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
14223         wrapper around it.  Expand the location once, rather than three
14224         times.
14225         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
14226         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
14227         wrappers.
14228         (optrecord_json_writer::optinfo_to_json): Likewise, in four
14229         places.  Fix some overlong lines.
14231 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
14233         * config/aarch64/aarch64-simd.md
14234         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
14235         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
14236         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
14237         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
14238         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
14239         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
14240         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
14241         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
14242         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
14244 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
14246         PR middle-end/86627
14247         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
14248         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
14249         and abs_d == d, do the power of two handling if profitable.
14251 2018-07-24  Richard Biener  <rguenther@suse.de>
14253         * match.pd: Add BIT_FIELD_REF canonicalizations.
14255 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14257         PR c/86617
14258         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
14260 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14262         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
14263         terminated STRING_CST object.
14265 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14267         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
14269 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
14271         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
14272         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
14273         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
14274         the elements into a register.
14275         (rs6000_split_v4si_init_di_reg): Delete.
14276         (rs6000_split_v4si_init): Delete.
14277         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
14278         (vsx_init_v4si): Rewrite as a define_expand.
14280 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
14282         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
14283         zero_extend argument from memory): New.
14285 2018-07-22  Martin Sebor  <msebor@redhat.com>
14287         PR bootstrap/86621
14288         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
14289         diagnosing calls with unknown arguments unless -Walloca-larger-than
14290         is restricted to less than PTRDIFF_MAX bytes.
14292 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
14294         * doc/gcov.texi (Invoking Gcov): Editorial changes.
14296 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
14298         * pretty-print.c (text_info::set_location): Remove redundant
14299         "line_table" parameter from call to rich_location::set_range.
14301 2018-07-20  Martin Sebor  <msebor@redhat.com>
14303         PR middle-end/82063
14304         * builtins.c (expand_builtin_alloca): Adjust.
14305         * calls.c (alloc_max_size): Simplify.
14306         * cgraphunit.c (cgraph_node::expand): Adjust.
14307         * common.opt (larger_than_size, warn_frame_larger_than): Remove
14308         variables.
14309         (frame_larger_than_size): Same.
14310         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
14311         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
14312         Initialize.
14313         * doc/invoke.texi (GCC Command Options): Document option arguments.
14314         Explain byte-size arguments and suffixes.
14315         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
14316         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
14317         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
14318         * doc/options.texi (UInteger): Expand.
14319         (Host_Wide_Int, ByteSize): Document new properties.
14320         * final.c (final_start_function_1): Include sizes in an error message.
14321         * function.c (frame_offset_overflow): Same.
14322         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
14323         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
14324         Diagnose unbounded alloca calls only for limits of less than
14325         PTRDIFF_MAX.
14326         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
14327         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
14328         for alloca(0).
14329         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
14330         only for limits of less than PTRDIFF_MAX.
14331         * langhooks-def.h (lhd_handle_option): Change function argument
14332         to HOST_WIDE_INT.
14333         * langhooks.c (lhd_handle_option): Same.
14334         * langhooks.h (handle_option): Same.
14335         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
14336         ByteSize flags.
14337         (var_type, var_type_struct): Same.
14338         (var_set): Handle ByteSize flag.
14339         * optc-gen.awk: Add comments to output to ease debugging.  Make
14340         use of HOST_WIDE_INT where appropriate.
14341         * opts-gen-save.awk:  Use %lx to format unsigned long.
14342         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
14343         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
14344         arguments.  Parse bytes-size suffixes.
14345         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
14346         (enum_value_to_arg): Same.
14347         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
14348         (handle_option): Adjust.
14349         (generate_option): Change function argument to HOST_WIDE_INT.
14350         (cmdline_handle_error): Adjust.
14351         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
14352         (set_option): Change function argument to HOST_WIDE_INT.
14353         (option_enabled): Handle cl_host_wide_int.
14354         (get_option_state): Handle CLVC_SIZE.
14355         (control_warning_option): Same.
14356         * opts.c (common_handle_option): Change function argument to
14357         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
14358         OPT_Wvla_larger_than_.
14359         * opts.h (enum cl_var_type): Add an enumerator.
14360         * stor-layout.c (layout_decl): Print a more meaningful warning.
14361         * toplev.c (output_stack_usage): Adjust.
14363 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
14365         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
14366         call to inline_expand_builtin_string_cmp.
14367         (expand_builtin_strcmp): Likewise.
14368         (expand_builtin_strncmp): Likewise.
14369         (inline_string_cmp): Delete the last parameter, change char_type_node
14370         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
14371         two operands.
14372         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
14373         the inlining expansion on target where the type of the call has same or
14374         narrower precision than unsigned char.
14376 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
14378         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
14379         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
14380         * common.opt (fsave-optimization-record): New option.
14381         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
14382         * doc/invoke.texi (-fsave-optimization-record): New option.
14383         * dumpfile.c: Include "optinfo-emit-json.h".
14384         (struct kv_pair): Move to coretypes.h.
14385         (optgroup_options): Make non-static.
14386         (dump_context::end_scope): Call
14387         optimization_records_maybe_pop_dump_scope.
14388         * dumpfile.h (optgroup_options): New decl.
14389         * json.cc: New file.
14390         * json.h: New file.
14391         * optinfo-emit-json.cc: New file.
14392         * optinfo-emit-json.h: New file.
14393         * optinfo.cc: Include "optinfo-emit-json.h".
14394         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
14395         (optinfo_enabled_p): Check optimization_records_enabled_p.
14396         (optinfo_wants_inlining_info_p): Likewise.
14397         * optinfo.h: Update comment.
14398         * profile-count.c (profile_quality_as_string): New function.
14399         * profile-count.h (profile_quality_as_string): New decl.
14400         (profile_count::quality): New accessor.
14401         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
14402         and optinfo_emit_json_cc_tests.
14403         * selftest.h (selftest::json_cc_tests): New decl.
14404         (selftest::optinfo_emit_json_cc_tests): New decl.
14405         * toplev.c: Include "optinfo-emit-json.h".
14406         (compile_file): Call optimization_records_finish.
14407         (do_compile): Call optimization_records_start.
14408         * tree-ssa-live.c: Include optinfo.h.
14409         (remove_unused_scope_block_p): Retain inlining information if
14410         optinfo_wants_inlining_info_p returns true.
14412 2018-07-20  Richard Biener  <rguenther@suse.de>
14414         PR debug/86585
14415         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
14416         to cover -flto-partition=none.
14418 2018-07-20  Martin Liska  <mliska@suse.cz>
14420         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
14421         (get_decl_source_range): Remove unused function.
14423 2018-07-20  Richard Biener  <rguenther@suse.de>
14425         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
14426         (struct vn_phi_s): Likewise.
14427         (struct vn_reference_s): Likewise.
14428         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
14429         for searching the slot of an entry known to be in the hash itself.
14430         (vn_phi_hasher::equal): Likewise.
14431         (vn_reference_hasher::equal): Likewise.
14432         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
14433         globals.
14434         (optimistic_info, current_info): Remove, keeping only valid_info.
14435         (vn_reference_lookup_1): Remove fallback lookup.
14436         (vn_reference_lookup_2): Likewise.
14437         (vn_nary_op_lookup_1): Likewise.
14438         (vn_phi_lookup): Likewise.
14439         (vn_nary_build_or_lookup_1): Make sure to not chain the built
14440         hash element.
14441         (vn_reference_insert): Adjust, chain the inserted hash element
14442         at last_inserted_ref.
14443         (vn_reference_insert_pieces): Likewise.
14444         (visit_reference_op_call): Likewise.
14445         (vn_nary_op_insert_into): Chain the inserted hash element at
14446         last_inserted_nary.
14447         (vn_nary_op_insert_pieces): Adjust.
14448         (vn_nary_op_insert): Likewise.
14449         (vn_nary_op_insert_stmt): Likewise.
14450         (vn_phi_insert): Adjust, chain the inserted hash element at
14451         last_inserted_phi.
14452         (process_scc): Remove clearing and copying the optimistic
14453         table.  Instead remove elements inserted during an optimistic
14454         iteration from the single table we maintain.
14455         (init_scc_vn): Adjust.
14456         (free_scc_vn): Likewise.
14457         (sccvn_dom_walker::record_cond): Likewise.
14458         (sccvn_dom_walker::after_dom_children): Likewise.
14460 2018-07-19  Martin Sebor  <msebor@redhat.com>
14462         PR tree-optimization/84047
14463         PR tree-optimization/83776
14464         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
14465         (check_array_bounds): Call it.
14467 2018-07-19  Martin Sebor  <msebor@redhat.com>
14469         * align.h (align_flags): Use member initialization.
14471 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
14473         * Makefile.in (OBJS): Add optinfo.o.
14474         * coretypes.h (class symtab_node): New forward decl.
14475         (struct cgraph_node): New forward decl.
14476         (class varpool_node): New forward decl.
14477         * dump-context.h: New file.
14478         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
14479         "tree-pass.h".
14480         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
14481         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
14482         (set_alt_dump_file): Likewise.
14483         (dump_context::~dump_context): New dtor.
14484         (dump_gimple_stmt): Move implementation to...
14485         (dump_context::dump_gimple_stmt): ...this new member function.
14486         Add the stmt to any pending optinfo, creating one if need be.
14487         (dump_gimple_stmt_loc): Move implementation to...
14488         (dump_context::dump_gimple_stmt_loc): ...this new member function.
14489         Start a new optinfo and add the stmt to it.
14490         (dump_gimple_expr): Move implementation to...
14491         (dump_context::dump_gimple_expr): ...this new member function.
14492         Add the stmt to any pending optinfo, creating one if need be.
14493         (dump_gimple_expr_loc): Move implementation to...
14494         (dump_context::dump_gimple_expr_loc): ...this new member function.
14495         Start a new optinfo and add the stmt to it.
14496         (dump_generic_expr): Move implementation to...
14497         (dump_context::dump_generic_expr): ...this new member function.
14498         Add the tree to any pending optinfo, creating one if need be.
14499         (dump_generic_expr_loc): Move implementation to...
14500         (dump_context::dump_generic_expr_loc): ...this new member
14501         function.  Add the tree to any pending optinfo, creating one if
14502         need be.
14503         (dump_printf): Move implementation to...
14504         (dump_context::dump_printf_va): ...this new member function.  Add
14505         the text to any pending optinfo, creating one if need be.
14506         (dump_printf_loc): Move implementation to...
14507         (dump_context::dump_printf_loc_va): ...this new member function.
14508         Start a new optinfo and add the stmt to it.
14509         (dump_dec): Move implementation to...
14510         (dump_context::dump_dec): ...this new member function.  Add the
14511         value to any pending optinfo, creating one if need be.
14512         (dump_context::dump_symtab_node): New member function.
14513         (dump_context::get_scope_depth): New member function.
14514         (dump_context::begin_scope): New member function.
14515         (dump_context::end_scope): New member function.
14516         (dump_context::ensure_pending_optinfo): New member function.
14517         (dump_context::begin_next_optinfo): New member function.
14518         (dump_context::end_any_optinfo): New member function.
14519         (dump_context::s_current): New global.
14520         (dump_context::s_default): New global.
14521         (dump_scope_depth): Delete global.
14522         (dumpfile_ensure_any_optinfo_are_flushed): New function.
14523         (dump_symtab_node): New function.
14524         (get_dump_scope_depth): Reimplement in terms of dump_context.
14525         (dump_begin_scope): Likewise.
14526         (dump_end_scope): Likewise.
14527         (selftest::temp_dump_context::temp_dump_context): New ctor.
14528         (selftest::temp_dump_context::~temp_dump_context): New dtor.
14529         (selftest::verify_item): New function.
14530         (ASSERT_IS_TEXT): New macro.
14531         (ASSERT_IS_TREE): New macro.
14532         (ASSERT_IS_GIMPLE): New macro.
14533         (selftest::test_capture_of_dump_calls): New test.
14534         (selftest::dumpfile_c_tests): Call it.
14535         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
14536         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
14537         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
14538         descriptive comment.
14539         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
14540         (dump_node, dump_bb): Move these unrelated decls.
14541         (class dump_manager): Add leading comment.
14542         * optinfo.cc: New file.
14543         * optinfo.h: New file.
14545 2018-07-19  Michael Collison  <michael.collison@arm.com>
14546             Richard Henderson <rth@redhat.com>
14548         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
14549         (subti): Handle op1 zero.
14550         (subvti4, usub4ti4): New.
14551         (*sub<GPI>3_compare1_imm): New.
14552         (sub<GPI>3_carryinCV): New.
14553         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
14554         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
14556 2018-07-19  Michael Collison  <michael.collison@arm.com>
14557             Richard Henderson <rth@redhat.com>
14559         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
14560         (addti3): Create simpler code if low part is already known to be 0.
14561         (addvti4, uaddvti4): New.
14562         (*add<GPI>3_compareC_cconly_imm): New.
14563         (*add<GPI>3_compareC_cconly): New.
14564         (*add<GPI>3_compareC_imm): New.
14565         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
14566         handle constants within this pattern..
14567         (*add<GPI>3_compareV_cconly_imm): New.
14568         (*add<GPI>3_compareV_cconly): New.
14569         (*add<GPI>3_compareV_imm): New.
14570         (add<GPI>3_compareV): New.
14571         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
14572         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
14573         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
14574         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
14575         with 'comparison' operator.
14576         (*add<GPI>3_compareV_cconly_imm): Ditto.
14577         (*add<GPI>3_compareV_cconly): Ditto.
14578         (*add<GPI>3_compareV_imm): Ditto.
14579         (add<GPI>3_compareV): Ditto.
14580         (add<mode>3_carryinC): Ditto.
14581         (*add<mode>3_carryinC_zero): Ditto.
14582         (*add<mode>3_carryinC): Ditto.
14583         (add<mode>3_carryinV): Ditto.
14584         (*add<mode>3_carryinV_zero): Ditto.
14585         (*add<mode>3_carryinV): Ditto.
14587 2018-07-19  Michael Collison  <michael.collison@arm.com>
14588             Richard Henderson <rth@redhat.com>
14590         * config/aarch64/aarch64-modes.def (CC_V): New.
14591         * config/aarch64/aarch64-protos.h
14592         (aarch64_addti_scratch_regs): Declare
14593         (aarch64_subvti_scratch_regs): Declare.
14594         (aarch64_expand_subvti): Declare.
14595         (aarch64_gen_unlikely_cbranch): Declare
14596         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
14597         for signed overflow using CC_Vmode.
14598         (aarch64_get_condition_code_1): Handle CC_Vmode.
14599         (aarch64_gen_unlikely_cbranch): New function.
14600         (aarch64_addti_scratch_regs): New function.
14601         (aarch64_subvti_scratch_regs): New function.
14602         (aarch64_expand_subvti): New function.
14604 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14606         * config/aarch64/aarch64-option-extensions.def: New entry for profile
14607         extension.
14608         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
14609         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
14610         extension.
14612 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14614         PR target/83009
14615         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14616         address check not strict.
14618 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14620         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
14621         Umq with Umn.
14622         (store_pair_lanes<mode>): Likewise.
14623         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
14624         enum value 'ADDR_QUERY_LDP_STP_N'.
14625         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
14626         (aarch64_print_address_internal): Add declaration.
14627         (aarch64_print_ldpstp_address): Remove.
14628         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
14629         (aarch64_print_operand): Change printing of 'y'.
14630         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
14631         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
14632         'true' rather than '1'.
14633         * config/aarch64/constraints.md (Uml): Likewise.
14634         (Uml): Rename to Umn.
14635         (Umq): Remove.
14637 2018-07-19  Richard Biener  <rguenther@suse.de>
14639         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
14640         a trailing array.
14641         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
14642         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
14643         (vn_reference_hasher): Likewise.
14644         (struct vn_tables_s): Remove obstack and alloc-pool members.
14645         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
14646         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
14647         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
14648         (vn_reference_insert_pieces): Likewise.
14649         (alloc_vn_nary_op_noinit): Adjust.
14650         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
14651         (vn_phi_eq): Adjust.
14652         (shared_lookup_phiargs): Remove.
14653         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
14654         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
14655         (visit_reference_op_call): Likewise.
14656         (copy_nary, copy_phi, copy_reference): Remove.
14657         (process_scc): Rewind the obstack when iterating.  Do not
14658         copy the elements to valid_info but just move them from one
14659         hashtable to the other.
14660         (allocate_vn_table): Adjust.
14661         (free_vn_table): Likewise.
14662         (init_scc_vn): Likewise.
14663         (free_scc_vn): Likewise.
14665 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
14667         PR target/86560
14668         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
14669         indirect_return as function type attribute.
14670         (ix86_attribute_table): Change indirect_return to function
14671         type attribute.
14672         * doc/extend.texi: Update indirect_return attribute.
14674 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
14676         * wide-int.h (widest2_int): New.
14677         * gimple-fold.c (arith_overflowed_p): Use it.
14678         * tree.h (widest2_int_cst): New.
14679         * tree-vrp.c (wide_int_binop_overflow): Rename from
14680         vrp_int_const_binop.
14681         Rewrite to work on trees.
14682         (extract_range_from_multiplicative_op_1): Abstract code to...
14683         (wide_int_range_min_max): ...here.
14684         (wide_int_range_cross_product): ...and here.
14685         (extract_range_from_binary_expr_1): Abstract overflow code to...
14686         (wide_int_range_mult_wrapping): ...here.
14687         * tree-vrp.h (wide_int_range_cross_product): New.
14688         (wide_int_range_mult_wrapping): New.
14690 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
14691             Julia Koval  <julia.koval@intel.com>
14693         * config/i386/x86-tune-costs.h (skylake_memcpy,
14694         skylake_memset): Replace rep_prefix with unrolling for size 512.
14696 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
14698         PR middle-end/86544
14699         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
14700         comparision with EQ_EXPR in last stmt.
14702 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14704         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
14705         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
14706         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
14707         previously known as "PowerPC AltiVec Built-in Functions".  Move
14708         some material to new subsubsections "PowerPC AltiVec Built-in
14709         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
14710         ISA 2.07".
14711         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
14712         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
14713         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
14714         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
14716 2018-07-18  Richard Biener  <rguenther@suse.de>
14718         PR tree-optimization/86557
14719         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
14720         EXACT_DIV_EXPR.
14722 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
14724         * config/s390/s390.c (s390_function_profiler): Generate CFI.
14726 2018-07-17  Jeff Law  <law@redhat.com>
14728         * config/arm/arm.c (get_label_padding): Update for recent
14729         changes to label_to_alignment.
14731         PR tree-optimization/86010
14732         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
14734         * config/mips/mips.c (vr4130_align_insns): Update for recent
14735         changes to label_to_alignment.
14737         * config/frv/frv.c (frv_label_align): Update for recent changes
14738         to label_to_alignment.
14740         * config/nios2/nios2.c (nios2_label_align): Update for recent
14741         changes which dropped ALIGN_LABELS_LOG.
14743 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
14745         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
14746         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
14748 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
14750         * config/arc/arc.c (arc_label_align): Use align_labels instead of
14751         deprecated align_labels_log.
14753 2018-07-17  Richard Biener  <rguenther@suse.de>
14755         PR lto/86456
14756         * dwarf2out.c (init_sections_and_labels): Always generate
14757         a debug_line_str_section for early LTO debug.
14758         (dwarf2out_finish): Reset debug_line_str_hash output early.
14759         Bump counter for extra dwarf5 .debug_loc labels to not conflict
14760         with fat LTO part.
14761         (dwarf2out_early_finish): Output debug_line_str.
14763 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
14765         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
14766         index register on z196 or later.
14768 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
14770         * config/s390/s390.c (s390_default_align): Set default function
14771         alignment to 16.
14772         (s390_override_options_after_change): Call s390_default align.
14773         (s390_option_override_internal): Call s390_default_align.
14774         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
14776 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
14778         PR middle-end/86542
14779         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
14780         to _looptemp_ clauses, other than the first two.
14782 2018-07-17  Martin Liska  <mliska@suse.cz>
14784         * opts.c: Do not enable OPT_falign_* for -Os.
14786 2018-07-17  Martin Liska  <mliska@suse.cz>
14788         * align.h (MAX_CODE_ALIGN): New.
14789         (MAX_CODE_ALIGN_VALUE): New.
14790         * common/config/i386/i386-common.c (ix86_handle_option):
14791         (MAX_CODE_ALIGN): Moved to align.h.
14792         * final.c (MAX_CODE_ALIGN): Likewise.
14793         * opts.c (parse_and_check_align_values):
14794         (MAX_CODE_ALIGN): Likewise.
14795         (MAX_CODE_ALIGN_VALUE): Likewise.
14797 2018-07-17  Martin Liska  <mliska@suse.cz>
14799         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
14800         in order to fulfil coding style.
14801         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
14802         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14803         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
14804         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
14805         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
14806         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
14807         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
14808         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
14809         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14811 2018-07-17  Martin Liska  <mliska@suse.cz>
14813         * align.h: New file.
14814         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
14815         directly.
14816         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
14817         align_flags of label_to_alignment.
14818         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
14819         align_flags class.
14820         * config/m68k/m68k.c: Do not use removed align_labels_value and
14821         align_loops_value.
14822         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
14823         (LOOP_ALIGN): Likewise.
14824         (LABEL_ALIGN): Likewise.
14825         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
14826         Remove not used macro.
14827         (rs6000_loop_align): Change return type to align_flags.
14828         (rs6000_loop_align_max_skip): Remove.
14829         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
14830         Change return type to align_flags.
14831         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
14832         Remove not used macro.
14833         (rs6000_loop_align):  Change return type to align_flags.
14834         (rs6000_loop_align_max_skip): Remove.
14835         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
14836         * config/rx/rx-protos.h (rx_align_for_label): Make it
14837         static function.
14838         * config/rx/rx.c (rx_align_for_label): Change return type
14839         to align_flags.
14840         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
14841         macro definitions.
14842         into align_flags class.
14843         (LABEL_ALIGN): Likewise.
14844         (LOOP_ALIGN): Likewise.
14845         * config/s390/s390.c (s390_label_align): Use align_flags
14846         class member.
14847         (s390_asm_output_function_label): Likewise.
14848         * config/sh/sh.c (sh_override_options_after_change):
14849         Use align_flags class directly without macros.
14850         (find_barrier): Likewise.
14851         (barrier_align): Likewise.
14852         (sh_loop_align): Likewise.
14853         * config/spu/spu.c (spu_option_override):
14854         Use align_flags_tuple::get_value instead of removed macros.
14855         (spu_sched_init): Likewise.
14856         * config/spu/spu.h (GTY): Likewise.
14857         * config/visium/visium.c (visium_option_override):
14858         Set "8" as default secondary alignment.
14859         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
14860         in order to guarantee secondary alignment of 8.
14861         * coretypes.h: Include align.h header file.
14862         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
14863         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
14864         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
14865         * doc/tm.texi.in: Likewise.
14866         * final.c (struct label_alignment): Remove not used structure.
14867         (LABEL_ALIGN): Change type to align_flags.
14868         (LOOP_ALIGN): Likewise.
14869         (JUMP_ALIGN): Likewise.
14870         (default_loop_align_max_skip): Remove.
14871         (default_label_align_max_skip): Likewise.
14872         (default_jump_align_max_skip): Likewise.
14873         (default_label_align_after_barrier_max_skip):
14874         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
14875         (LABEL_TO_MAX_SKIP): Remove.
14876         (label_to_alignment): Return align_flags type instead of integer.
14877         (label_to_max_skip): Remove.
14878         (align_fuzz): Use align_flags type.
14879         (compute_alignments): Use align_flags type and use align_flags::max
14880         to combine multiple alignments.
14881         (grow_label_align): Grow vec instead of C array.
14882         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
14883         (shorten_branches):  Use align_flags type and use align_flags::max
14884         to combine multiple alignments.
14885         (final_scan_insn_1): Remove usage of secondary alignment that comes
14886         from label alignment, but instead use proper secondary alignment
14887         which is computed in grow_label_align.
14888         * flags.h (struct align_flags_tuple): Move to align.h.
14889         (struct align_flags): Likewise.
14890         (state_align_loops): Rename to align_loops.
14891         (state_align_jumps): Rename to align_jumps.
14892         (state_align_labels): Rename to align_labels.
14893         (state_align_functions): Rename to align_functions.
14894         (align_loops_log): Remove.
14895         (align_jumps_log): Remove.
14896         (align_labels_log): Remove.
14897         (align_functions_log): Remove.
14898         (align_loops_max_skip): Remove.
14899         (align_jumps_max_skip): Remove.
14900         (align_labels_max_skip): Remove.
14901         (align_functions_max_skip): Remove.
14902         (align_loops_value): Remove.
14903         (align_jumps_value): Remove.
14904         (align_labels_value): Remove.
14905         (align_functions_value): Remove.
14906         * output.h (label_to_alignment): Change return type to align_flags.
14907         (label_to_max_skip): Remove.
14908         * target.def: Remove loop_align_max_skip, label_align_max_skip,
14909         jump_align_max_skip macros.
14910         * targhooks.h (default_loop_align_max_skip): Remove.
14911         (default_label_align_max_skip): Likewise.
14912         (default_jump_align_max_skip): Likewise.
14913         (default_label_align_after_barrier_max_skip): Remove.
14914         * toplev.c (read_log_maxskip): Use ::normalize function.
14915         (parse_N_M): Remove not used argument and also call ::normalize.
14916         (parse_alignment_opts): Do not pass unused arguments.
14917         * varasm.c (assemble_start_function): Use directly align_functions
14918         instead of removed macros.
14919         * system.h: Do not poison removed macros.
14921 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
14923         PR middle-end/86539
14924         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
14925         and cond temporaries don't have reference type if iterator has
14926         pointer type.  For init use &for_pre_body instead of pre_p if
14927         for_pre_body is non-empty.
14929 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
14931         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
14932         double-double modes to SFmode directly directly.
14933         (trunc<mode>sf2_fprs): Delete.
14935 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
14937         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
14938         for conversions between IFmode and the decimal floating point modes.
14939         (init_float128_ieee): Use the correct names for conversions between
14940         KFmode and the decimal floating point modes.
14942 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
14944         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
14945         for the conversions between TDmode and IFmode.
14946         (init_float128_ieee): Use more correct names for the conversions
14947         between TDmode and KFmode.
14949 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
14951         PR tree-optimization/86526
14952         * builtins.c (expand_builtin_memcmp): Formatting fixes.
14953         (inline_expand_builtin_string_cmp): Likewise.
14954         (inline_string_cmp): Likewise.  Use c_readstr instead of
14955         builtin_memcpy_read_str.  Add unit_mode temporary.
14957 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14959         PR middle-end/86528
14960         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
14961         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
14963 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14965         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
14966         Alphabetize prototypes of built-in functions, separating out
14967         built-in functions that are listed in this section but should be
14968         described elsewhere.
14970 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
14972         PR target/86511
14973         * expmed.c (emit_store_flag): Do not emit setcc followed by a
14974         conditional move when trapping comparison was split to a
14975         non-trapping one (and vice versa).
14977 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
14979         * config/s390/s390.c (s390_function_profiler): Generate nops
14980         instead of profiler call sequences.
14981         * config/s390/s390.opt: Add the new option.
14983 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
14985         * config/s390/s390.c (s390_function_profiler): Generate
14986         __mcount_loc section.
14987         * config/s390/s390.opt: Add the new option.
14989 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
14991         * common.opt: Add the new warning.
14992         * config/s390/s390.c (s390_function_profiler): Emit "brasl
14993         %r0,__fentry__" when -mfentry is specified.
14994         (s390_option_override_internal): Disallow -mfentry for 31-bit
14995         CPUs.
14996         * config/s390/s390.opt: Add the new option.
14998 2018-07-16  Richard Biener  <rguenther@suse.de>
15000         PR lto/86523
15001         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
15002         for function-local FUNCTION_DECL and RESULT_DECL immediately.
15004 2018-07-16  Martin Liska  <mliska@suse.cz>
15006         PR ipa/86529
15007         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
15008         to ::get_create.
15010 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
15012         * config/arc/arcHS.md: Update ARCHS scheduling rules.
15014 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
15016         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
15017         for ARCHS4x.
15018         * config/arc/arc-cpus.def (hs4x): New cpu.
15019         (hs4xd): Likewise.
15020         * config/arc/arc-tables.opt: Regenerate.
15021         * config/arc/arc.c (arc_sched_issue_rate): New function.
15022         (TARGET_SCHED_ISSUE_RATE): Define.
15023         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
15024         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
15025         fpu_cvt.
15026         (attr tune): Add ARCHS4x tune values.
15027         (attr tune_dspmpy): Define.
15028         (*tst): Correct instruction type.
15029         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
15030         * config/arc/arcHS4x.md: New file.
15031         * config/arc/fpu.md: Update instruction type attributes.
15032         * config/arc/t-multilib: Regenerate.
15034 2018-07-16  Tom de Vries  <tdevries@suse.de>
15036         PR debug/86455
15037         * var-tracking.c (vt_initialize): Fix pre_dec handling.
15039 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
15041         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
15042         early clobber.
15044 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
15046         PR tree-optimization/86514
15047         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
15048         conversion to a boolean type from a type with greater precision.
15050 2018-07-16  Tom de Vries  <tdevries@suse.de>
15052         * var-tracking.c (vt_initialize): Print adjusted insn slim if
15053         dump_flags request TDF_SLIM.
15055 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
15057         * fold-const.c (int_const_binop_1): Abstract...
15058         (wide_int_binop): ...wide int code here.
15059         (poly_int_binop): ...poly int code here.
15060         Abstract the rest of int_const_binop_1 into int_const_binop.
15061         * fold-const.h (wide_int_binop): New.
15062         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
15063         Remove useless PLUS/MINUS_EXPR case.
15064         (zero_nonzero_bits_from_vr): Move wide int code...
15065         (zero_nonzero_bits_from_bounds): ...here.
15066         (extract_range_from_binary_expr_1): Move mask optimization code...
15067         (range_easy_mask_min_max): ...here.
15068         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
15069         (range_easy_mask_min_max): New.
15071 2018-07-15  Jeff Law  <law@redhat.com>
15073         PR target/85993
15074         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
15075         block.
15077 2018-07-14  Jim Wilson  <jimw@sifive.com>
15079         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
15081 2018-07-14  Paul Koning  <ni1d@arrl.net>
15083         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
15085 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
15087         * lto-streamer-out.c (copy_function_or_variable): Dump info about
15088         copying section.
15090 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
15091             Steve Munroe  <munroesj52@gmail.com>
15093         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
15094         (_mm_andnot_si128): Likewise.
15095         (_mm_or_si128): Likewise.
15096         (_mm_xor_si128): Likewise.
15098 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
15100         PR middle-end/78809
15101         * builtins.c (expand_builtin_memcmp): Inline the calls first
15102         when result_eq is false.
15103         (expand_builtin_strcmp): Inline the calls first.
15104         (expand_builtin_strncmp): Likewise.
15105         (inline_string_cmp): New routine. Expand a string compare
15106         call by using a sequence of char comparison.
15107         (inline_expand_builtin_string_cmp): New routine. Inline expansion
15108         a call to str(n)cmp/memcmp.
15109         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
15110         New option.
15111         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
15113 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
15115         * config/arm/driver-arm.c: Include arm-native.h.
15116         (host_detect_local_cpu): Use auto-generated data tables.
15117         (vendors, arm_cpu_table): Delete.  Move part information to ...
15118         * config/arm/arm-cpus.in: ... here.
15119         * config/arm/parsecpu.awk (gen_native): New function.
15120         (vendor, part): New CPU fields.
15121         (END): Add support for building the native CPU detection tables.
15122         * config/arm/t-arm (arm-native.h): Add build rule.
15123         (driver-arm.o): Add dependency on arm-native.h.
15125 2018-07-13  Richard Biener  <rguenther@suse.de>
15127         PR middle-end/85974
15128         * match.pd (addr1 - addr2): Allow either of the operand to
15129         have a conversion.
15131 2018-07-13  Tom de Vries  <tdevries@suse.de>
15133         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
15134         in remap_ssa_name.
15136 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
15138         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
15139         arrays instead of numbered variables.
15141 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
15143         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
15144         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
15146 2018-07-13  Richard Biener  <rguenther@suse.de>
15148         PR debug/86452
15149         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
15150         instead of get_context_die.
15152 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
15153             Richard Biener  <rguenther@suse.de>
15155         PR middle-end/86489
15156         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
15157         that the loop latch destination where phi is defined.
15159 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
15161         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
15162         (riscv_expand_epilogue): Add assertion to check interrupt mode.
15163         (riscv_set_current_function): Extract getting interrupt type to new
15164         function.
15165         (riscv_get_interrupt_type): New function.
15166         (riscv_merge_decl_attributes): New function, checking interrupt type is
15167         same.
15168         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
15170 2018-07-12  Paul Koning  <ni1d@arrl.net>
15172         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
15173         directive.
15175 2018-07-12  Paul Koning  <ni1d@arrl.net>
15177         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
15178         zero reference, add doloop_end instead.
15179         * doc/md.texi (decrement_and_branch_until_zero): Remove.
15180         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
15181         detail for doloop_end.
15183 2018-07-12  Martin Sebor  <msebor@redhat.com>
15185         PR c/86453
15186         * attribs.c (decl_attributes): Reject conflicting attributes before
15187         calling attribute handlers.
15189 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
15191         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
15192          parameter.
15193         (gcc::dump_manager::get_dump_file_name): likewise.
15194         (dump_begin): Likewise.
15195         * dumpfile.h (dump_begin): Update prototype.
15196         (gcc::dump_manager::get_dump_file_name,
15197         gcc::dump_manager::get_dump_file_name): Update prototype.
15199 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15201         * internal-fn.h (vectorizable_internal_fn_p): New function.
15202         * tree-vect-slp.c (compatible_calls_p): Likewise.
15203         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
15204         to internal functions.
15205         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
15207 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15209         * fold-const.h (inverse_conditions_p): Declare.
15210         * fold-const.c (inverse_conditions_p): New function.
15211         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
15212         that test the inverse condition of a conditional internal function.
15213         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
15214         * internal-fn.c (internal_fn_mask_index): Handle conditional
15215         internal functions.
15216         (vectorized_internal_fn_supported_p): New function.
15217         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
15218         (any_pred_load_store): Replace with...
15219         (need_to_predicate): ...this new variable.
15220         (redundant_ssa_names): New variable.
15221         (ifcvt_can_use_mask_load_store): Move initial checks to...
15222         (ifcvt_can_predicate): ...this new function.  Handle tree codes
15223         for which a conditional internal function exists.
15224         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
15225         instead of ifcvt_can_use_mask_load_store.  Update after variable
15226         name change.
15227         (predicate_load_or_store): New function, split out from
15228         predicate_mem_writes.
15229         (check_redundant_cond_expr): New function.
15230         (value_available_p): Likewise.
15231         (predicate_rhs_code): Likewise.
15232         (predicate_mem_writes): Rename to...
15233         (predicate_statements): ...this.  Use predicate_load_or_store
15234         and predicate_rhs_code.
15235         (combine_blocks, tree_if_conversion): Update after above name changes.
15236         (ifcvt_local_dce): Handle redundant_ssa_names.
15237         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
15238         general conditional functions.
15239         * tree-vect-stmts.c (vectorizable_call): Likewise.
15241 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15242             Alan Hayward  <alan.hayward@arm.com>
15243             David Sherwood  <david.sherwood@arm.com>
15245         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
15246         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
15247         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
15248         plus and minus and convert them into IFN_COND_FMA-based sequences.
15249         (convert_mult_to_fma): Handle conditional plus and minus.
15251 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15253         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
15254         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
15255         (cond_fnms_optab): New optabs.
15256         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
15257         internal functions.
15258         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
15259         * internal-fn.h (get_conditional_internal_fn): Declare.
15260         (get_unconditional_internal_fn): Likewise.
15261         * internal-fn.c (cond_ternary_direct): New macro.
15262         (expand_cond_ternary_optab_fn): Likewise.
15263         (direct_cond_ternary_optab_supported_p): Likewise.
15264         (FOR_EACH_COND_FN_PAIR): Likewise.
15265         (get_conditional_internal_fn): New function.
15266         (get_unconditional_internal_fn): Likewise.
15267         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
15268         (gimple_match_op::gimple_match_op): Add a new overload for 5
15269         operands.
15270         (gimple_match_op::set_op): Likewise.
15271         (gimple_resimplify5): Declare.
15272         * genmatch.c (decision_tree::gen): Generate simplifications for
15273         5 operands.
15274         * gimple-match-head.c (gimple_simplify): Define an overload for
15275         5 operands.  Handle calls with 5 arguments in the top-level overload.
15276         (convert_conditional_op): Handle conversions from unconditional
15277         internal functions to conditional ones.
15278         (gimple_resimplify5): New function.
15279         (build_call_internal): Pass a fifth operand.
15280         (maybe_push_res_to_seq): Likewise.
15281         (try_conditional_simplification): Try converting conditional
15282         internal functions to unconditional internal functions.
15283         Handle 3-operand unconditional forms.
15284         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
15285         Define ternary equivalents of the current rules for binary conditional
15286         internal functions.
15287         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
15288         ternary operations.
15289         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
15290         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
15291         (optab): Handle them.
15292         (SVE_COND_FP_TERNARY): New int iterator.
15293         (sve_fmla_op, sve_fmad_op): New int attributes.
15294         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
15295         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
15296         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
15298 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15300         * target.def (preferred_else_value): New target hook.
15301         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
15302         * doc/tm.texi: Regenerate.
15303         * targhooks.h (default_preferred_else_value): Declare.
15304         * targhooks.c (default_preferred_else_value): New function.
15305         * internal-fn.h (conditional_internal_fn_code): Declare.
15306         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
15307         (get_conditional_internal_fn): Use it.
15308         (conditional_internal_fn_code): New function.
15309         * gimple-match.h (gimple_match_cond): New struct.
15310         (gimple_match_op): Add a cond member function.
15311         (gimple_match_op::gimple_match_op): Update all forms to take a
15312         gimple_match_cond.
15313         * genmatch.c (expr::gen_transform): Use the same condition as res_op
15314         for the suboperation, but don't specify a particular else_value.
15315         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
15316         (visit_nary_op, visit_reference_op_load): Pass
15317         gimple_match_cond::UNCOND to the gimple_match_op constructor.
15318         * gimple-match-head.c: Include tree-eh.h
15319         (convert_conditional_op): New function.
15320         (maybe_resimplify_conditional_op): Likewise.
15321         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
15322         (gimple_resimplify2): Likewise.
15323         (gimple_resimplify3): Likewise.
15324         (gimple_resimplify4): Likewise.
15325         (maybe_push_res_to_seq): Return null for conditional operations.
15326         (try_conditional_simplification): New function.
15327         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
15328         constructor.
15329         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
15330         IFN_COND_* call.
15331         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
15332         function.
15333         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
15335 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
15337         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15338         DECL_FCONTEXT
15339         (hash_tree): Do not hash DECL_FCONTEXT
15340         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
15341         Do not stream DECL_FCONTEXT.
15342         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
15343         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
15345 2018-07-12  Richard Biener  <rguenther@suse.de>
15347         PR debug/86462
15348         * dwarf2out.c (gen_block_die): Only output blocks when they have
15349         at least one !DECL_IGNORED_P variable.
15351 2018-07-12  Richard Biener  <rguenther@suse.de>
15353         PR target/84829
15354         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
15355         Remove -mieee-fp handling.
15357 2018-07-12  Richard Biener  <rguenther@suse.de>
15359         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
15360         left-over from last patch.
15362 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
15364         PR tree-optimization/86492
15365         * gimple-ssa-store-merging.c
15366         (imm_store_chain_info::coalesce_immediate_stores): Call
15367         check_no_overlap even for the merge_overlapping case.  Formatting fix.
15369 2018-07-12  Richard Biener  <rguenther@suse.de>
15371         PR middle-end/86479
15372         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
15373         move possibly trapping operations into the conditional.
15375 2018-07-12  Richard Biener  <rguenther@suse.de>
15377         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
15378         (vn_lookup_simplify_result): Remove recursion limit applied
15379         here.
15380         (vn_nary_build_or_lookup_1): Adjust.
15381         (try_to_simplify): Likewise.
15382         * gimple-match-head.c (gimple_resimplify1): Instead apply one
15383         here.
15384         (gimple_resimplify2): Likewise.
15385         (gimple_resimplify3): Likewise.
15386         (gimple_resimplify4): Likewise.
15388 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
15390         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
15391         Use __mmask64 type instead of __mmask8 for __M argument.
15392         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
15393         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
15394         __U argument.
15395         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
15396         __mmask16 for __M argument.
15397         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
15398         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
15399         to __mmask16 instead of __mmask8.
15400         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
15401         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
15402         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
15403         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
15404         instead of __mmask16 for __U argument.
15405         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
15406         __mmask16 instead of __mmask8 for __U argument.
15407         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
15408         __U argument.
15409         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
15410         __mmask16.
15411         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
15412         argument.
15413         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
15414         __U argument.
15415         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
15416         __mmask16.
15417         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
15418         of __mmask16.
15419         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
15420         __U argument.
15421         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
15422         __U argument.
15423         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
15424         __U argument.
15425         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
15426         __U argument.
15427         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
15428         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
15429         return type as well as __M argument type and all casts from __mmask8
15430         to __mmask32.
15431         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
15432         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
15433         return type as well as __M argument type and all casts from __mmask8
15434         to __mmask16.
15435         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
15436         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
15437         return type as well as __M argument type and all casts from __mmask8
15438         to __mmask32.
15439         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
15440         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
15441         return type as well as __M argument type and all casts from __mmask8
15442         to __mmask16.
15443         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
15444         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
15445         __mmask16.
15447 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
15449         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
15450         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
15451         for __U argument.
15453 2018-07-11  Paul Koning  <ni1d@arrl.net>
15455         * doc/md.texi (define_subst): Document how multiple occurrences of
15456         the same argument in the replacement pattern are handled.
15458 2018-07-11  Paul Koning  <ni1d@arrl.net>
15460         * doc/extend.texi (Common Variable Attributes): Move "mode" into
15461         alphabetical order.
15462         (Common Type Attributes): Add "mode" attribute.
15464 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
15466         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
15467         stream DECL_ORIGINAL_TYPE.
15468         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
15469         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
15470         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
15471         Do not walk original type.
15472         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
15473         external decls.
15474         (write_ts_decl_non_common_tree_pointers): Do not stream
15475         DECL_ORIGINAL_TYPE
15476         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
15477         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
15479 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
15481         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
15482         thread twice from the same starting edge.
15484 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
15486         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
15487         * gimple.c (gimple_call_nonnull_result_p): ...here...
15488         (gimple_call_nonnull_arg): ...and here.
15489         * gimple.h (gimple_call_nonnull_result_p): New.
15490         (gimple_call_nonnull_arg): New.
15492 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
15494         * config/arm/arm-cpus.in: Move information from fpu field of each
15495         cpu definition to the isa field.
15496         * config/arm/parsecpu.awk (fpu): Delete match rule.
15497         (gen_comm_data): Don't add bits from the CPU's FPU entry.
15499 2018-07-11  Richard Biener  <rguenther@suse.de>
15501         PR debug/86457
15502         * dwarf2out.c (init_sections_and_labels): Use
15503         output_asm_line_debug_info consistently.
15504         (dwarf2out_early_finish): Likewise.
15505         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
15506         type units.
15508 2018-07-11  Richard Biener  <rguenther@suse.de>
15510         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
15511         Rework father_bb setting in a way to avoid propagating constants
15512         multiple times on a loop body.
15514 2018-07-10  Mark Wielaard  <mark@klomp.org>
15516         PR debug/86459
15517         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
15519 2018-07-10  Richard Biener  <rguenther@suse.de>
15521         * hash-map.h (hash_map::iterator::operator*): Return
15522         references to key and value.
15524 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
15526         PR c++/86443
15527         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
15528         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
15529         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
15530         (gimplify_omp_for): For composite loops, move outer
15531         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
15532         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
15533         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
15534         TREE_LIST for both the original class iterator and the "last" helper
15535         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
15536         loop, remember has_decl_expr from outer composite loops for the
15537         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
15539 2018-07-09  Martin Sebor  <msebor@redhat.com>
15541         PR middle-end/77357
15542         PR middle-end/86428
15543         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
15544         accessing implicitly initialized array elements.
15545         * expr.c (string_constant): Handle string initializers of
15546         character arrays within aggregates.
15547         * gimple-fold.c (fold_array_ctor_reference): Add argument.
15548         Store element offset.  As a special case, handle zero size.
15549         (fold_nonarray_ctor_reference): Same.
15550         (fold_ctor_reference): Add argument.  Store subobject offset.
15551         * gimple-fold.h (fold_ctor_reference): Add argument.
15553 2018-07-09  Paul Koning  <ni1d@arrl.net>
15555         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
15556         (pdp11_insn_cost): New function.
15557         (pdp11_md_asm_adjust): New function.
15558         (TARGET_INVALID_WITHIN_DOLOOP): Define.
15559         (pdp11_rtx_costs): Update to match machine better.
15560         (output_addr_const_pdp11): Correct format mismatch warnings.
15561         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
15562         * config/pdp11/pdp11.md: General change to add base_cost and/or
15563         length attributes for use by new pdp11_insn_cost function.
15564         (MIN_BRANCH): Correct definition.
15565         (MIN_SOB): Ditto.
15566         (doloop_end): Use standard pattern name for looping pattern.
15567         (doloop_end_nocc): New.
15568         (movsf): Add another constraint alternative.
15569         (zero_extendqihi2): Add constraint alternatives for not in place
15570         extend.
15571         (zero_extendhisi2): Remove.
15572         (shift patterns): Add CC handling variants.
15573         (bswaphi2): New.
15574         (bswapsi2): New.
15575         (rothi3): New.
15576         (define_peephole2): New peephole to recognize mov that sets CC for
15577         subsequent test.
15579 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15581         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
15582         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
15583         wi::add.
15585 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
15587         PR c/86420
15588         * real.c (real_nextafter): Return true if result is denormal.
15590 2018-07-09  Martin Liska  <mliska@suse.cz>
15592         * common.opt: Add back wrongly removed attribute.
15594 2018-07-09  Richard Biener  <rguenther@suse.de>
15596         PR debug/86413
15597         * dwarf2out.c (gen_block_die): For an early generated DIE
15598         always output high/low PC attributes.
15600 2018-07-09  Tom de Vries  <tdevries@suse.de>
15602         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
15603         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
15604         onto VAR_DECL with abstract origin.
15606 2018-07-07  Jim Wilson  <jimw@sifive.com>
15608         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
15610 2018-07-07  Tom de Vries  <tdevries@suse.de>
15612         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
15614 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
15616         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
15617         overflow_type.
15618         (combine_bound): Use wide-int overflow calculation instead of
15619         rolling our own.
15620         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
15621         overflow_type.
15622         * fold-const.c (int_const_binop_2): Same.
15623         (extract_muldiv_1): Same.
15624         (fold_div_compare): Same.
15625         (fold_abs_const): Same.
15626         * match.pd: Same.
15627         * poly-int.h (add): Same.
15628         (sub): Same.
15629         (neg): Same.
15630         (mul): Same.
15631         * predict.c (predict_iv_comparison): Same.
15632         * profile-count.c (slow_safe_scale_64bit): Same.
15633         * simplify-rtx.c (simplify_const_binary_operation): Same.
15634         * tree-chrec.c (tree_fold_binomial): Same.
15635         * tree-data-ref.c (split_constant_offset_1): Same.
15636         * tree-if-conv.c (idx_within_array_bound): Same.
15637         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
15638         * tree-ssa-phiopt.c (minmax_replacement): Same.
15639         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
15640         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
15641         * vr-values.c (vr_values::adjust_range_with_scev): Same.
15642         * wide-int.cc (wi::add_large): Same.
15643         (wi::mul_internal): Same.
15644         (wi::sub_large): Same.
15645         (wi::divmod_internal): Same.
15646         * wide-int.h: Change overflow type to overflow_type for neg, add,
15647         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
15648         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
15649         mul_internal, divmod_internal.
15650         (overflow_type): New enum.
15651         (accumulate_overflow): New.
15653 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
15655         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
15656         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
15658 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
15660         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
15661         argument is checked for zero before entering loop, avoid checking again.
15663 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
15665         * gimplify.h (generic_expr_could_trap_p): Set as global function.
15666         * gimplify.h (generic_expr_could_trap_p): Likwise.
15667         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
15669 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
15671         PR tree-optimization/86401
15672         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
15673         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
15674         (fold_bit_and_mask): ... here.  New helper function for match.pd.
15675         * fold-const.h (fold_bit_and_mask): Declare.
15676         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
15678 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
15680         PR target/86324
15681         * target.def (translate_mode_attribute): New hook.
15682         * targhooks.h (default_translate_mode_attribute): Declare.
15683         * targhooks.c (default_translate_mode_attribute): New function.
15684         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
15685         * doc/tm.texi: Regenerate.
15686         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
15687         (rs6000_translate_mode_attribute): New function.
15689 2018-07-06  Paul Koning  <ni1d@arrl.net>
15691         * doc/md.texi (define_split): Document DONE and FAIL.
15692         (define_peephole2): Ditto.
15694 2018-07-05  Jeff Law  <law@redhat.com>
15696         PR tree-optimization/86010
15697         * tree-ssa-dse.c (compute_trims): More aggressively trim at
15698         both the head and tail of mem* and str* calls.
15700 2018-07-05  Jim Wilson  <jimw@sifive.com>
15702         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
15704 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
15706         * config/aarch64/aarch64-simd.md: correct flags text for
15707         MIN_EXPR replacement.
15709 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
15711         * configure: Regenerated.
15713 2018-07-05  Carl Love  <cel@us.ibm.com>
15715         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
15716         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
15717         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
15718         VSX_BUILTIN_DOUBLEL_V4SF.
15720 2018-07-05  Martin Sebor  <msebor@redhat.com>
15722         PR c++/86400
15723         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
15724         than its domain to compute its the upper bound of a char array.
15726 2018-07-05  Nathan Sidwell  <nathan@acm.org>
15728         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
15729         * doc/cpp.texi: Update comment.
15730         * doc/tm.texi: Rebuilt.
15731         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
15732         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
15733         * doc/extend.texi (Backwards Compatibility): Clarify it is system
15734         headers affected by extern "C".
15735         * system.h: Poison NO_IMPLICIT_EXTERN_C.
15736         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
15737         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
15738         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
15739         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
15740         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
15741         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
15742         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
15743         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
15744         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
15745         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
15746         config/sparc/sp64-elf.h, config/spu/spu.h,
15747         config/stormy16/stormy16.h, config/v850/v850.h,
15748         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
15749         define NO_IMPLICIT_EXTERN_C.
15750         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
15752 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
15754         PR target/84711
15755         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
15756         instead of GET_MODE_SIZE when comparing Units.
15758 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
15760         PR target/84711
15761         * rtlanal.c (set_noop_p): Constrain on mode change,
15762         include hard-reg-set.h
15764 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
15766         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
15768 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
15770         Revert
15771         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
15773         PR sanitizer/84250
15774         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
15775         libasan.
15776         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
15778 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
15780         PR sanitizer/84250
15781         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
15782         libasan.
15783         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
15785 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
15787         PR middle-end/86380
15788         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
15790 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
15792         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
15793         neg_*_op* variables.
15795 2018-07-04  Martin Liska  <mliska@suse.cz>
15797         * tree-switch-conversion.c: Define
15798         max_ratio_for_speed and max_ratio_for_size constants.
15800 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
15801             Martin Liska  <mliska@suse.cz>
15803         PR middle-end/66240
15804         PR target/45996
15805         PR c/84100
15806         * common.opt: Rename align options with 'str_' prefix.
15807         * common/config/i386/i386-common.c (set_malign_value): New
15808         function.
15809         (ix86_handle_option): Use it to set -falign-* options/
15810         * config/aarch64/aarch64-protos.h (struct tune_params): Change
15811         type from int to string.
15812         * config/aarch64/aarch64.c: Update default values from int
15813         to string.
15814         * config/alpha/alpha.c (alpha_override_options_after_change):
15815         Likewise.
15816         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
15817         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15818         max skip conditionally.
15819         * config/i386/freebsd.h (SUBALIGN_LOG): New.
15820         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15821         max skip conditionally.
15822         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15823         max skip conditionally.
15824         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
15825         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15826         max skip conditionally.
15827         * config/i386/i386.c (struct ptt): Change type from int to
15828         string.
15829         (ix86_default_align): Set default values.
15830         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
15831         max skip conditionally.
15832         * config/i386/iamcu.h (SUBALIGN_LOG): New.
15833         (ASM_OUTPUT_MAX_SKIP_ALIGN):
15834         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
15835         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15836         max skip conditionally.
15837         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
15838         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
15839         * config/i386/x86-64.h (SUBALIGN_LOG): New.
15840         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15841         max skip conditionally.
15842         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15843         * config/ia64/ia64.c (ia64_option_override): Set default values
15844         for alignment options.
15845         * config/m68k/m68k.c: Handle new str_align_* options.
15846         * config/mips/mips.c (mips_set_compression_mode): Change
15847         type of constants.
15848         (mips_option_override): Set default values for options.
15849         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
15850         Likewise.
15851         * config/rs6000/rs6000.c (rs6000_option_override_internal):
15852         Likewise.
15853         * config/rx/rx.c (rx_option_override): Likewise.
15854         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
15855         (LABEL_ALIGN): Use align_labels_log.
15856         (LOOP_ALIGN): Use align_loops_align.
15857         * config/s390/s390.c (s390_asm_output_function_label): Use new
15858         macros.
15859         * config/sh/sh.c (sh_override_options_after_change):
15860         Change type of constants.
15861         * config/spu/spu.c (spu_sched_init): Likewise.
15862         * config/sparc/sparc.c (sparc_option_override): Set default
15863         values for options.
15864         * config/visium/visium.c (visium_option_override): Likewise.
15865         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
15866         emit p2align format with last argument if it's not needed.
15867         * doc/invoke.texi: Document extended format of -falign-*.
15868         * final.c: Use align_labels alignment.
15869         * flags.h (struct target_flag_state): Change type to use
15870         align_flags.
15871         (struct align_flags_tuple): New.
15872         (struct align_flags): Likewise.
15873         (align_loops_log): Redefine macro to use new types.
15874         (align_loops_max_skip): Redefine macro to use new types.
15875         (align_jumps_log): Redefine macro to use new types.
15876         (align_jumps_max_skip): Redefine macro to use new types.
15877         (align_labels_log): Redefine macro to use new types.
15878         (align_labels_max_skip): Redefine macro to use new types.
15879         (align_functions_log): Redefine macro to use new types.
15880         (align_loops): Redefine macro to use new types.
15881         (align_jumps): Redefine macro to use new types.
15882         (align_labels): Redefine macro to use new types.
15883         (align_functions): Redefine macro to use new types.
15884         (align_functions_max_skip): Redefine macro to use new types.
15885         (align_loops_value): New macro.
15886         (align_jumps_value): New macro.
15887         (align_labels_value): New macro.
15888         (align_functions_value): New macro.
15889         * function.c (invoke_set_current_function_hook): Propagate
15890         alignment values from flags to global variables default in
15891         topleev.h.
15892         * ipa-icf.c (sem_function::equals_wpa): Use
15893         cl_optimization_option_eq instead of memcmp.
15894         * lto-streamer.h (cl_optimization_stream_out): Support streaming
15895         of string types.
15896         (cl_optimization_stream_in): Likewise.
15897         * optc-save-gen.awk: Support strings in cl_optimization.
15898         * opth-gen.awk: Likewise.
15899         * opts.c (finish_options): Remove error checking of invalid
15900         value ranges.
15901         (MAX_CODE_ALIGN): Remove.
15902         (MAX_CODE_ALIGN_VALUE): Likewise.
15903         (parse_and_check_align_values): New function.
15904         (check_alignment_argument): Likewise.
15905         (common_handle_option): Use check_alignment_argument.
15906         * opts.h (parse_and_check_align_values): Declare.
15907         * toplev.c (init_alignments): Remove.
15908         (read_log_maxskip): New.
15909         (parse_N_M): Likewise.
15910         (parse_alignment_opts): Likewise.
15911         (backend_init_target): Remove usage of init_alignments.
15912         * toplev.h (parse_alignment_opts): Declare.
15913         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
15914         argument.
15915         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
15916         * tree.c (cl_option_hasher::equal): New.
15917         * varasm.c: Use new global macros.
15919 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
15921         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
15922         Use a simpler align directive also if MAXSKIP = ALIGN-1.
15923         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15924         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15925         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15926         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15927         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
15928         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
15929         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
15930         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15931         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15932         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15933         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15935 2018-07-04  Martin Liska  <mliska@suse.cz>
15936             Jonathan Wakely  <jwakely@redhat.com>
15938         * coverage.c: Use correct type.
15939         * doc/invoke.texi: Language correction.
15941 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
15943         PR target/85620
15944         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
15945         ENDBRANCH for non-tail call which may return via indirect branch.
15946         * doc/extend.texi: Document indirect_return attribute.
15948 2018-07-03  Martin Sebor  <msebor@redhat.com>
15950         PR tree-optimization/86274
15951         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
15952         precondition.
15953         (format_floating): Correct handling of infinities and NaNs.
15955 2018-07-03  Martin Sebor  <msebor@redhat.com>
15957         * print-tree.c (print_real_cst): New function.
15958         (print_node_brief): Call it.
15959         (print_node): Ditto.
15961 2018-07-03  Jeff Law  <law@redhat.com>
15963         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
15964         into a single pattern.
15966         * config/h8300/h8300.md (ors code_iterator): New.
15967         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
15968         a single pattern and single splitter.
15969         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
15970         (iorqi3_1, xorqi3_1): Likewise.
15971         (iorqi3, xorqi3 expanders): Similarly.
15973         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
15974         (movmd_internal) into a single pattern using the P mode iterator.
15975         (movmd splitters): Similarly.
15976         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
15977         (movsd splitters): Similarly.
15979         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
15980         ADDB, ADDW and ADDL into a single ADD attribute which selects the
15981         right table based on the size of the operand.
15982         * config/h8300/h8300.md (length_table): Corresponding changes. All
15983         references to "addb", "addw" and "addl" changed to "add".
15984         (btst patterns): Merge two variants into a single pattern.
15985         (tstqi, tsthi): Likewise.
15986         (addhi3_incdec, addsi3_incdec): Likewise.
15987         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
15988         (mulhi3, mulsi3): Likewise.
15989         (udivhi3, udivsi3): Likewise.
15990         (divhi3, divsi3): Likewise.
15991         (andorqi3, andorhi3, andorsi3): Likewise.
15993 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
15995         PR target/85694
15996         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
15997         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
15999 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16001         PR tree-optimization/85694
16002         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
16003         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
16004         UNSPEC_URHADD.
16005         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
16006         (<u>avg<mode>3_ceil): New patterns.
16008 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
16010         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
16011         scan-tree-dump directive.
16012         * gcc.dg/vect/slp-perm-2.c: Likewise.
16013         * gcc.dg/vect/slp-perm-3.c: Likewise.
16014         * gcc.dg/vect/slp-perm-5.c: Likewise.
16015         * gcc.dg/vect/slp-perm-6.c: Likewise.
16016         * gcc.dg/vect/slp-perm-7.c: Likewise.
16017         * gcc.dg/vect/slp-perm-8.c: Likewise.
16019 2018-07-03  Marek Polacek  <polacek@redhat.com>
16021         PR middle-end/86202
16022         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
16024 2018-07-03  Richard Biener  <rguenther@suse.de>
16026         PR ipa/86389
16027         * tree-ssa-structalias.c (find_func_clobbers): Properly
16028         handle indirect calls.
16030 2018-07-03  Jeff Law  <law@redhat.com>
16032         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
16033         (shifts): New code iterator.
16034         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
16035         expander.  Fix HImode handling on H8/SX.
16036         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
16037         (subqi3, subhi3, subsi3 expanders): Likewise.
16038         (andqi3, andhi3, andsi3 expanders): Likewise.
16039         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
16040         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
16041         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
16042         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
16043         (zero_extendqihi2, zero_extendqisi2): Likewise.
16044         (extendqihi2, extendqisi2): Likewise.
16045         (rotlqi3, rotlhi3, rotlsi3): Likewise.
16046         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
16047         (rotlqi3_1, rotlhi3_1): Likewise.
16048         (logicalhi3_sn, logicalsi3_sn): Likewise.
16049         (logicalhi3, logicalsi3): Likewise.
16051 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16053         * tree-vect-patterns.c (vect_recog_rotate_pattern)
16054         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
16055         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
16056         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
16057         type to append_pattern_def_seq instead of creating a stmt_vec_info
16058         directly.
16059         (build_mask_conversion): Likewise.  Remove vinfo argument.
16060         (vect_add_conversion_to_patterm): Likewise, renaming to...
16061         (vect_add_conversion_to_pattern): ...this.
16062         (vect_recog_mask_conversion_pattern): Update call to
16063         build_mask_conversion.  Pass the vector type to
16064         append_pattern_def_seq here too.
16065         (vect_recog_gather_scatter_pattern): Update call to
16066         vect_add_conversion_to_pattern.
16068 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16070         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
16071         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16072         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
16073         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
16074         STMT_VINFO_PATTERN_DEF_SEQ to null here.
16075         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
16076         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
16077         append_pattern_def_seq instead of new_pattern_def_seq.
16078         (vect_recog_divmod_pattern): Do both of the above.
16079         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
16080         is null.
16082 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16084         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
16085         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
16086         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
16087         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
16088         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
16089         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
16090         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
16091         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
16092         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
16093         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
16094         parameter with a single stmt_vec_info.
16095         (vect_recog_func_ptr): Likewise.
16096         (vect_recog_gather_scatter_pattern): Likewise, folding in...
16097         (vect_try_gather_scatter_pattern): ...this.
16098         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
16099         the stmt_vec_info of the statement to be matched.  Don't clear
16100         STMT_VINFO_RELATED_STMT.
16101         (vect_pattern_recog): Update call accordingly.
16103 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16105         PR tree-optimization/85694
16106         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
16107         (uavgM3_ceil): Document new optabs.
16108         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
16109         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
16110         functions.
16111         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
16112         (savg_ceil_optab): New optabs.
16113         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
16114         (vect_vect_recog_func_ptrs): Add it.
16115         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
16116         constant directly from the associated lhs.
16118 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16120         * tree-vect-patterns.c (vect_split_statement): New function.
16121         (vect_convert_input): Use it to try to split an existing cast.
16123 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16125         * poly-int.h (print_hex): New function.
16126         * dumpfile.h (dump_dec, dump_hex): Declare.
16127         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
16128         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
16129         min_input_precision, operation_precision and operation_sign.
16130         * tree-vect-patterns.c (vect_get_range_info): New function.
16131         (vect_same_loop_or_bb_p, vect_single_imm_use)
16132         (vect_operation_fits_smaller_type): Delete.
16133         (vect_look_through_possible_promotion): Add an optional
16134         single_use_p parameter.
16135         (vect_recog_over_widening_pattern): Rewrite to use new
16136         stmt_vec_info infomration.  Handle one operation at a time.
16137         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
16138         (vect_truncatable_operation_p, vect_set_operation_type)
16139         (vect_set_min_input_precision): New functions.
16140         (vect_determine_min_output_precision_1): Likewise.
16141         (vect_determine_min_output_precision): Likewise.
16142         (vect_determine_precisions_from_range): Likewise.
16143         (vect_determine_precisions_from_users): Likewise.
16144         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
16145         (vect_vect_recog_func_ptrs): Put over_widening first.
16146         Add cast_forwprop.
16147         (vect_pattern_recog): Call vect_determine_precisions.
16149 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16151         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
16152         statements that have been replaced by further pattern statements.
16153         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
16155 2018-07-03  Richard Biener  <rguenther@suse.de>
16157         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
16158         always set *dt.  Dump vectype in vectype overload.
16159         * dumpfile.h (dump_gimple_expr): New function.
16160         (dump_gimple_expr_loc): Likewise.
16161         * dumpfile.c (dump_gimple_expr): New function.
16162         (dump_gimple_expr_loc): Likewise.
16164 2018-07-02  Jeff Law  <law@redhat.com>
16166         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
16167         the H8/300, H8/300H and H8/S variants into a single pattern.
16168         (movhi_h8300, movqi_h8300hs): Similarly.
16169         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
16170         (QHI mode iterator): New.
16172         * config/h8300/h8300.md: Remove trailing whitespace.
16174 2018-07-02  Jim Wilson  <jimw@sifive.com>
16176         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
16177         instead of emit_insn for interrupt returns.
16178         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
16179         (riscv_sret, riscv_uret): Likewise.
16181 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
16183         * pretty-print.c (selftest::test_pp_format): Move save and restore
16184         of quotes to class auto_fix_quotes, and add an instance.
16185         * selftest.c: Include "intl.h".
16186         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
16187         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
16188         * selftest.h (selftest::auto_fix_quotes): New class.
16190 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
16192         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
16193         (aarch64_sve_prepare_conditional_op): Remove.
16194         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
16195         Allow aarch64_simd_reg_or_zero as select operand; remove
16196         the aarch64_sve_prepare_conditional_op call.
16197         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
16198         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
16199         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
16200         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
16201         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
16202         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
16203         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
16204         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
16205         and a splitters to match all of the *_any patterns.
16206         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
16208         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
16209         (SVE_COND_FP_BINARY_REV): Remove.
16210         (sve_int_op_rev, sve_fp_op_rev): New.
16211         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
16212         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
16213         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
16214         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
16215         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
16216         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
16217         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
16218         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
16219         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
16221         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
16222         Remove match_dup 1 from the inner unspec.
16223         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
16225         * config/aarch64/aarch64.md (movprfx): New attr.
16226         (length): Default movprfx to 8.
16227         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
16228         (*madd<SVE_I>, *msub<SVE_I): Likewise.
16229         (*<su>mul<SVE_I>3_highpart): Likewise.
16230         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
16231         (*v<ASHIFT><SVE_I>3): Likewise.
16232         (*<su><MAXMIN><SVE_I>3): Likewise.
16233         (*<su><MAXMIN><SVE_F>3): Likewise.
16234         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
16235         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
16236         (*div<SVE_F>4): Likewise.
16238 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
16240         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
16241         in dump string.
16243 2018-07-02  Richard Biener  <rguenther@suse.de>
16245         PR tree-optimization/86363
16246         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
16247         memset argument refers to a non-variable address.
16249 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
16251         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
16252         {PLUS,MINUS}_EXPR code to...
16253         (adjust_symbolic_bound): ...here,
16254         (combine_bound): ...here,
16255         (set_value_range_with_overflow): ...and here.
16257 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
16259         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
16260         code...
16261         (extract_range_from_abs_expr): ...here.
16263 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
16265         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
16266         -fno-omit-frame-pointer when not optimizing.
16268 2018-07-02  Martin Liska  <mliska@suse.cz>
16270         PR ipa/86279
16271         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
16272         (propagate_nothrow): Likewise.
16274 2018-07-02  Martin Liska  <mliska@suse.cz>
16276         PR ipa/86323
16277         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
16279 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
16281         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
16282         function in r262149, changing "loc" param from source_location to
16283         const dump_location_t &.
16284         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
16285         declaration, as above.
16287 2018-07-01  Paul Koning  <ni1d@arrl.net>
16289         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
16290         -munit-asm, -mgnu-asm, -mdec-asm.
16291         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
16292         (pdp11_output_labelref): New.
16293         (pdp11_output_def): New.
16294         (pdp11_output_addr_vec_elt): New.
16295         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
16296         %# and %@ format codes.
16297         (pdp11_option_override): New.
16298         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
16299         (pdp11_output_ident): New.
16300         (pdp11_asm_named_section): New.
16301         (pdp11_asm_init_sections): New.
16302         (pdp11_file_start): New.
16303         (pdp11_file_end): New.
16304         (output_ascii): Use .ascii/.asciz for -mdec-asm.
16305         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
16306         %o, like %c but octal.
16307         (pdp11_option_override): New.
16308         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
16309         -mdec-asm.
16310         (DATA_SECTION_ASM_OP): Ditto.
16311         (READONLY_DATA_SECTION_ASM_OP): New.
16312         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
16313         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
16314         (ASM_OUTPUT_LABELREF): Ditto.
16315         (ASM_OUTPUT_DEF): Ditto.
16316         (ASM_OUTPUT_EXTERNAL): New.
16317         (ASM_OUTPUT_SOURCE_FILENAME): New.
16318         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
16319         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
16320         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
16321         %# and %@ format codes.
16322         * config/pdp11/pdp11.opt (mgnu-asm): New.
16323         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
16324         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
16325         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
16327 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
16329         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
16330         dereferencing path[] beyond its length.
16331         (debug_path): New.
16332         (debug_all_paths): New.
16333         (rewire_first_differing_edge): New.
16334         (adjust_paths_after_duplication): New.
16335         (duplicate_thread_path): Call adjust_paths_after_duplication.
16336         Add new argument.
16337         (thread_through_all_blocks): Add new argument to
16338         duplicate_thread_path.
16340 2018-06-30  Jim Wilson  <jimw@sifive.com>
16342         * config/riscv/predicates.md (p2m1_shift_operand): New.
16343         (high_mask_shift_operand): New.
16344         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
16345         pattern using p2m1_shift_operand.
16346         (lshsi3_zero_extend_3+2): New combiner pattern using
16347         high_mask_shift_operand.
16349 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16351         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
16352         split out from...
16353         (vect_recog_rotate_pattern): ...here.
16354         (vect_convert_input): Try to insert casts of invariants in the
16355         preheader.
16356         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
16357         preheader to be empty.
16359 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16361         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
16362         vector type.  If given, install it in the new statement's
16363         STMT_VINFO_VECTYPE.
16364         (vect_element_precision): New function.
16365         (vect_unpromoted_value): New struct.
16366         (vect_unpromoted_value::vect_unpromoted_value): New function.
16367         (vect_unpromoted_value::set_op): Likewise.
16368         (vect_look_through_possible_promotion): Likewise.
16369         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
16370         (vect_widened_op_tree, vect_convert_input): Likewise.
16371         (vect_convert_inputs, vect_convert_output): Likewise.
16372         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
16373         to handle the optional cast of the multiplication result and
16374         vect_widened_op_tree to detect the widened multiplication itself.
16375         Do not require the input and output of promotion casts to have
16376         the same sign, but base the signedness of the operation on the
16377         input rather than the result.  If the pattern includes two
16378         promotions, check that those promotions have the same sign.
16379         Do not restrict the MULT_EXPR handling to a double-width result;
16380         handle quadruple-width results and wider.  Use vect_convert_inputs
16381         to convert the inputs to the common type.
16382         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
16383         to handle the optional cast of the ABS result.  Also allow a sign
16384         change or a sign extension between the ABS and MINUS.
16385         Use vect_widened_op_tree to detect the widened subtraction and use
16386         vect_convert_inputs to convert the inputs to the common type.
16387         (vect_handle_widen_op_by_const): Delete.
16388         (vect_recog_widen_op_pattern): New function.
16389         (vect_recog_widen_mult_pattern): Use it.
16390         (vect_recog_widen_shift_pattern): Likewise.
16391         (vect_recog_widen_sum_pattern): Use
16392         vect_look_through_possible_promotion to handle the promoted
16393         PLUS_EXPR operand.
16395 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16397         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
16398         the containing gimple_seq *.
16399         * gimple-iterator.h (gsi_for_stmt): Declare it.
16400         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
16401         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
16402         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
16403         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
16404         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
16405         checks.
16406         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
16407         split out from...
16408         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
16409         statement being replaced is part of an existing pattern
16410         definition sequence, inserting the new pattern statements before
16411         the original one.
16412         (vect_pattern_recog_1): Don't return a bool.  If the statement
16413         is already part of a pattern, instead apply pattern matching
16414         to the pattern definition statements.  Don't clear the
16415         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
16416         (vect_pattern_recog): Don't break after the first match;
16417         continue processing the pattern definition statements instead.
16418         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
16420 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16422         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16423         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16424         (vect_recog_widen_sum_pattern): Use it.
16426 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16428         * tree-vect-loop.c (vectorizable_reduction): Assert that the
16429         phi is not a pattern statement and has not been replaced by
16430         a pattern statement.
16431         * tree-vect-patterns.c (type_conversion_p): Don't check
16432         STMT_VINFO_IN_PATTERN_P.
16433         (vect_recog_vector_vector_shift_pattern): Likewise.
16434         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
16435         the pattern statement rather than the original statement; check
16436         directly for a WIDEN_MULT_EXPR here.
16437         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
16438         vect_is_simple_use to return the pattern statement rather
16439         than the original statement; use is_pattern_stmt_p to check
16440         for such a pattern statement.
16441         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
16442         to return the pattern statement rather than the original statement;
16443         don't do the same transformation here.
16444         (vect_is_simple_use): If the defining statement has been replaced
16445         by a pattern statement, return the pattern statement instead.
16446         Remove the corresponding (local) transformation from the vectype
16447         overload.
16449 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16451         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
16452         end and default to null.
16453         * tree-vect-loop.c (vect_create_epilog_for_reduction)
16454         (vectorizable_reduction): Update calls accordingly, dropping the
16455         gimple ** argument if the passed-back statement isn't needed.
16456         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
16457         (vect_recog_rotate_pattern): Likewise.
16458         (vect_recog_mask_conversion_pattern): Likewise.
16459         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
16460         (vect_mask_constant_operand_p): Likewise.
16461         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
16462         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
16463         (get_group_load_store_type, get_load_store_type): Likewise.
16464         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
16465         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
16466         (vectorizable_conversion, vectorizable_assignment): Likewise.
16467         (vectorizable_shift, vectorizable_operation): Likewise.
16468         (vectorizable_store, vect_is_simple_cond): Likewise.
16469         (vectorizable_condition, vectorizable_comparison): Likewise.
16470         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
16471         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
16472         and move it to the end.  Cope with null def_stmt_outs.
16474 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16476         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
16478 2018-06-29  Jeff Law  <law@redhat.com>
16480         * config/v850/v850.c (v850_legitimate_address_p): Handle large
16481         displacements for TARGET_V850E2V3 and newer.
16482         (TARGET_LRA_P): Remove.  Defaults to LRA now.
16483         * config/v850/v850.md (sign23byte_load): Remove.
16484         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
16485         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
16487 2018-06-29  Martin Liska  <mliska@suse.cz>
16489         PR lto/85759
16490         * coverage.c (coverage_init): Mangle full path name.
16491         * doc/invoke.texi: Document the change.
16492         * gcov-io.c (mangle_path): New.
16493         * gcov-io.h (mangle_path): Likewise.
16494         * gcov.c (mangle_name): Use mangle_path for path mangling.
16496 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16498         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
16499         if starting source register is not even.
16501 2018-06-29  Martin Liska  <mliska@suse.cz>
16503         PR tree-optimization/86263
16504         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
16505         Make edge redirection.
16507 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
16509         * dumpfile.c (dump_loc): Add indentation based on scope depth.
16510         (dump_scope_depth): New variable.
16511         (get_dump_scope_depth): New function.
16512         (dump_begin_scope): New function.
16513         (dump_end_scope): New function.
16514         * dumpfile.h (get_dump_scope_depth): New declaration.
16515         (dump_begin_scope): New declaration.
16516         (dump_end_scope): New declaration.
16517         (class auto_dump_scope): New class.
16518         (AUTO_DUMP_SCOPE): New macro.
16519         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
16520         AUTO_DUMP_SCOPE.
16522 2018-06-29  Richard Biener  <rguenther@suse.de>
16524         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
16525         compute_all_dependences succeeds.
16526         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
16527         exceed --param loop-max-datarefs-for-datadeps.
16529 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
16531         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
16533 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
16535         PR target/86348
16536         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
16537         alternative 0 in preferred_for_speed attribute.
16539 2018-06-28  Paul Koning  <ni1d@arrl.net>
16541         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
16542         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
16543         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
16544         * config/pdp11/pdp11.md: Correct "length" attribute calculation
16545         for shift insn patterns.
16547 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
16549         * cgraph.c (cgraph_node::get_body): Replace assignments to
16550         "dump_file" with calls to set_dump_file.
16551         * dumpfile.c (alt_dump_file): Make static, and group with...
16552         (alt_flags): ...this definition.
16553         (dumps_are_enabled): New variable.
16554         (refresh_dumps_are_enabled): New function.
16555         (set_dump_file): New function.
16556         (set_alt_dump_file): New function.
16557         (gcc::dump_manager::dump_start): Replace assignments to
16558         "dump_file" and "alt_dump_file" with calls to set_dump_file and
16559         set_alt_dump_file.
16560         (gcc::dump_manager::dump_finish): Likewise.
16561         * dumpfile.h (alt_dump_file): Delete decl.
16562         (dumps_are_enabled): New variable decl.
16563         (set_dump_file): New function decl.
16564         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
16565         global.
16566         * tree-nested.c (lower_nested_functions): Replace assignments to
16567         "dump_file" with calls to set_dump_file.
16569 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
16571         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
16572         goto_locus of each outgoing edge of each basic block.
16574 2018-06-28  Richard Biener  <rguenther@suse.de>
16576         * dwarf2out.c (decl_scope_table): Remove.
16577         (push_decl_scope): Likewise.
16578         (pop_decl_scope): Likewise.
16579         (gen_type_die_for_member): Do not call push/pop_decl_scope.
16580         (gen_struct_or_union_type_die): Likewise.
16581         (gen_tagged_type_die): Likewise.
16582         (dwarf2out_init): Do not initialize decl_scope_table.
16583         (dwarf2out_c_finalize): Do not free it.
16585 2018-06-28  Richard Biener  <rguenther@suse.de>
16587         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
16588         deciding whether to not re-use a DIE.
16590 2018-06-28  Richard Biener  <rguenther@suse.de>
16592         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
16593         DW_AT_abstract_origin attribute.
16595 2018-06-28  Martin Liska  <mliska@suse.cz>
16597         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
16598         Use newly introduced constants.
16599         * tree-switch-conversion.h (struct jump_table_cluster):
16600         Define max_ratio_for_size and max_ratio_for_speed.
16602 2018-06-28  Martin Liska  <mliska@suse.cz>
16604         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
16605         Add new checking assert to catch invalid state.
16606         (jump_table_cluster::can_be_handled): Handle single case
16607         clusters.
16608         (jump_table_cluster::is_beneficial): Bail out for such case.
16609         (bit_test_cluster::find_bit_tests):
16610         Add new checking assert to catch invalid state.
16611         (bit_test_cluster::can_be_handled): Handle single case
16612         clusters.
16613         (bit_test_cluster::is_beneficial): Bail out for such case.
16614         (switch_decision_tree::analyze_switch_statement):
16615         Fix comment.
16617 2018-06-28  Martin Liska  <mliska@suse.cz>
16619         * common.opt: Introduce -completion option.
16620         * gcc.c (driver_handle_option): Handle it.
16621         (driver::main): Print completions if completion
16622         is set.
16623         * opt-suggestions.c (option_proposer::get_completions):
16624         New function.
16625         (option_proposer::suggest_completion): Likewise.
16626         (option_proposer::find_param_completions): Likewise.
16627         (verify_autocompletions): Likewise.
16628         (test_completion_valid_options): Likewise.
16629         (test_completion_valid_params): Likewise.
16630         (in_completion_p): Likewise.
16631         (empty_completion_p): Likewise.
16632         (test_completion_partial_match): Likewise.
16633         (test_completion_garbage): Likewise.
16634         (opt_proposer_c_tests): Likewise.
16635         * opt-suggestions.h: Declare new functions.
16636         * opts.c (common_handle_option): Handle OPT__completion_.
16637         * selftest-run-tests.c (selftest::run_tests): Add
16638         opt_proposer_c_tests.
16639         * selftest.c (assert_str_startswith): New.
16640         * selftest.h (assert_str_startswith): Likewise.
16641         (opt_proposer_c_tests): New.
16642         (ASSERT_STR_STARTSWITH): Likewise.
16644 2018-06-28  Martin Liska  <mliska@suse.cz>
16646         * Makefile.in: Add opt-suggestions.o.
16647         * gcc-main.c: Include opt-suggestions.h.
16648         * gcc.c (driver::driver): Likewise.
16649         (driver::~driver): Remove m_option_suggestions.
16650         (driver::build_option_suggestions): Moved to option_proposer.
16651         (driver::suggest_option): Likewise.
16652         (driver::handle_unrecognized_options): Use option_proposer.
16653         * gcc.h (class driver): Add new memver m_option_proposer.
16654         * opt-suggestions.c: New file.
16655         * opt-suggestions.h: New file.
16657 2018-06-28  Martin Liska  <mliska@suse.cz>
16659         * vec.h (class auto_string_vec): New (moved from auto_argvec).
16660         (auto_string_vec::~auto_string_vec): Likewise.
16662 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
16664         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
16665         prevent_decl_creation_for_types fields up and add reset_location field.
16666         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
16667         statement if id->reset_location is true.
16668         (copy_edges_for_bb): Do not set goto_locus on the new edges if
16669         id->reset_location is true.
16670         (copy_phis_for_bb): Force input_location on the arguments if
16671         id->reset_location is true.
16672         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
16673         is set on the function to be inlined.
16675 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
16677         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
16679 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
16681         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
16682         registers for Pmode.
16683         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
16684         hard registers for the clobbered pseudo.
16686 2018-06-27  Paul Koning  <ni1d@arrl.net>
16688         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
16689         mutually exclusive options.
16690         * config/pdp11/constraints.md (h): New constraint.
16691         (O): Update definition to match shift code generation.
16692         (D): New constraint.
16693         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
16694         (CCFP): Remove.
16695         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
16696         function.
16697         (output_jump): Change arguments.
16698         (pdp11_fixed_cc_regs): New function.
16699         (pdp11_cc_mode): Ditto.
16700         (pdp11_expand_shift): Ditto.
16701         (pdp11_assemble_shift): Ditto.
16702         (pdp11_small_shift): Ditto.
16703         (pdp11_branch_cost): Remove.
16704         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
16705         from output.
16706         (pdp11_register_move_cost): Update for CC registers.
16707         (pdp11_rtx_costs): Add case for LSHIFTRT.
16708         (pdp11_output_jump): Add CCNZ mode conditional branches.
16709         (notice_update_cc_on_set): Remove.
16710         (pdp11_cc_mode): New function.
16711         (simple_memory_operand): Correct pre/post decrement case.
16712         (no_side_effect_operand): New function.
16713         (pdp11_regno_reg_class): Add CC_REGS class.
16714         (pdp11_fixed_cc_regs): New function.
16715         (pdp11_small_shift): New function.
16716         (pdp11_expand_shift): New function to expand shift insns.
16717         (pdp11_assemble_shift): New function to output shifts.
16718         (pdp11_branch_cost): Remove.
16719         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
16720         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
16721         (WCHAR_TYPE): Ditto.
16722         (PTRDIFF_TYPE): Ditto.
16723         (ADJUST_INSN_LENGTH): New macro.
16724         (FIXED_REGISTERS): Add CC registers.
16725         (CALL_USED_REGISTERS): Ditto.
16726         (reg_class): Ditto.
16727         (REG_CLASS_NAMES): Ditto.
16728         (REG_CLASS_CONTENTS): Ditto.
16729         (SELECT_CC_MODE): Use new function.
16730         (TARGET_FLAGS_REGNUM): New macro.
16731         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
16732         (cc0_reg_rtx): Remove.
16733         (CC_STATUS_MDEP): Remove.
16734         (CC_STATUS_MDEFP_INIT): Remove.
16735         (CC_IN_FPU): Remove.
16736         (NOTICE_UPDATE_CC): Remove.
16737         (REGISTER_NAMES): Add CC registers.
16738         (BRANCH_COST): Change to constant 1.
16739         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
16740         handling.
16741         * config/pdp11/pdp11.opt (mbcopy): Remove.
16742         (mbcopy-builtin): Remove.
16743         (mbranch-cheap): Remove.
16744         (mbranch-expensive): Remove.
16745         * config/pdp11/predicates.md (expand_shift_operand): Update to
16746         match shift code generation.
16747         (ccnz_operator): New predicate.
16748         * doc/invoke.texi (PDP-11 Options): Remove deleted options
16749         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
16750         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
16751         exclusive options.
16752         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
16753         description of O constraint.
16755 2018-06-27  Jeff Law  <law@redhat.com>
16756             Austin Law  <austinklaw@gmail.com>
16758         * config/v850/v850.md (addsi3_set_flags): New pattern.
16759         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
16760         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
16761         (zero_extendhisi2_v850_set_flags): Likewise.
16762         (zero_extendqisi2_v850_set_flags): Likewise.
16763         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
16764         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
16765         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
16767         * config/v850/v850-protos.h (notice_update_cc): Remove.
16768         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
16769         (v850_print_operand): Handle 'D' and "d".
16770         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
16771         Add handling of arithmetic/logical operations compared against zero.
16772         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
16773         Do not look at v850_compare_op, instead get mode from last argument.
16774         (v850_gen_compare): Remove
16775         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
16776         after reload for prologue insns.
16777         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
16778         patterns.
16779         (construct_save_jarl): Likewise.
16780         (TARGET_FLAGS_REGNUM): Define.
16781         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
16782         (NOTICE_UPDATE_CC): Remove.
16783         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
16784         than cc0.  Conditionalize on reload_completed.
16785         (cmpsi_insn, setfcc_insn): Likewise.
16786         (tst1 splitter): Turn into define_and_split which sets the flags
16787         after reload.
16788         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
16789         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
16790         (cstoresf4, cstoredf4): Clobber the flags.
16791         (cmpsi, cmpsf, cmpdf): Remove expanders.
16792         (setf_insn): Remove pattern.
16793         (addsi3): Turn into define_and_split which clobbers the flags after
16794         reload and a suitable pattern (addsi3_clobber_flags) for use after
16795         reload.
16796         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
16797         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
16798         (ashrsi3, ashrsi3_v850e2): Likewise.
16799         (bins): Clobber the flags.
16800         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
16801         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
16802         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
16803         (call_value_internal_short, call_value_internal_long): Likewise.
16804         (callt_save_interrupt, callt_return_interrupt): Likewise.
16805         (save_interrupt, return_interrupt): Likewise.
16806         (callt_save_all_interrupt, save_all_interrupt): Likewise.
16807         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
16808         (restore_all_interrupt, _restore_all_interrupt): Likewise.
16809         (All FP comparisons): Only allow after reload has completed.
16810         (trfsr): Likewise.
16811         (divh, divhu): Tweak output template.
16812         (branch_z_normal, branch_z_invert): Remove
16813         (branch_nz_normal, branch_nz_invert): Likewise.
16814         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
16816         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
16817         * config/v850/v850.c (notice_update_cc): Remove.
16818         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
16819         (CC_NO_CARRY): Likewise.
16820         (NOTICE_UPDATE_CC): Define to nothing.
16821         * config/v850/v850.md: Remove block comment on cc0 handling
16822         Remove "cc" attribute from all patterns.  Remove cc_status handling
16823         from all patterns.  Minor formatting fixes.
16825 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16827         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
16828         (cortex-a76.cortex-a55): Likewise.
16829         * config/aarch64/aarch64-tune.md: Regenerate.
16830         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
16831         cortex-a76.cortex-a55.
16833 2018-06-27  Jeff Law  <law@redhat.com>
16835         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
16836         (MULTILIB_DIRNAMES): Similarly.
16838 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
16840         * gimple.h (gimple_return_retbnd): Delete.
16841         (gimple_return_set_retbnd): Likewise.
16842         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
16843         gimple_return_set_retbnd.
16844         * gimple-pretty-print.c (dump_gimple_return): Remove call to
16845         gimple_return_retbnd and adjust.
16846         * tree-inline.h (struct copy_body_data): Remove retbnd field.
16847         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
16848         Explicitly return NULL in a couple more cases.  Move assertion
16849         on debug statements and remove unreachable code.
16850         (reset_debug_binding): Do not test id->retbnd.
16851         (expand_call_inline): Do not set it.
16853 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
16855         * configure.ac: Add --disable-gcov option.
16856         * configure: Regenerate.
16857         * Makefile.in: Honour @enable_gcov@.
16858         * doc/install.texi: Document --disable-gcov.
16860 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16862         * config/arm/arm-cpus.in (cortex-a76): New entry.
16863         (cortex-a76.cortex-a55): Likewise.
16864         * config/arm/arm-tables.opt: Regenerate.
16865         * config/arm/arm-tune.md: Likewise.
16866         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
16867         * doc/invoke.texi (ARM Options): Document cortex-a76 and
16868         cortex-a76.cortex-a55.
16870 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
16872         PR target/85769
16873         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
16875 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
16877         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
16878         comment.
16879         (EPILOGUE_USES): Likewise.
16881 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
16883         * tree-inline.c (remap_location): New function extracted from...
16884         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
16885         (copy_phis_for_bb): ...here.  Call remap_location.
16886         (copy_cfg_body): Adjust call to copy_edges_for_bb.
16888 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
16890         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
16891         unaligned vsx for 16B memset.
16893 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
16895         PR target/86285
16896         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
16897         ieee128_float_type_node to long_double_type_node unless
16898         TARGET_LONG_DOUBLE_128 is set.
16900 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
16902         * cfgloop.c (get_loop_location): Convert return type from
16903         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
16904         by implicit construction from rtx_insn *, and using
16905         dump_user_location_t::from_function_decl for the fallback case.
16906         * cfgloop.h (get_loop_location): Convert return type from
16907         location_t to dump_user_location_t.
16908         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
16909         dump_printf_loc to pass in a dump_location_t rather than a
16910         location_t, via the gimple stmt.
16911         * coverage.c (get_coverage_counts): Update calls to
16912         dump_printf_loc to pass in dump_location_t rather than a
16913         location_t.
16914         * doc/optinfo.texi (Dump types): Convert example of
16915         dump_printf_loc from taking "locus" to taking "insn".  Update
16916         description of the "_loc" calls to cover dump_location_t.
16917         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
16918         "selftest.h".
16919         (dump_user_location_t::dump_user_location_t): New constructors,
16920         from gimple *stmt and rtx_insn *.
16921         (dump_user_location_t::from_function_decl): New function.
16922         (dump_loc): Make static.
16923         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
16924         const dump_location_t &.
16925         (dump_generic_expr_loc): Delete.
16926         (dump_printf_loc): Convert param "loc" from location_t to
16927         const dump_location_t &.
16928         (selftest::test_impl_location): New function.
16929         (selftest::dumpfile_c_tests): New function.
16930         * dumpfile.h: Include "profile-count.h".
16931         (class dump_user_location_t): New class.
16932         (struct dump_impl_location_t): New struct.
16933         (class dump_location_t): New class.
16934         (dump_printf_loc): Convert 2nd param from source_location to
16935         const dump_location_t &.
16936         (dump_generic_expr_loc): Delete.
16937         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
16938         const dump_location_t &.
16939         * gimple-fold.c (fold_gimple_assign): Update call to
16940         dump_printf_loc to pass in a dump_location_t rather than a
16941         location_t, via the gimple stmt.
16942         (gimple_fold_call): Likewise.
16943         * gimple-loop-interchange.cc
16944         (loop_cand::analyze_iloop_reduction_var): Update for change to
16945         check_reduction_path.
16946         (tree_loop_interchange::interchange): Update for change to
16947         find_loop_location.
16948         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
16949         change in return-type of find_loop_location.
16950         (graphite_regenerate_ast_isl): Likewise.
16951         * graphite-optimize-isl.c (optimize_isl): Likewise.
16952         * graphite.c (graphite_transform_loops): Likewise.
16953         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
16954         pass in a dump_location_t rather than a location_t, via the
16955         gimple stmt.
16956         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16957         * ipa.c (walk_polymorphic_call_targets): Likewise.
16958         * loop-unroll.c (report_unroll): Convert "locus" param from
16959         location_t to dump_location_t.
16960         (decide_unrolling): Update for change to get_loop_location's
16961         return type.
16962         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
16963         location_t to dump_user_location_t.
16964         (grid_find_single_omp_among_assignments_1): Updates calls to
16965         dump_printf_loc to pass in a dump_location_t rather than a
16966         location_t, via the gimple stmt.
16967         (grid_parallel_clauses_gridifiable): Convert "tloc" from
16968         location_t to dump_location_t.  Updates calls to dump_printf_loc
16969         to pass in a dump_location_t rather than a location_t, via the
16970         gimple stmt.
16971         (grid_inner_loop_gridifiable_p): Likewise.
16972         (grid_dist_follows_simple_pattern): Likewise.
16973         (grid_gfor_follows_tiling_pattern): Likewise.
16974         (grid_target_follows_gridifiable_pattern): Likewise.
16975         (grid_attempt_target_gridification): Convert initialization
16976         of local "grid" from memset to zero-initialization; FIXME: does
16977         this require C++11?  Update call to dump_printf_loc to pass in a
16978         optinfo_location rather than a location_t, via the gimple stmt.
16979         * profile.c (read_profile_edge_counts): Updates call to
16980         dump_printf_loc to pass in a dump_location_t rather than a
16981         location_t
16982         (compute_branch_probabilities): Likewise.
16983         * selftest-run-tests.c (selftest::run_tests): Call
16984         dumpfile_c_tests.
16985         * selftest.h (dumpfile_c_tests): New decl.
16986         * tree-loop-distribution.c (pass_loop_distribution::execute):
16987         Update for change in return type of find_loop_location.
16988         * tree-parloops.c (parallelize_loops): Likewise.
16989         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
16990         "locus" from location_t to dump_user_location_t.
16991         (canonicalize_loop_induction_variables): Likewise.
16992         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
16993         for change in return type of find_loop_location.
16994         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
16995         to dump_printf_loc to pass in a dump_location_t rather than a
16996         location_t, via the stmt.
16997         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
16998         Likewise.
16999         * tree-vect-loop-manip.c (find_loop_location): Convert return
17000         type from source_location to dump_user_location_t.
17001         (vect_do_peeling): Update for above change.
17002         (vect_loop_versioning): Update for change in type of
17003         vect_location.
17004         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
17005         from location_t to dump_user_location_t.
17006         (vect_estimate_min_profitable_iters): Update for change in type
17007         of vect_location.
17008         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
17009         location_t to dump_location_t.
17010         (vect_slp_bb): Update for change in type of vect_location.
17011         * tree-vectorizer.c (vect_location): Convert from source_location
17012         to dump_user_location_t.
17013         (try_vectorize_loop_1): Update for change in vect_location's type.
17014         (vectorize_loops): Likewise.
17015         (increase_alignment): Likewise.
17016         * tree-vectorizer.h (vect_location): Convert from source_location
17017         to dump_user_location_t.
17018         (find_loop_location): Convert return type from source_location to
17019         dump_user_location_t.
17020         (check_reduction_path): Convert 1st param from location_t to
17021         dump_user_location_t.
17022         * value-prof.c (check_counter): Update call to dump_printf_loc to
17023         pass in a dump_user_location_t rather than a location_t; update
17024         call to error_at for change in type of "locus".
17025         (check_ic_target): Update call to dump_printf_loc to
17026         pass in a dump_user_location_t rather than a location_t, via the
17027         call_stmt.
17029 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
17031         * config/s390/s390.h (enum processor_flags): Do not use
17032         default tune parameter when -march was specified.
17034 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
17036         PR target/86314
17037         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
17038         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
17039         operands.
17041 2018-06-26  Richard Biener  <rguenther@suse.de>
17043         PR tree-optimization/86287
17044         PR bootstrap/86316
17045         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
17046         (vect_analyze_loop): Initialize n_stmts.
17048 2018-06-26  Richard Biener  <rguenther@suse.de>
17050         PR middle-end/86271
17051         * fold-const.c (fold_convertible_p): Pointer extension
17052         isn't valid.
17054 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
17056         PR debug/86064
17057         * dwarf2out.c (loc_list_has_views): Adjust comments.
17058         (dw_loc_list): Split single cross-partition range with
17059         nonzero locview.
17061 2018-06-25  Jeff Law  <law@redhat.com>
17063         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
17064         on -mbig-switch by default.
17066         * config/v850/predicates.md (const_float_1_operand): Fix match_code
17067         test.
17068         (const_float_0_operand): Remove unused predicate.
17069         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
17070         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
17071         (recipsf2): New expander.  Original pattern now called
17072         (recipsf2_insn).
17073         (recipdf2, recipdf2_insn): Similarly.
17074         (rsqrtsf2, rsqrtsf2_insn): Similarly
17075         (rsqrtdf2, rsqrtdf2_insn): Similarly
17077 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
17079         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
17080         Simplify logic for FreeBSD (twice).
17082 2018-06-25  Martin Sebor  <msebor@redhat.com>
17084         PR tree-optimization/86204
17085         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
17086         a strnlen result if it's less than the length of the string.
17088 2018-06-25  Martin Sebor  <msebor@redhat.com>
17090         PR tree-optimization/85700
17091         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
17092         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
17093         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
17095 2018-06-25  Martin Sebor  <msebor@redhat.com>
17097         * doc/extend.texi (Zero-length arrays): Update and clarify.
17099 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
17101         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
17102         added IEEE/IBM long double multilib support on PowerPC little
17103         endian Linux systems.
17104         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
17105         (MULTILIB_DEFAULTS): Likewise.
17106         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17107         Likewise.
17108         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
17109         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
17110         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
17112 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
17114         PR middle-end/86311
17115         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
17116         (REORDER_45): Likewise.
17118 2018-06-25  Jeff Law  <law@redhat.com>
17120         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
17121         dividend to 32 bits.  Adjust length.
17122         (udivmodhi4): Cleanup output template.  Fix length.
17124 2018-06-25  Carl Love  <cel@us.ibm.com>
17126         * config/rs6000/vsx.md: Change word selector to prefered location.
17128 2018-06-25  Richard Biener  <rguenther@suse.de>
17130         PR tree-optimization/86304
17131         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
17132         epilogue-if-converted loops as well.
17134 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
17136         * lto-section-out.c (lto_begin_section): Do not print section
17137         name for noaddr and unnumbered dumps.
17139 2018-06-25  Richard Biener  <rguenther@suse.de>
17141         * tree-vectorizer.h (struct vec_info_shared): New structure
17142         with parts split out from struct vec_info and loop_nest from
17143         struct _loop_vec_info.
17144         (struct vec_info): Adjust accordingly.
17145         (struct _loop_vec_info): Likewise.
17146         (LOOP_VINFO_LOOP_NEST): Adjust.
17147         (LOOP_VINFO_DATAREFS): Likewise.
17148         (LOOP_VINFO_DDRS): Likewise.
17149         (struct _bb_vec_info): Likewise.
17150         (BB_VINFO_DATAREFS): Likewise.
17151         (BB_VINFO_DDRS): Likewise.
17152         (struct _stmt_vec_info): Add dr_aux member.
17153         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
17154         (DR_MISALIGNMENT_UNINITIALIZED): New.
17155         (set_dr_misalignment): Adjust.
17156         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
17157         (vect_analyze_loop): Adjust prototype.
17158         (vect_analyze_loop_form): Likewise.
17159         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
17160         Compute dependences lazily.
17161         (vect_record_base_alignments): Use shared datarefs/ddrs.
17162         (vect_verify_datarefs_alignment): Likewise.
17163         (vect_analyze_data_refs_alignment): Likewise.
17164         (vect_analyze_data_ref_accesses): Likewise.
17165         (vect_analyze_data_refs): Likewise.
17166         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
17167         constructor parameter for shared part.
17168         (vect_analyze_loop_form): Pass in shared part and adjust.
17169         (vect_analyze_loop_2): Pass in storage for the number of
17170         stmts.  Move loop nest finding to the caller.  Compute
17171         datarefs lazily.
17172         (vect_analyze_loop): Pass in shared part.
17173         (vect_transform_loop): Verify shared datarefs are unchanged.
17174         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
17175         constructor parameter for shared part.
17176         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
17177         (vect_slp_bb): Verify shared datarefs are unchanged before
17178         transform.
17179         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
17180         change.
17181         (new_stmt_vec_info): Initialize DR_AUX misalignment to
17182         DR_MISALIGNMENT_UNINITIALIZED.
17183         * tree-vectorizer.c (vec_info::vec_info): Add constructor
17184         parameter for shared part.
17185         (vec_info::~vec_info): Adjust.
17186         (vec_info_shared::vec_info_shared): New.
17187         (vec_info_shared::~vec_info_shared): Likewise.
17188         (vec_info_shared::save_datarefs): Likewise.
17189         (vec_info_shared::check_datarefs): Likewise.
17190         (try_vectorize_loop_1): Construct shared part live for analyses
17191         of a single loop for multiple vector sizes.
17192         * tree-parloops.c (gather_scalar_reductions): Adjust.
17194 2018-06-25  Richard Biener  <rguenther@suse.de>
17196         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
17197         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
17198         (vect_analyze_data_refs): Remove similar code from here and
17199         simplify accordingly.
17201 2018-06-25  Richard Biener  <rguenther@suse.de>
17203         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
17204         for reverse storage order accesses rather than asserting
17205         they cannot happen here.
17207 2018-06-25  Tom de Vries  <tdevries@suse.de>
17209         PR debug/86257
17210         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
17211         Use data16 instead of .byte for insn prefix.
17213 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
17215         PR C++/86082
17216         * parser.c (make_char_string_pack): Pass this literal chars
17217         through cpp_interpret_string.
17218         (cp_parser_userdef_numeric_literal): Check the result of
17219         make_char_string_pack.
17221 2018-06-24  Maya Rashish  <coypu@sdf.org>
17223         * ginclude/stddef.h: Simplify conditions around avoiding
17224         re-definition of __size_t.
17226 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17228         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
17229         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
17231 2018-06-22  Maya Rashish  <coypu@sdf.org>
17233         * doc/invoke.texi (mno-fancy-math-387): Update for changes
17234         made to OpenBSD and NetBSD through the years.
17236 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17238         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
17239         behavior of vec_pack (vector double, vector double) to match
17240         behavior of vec_float2 (vector double, vector double).
17242 2018-06-22  Olivier Hainque  <hainque@adacore.com>
17244         * gimplify.c (gimplify_function_tree): Prevent creation
17245         of a trampoline for the address of the current function
17246         passed to entry/exit instrumentation hooks.
17248 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
17250         PR target/86222
17251         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
17252         correctly.
17254 2018-06-22  Martin Liska  <mliska@suse.cz>
17256         PR tree-optimization/86263
17257         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17258         Bail out if is_enabled is false.
17259         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
17260         New declaration.
17261         (jump_table_cluster::is_enabled): New function.
17263 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17265         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
17266         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
17267         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
17268         (lto_input_ts_binfo_tree_pointers): Likewise.
17269         * tree-streamer-out.c (streamer_write_tree_bitfields,
17270         write_ts_binfo_tree_pointers): Likewise.
17271         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
17273 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17275         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
17277 2018-06-22  Martin Liska  <mliska@suse.cz>
17279         * symbol-summary.h (get): Make it pure and inline move
17280         functionality from ::get function.
17281         (get): Remove and inline into ::get and ::get_create.
17282         (get_create): Move code from ::get function.
17284 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17286         PR target/85994
17287         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
17288         -x assembler-with-cpp.
17290 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17292         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
17293         _FILE_OFFSET_BITS=64 for C++.
17295 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
17297         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
17298         conversion insn that shows up when pr85657-3.c is compiled using
17299         IEEE 128-bit long double.
17300         (neg<mode>2_internal): Use the correct mode to check whether the
17301         mode is IBM extended.
17302         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
17303         multiply and divide external functions from being created more
17304         than once.
17306 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17308         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
17309         functions.
17310         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
17311         the edge can be forwarded.
17312         (cfg_layout_merge_blocks): Likewise.
17314 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17316         * except.c (finish_eh_generation): Commit edge insertions only after
17317         the EH edges have been redirected from post-landing to landing pads.
17319 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17321         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
17322         create_tmp_var_for to create the FRAME decl.
17323         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
17325 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17327         * tree-inline.c (copy_edges_for_bb): Minor tweak.
17328         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
17329         debug statement when resetting its value.
17330         (expand_call_inline): Copy the locus of the call onto the assignment
17331         of the return value, if any.  Use local variable in more cases.
17333 2018-06-21  Martin Liska  <mliska@suse.cz>
17335         * ipa-pure-const.c (propagate_nothrow): Use
17336         funct_state_summaries->get.
17337         (dump_malloc_lattice): Likewise.
17338         (propagate_malloc): Likewise.
17340 2018-06-21  Richard Biener  <rguenther@suse.de>
17342         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
17343         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
17344         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
17345         comment.
17346         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
17347         BLOCK_ABSTRACT_ORIGIN unconditionally.
17349 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
17351         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
17352         deleting it.
17353         * ipa-reference.c (ipa_reference_c_finalize): Delete
17354         ipa_ref_opt_sum_summaries and set it to NULL.
17356 2018-06-21  Tom de Vries  <tdevries@suse.de>
17358         PR tree-optimization/85859
17359         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
17360         test with comment from bb_no_side_effects_p.
17362 2018-06-21  Richard Biener  <rguenther@suse.de>
17364         PR tree-optimization/86232
17365         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
17366         max for constant niter.
17368 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17370         * config/aarch64/aarch64-simd.md
17371         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
17373 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17375         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
17376         Make opernads of the unspec commutative.
17378 2018-06-21  Richard Biener  <rguenther@suse.de>
17380         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
17381         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
17382         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
17383         (vect_analyze_data_ref_dependence): Re-order checks to deal with
17384         NULL DR_STEP.
17385         (vect_record_base_alignments): Do not record base alignment
17386         for gathers or scatters.
17387         (vect_compute_data_ref_alignment): Drop return value that is always
17388         true.  Bail out early for gathers or scatters.
17389         (vect_enhance_data_refs_alignment): Bail out early for gathers
17390         or scatters.
17391         (vect_find_same_alignment_drs): Likewise.
17392         (vect_analyze_data_refs_alignment): Remove dead code.
17393         (vect_slp_analyze_and_verify_node_alignment): Likewise.
17394         (vect_analyze_data_refs): For possible gathers or scatters do
17395         not create an alternate DR, just check their possible validity
17396         and mark them.  Adjust DECL_NONALIASED handling to not rely
17397         on DR_BASE_ADDRESS.
17398         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
17399         update inits of gathers or scatters.
17400         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
17401         Also copy gather/scatter flag to pattern vinfo.
17403 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17405         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
17406         behavior of vec_packsu (vector unsigned long long, vector unsigned
17407         long long) to match behavior of vec_packs with same signature.
17409 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
17410             Thomas Schwinge <thomas@codesourcery.com>
17411             Cesar Philippidis  <cesar@codesourcery.com>
17413         * gimplify.c (gimplify_scan_omp_clauses): Add support for
17414         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
17415         (gimplify_adjust_omp_clauses): Likewise.
17416         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
17417         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
17418         (gimplify_omp_target_update): Update handling of acc update and
17419         enter/exit data.
17420         * omp-low.c (install_var_field): Remove unused parameter
17421         base_pointers_restrict.
17422         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
17423         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
17424         FINALIZE}
17425         (omp_target_base_pointers_restrict_p): Delete.
17426         (scan_omp_target): Update call to scan_sharing_clauses.
17427         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
17428         FINALIZE}.
17429         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
17430         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
17431         (convert_local_omp_clauses): Likewise.
17432         * tree-pretty-print.c (dump_omp_clause): Likewise.
17433         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
17434         FINALIZE}.
17435         (omp_clause_code_name): Likewise.
17437 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
17439         PR debug/86194
17440         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
17441         be narrowed.
17443         PR tree-optimization/86231
17444         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
17445         anti-range don't overwrite *vr0min before using it to compute *vr0max.
17447 2018-06-20  Tom de Vries  <tdevries@suse.de>
17449         PR tree-optimization/86097
17450         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
17451         iv type if signedness of iv type is not the same as that of *nit.
17453 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
17455         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
17456         EDGE_EH edges, verify they are all EDGE_EH.
17458 2018-06-20  Maya Rashish  <coypu@sdf.org>
17460         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
17462 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17464         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
17465         * config/aarch64/aarch64.c (xgene1_tunings): Add
17466         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
17467         (aarch64_mode_valid_for_sched_fusion_p):
17468         Allow 16-byte modes.
17469         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
17470         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
17471         128-bit modes.
17472         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
17473         New pattern.
17474         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
17475         * config/aarch64/iterators.md (VQ2): New mode iterator.
17477 2018-06-20  Martin Liska  <mliska@suse.cz>
17479         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
17480         Change default ratio from 10 to 8.
17482 2018-06-20  Martin Liska  <mliska@suse.cz>
17484         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17485         New.
17486         (bit_test_cluster::find_bit_tests): Likewise.
17487         (switch_decision_tree::analyze_switch_statement): Find clusters.
17488         * tree-switch-conversion.h (struct jump_table_cluster): Document
17489         hierarchy.
17491 2018-06-20  Martin Liska  <mliska@suse.cz>
17493         * tree-switch-conversion.c (switch_conversion::collect):
17494         Record m_uniq property.
17495         (switch_conversion::expand): Bail out for special conditions.
17496         (group_cluster::~group_cluster): New.
17497         (group_cluster::group_cluster): Likewise.
17498         (group_cluster::dump): Likewise.
17499         (jump_table_cluster::emit): New.
17500         (switch_decision_tree::fix_phi_operands_for_edges): New.
17501         (struct case_node): Remove struct.
17502         (jump_table_cluster::can_be_handled): New.
17503         (case_values_threshold): Moved to header.
17504         (reset_out_edges_aux): Likewise.
17505         (jump_table_cluster::is_beneficial): New.
17506         (bit_test_cluster::can_be_handled): Likewise.
17507         (add_case_node): Remove.
17508         (bit_test_cluster::is_beneficial): New.
17509         (case_bit_test::cmp): New.
17510         (bit_test_cluster::emit): New.
17511         (expand_switch_as_decision_tree_p): Remove.
17512         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
17513         (fix_phi_operands_for_edge): Likewise.
17514         (switch_decision_tree::analyze_switch_statement): New.
17515         (compute_cases_per_edge): Move ...
17516         (switch_decision_tree::compute_cases_per_edge): ... here.
17517         (try_switch_expansion): Likewise.
17518         (switch_decision_tree::try_switch_expansion): Likewise.
17519         (record_phi_operand_mapping): Likewise.
17520         (switch_decision_tree::record_phi_operand_mapping): Likewise.
17521         (emit_case_decision_tree): Likewise.
17522         (switch_decision_tree::emit): Likewise.
17523         (balance_case_nodes): Likewise.
17524         (switch_decision_tree::balance_case_nodes): Likewise.
17525         (dump_case_nodes): Likewise.
17526         (switch_decision_tree::dump_case_nodes): Likewise.
17527         (emit_jump): Likewise.
17528         (switch_decision_tree::emit_jump): Likewise.
17529         (emit_cmp_and_jump_insns): Likewise.
17530         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
17531         (emit_case_nodes): Likewise.
17532         (switch_decision_tree::emit_case_nodes): Likewise.
17533         (conditional_probability): Remove.
17534         * tree-switch-conversion.h (enum cluster_type): New.
17535         (PRINT_CASE): New.
17536         (struct cluster): Likewise.
17537         (cluster::cluster): Likewise.
17538         (struct simple_cluster): Likewise.
17539         (simple_cluster::simple_cluster): Likewise.
17540         (struct group_cluster): Likewise.
17541         (struct jump_table_cluster): Likewise.
17542         (struct bit_test_cluster): Likewise.
17543         (struct min_cluster_item): Likewise.
17544         (struct case_tree_node): Likewise.
17545         (case_tree_node::case_tree_node): Likewise.
17546         (jump_table_cluster::case_values_threshold): Likewise.
17547         (struct case_bit_test): Likewise.
17548         (struct switch_decision_tree): Likewise.
17549         (struct switch_conversion): Likewise.
17550         (switch_decision_tree::reset_out_edges_aux): Likewise.
17552 2018-06-20  Martin Liska  <mliska@suse.cz>
17554         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
17555         (hoist_edge_and_branch_if_true): Likewise.
17556         (expand_switch_using_bit_tests_p): Likewise.
17557         (struct case_bit_test): Likewise.
17558         (case_bit_test_cmp): Likewise.
17559         (emit_case_bit_tests): Likewise.
17560         (switch_conversion::switch_conversion): New class.
17561         (struct switch_conv_info): Remove old struct.
17562         (collect_switch_conv_info): More to ...
17563         (switch_conversion::collect): ... this.
17564         (check_range): Likewise.
17565         (switch_conversion::check_range): Likewise.
17566         (check_all_empty_except_final): Likewise.
17567         (switch_conversion::check_all_empty_except_final): Likewise.
17568         (check_final_bb): Likewise.
17569         (switch_conversion::check_final_bb): Likewise.
17570         (create_temp_arrays): Likewise.
17571         (switch_conversion::create_temp_arrays): Likewise.
17572         (free_temp_arrays): Likewise.
17573         (gather_default_values): Likewise.
17574         (switch_conversion::gather_default_values): Likewise.
17575         (build_constructors): Likewise.
17576         (switch_conversion::build_constructors): Likewise.
17577         (constructor_contains_same_values_p): Likewise.
17578         (switch_conversion::contains_same_values_p): Likewise.
17579         (array_value_type): Likewise.
17580         (switch_conversion::array_value_type): Likewise.
17581         (build_one_array): Likewise.
17582         (switch_conversion::build_one_array): Likewise.
17583         (build_arrays): Likewise.
17584         (switch_conversion::build_arrays): Likewise.
17585         (gen_def_assigns): Likewise.
17586         (switch_conversion::gen_def_assigns): Likewise.
17587         (prune_bbs): Likewise.
17588         (switch_conversion::prune_bbs): Likewise.
17589         (fix_phi_nodes): Likewise.
17590         (switch_conversion::fix_phi_nodes): Likewise.
17591         (gen_inbound_check): Likewise.
17592         (switch_conversion::gen_inbound_check): Likewise.
17593         (process_switch): Use the newly created class.
17594         (switch_conversion::expand): New.
17595         (switch_conversion::~switch_conversion): New.
17596         * tree-switch-conversion.h: New file.
17598 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17600         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
17601         tree-vect-patterns.c.
17602         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
17603         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
17604         (vect_recog_sad_pattern): Likewise.
17605         (vect_recog_widen_sum_pattern): Likewise.
17606         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
17607         (vect_recog_widen_shift_pattern): Remove the type_in argument.
17608         (vect_recog_rotate_pattern): Likewise.
17609         (vect_recog_mult_pattern): Likewise.
17610         (vect_recog_vector_vector_shift_pattern): Likewise.
17611         (vect_recog_divmod_pattern): Likewise.
17612         (vect_recog_mixed_size_cond_pattern): Likewise.
17613         (vect_recog_bool_pattern): Likewise.
17614         (vect_recog_mask_conversion_pattern): Likewise.
17615         (vect_try_gather_scatter_pattern): Likewise.
17616         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
17617         (vect_recog_over_widening_pattern): Likewise.
17618         (vect_recog_gather_scatter_pattern): Likewise.
17619         (vect_recog_func_ptr): Move from tree-vectorizer.h
17620         (vect_vect_recog_func_ptrs): Move further down the file.
17621         (vect_recog_func): Likewise.  Remove the third argument.
17622         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
17623         (vect_pattern_recog_1): Expect the pattern function to do any
17624         necessary target tests.  Also expect it to provide a vector type.
17625         Remove the type_in handling.
17627 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17629         * tree-vect-patterns.c (vect_pattern_detected): New function.
17630         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
17631         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
17632         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
17633         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
17634         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
17635         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
17636         (vect_recog_mask_conversion_pattern)
17637         (vect_try_gather_scatter_pattern): Likewise.
17639 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17641         * tree-vect-patterns.c (vect_get_internal_def): New function.
17642         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17643         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
17644         (search_type_for_mask_1): Use it.
17646 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17648         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
17649         redundant WIDEN_SUM_EXPR handling.
17650         (vect_recog_sad_pattern): Likewise.
17652 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17654         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
17655         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
17656         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
17657         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
17658         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
17660 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17662         * tree-vect-stmts.c (vectorizable_call): Make sure that we
17663         use the stmt_vec_info of the original bb statement for the
17664         new zero assignment, even if the call is part of a pattern.
17666 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17668         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
17669         that the sequence is attached to the original statement rather
17670         than the pattern statement.
17671         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
17672         PATTERN_DEF_SEQ from the original statement rather than
17673         the main pattern statement.
17674         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
17675         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
17676         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
17678 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17680         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
17681         definition statements before the early exit for statements that aren't
17682         live or relevant.
17683         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
17684         split out from...
17685         (vect_transform_loop): ...here.  Process pattern definition
17686         statements without first checking whether the main pattern
17687         statement is live or relevant.
17689 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
17691         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
17692         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
17694 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
17696         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
17697         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
17698         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
17699         (expand_block_compare): Change select_block_compare_mode call.
17700         (expand_strncmp_align_check): Use new functions, fix comment.
17701         (emit_final_str_compare_gpr): New function.
17702         (expand_strn_compare): Refactor and clean up code.
17703         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
17705 2018-06-19  Tony Reix  <tony.reix@atos.com>
17706             Damien Bergamini  <damien.bergamini@atos.com>
17707             David Edelsohn  <dje.gcc@gmail.com>
17709         * collect2.c (static_obj): New variable.
17710         (static_libs): New variable.
17711         (is_in_list): Uncomment declaration.
17712         (main): Track AIX libraries linked statically.
17713         (is_in_list): Uncomment definition.
17714         (scan_prog_file): Don't add AIX shared libraries initializer
17715         to constructor list if linking statically.
17717 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
17719         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
17720         constant.
17721         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
17723 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
17725         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
17726         blocks.
17728 2018-06-19  Martin Liska  <mliska@suse.cz>
17730         * config/i386/i386.c (ix86_can_inline_p): Do not use
17731         ipa_fn_summaries::get_create.
17732         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
17733         get.
17734         (devirtualization_time_bonus): Likewise.
17735         (ipcp_propagate_stage): Likewise.
17736         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
17737         (edge_set_predicate): Likewise.
17738         (evaluate_conditions_for_known_args): Likewise.
17739         (evaluate_properties_for_edge): Likewise.
17740         (ipa_call_summary::reset): Tranform to ...
17741         (ipa_call_summary::~ipa_call_summary): ... this.
17742         (ipa_fn_summary::reset): Transform to ...
17743         (ipa_fn_summary::~ipa_fn_summary): ... this.
17744         (ipa_fn_summary_t::remove): Rename to ...
17745         (ipa_fn_summary_t::remove_callees): ... this.
17746         (ipa_fn_summary_t::duplicate): Use placement new
17747         instead of memory copy.
17748         (ipa_call_summary_t::duplicate): Likewise.
17749         (ipa_call_summary_t::remove): Remove.
17750         (dump_ipa_call_summary): Change get_create to get.
17751         (ipa_dump_fn_summary): Dump only when summary exists.
17752         (analyze_function_body): Use symbol_summary::get instead
17753         of get_create.
17754         (compute_fn_summary): Likewise.
17755         (estimate_edge_devirt_benefit): Likewise.
17756         (estimate_edge_size_and_time): Likewise.
17757         (inline_update_callee_summaries): Likewise.
17758         (remap_edge_change_prob): Likewise.
17759         (remap_edge_summaries): Likewise.
17760         (ipa_merge_fn_summary_after_inlining): Likewise.
17761         (write_ipa_call_summary): Likewise.
17762         (ipa_fn_summary_write): Likewise.
17763         (ipa_free_fn_summary): Likewise.
17764         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
17765         (struct ipa_call_summary): Likewise.
17766         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
17767         of get_create.
17768         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
17769         (estimate_size_after_inlining): Likewise.
17770         (estimate_growth): Likewise.
17771         (growth_likely_positive): Likewise.
17772         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
17773         (inline_call): Likewise.
17774         * ipa-inline.c (caller_growth_limits): Likewise.
17775         (can_inline_edge_p): Likewise.
17776         (can_inline_edge_by_limits_p): Likewise.
17777         (compute_uninlined_call_time): Likewise.
17778         (compute_inlined_call_time): Likewise.
17779         (want_inline_small_function_p): Likewise.
17780         (edge_badness): Likewise.
17781         (update_caller_keys): Likewise.
17782         (update_callee_keys): Likewise.
17783         (inline_small_functions): Likewise.
17784         (inline_to_all_callers_1): Likewise.
17785         (dump_overall_stats): Likewise.
17786         (early_inline_small_functions): Likewise.
17787         (early_inliner): Likewise.
17788         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
17789         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
17790         * ipa-pure-const.c (malloc_candidate_p): Likewise.
17791         * ipa-split.c (execute_split_functions): Likewise.
17792         * symbol-summary.h: Likewise.
17793         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
17795 2018-06-19  Richard Biener  <rguenther@suse.de>
17797         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
17798         (vectorize_loops): ... here.  Fix dbgcnt handling.
17799         (try_vectorize_loop): Wrap try_vectorize_loop_1.
17801 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
17803         PR target/86197
17804         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
17805         ieee128 argument takes up only one (vector) register, not two (floating
17806         point) registers.
17808 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
17810         * gimplify.c (gimplify_init_constructor): Really never clear for an
17811         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
17813 2018-06-19  Richard Biener  <rguenther@suse.de>
17815         PR tree-optimization/86179
17816         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
17817         after failed recognition.
17819 2018-06-18  Martin Sebor  <msebor@redhat.com>
17821         PR middle-end/85602
17822         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
17823         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
17824         Handle integer subtraction.
17825         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
17826         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
17828 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
17830         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
17831         param from rtx to rtx_insn *.
17832         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
17833         param.
17834         (frv_ifcvt_modify_insn): Likwise.
17835         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
17836         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
17837         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
17838         as_a <rtx_insn *> cast to local "unprotected_region" once
17839         it's been established that it's not NULL or pc_rtx.
17840         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
17841         param "sethi" from rtx to rtx_insn *.
17842         (nds32_group_float_insns): Likewise for param "insn".
17843         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
17844         param.
17845         (vax_output_int_subtract): Likewise.
17846         * config/vax/vax.c (vax_output_int_add): Likewise for param
17847         "insn".
17848         (vax_output_int_subtract): Likewise.
17849         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
17850         (emit_pattern_after): Likewise for param "after".
17851         (emit_insn_after): Likewise.
17852         (emit_jump_insn_after): Likewise.
17853         (emit_call_insn_after): Likewise.
17854         (emit_debug_insn_after): Likewise.
17855         (emit_pattern_before): Likewise for param "before".
17856         (emit_insn_before): Likewise.
17857         (emit_jump_insn_before): Likewise.
17858         * final.c (get_insn_template): Likewise for param "insn", removing
17859         a cast.
17860         * output.h (get_insn_template): Likewise for 2nd param.
17861         * rtl.h (emit_insn_before): Likewise.
17862         (emit_jump_insn_before): Likewise.
17863         (emit_debug_insn_before_noloc): Likewise.
17864         (emit_insn_after): Likewise.
17865         (emit_jump_insn_after): Likewise.
17866         (emit_call_insn_after): Likewise.
17867         (emit_debug_insn_after): Likewise.
17868         (set_insn_deleted): Likewise for param.
17870 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
17872         PR target/85358
17873         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
17874         floating point modes, so that IFmode is numerically greater than
17875         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
17876         to declare the ordering.  This prevents IFmode from being
17877         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
17878         machine.  Include rs6000-modes.h to share the fractional values
17879         between genmodes* and the rest of the compiler.
17880         (IFmode): Likewise.
17881         (KFmode): Likewise.
17882         (TFmode): Likewise.
17883         * config/rs6000/rs6000-modes.h: New file.
17884         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
17885         meaning of rs6000_long_double_size so that 126..128 selects an
17886         appropriate 128-bit floating point type.
17887         (rs6000_option_override_internal): Likewise.
17888         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
17889         (TARGET_LONG_DOUBLE_128): Change the meaning of
17890         rs6000_long_double_size so that 126..128 selects an appropriate
17891         128-bit floating point type.
17892         (LONG_DOUBLE_TYPE_SIZE): Update comment.
17893         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
17894         source and destination to match the standard usage.
17895         (truncifkf2): Likewise.
17896         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
17897         ISA 2.07 to use an explicit clobber, instead of passing in a
17898         temporary.
17899         (copysign<mode>3_soft): Likewise.
17901 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
17903         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
17904         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
17905         (vect_slp_analyze_instance_dependence): Likewise.
17906         (vect_enhance_data_refs_alignment): Likewise.
17907         (vect_analyze_data_refs_alignment): Likewise.
17908         (vect_slp_analyze_and_verify_instance_alignment
17909         (vect_analyze_data_ref_accesses): Likewise.
17910         (vect_prune_runtime_alias_test_list): Likewise.
17911         (vect_analyze_data_refs): Likewise.
17912         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
17913         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17914         (vect_analyze_scalar_cycles_1): Likewise.
17915         (vect_get_loop_niters): Likewise.
17916         (vect_analyze_loop_form_1): Likewise.
17917         (vect_update_vf_for_slp): Likewise.
17918         (vect_analyze_loop_operations): Likewise.
17919         (vect_analyze_loop): Likewise.
17920         (vectorizable_induction): Likewise.
17921         (vect_transform_loop): Likewise.
17922         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
17923         * tree-vect-slp.c (vect_analyze_slp): Likewise.
17924         (vect_make_slp_decision): Likewise.
17925         (vect_detect_hybrid_slp): Likewise.
17926         (vect_slp_analyze_operations): Likewise.
17927         (vect_slp_bb): Likewise.
17928         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
17929         (vectorizable_bswap): Likewise.
17930         (vectorizable_call): Likewise.
17931         (vectorizable_simd_clone_call): Likewise.
17932         (vectorizable_conversion): Likewise.
17933         (vectorizable_assignment): Likewise.
17934         (vectorizable_shift): Likewise.
17935         (vectorizable_operation): Likewise.
17936         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
17938 2018-06-18  Martin Sebor  <msebor@redhat.com>
17940         PR tree-optimization/81384
17941         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
17942         * builtins.c (expand_builtin_strnlen): New function.
17943         (expand_builtin): Call it.
17944         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
17945         * builtins.def (BUILT_IN_STRNLEN): New.
17946         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
17947         Warn for bounds in excess of maximum object size.
17948         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
17949         single-value ranges.  Handle strnlen.
17950         (handle_builtin_strlen): Handle strnlen.
17951         (strlen_check_and_optimize_stmt): Same.
17952         * doc/extend.texi (Other Builtins): Document strnlen.
17954 2018-06-18  Maya Rashish  <coypu@sdf.org>
17956         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
17957         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
17958         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
17960         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
17961         here to ...
17962         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
17964 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17966         * tree.c (escaped_string::escape): Replace cast to char * by
17967         const_cast<char *> (unescaped).
17969 2018-06-18  Nick Clifton  <nickc@redhat.com>
17971         PR 84195
17972         * tree.c (escaped_string): New class.  Converts an unescaped
17973         string into its escaped equivalent.
17974         (warn_deprecated_use): Use the new class to convert the
17975         deprecation message, if present.
17976         (test_escaped_strings): New self test.
17977         (test_c_tests): Add test_escaped_strings.
17978         * doc/extend.texi (deprecated): Add a note that the
17979         deprecation message is affected by the -fmessage-length
17980         option, and that control characters will be escaped.
17981         (#pragma GCC error): Document this pragma.
17982         (#pragma GCC warning): Likewise.
17983         * doc/invoke.texi (-fmessage-length): Document this option's
17984         effect on the #warning and #error preprocessor directives and
17985         the deprecated attribute.
17987 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
17989         * tree.c (decl_value_expr_lookup): Revert latest change.
17990         (decl_value_expr_insert): Likewise.
17992 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
17994         * gimplify.c (nonlocal_vlas): Delete.
17995         (nonlocal_vla_vars): Likewise.
17996         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
17997         referenced VLAs.
17998         (gimplify_body): Do not create and destroy nonlocal_vlas.
17999         * tree-nested.c: Include diagnostic.h.
18000         (use_pointer_in_frame): Tweak.
18001         (lookup_field_for_decl): Add assertion and declare the transformation.
18002         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
18003         internal error when the reference is in a wrong context.  Do not
18004         create a debug decl by default.
18005         (note_nonlocal_block_vlas): Delete.
18006         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
18007         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
18008         create a debug decl by default.
18009         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
18010         call is in a wrong context.
18011         (fixup_vla_decls): New function.
18012         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
18013         debug variables were created.
18014         * tree.c (decl_value_expr_lookup): Add checking assertion.
18015         (decl_value_expr_insert): Likewise.
18017 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
18019         PR middle-end/82479
18020         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
18021         * tree-scalar-evolution.c (interpret_expr): Likewise.
18022         (expression_expensive_p): Likewise.
18023         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
18024         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
18025         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
18026         (ssa_defined_by_minus_one_stmt_p): New.
18028 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
18030         PR middle-end/64946
18031         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
18032         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
18033         * dojump.c (do_jump): Likewise.
18034         * expr.c (expand_expr_real_2): Check operand type's sign.
18035         * fold-const.c (const_unop): Handle ABSU_EXPR.
18036         (fold_abs_const): Likewise.
18037         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
18038         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
18039         (strip_sign_op_1): Likesise.
18040         * match.pd: Add new pattern to generate ABSU_EXPR.
18041         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
18042         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18043         * tree-eh.c (operation_could_trap_helper_p): Likewise.
18044         * tree-inline.c (estimate_operator_cost): Likewise.
18045         * tree-pretty-print.c (dump_generic_node): Likewise.
18046         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
18047         * tree.def (ABSU_EXPR): New.
18049 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
18051         PR middle-end/86095
18052         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
18053         documented as preserved for backward compatibility only.
18054         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
18056         PR rtl-optimization/86108
18057         * bb-reorder.c (create_forwarder_block): Renamed to ...
18058         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
18059         jump from new landing pad to the second part.
18060         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
18061         Adjust callers.
18063 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
18065         PR middle-end/85878
18066         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
18067         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
18068         Only call store_expr for halves if the mode is the same.
18070         PR middle-end/86123
18071         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
18072         Fix up comment formatting.
18074 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18076         * typed-splay-tree.h (typed_splay_tree::remove): New function.
18077         (typed_splay_tree::closure,
18078         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
18079         (typed_splay_tree::typed_splay_tree,
18080         typed_splay_tree::operator =): Declared private.
18081         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
18082         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
18083         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
18084         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
18085         typed_splay_tree::splay_tree_splay,
18086         typed_splay_tree::splay_tree_foreach_helper,
18087         typed_splay_tree::splay_tree_insert,
18088         typed_splay_tree::splay_tree_remove,
18089         typed_splay_tree::splay_tree_lookup,
18090         typed_splay_tree::splay_tree_predecessor,
18091         typed_splay_tree::splay_tree_successor,
18092         typed_splay_tree::splay_tree_min,
18093         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
18094         (typed_splay_tree::root, typed_splay_tree::comp,
18095         typed_splay_tree::delete_key,
18096         typed_splay_tree::delete_value): New data members.
18097         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
18098         typed_splay_tree::remove.
18100 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
18102         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
18103         -mginv and -mno-ginv to the assembler.
18104         * config/mips/mips.opt (-mcrc): New option.
18105         (-mginv): Likewise.
18106         * doc/invoke.text (-mcrc): Document.
18107         (-mginv): Likewise.
18109 2018-06-15  Nick Clifton  <nickc@redhat.com>
18111         PR 84195
18112         * tree.c (escaped_string): New class.  Converts an unescaped
18113         string into its escaped equivalent.
18114         (warn_deprecated_use): Use the new class to convert the
18115         deprecation message, if present.
18116         (test_escaped_strings): New self test.
18117         (test_c_tests): Add test_escaped_strings.
18118         * doc/extend.texi (deprecated): Add a note that the
18119         deprecation message is affected by the -fmessage-length
18120         option, and that control characters will be escaped.
18121         (#pragma GCC error): Document this pragma.
18122         (#pragma GCC warning): Likewise.
18123         * doc/invoke.texi (-fmessage-length): Document this option's
18124         effect on the #warning and #error preprocessor directives and
18125         the deprecated attribute.
18127 2018-06-15  Richard Biener  <rguenther@suse.de>
18129         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
18130         here, also noting vector size used.
18131         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
18132         size used in MSG_OPTIMIZED_LOCATIONS dump.
18133         (pass_slp_vectorize::execute): Adjust.
18135 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
18137         PR target/85968
18138         * config/arc/arc.c (arc_return_address_register): Fix
18139         if-condition.
18141 2018-06-15  Richard Biener  <rguenther@suse.de>
18143         PR middle-end/86159
18144         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
18145         leave useless conversion stripping to force_gimple_operand_gsi.
18146         (gimplify_build2): Likewise.
18147         (gimplify_build1): Likewise.
18149 2018-06-15  Richard Biener  <rguenther@suse.de>
18151         PR middle-end/86076
18152         * tree-cfg.c (move_stmt_op): unshare invariant addresses
18153         before adjusting their block.
18155 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
18157         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
18158         multilibs for *-*-rtems*.
18159         * config/riscv/t-rtems: New file.
18161 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
18163         PR middle-end/86122
18164         * match.pd ((A +- CST1) +- CST2): Punt if last resort
18165         unsigned_type_for returns NULL.
18167         PR target/85945
18168         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
18169         subregs of multi-word pseudos unless the float mode has word size.
18171 2018-06-14  Richard Biener  <rguenther@suse.de>
18173         PR middle-end/86139
18174         * tree-vect-generic.c (build_word_mode_vector_type): Remove
18175         duplicate and harmful type_hash_canon.
18176         * tree.c (type_hash_canon): Assert we didn't find ourselves.
18178 2018-06-14  Richard Biener  <rguenther@suse.de>
18180         PR ipa/86124
18181         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
18182         NULL cgraph_node.
18184 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
18186         * config/rtems.h (STDINT_LONG32): Define.
18188 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
18189             Prachi Godbole  <prachi.godbole@imgtec.com>
18191         * config/mips/mips-cpus.def: Define P6600.
18192         * config/mips/mips-tables.opt: Regenerate.
18193         * config/mips/mips.c (mips_ucbranch_type): New enum.
18194         (mips_rtx_cost_data): Add support for P6600.
18195         (mips_issue_rate): Likewise.
18196         (mips_multipass_dfa_lookahead): Likewise.
18197         (mips_avoid_hazard): Likewise.
18198         (mips_reorg_process_insns): Likewise.
18199         (mips_classify_branch_p6600): New function.
18200         * config/mips/mips.h (TUNE_P6600): New define.
18201         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
18202         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
18203         * config/mips/mips.md: Include p6600.md.
18204         (processor): Add p6600.
18205         * config/mips/p6600.md: New file.
18206         * doc/invoke.texi: Add p6600 to supported architectures.
18208 2018-06-13  Martin Sebor  <msebor@redhat.com>
18210         PR tree-optimization/86114
18211         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
18212         of integer types.
18213         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
18215 2018-06-13  Richard Biener  <rguenther@suse.de>
18217         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
18218         Properly set vector type of the intermediate stmt.
18219         * tree-vect-stmts.c (vectorizable_operation): The destination
18220         var always has vectype_out type.
18222 2018-06-13  Jeff Law  <law@redhat.com>
18224         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
18225         integer 0 for argument to print_rtl_with_bb.
18226         (rl78_reorg): Likewise.
18228 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
18230         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
18231         from rtx to rtx_insn *.
18232         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
18233         "label".
18234         (add_sched_insns_for_speculation): Likewise for local "target",
18235         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
18236         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
18237         from rtx_insn ** to rtx_code_label **.
18238         (reorg_emit_nops): Likewise.
18239         (c6x_reorg): Likewise for local "call_labels".
18240         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
18241         rtx to rtx_insn *.
18242         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
18243         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
18244         the loops over LABEL_REFS.
18245         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
18246         braf_label.
18247         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
18248         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
18249         (split_branches): Strengthen local "olabel" from rtx to
18250         rtx_insn *, adding a safe_as_a cast.
18251         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
18252         to "rtx_insn *".
18253         (add_insn_after): Likewise for first two params.
18254         (add_insn_before): Likewise.
18255         (remove_insn): Likewise for param.
18256         (emit_pattern_before_noloc): Likewise for second and third params.
18257         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
18258         (emit_call_insn_before_noloc): Likewise.
18259         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
18260         to "rtx_insn *".
18261         (emit_barrier_before): Likewise.
18262         (emit_label_before): Strengthen "label" param from "rtx" to
18263         "rtx_code_label *".  Strengthen "before" param from "rtx" to
18264         "rtx_insn *".
18265         (emit_insn_after_1): Strengthen "after" param from "rtx" to
18266         "rtx_insn *".
18267         (emit_pattern_after_noloc): Likewise.
18268         (emit_insn_after_noloc): Likewise.
18269         (emit_jump_insn_after_noloc): Likewise.
18270         (emit_call_insn_after_noloc): Likewise.
18271         (emit_debug_insn_after_noloc): Likewise.
18272         (emit_barrier_after): Likewise.
18273         (emit_label_after): Likewise for both params.
18274         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
18275         "loc" param from "int" to "location_t".
18276         (emit_insn_after_setloc): Likewise.
18277         (emit_jump_insn_after_setloc): Likewise.
18278         (emit_call_insn_after_setloc): Likewise.
18279         (emit_debug_insn_after_setloc): Likewise.
18280         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
18281         "loc" param from "int" to "location_t".
18282         (emit_pattern_before): Convert NULL_RTX to NULL.
18283         (emit_insn_before_setloc): Convert "loc" param from "int" to
18284         "location_t".
18285         (emit_jump_insn_before_setloc): Likewise.
18286         (emit_call_insn_before_setloc): Likewise.
18287         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
18288         rtx_insn *.  Convert "loc" param from "int" to "location_t".
18289         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
18290         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
18291         Convert 3rd param from "int" to "location_t".
18292         (emit_barrier_before, emit_barrier_after, next_real_insn):
18293         Strengthen param from rtx to rtx_insn *.
18294         (emit_label_before): Strengthen 1st param from "rtx" to
18295         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
18296         "rtx_insn *".
18297         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
18298         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
18299         Strengthen 2nd param from "rtx" to "rtx_insn *".
18300         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
18301         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
18302         Likewise. Convert 3rd param from "int" to "location_t".
18303         (emit_label_after): Strengthen 1st param from "rtx" to
18304         "rtx_code_label *".
18305         (next_real_insn, remove_insn): Strengthen param from "rtx" to
18306         "rtx_insn *".
18307         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
18308         from "rtx" to "rtx_insn *".
18310 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
18312         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
18313         bodies streamed in with -Q.
18314         * dumpfile.c (dump_files): Add lto-stream-out dump file.
18315         * dumpfile.h (tree_dump_index): Add lto_stream_out.
18316         * gimple-streamer-out.c: Include gimple-pretty-print.h
18317         (output_bb): Dump stmts streamed.
18318         * lto-section-out.c: Include print-tree.h
18319         (lto_begin_section): Dump sections created.
18320         (lto_output_decl_index): Dump decl encoded.
18321         * lto-streamer-out.c: Include print-tree.h
18322         (create_output_block): Dump output block created.
18323         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
18324         (output_function): Dump function output.
18325         (output_constructor): Dump constructor streamed.
18326         (write_global_stream): Output indexes encoded.
18327         (produce_asm_for_decls): Dump streams encoded.
18328         * lto-streamer.c (streamer_dump_file): New global var.
18329         * lto-streamer.h (streamer_dump_file): Declare.
18330         * passes.c (ipa_write_summaries): Initialize streamer dump.
18331         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
18332         in.
18334 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
18336         PR target/86048
18337         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
18338         offsets for register save directives.  Emit a second batch of save
18339         directives, if need be, when the function accesses prior frames.
18341 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18343         * config/arc/fpu.md (fmasf4): Force operand to register.
18344         (fnmasf4): Likewise.
18346 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18348         * config/arc/arc-protos.h (arc_pad_return): Remove.
18349         * config/arc/arc.c (machine_function): Remove force_short_suffix
18350         and size_reason.
18351         (arc_print_operand): Adjust printing of '&'.
18352         (arc_verify_short): Remove conditional printing of short suffix.
18353         (arc_final_prescan_insn): Remove reference to size_reason.
18354         (pad_return): New function.
18355         (arc_reorg): Call pad_return.
18356         (arc_pad_return): Remove.
18357         (arc_init_machine_status): Remove reference to force_short_suffix.
18358         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
18359         (attr length): When attribute iscompact is true force to 2
18360         regardless; in the case of maybe check if we want to force the
18361         instruction to have 4 bytes length.
18362         (nopv): Change it to generate 4 byte long nop as well.
18363         (blockage): New pattern.
18364         (simple_return): Remove call to arc_pad_return.
18365         (p_return_i): Likewise.
18367 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18369         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
18371 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18373         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
18374         ARC cores.
18376 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18378         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
18379         for ARC700 and ARCv2.
18381 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
18383         PR target/86076
18384         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
18385         operands[2] instead of operands[1].
18388 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
18390         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
18391         case, check whether the outer register overlaps an unallocatable
18392         register, not just whether it fits the required class.
18394 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
18396         * poly-int.h (can_div_trunc_p): Add new overload in which all values
18397         are poly_ints.
18398         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
18399         (memrefs_conflict_p): Likewise.
18400         (init_alias_analysis): Likewise.
18401         * cfgexpand.c (expand_debug_expr): Likewise.
18402         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
18403         * cse.c (fold_rtx): Likewise.
18404         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
18405         * expr.c (emit_block_move_hints): Likewise.
18406         (clear_storage_hints, push_block, emit_push_insn): Likewise.
18407         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
18408         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
18409         (emit_group_store): Likewise.
18410         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
18411         to read the PRE/POST_MODIFY increment.
18412         * calls.c (store_one_arg): Use strip_offset.
18413         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
18414         poly_int_rtx_p.
18415         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
18416         by a VEC_SELECT.
18417         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
18418         (simplify_binary_operation_1): Extend CONST_INT handling to
18419         poly_int_rtx_p.
18420         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
18421         than a HOST_WIDE_INT.
18422         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
18423         poly_int64.
18424         (adjust_mems, add_stores): Update accodingly.
18425         (vt_canonicalize_addr): Track polynomial offsets.
18426         (emit_note_insn_var_location): Likewise.
18427         (vt_add_function_parameter): Likewise.
18428         (vt_initialize): Likewise.
18430 2018-06-12  Jeff Law  <law@redhat.com>
18432         * config.gcc (alpha*-*-freebsd*): Remove.
18433         * config/alpha/freebsd.h: Remove.
18435 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
18437         PR other/69968
18438         * spellcheck-tree.c (levenshtein_distance): Rename to...
18439         (get_edit_distance): ...this, and update for underlying renaming.
18440         * spellcheck-tree.h (levenshtein_distance): Rename to...
18441         (get_edit_distance): ...this.
18442         * spellcheck.c (levenshtein_distance): Rename to...
18443         (get_edit_distance): ...this.  Convert from Levenshtein distance
18444         to Damerau-Levenshtein distance by supporting transpositions of
18445         adjacent characters.  Rename "v1" to "v_next" and "v0" to
18446         "v_one_ago".
18447         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
18448         (selftest::test_edit_distance_unit_test_oneway): ...this, and
18449         update for underlying renaming.
18450         (selftest::levenshtein_distance_unit_test): Rename to...
18451         (selftest::test_get_edit_distance_unit): ...this, and update for
18452         underlying renaming.
18453         (selftest::test_find_closest_string): Add example from PR 69968
18454         where transposition helps
18455         (selftest::test_metric_conditions): Update for renaming.
18456         (selftest::test_metric_conditions): Likewise.
18457         (selftest::spellcheck_c_tests): Likewise.
18458         * spellcheck.h (levenshtein_distance): Rename both overloads to...
18459         (get_edit_distance): ...this.
18460         (best_match::consider): Update for renaming.
18462 2018-06-12  Martin Sebor  <msebor@redhat.com>
18464         PR tree-optimization/85259
18465         * builtins.c (compute_objsize): Handle constant offsets.
18466         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
18467         true iff a warning has been issued.
18468         * gimple.h (gimple_nonartificial_location): New function.
18469         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
18470         gimple_nonartificial_location and handle -Wno-system-headers.
18471         (handle_builtin_stxncpy): Same.
18473 2018-06-12  Martin Sebor  <msebor@redhat.com>
18475         PR c/85931
18476         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
18478 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
18480         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
18481         BUILTIN_VEC_XST entries for pointer to double and long long.
18483 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
18485         PR target/85990
18486         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
18487         Update comments.
18488         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
18489         Likewise.
18491 2018-06-12  Martin Liska  <mliska@suse.cz>
18493         * doc/options.texi: Document IntegerRange.
18495 2018-06-12  Martin Liska  <mliska@suse.cz>
18497         * config/i386/i386.opt: Make MPX-related options as Deprecated.
18498         * opt-functions.awk: Handle Deprecated flag.
18499         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
18500         and report error.
18501         (read_cmdline_option): Report warning for a deprecated option.
18502         * opts.h (struct cl_option): Add new field cl_deprecated.
18503         (CL_ERR_DEPRECATED): New.
18505 2018-06-12  Martin Liska  <mliska@suse.cz>
18507         * doc/options.texi: Document Deprecated option flag.
18509 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18511         * config/arc/arc-arch.h (arc_extras): New enum.
18512         (arc_cpu_t):Add field extra.
18513         (arc_cpu_types): Consider the extras.
18514         * config/arc/arc-cpus.def: Add extras info.
18515         * config/arc/arc-opts.h (processor_type): Consider extra field.
18516         * config/arc/arc.c (arc_override_options): Handle extra field.
18518 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18520         * config/arc/arc-arch.h: Update ARC_OPTX macro.
18521         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
18522         field.
18523         * config/arc/arc.c (arc_init): Update pic warning.
18524         (irq_range): Update irq range parsing warnings.
18525         (arc_override_options): Update various warning messages.
18526         (arc_handle_aux_attribute): Likewise.
18528 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
18530         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
18532 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18534         * doc/sourcebuild.texi: Document usage of line number 0 in verify
18535         compiler messages directives.
18537 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
18539         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
18540         * config/mips/mips-tables.opt: Regenerate.
18541         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
18542         mips64r6.
18543         * doc/invoke.texi: Document -march=i6500.
18545 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
18547         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
18548         (i6400_gpmul): Add cpu_unit.
18549         (i6400_gpdiv): Likewise.
18550         (i6400_msa_add_d): Update reservations.
18551         (i6400_msa_int_add) Likewise.
18552         (i6400_msa_short_logic3) Likewise.
18553         (i6400_msa_short_logic2) Likewise.
18554         (i6400_msa_short_logic) Likewise.
18555         (i6400_msa_move) Likewise.
18556         (i6400_msa_cmp) Likewise.
18557         (i6400_msa_short_float2) Likewise.
18558         (i6400_msa_div_d) Likewise.
18559         (i6400_msa_long_logic1) Likewise.
18560         (i6400_msa_long_logic2) Likewise.
18561         (i6400_msa_mult) Likewise.
18562         (i6400_msa_long_float2) Likewise.
18563         (i6400_msa_long_float4) Likewise.
18564         (i6400_msa_long_float5) Likewise.
18565         (i6400_msa_long_float8) Likewise.
18566         (i6400_fpu_fadd): Include frint type.
18567         (i6400_fpu_store): New define_insn_reservation.
18568         (i6400_fpu_load): Likewise.
18569         (i6400_fpu_move): Likewise.
18570         (i6400_fpu_fcmp): Likewise.
18571         (i6400_fpu_fmadd): Likewise.
18572         (i6400_int_mult): Include imul3nc type and update reservation.
18573         (i6400_int_div): Include idiv3 type and update reservation.
18574         (i6400_int_load): Update to check type not move_type.
18575         (i6400_int_store): Likewise.
18576         (i6400_int_prefetch): Set zero latency.
18578 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
18580         * gcc.c: Document new %@{...} sequence.
18581         (LINK_COMMAND_SPEC): Use it for the -L switches.
18582         (cpp_unique_options): Use it for the -I switches.
18583         (at_file_argbuf): New global variable.
18584         (in_at_file): Likewise.
18585         (alloc_args): Create at_file_argbuf.
18586         (clear_args): Truncate at_file_argbuf.
18587         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
18588         (open_at_file): New function.
18589         (close_at_file): Likewise.
18590         (create_at_file): Delete.
18591         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
18592         <'o'>: Likewise.
18593         <'@'>: New case.
18594         (validate_switches_from_spec): Deal with %@{...} sequence.
18595         (validate_switches): Likewise.
18596         (driver::finalize): Call clear_args.
18598 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
18600         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
18602 2018-06-11  Martin Sebor  <msebor@redhat.com>
18604         * doc/invoke.texi (-Wall): List -Wc++17-compat.
18605         (Wno-class-memaccess): Add @opindex.
18606         (Wno-templates, Wno-multiple-inheritance): Same.
18607         (Wno-virtual-inheritance, Wno-namespaces): Same.
18608         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
18609         (Wno-format-overflow, Wno-format-truncation): Same.
18610         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
18611         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
18612         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
18613         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
18614         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
18615         (Wno-misspelled-isr): Same.
18617 2018-06-11  Martin Sebor  <msebor@redhat.com>
18619         * PR tree-optimization/86083
18620         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
18622 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
18624         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
18626 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
18628         PR target/85755
18629         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
18630         on the correct operand.
18631         (*movdi_internal64): Ditto.
18633 2018-06-11  Martin Liska  <mliska@suse.cz>
18635         PR tree-optimization/86089
18636         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
18638 2018-06-11  Julia Koval  <julia.koval@intel.com>
18640         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
18641         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
18642         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
18644 2018-06-11  Olivier Hainque  <hainque@adacore.com>
18646         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
18647         for Ada with strict dwarf2.
18649 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
18651         PR target/85755
18652         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
18653         addresses.
18655 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
18657         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
18659 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
18661         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
18662         TARGET_ELF.
18664 2018-06-08  Martin Liska  <mliska@suse.cz>
18666         * tree-cfg.h (debug_function): Fix argument type to match
18667         implementation.
18669 2018-06-08  Martin Liska  <mliska@suse.cz>
18671         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
18672         Remove usage of MPX-related (and removed) fields.
18673         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
18675 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
18677         * cfg.c (debug): Use TDF_NONE rather than 0.
18678         * cfghooks.c (debug): Likewise.
18679         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
18680         (struct dump_option_value_info): Convert to...
18681         (struct kv_pair): ...this template type.
18682         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
18683         rather than 0.
18684         (optinfo_verbosity_options): Likewise.
18685         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
18686         OPTGROUP_NONE.
18687         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
18688         than int for "optgroup_flags" param.
18689         (dump_generic_expr_loc): Use dump_flags_t rather than int for
18690         "dump_kind" param.
18691         (dump_dec): Likewise.
18692         (dump_finish): Use TDF_NONE rather than 0.
18693         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
18694         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
18695         than 0.  Update for change to option_ptr.
18696         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
18697         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
18698         0.  Update for changes to optinfo_verbosity_options and
18699         optgroup_options.
18700         (opt_info_switch_p): Convert optgroup_flags from int to
18701         optgroup_flags_t.
18702         (dump_basic_block): Use dump_flags_t rather than int
18703         for "dump_kind" param.
18704         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
18705         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
18706         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
18707         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
18708         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
18709         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
18710         TDF_NONE): Convert from macros to...
18711         (enum dump_flag): ...this new enum.
18712         (dump_flags_t): Update to use enum.
18713         (operator|, operator&, operator~, operator|=, operator&=):
18714         Implement for dump_flags_t.
18715         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
18716         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
18717         Convert from macros to...
18718         (enum optgroup_flag): ...this new enum.
18719         (optgroup_flags_t): New typedef.
18720         (operator|, operator|=): Implement for optgroup_flags_t.
18721         (struct dump_file_info): Convert field "alt_flags" to
18722         dump_flags_t.  Convert field "optgroup_flags" to
18723         optgroup_flags_t.
18724         (dump_basic_block): Use dump_flags_t rather than int for param.
18725         (dump_generic_expr_loc): Likewise.
18726         (dump_dec): Likewise.
18727         (dump_register): Convert param "optgroup_flags" to
18728         optgroup_flags_t.
18729         (opt_info_enable_passes): Likewise.
18730         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
18731         than 0.
18732         * gimple-pretty-print.c (debug): Likewise.
18733         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
18734         (merged_store_group::apply_stores): Likewise.
18735         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
18736         * gimple.c (verify_gimple_pp): Likewise.
18737         * graphite-poly.c (print_pbb_body): Likewise.
18738         * passes.c (pass_manager::register_one_dump_file): Convert
18739         local "optgroup_flags" to optgroup_flags_t.
18740         * print-tree.c (print_node): Use TDF_NONE rather than 0.
18741         (debug): Likewise.
18742         (debug_body): Likewise.
18743         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
18744         to optgroup_flags_t.
18745         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
18746         than 0.
18747         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
18748         (convert_mult_to_fma): Likewise.
18749         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
18750         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
18751         * tree-vect-data-refs.c (dump_lower_bound): Convert param
18752         "dump_kind" to dump_flags_t.
18754 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
18756         * config/rs6000/rs6000.c (min, max): Delete.
18758 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
18760         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
18761         -mabi=spe and -mabi=no-spe.
18763 2018-06-08  Martin Liska  <mliska@suse.cz>
18765         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
18766         where we expect an existing summary.
18768 2018-06-08  Martin Liska  <mliska@suse.cz>
18770         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
18771         * ipa-inline.h (estimate_edge_growth): Likewise.
18773 2018-06-08  Martin Liska  <mliska@suse.cz>
18775         * cgraph.c (function_version_hasher::hash): Use
18776         cgraph_node::get_uid ().
18777         (function_version_hasher::equal):
18778         * cgraph.h (cgraph_node::get_uid): New method.
18779         * ipa-inline.c (update_caller_keys): Use
18780         cgraph_node::get_uid ().
18781         (update_callee_keys): Likewise.
18782         * ipa-utils.c (searchc): Likewise.
18783         (ipa_reduced_postorder): Likewise.
18784         * lto-cgraph.c (input_node): Likewise.
18785         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
18786         * symbol-summary.h (symtab_insertion): Likewise.
18787         (symtab_removal): Likewise.
18788         (symtab_duplication): Likewise.
18789         * tree-pretty-print.c (dump_function_header): Likewise.
18790         * tree-sra.c (convert_callers_for_node): Likewise.
18792 2018-06-08  Martin Liska  <mliska@suse.cz>
18794         * cgraph.c (symbol_table::create_edge): Always assign a new
18795         unique number.
18796         (symbol_table::free_edge): Do not recycle numbers.
18797         * cgraph.h (cgraph_edge::get): New method.
18798         * symbol-summary.h (symtab_removal): Use it.
18799         (symtab_duplication): Likewise.
18800         (call_summary::hashable_uid): Remove.
18802 2018-06-08  Martin Liska  <mliska@suse.cz>
18804         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
18805         (initialize_growth_caches): Remove.
18806         (free_growth_caches): Likewise.
18807         (do_estimate_edge_time): Use edge_growth_cache.
18808         (do_estimate_edge_size): Likewise.
18809         (do_estimate_edge_hints): Likewise.
18810         * ipa-inline.c (reset_edge_caches): Likewise.
18811         (recursive_inlining): Likewise.
18812         (inline_small_functions): Likewise.
18813         * ipa-inline.h (initialize_growth_caches): Remove.
18814         (estimate_edge_size): Likewise.
18815         (estimate_edge_time): Likewise.
18816         (estimate_edge_hints): Likewise.
18817         (reset_edge_growth_cache): Likewise.
18818         * symbol-summary.h (call_summary::remove): New method.
18820 2018-06-08  Martin Liska  <mliska@suse.cz>
18822         * ipa-cp.c (class edge_clone_summary): New summary.
18823         (grow_edge_clone_vectors): Remove.
18824         (ipcp_edge_duplication_hook): Remove.
18825         (class edge_clone_summary_t): New call_summary class.
18826         (ipcp_edge_removal_hook): Remove.
18827         (edge_clone_summary_t::duplicate): New function.
18828         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
18829         (create_specialized_node): Likewise.
18830         (ipcp_driver): Initialize edge_clone_summaries and do not
18831         register hooks.
18833 2018-06-08  Martin Liska  <mliska@suse.cz>
18835         * symbol-summary.h (get): New function.
18836         (call_summary::m_initialize_when_cloning): New class member.
18838 2018-06-08  Martin Liska  <mliska@suse.cz>
18840         * cgraph.c (cgraph_node::remove): Do not recycle uid.
18841         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
18842         (symbol_table::allocate_cgraph_symbol): Do not set uid.
18843         * passes.c (uid_hash_t): Record removed_nodes by their uids.
18844         (remove_cgraph_node_from_order): Use the removed_nodes set.
18845         (do_per_function_toporder): Likwise.
18846         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
18847         instead of summary_uid.
18848         (symtab_removal): Likewise.
18849         (symtab_duplication): Likewise.
18851 2018-06-08  Martin Liska  <mliska@suse.cz>
18853         * ipa-cp.c (ipcp_store_bits_results): Use
18854         ipcp_transformation_sum.
18855         (ipcp_store_vr_results): Likewise.
18856         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
18857         to ...
18858         (ipcp_transformation_initialize): ... this.
18859         (ipa_set_node_agg_value_chain):
18860         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
18861         (write_ipcp_transformation_info): Likewise.
18862         (read_ipcp_transformation_info): Likewise.
18863         (ipcp_update_bits): Likewise.
18864         (ipcp_update_vr): Likewise.
18865         (ipcp_transform_function): Likewise.
18866         * ipa-prop.h: Rename ipcp_transformation_summary to
18867         ipcp_transformation.
18868         (class ipcp_transformation_t): New function summary.
18869         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
18870         (ipa_get_agg_replacements_for_node): Likewise.
18872 2018-06-08  Martin Liska  <mliska@suse.cz>
18874         * ipa-pure-const.c (struct funct_state_d): Do it class instead
18875         of struct.
18876         (class funct_state_summary_t): New function_summary class.
18877         (has_function_state): Remove.
18878         (get_function_state): Likewise.
18879         (set_function_state): Likewise.
18880         (add_new_function): Likewise.
18881         (funct_state_summary_t::insert): New function.
18882         (duplicate_node_data): Remove.
18883         (remove_node_data): Remove.
18884         (funct_state_summary_t::duplicate): New function.
18885         (register_hooks): Create new funct_state_summaries.
18886         (pure_const_generate_summary): Use it.
18887         (pure_const_write_summary): Likewise.
18888         (pure_const_read_summary): Likewise.
18889         (propagate_pure_const): Likewise.
18890         (propagate_nothrow): Likewise.
18891         (dump_malloc_lattice): Likewise.
18892         (propagate_malloc): Likewise.
18893         (execute): Do not register hooks, just remove summary
18894         instead.
18895         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
18896         constructor.
18898 2018-06-08  Martin Liska  <mliska@suse.cz>
18900         * ipa-reference.c (remove_node_data): Remove.
18901         (duplicate_node_data): Likewise.
18902         (class ipa_ref_var_info_summary_t): New class.
18903         (class ipa_ref_opt_summary_t): Likewise.
18904         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
18905         (get_reference_optimization_summary): Use
18906         ipa_ref_opt_sum_summaries.
18907         (set_reference_vars_info): Remove.
18908         (set_reference_optimization_summary): Likewise.
18909         (ipa_init): Create summaries.
18910         (init_function_info): Use function summary.
18911         (ipa_ref_opt_summary_t::duplicate): New function.
18912         (ipa_ref_opt_summary_t::remove): New function.
18913         (get_read_write_all_from_node): Fix GNU coding style.
18914         (propagate): Use function summary.
18915         (write_node_summary_p): Fix GNU coding style.
18916         (stream_out_bitmap): Likewise.
18917         (ipa_reference_read_optimization_summary): Use function summary.
18918         (ipa_reference_c_finalize): Do not release hooks.
18920 2018-06-08  Martin Liska  <mliska@suse.cz>
18922         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
18923         (analyze_function_body): Extract multiple calls of get_create.
18924         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
18925         * ipa-inline.c (recursive_inlining): Use ::get method.
18926         * ipa-inline.h (estimate_edge_growth): Likewise.
18928 2018-06-08  Martin Liska  <mliska@suse.cz>
18930         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
18931         HSA_INVALID.
18932         (hsa_function_summary::hsa_function_summary): Use the new enum
18933         value.
18934         (hsa_gpu_implementation_p): Use hsa_summaries::get.
18935         * hsa-gen.c (hsa_get_host_function): Likewise.
18936         (get_brig_function_name): Likewise.
18937         * ipa-hsa.c (process_hsa_functions): Likewise.
18938         (ipa_hsa_write_summary): Likewise.
18939         * symbol-summary.h (symtab_duplication): Use ::get function/
18940         (get): New function.
18942 2018-06-08  Martin Liska  <mliska@suse.cz>
18944         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
18945         of get.
18946         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
18947         (hsa_register_kernel): Likewise.
18948         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
18949         * hsa-gen.c (hsa_get_host_function): Likewise.
18950         (get_brig_function_name): Likewise.
18951         (generate_hsa): Likewise.
18952         (pass_gen_hsail::execute): Likewise.
18953         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
18954         (devirtualization_time_bonus): Likewise.
18955         (ipcp_propagate_stage): Likewise.
18956         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
18957         (edge_set_predicate): Likewise.
18958         (evaluate_conditions_for_known_args): Likewise.
18959         (evaluate_properties_for_edge): Likewise.
18960         (ipa_fn_summary::reset): Likewise.
18961         (ipa_fn_summary_t::duplicate): Likewise.
18962         (dump_ipa_call_summary): Likewise.
18963         (ipa_dump_fn_summary): Likewise.
18964         (analyze_function_body): Likewise.
18965         (compute_fn_summary): Likewise.
18966         (estimate_edge_devirt_benefit): Likewise.
18967         (estimate_edge_size_and_time): Likewise.
18968         (estimate_calls_size_and_time): Likewise.
18969         (estimate_node_size_and_time): Likewise.
18970         (inline_update_callee_summaries): Likewise.
18971         (remap_edge_change_prob): Likewise.
18972         (remap_edge_summaries): Likewise.
18973         (ipa_merge_fn_summary_after_inlining): Likewise.
18974         (ipa_update_overall_fn_summary): Likewise.
18975         (read_ipa_call_summary): Likewise.
18976         (inline_read_section): Likewise.
18977         (write_ipa_call_summary): Likewise.
18978         (ipa_fn_summary_write): Likewise.
18979         (ipa_free_fn_summary): Likewise.
18980         * ipa-hsa.c (process_hsa_functions): Likewise.
18981         (ipa_hsa_write_summary): Likewise.
18982         (ipa_hsa_read_section): Likewise.
18983         * ipa-icf.c (sem_function::merge): Likewise.
18984         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
18985         (do_estimate_edge_time): Likewise.
18986         (estimate_size_after_inlining): Likewise.
18987         (estimate_growth): Likewise.
18988         (growth_likely_positive): Likewise.
18989         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
18990         (inline_call): Likewise.
18991         * ipa-inline.c (caller_growth_limits): Likewise.
18992         (can_inline_edge_p): Likewise.
18993         (can_inline_edge_by_limits_p): Likewise.
18994         (compute_uninlined_call_time): Likewise.
18995         (compute_inlined_call_time): Likewise.
18996         (want_inline_small_function_p): Likewise.
18997         (edge_badness): Likewise.
18998         (update_caller_keys): Likewise.
18999         (update_callee_keys): Likewise.
19000         (recursive_inlining): Likewise.
19001         (inline_small_functions): Likewise.
19002         (inline_to_all_callers_1): Likewise.
19003         (dump_overall_stats): Likewise.
19004         (early_inline_small_functions): Likewise.
19005         (early_inliner): Likewise.
19006         * ipa-inline.h (estimate_edge_growth): Likewise.
19007         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
19008         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
19009         * ipa-prop.h (IPA_NODE_REF): Likewise.
19010         (IPA_EDGE_REF): Likewise.
19011         * ipa-pure-const.c (malloc_candidate_p): Likewise.
19012         (propagate_malloc): Likewise.
19013         * ipa-split.c (execute_split_functions): Likewise.
19014         * symbol-summary.h: Rename get to get_create.
19015         (get): Likewise.
19016         (get_create): Likewise.
19017         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
19019 2018-06-08  Martin Liska  <mliska@suse.cz>
19021         * symbol-summary.h (release): Move definition out of class
19022         declaration.
19023         (symtab_removal): Likewise.
19024         (symtab_duplication): Likewise.
19026 2018-06-08  Martin Liska  <mliska@suse.cz>
19028         * symbol-summary.h (function_summary): Move constructor
19029         implementation out of class declaration.
19030         (release): Likewise.
19031         (symtab_insertion): Likewise.
19032         (symtab_removal): Likewise.
19033         (symtab_duplication): Likewise.
19034         (get): Likewise.
19036 2018-06-08  Martin Liska  <mliska@suse.cz>
19038         * Makefile.in: Remove support for MPX (macros, related functions,
19039         fields in cgraph_node, ...).
19040         * builtin-types.def (BT_BND): Likewise.
19041         (BT_FN_BND_CONST_PTR): Likewise.
19042         (BT_FN_CONST_PTR_BND): Likewise.
19043         (BT_FN_VOID_PTR_BND): Likewise.
19044         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
19045         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
19046         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
19047         (expand_builtin_mempcpy_with_bounds): Likewise.
19048         (expand_builtin_memset_with_bounds): Likewise.
19049         (expand_builtin_memset_args): Likewise.
19050         (std_expand_builtin_va_start): Likewise.
19051         (expand_builtin): Likewise.
19052         (expand_builtin_with_bounds): Likewise.
19053         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
19054         (DEF_LIB_BUILTIN_CHKP): Likewise.
19055         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
19056         (DEF_CHKP_BUILTIN): Likewise.
19057         (BUILT_IN_MEMCPY): Likewise.
19058         (BUILT_IN_MEMMOVE): Likewise.
19059         (BUILT_IN_MEMPCPY): Likewise.
19060         (BUILT_IN_MEMSET): Likewise.
19061         (BUILT_IN_STPCPY): Likewise.
19062         (BUILT_IN_STRCAT): Likewise.
19063         (BUILT_IN_STRCHR): Likewise.
19064         (BUILT_IN_STRCPY): Likewise.
19065         (BUILT_IN_STRLEN): Likewise.
19066         (BUILT_IN_MEMCPY_CHK): Likewise.
19067         (BUILT_IN_MEMMOVE_CHK): Likewise.
19068         (BUILT_IN_MEMPCPY_CHK): Likewise.
19069         (BUILT_IN_MEMSET_CHK): Likewise.
19070         (BUILT_IN_STPCPY_CHK): Likewise.
19071         (BUILT_IN_STRCAT_CHK): Likewise.
19072         (BUILT_IN_STRCPY_CHK): Likewise.
19073         * calls.c (store_bounds): Likewise.
19074         (emit_call_1): Likewise.
19075         (special_function_p): Likewise.
19076         (maybe_warn_nonstring_arg): Likewise.
19077         (initialize_argument_information): Likewise.
19078         (finalize_must_preallocate): Likewise.
19079         (compute_argument_addresses): Likewise.
19080         (expand_call): Likewise.
19081         * cfgexpand.c (expand_call_stmt): Likewise.
19082         (expand_return): Likewise.
19083         (expand_gimple_stmt_1): Likewise.
19084         (pass_expand::execute): Likewise.
19085         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
19086         (cgraph_node::remove): Likewise.
19087         (cgraph_node::dump): Likewise.
19088         (cgraph_node::verify_node): Likewise.
19089         * cgraph.h (chkp_function_instrumented_p): Likewise.
19090         (symtab_node::get_alias_target): Likewise.
19091         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
19092         (cgraph_local_p): Likewise.
19093         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
19094         (cgraph_edge::rebuild_references): Likewise.
19095         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
19096         (walk_polymorphic_call_targets): Likewise.
19097         (cgraph_node::expand_thunk): Likewise.
19098         (symbol_table::output_weakrefs): Likewise.
19099         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
19100         (ix86_handle_option): Likewise.
19101         * config/i386/constraints.md: Likewise.
19102         * config/i386/i386-builtin-types.def (BND): Likewise.
19103         (VOID): Likewise.
19104         (PVOID): Likewise.
19105         (ULONG): Likewise.
19106         * config/i386/i386-builtin.def (BDESC_END): Likewise.
19107         (BDESC_FIRST): Likewise.
19108         (BDESC): Likewise.
19109         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
19110         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
19111         * config/i386/i386.c (enum reg_class): Likewise.
19112         (ix86_target_string): Likewise.
19113         (ix86_option_override_internal): Likewise.
19114         (ix86_conditional_register_usage): Likewise.
19115         (ix86_valid_target_attribute_inner_p): Likewise.
19116         (ix86_set_indirect_branch_type): Likewise.
19117         (ix86_set_current_function): Likewise.
19118         (ix86_function_arg_regno_p): Likewise.
19119         (init_cumulative_args): Likewise.
19120         (ix86_function_arg_advance): Likewise.
19121         (ix86_function_arg): Likewise.
19122         (ix86_pass_by_reference): Likewise.
19123         (ix86_function_value_regno_p): Likewise.
19124         (ix86_function_value_1): Likewise.
19125         (ix86_function_value_bounds): Likewise.
19126         (ix86_return_in_memory): Likewise.
19127         (ix86_setup_incoming_vararg_bounds): Likewise.
19128         (ix86_va_start): Likewise.
19129         (indirect_thunk_need_prefix): Likewise.
19130         (print_reg): Likewise.
19131         (ix86_print_operand): Likewise.
19132         (ix86_expand_call): Likewise.
19133         (ix86_output_function_return): Likewise.
19134         (reg_encoded_number): Likewise.
19135         (BDESC_VERIFYS): Likewise.
19136         (ix86_init_mpx_builtins): Likewise.
19137         (ix86_init_builtins): Likewise.
19138         (ix86_emit_cmove): Likewise.
19139         (ix86_emit_move_max): Likewise.
19140         (ix86_expand_builtin): Likewise.
19141         (ix86_builtin_mpx_function): Likewise.
19142         (ix86_get_arg_address_for_bt): Likewise.
19143         (ix86_load_bounds): Likewise.
19144         (ix86_store_bounds): Likewise.
19145         (ix86_load_returned_bounds): Likewise.
19146         (ix86_store_returned_bounds): Likewise.
19147         (ix86_class_likely_spilled_p): Likewise.
19148         (ix86_hard_regno_mode_ok): Likewise.
19149         (x86_order_regs_for_local_alloc): Likewise.
19150         (ix86_mitigate_rop): Likewise.
19151         (ix86_bnd_prefixed_insn_p): Likewise.
19152         (ix86_mpx_bound_mode): Likewise.
19153         (ix86_make_bounds_constant): Likewise.
19154         (ix86_initialize_bounds): Likewise.
19155         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
19156         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
19157         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
19158         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
19159         (TARGET_CHKP_BOUND_MODE): Likewise.
19160         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
19161         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
19162         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
19163         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
19164         * config/i386/i386.h (TARGET_MPX): Likewise.
19165         (TARGET_MPX_P): Likewise.
19166         (VALID_BND_REG_MODE): Likewise.
19167         (FIRST_BND_REG): Likewise.
19168         (LAST_BND_REG): Likewise.
19169         (enum reg_class): Likewise.
19170         (BND_REG_P): Likewise.
19171         (BND_REGNO_P): Likewise.
19172         (BNDmode): Likewise.
19173         (ADJUST_INSN_LENGTH): Likewise.
19174         * config/i386/i386.md: Likewise.
19175         * config/i386/i386.opt: Likewise.
19176         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
19177         (defined): Likewise.
19178         (LINK_MPX): Likewise.
19179         (MPX_SPEC): Likewise.
19180         (LIBMPX_SPEC): Likewise.
19181         (LIBMPXWRAPPERS_SPEC): Likewise.
19182         (CHKP_SPEC): Likewise.
19183         * config/i386/predicates.md: Likewise.
19184         * dbxout.c (dbxout_type): Likewise.
19185         * doc/extend.texi: Likewise.
19186         * doc/invoke.texi: Likewise.
19187         * doc/md.texi: Likewise.
19188         * doc/tm.texi: Likewise.
19189         * doc/tm.texi.in: Likewise.
19190         * dwarf2out.c (is_base_type): Likewise.
19191         (gen_formal_types_die): Likewise.
19192         (gen_subprogram_die): Likewise.
19193         (gen_type_die_with_usage): Likewise.
19194         (gen_decl_die): Likewise.
19195         (dwarf2out_late_global_decl): Likewise.
19196         * expr.c (expand_assignment): Likewise.
19197         (emit_storent_insn): Likewise.
19198         (store_expr_with_bounds): Likewise.
19199         (store_expr): Likewise.
19200         (expand_expr_real_1): Likewise.
19201         * expr.h (store_expr_with_bounds): Likewise.
19202         * function.c (use_register_for_decl): Likewise.
19203         (struct bounds_parm_data): Likewise.
19204         (assign_parms_augmented_arg_list): Likewise.
19205         (assign_parm_find_entry_rtl): Likewise.
19206         (assign_parm_is_stack_parm): Likewise.
19207         (assign_parm_load_bounds): Likewise.
19208         (assign_bounds): Likewise.
19209         (assign_parms): Likewise.
19210         (expand_function_start): Likewise.
19211         * gcc.c (CHKP_SPEC): Likewise.
19212         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
19213         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
19214         (wrestrict_dom_walker::check_call): Likewise.
19215         * gimple.c (gimple_build_call_from_tree): Likewise.
19216         * gimple.h (enum gf_mask): Likewise.
19217         (gimple_call_with_bounds_p): Likewise.
19218         (gimple_call_set_with_bounds): Likewise.
19219         * gimplify.c (gimplify_init_constructor): Likewise.
19220         * ipa-cp.c (initialize_node_lattices): Likewise.
19221         (propagate_constants_across_call): Likewise.
19222         (find_more_scalar_values_for_callers_subset): Likewise.
19223         * ipa-hsa.c (process_hsa_functions): Likewise.
19224         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
19225         * ipa-icf.c (sem_function::merge): Likewise.
19226         * ipa-inline.c (early_inliner): Likewise.
19227         * ipa-pure-const.c (warn_function_noreturn): Likewise.
19228         (warn_function_cold): Likewise.
19229         (propagate_pure_const): Likewise.
19230         * ipa-ref.h (enum GTY): Likewise.
19231         * ipa-split.c (find_retbnd): Likewise.
19232         (consider_split): Likewise.
19233         (split_function): Likewise.
19234         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
19235         * ipa.c (walk_polymorphic_call_targets): Likewise.
19236         (symbol_table::remove_unreachable_nodes): Likewise.
19237         (process_references): Likewise.
19238         (cgraph_build_static_cdtor_1): Likewise.
19239         * lto-cgraph.c (lto_output_node): Likewise.
19240         (output_refs): Likewise.
19241         (compute_ltrans_boundary): Likewise.
19242         (input_overwrite_node): Likewise.
19243         (input_node): Likewise.
19244         (input_cgraph_1): Likewise.
19245         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
19246         * passes.c (pass_manager::execute_early_local_passes): Likewise.
19247         (class pass_chkp_instrumentation_passes): Likewise.
19248         (make_pass_chkp_instrumentation_passes): Likewise.
19249         * passes.def: Likewise.
19250         * rtl.h (struct GTY): Likewise.
19251         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
19252         * stor-layout.c (layout_type): Likewise.
19253         * symtab.c: Likewise.
19254         * target.def: Likewise.
19255         * targhooks.c (default_chkp_bound_type): Likewise.
19256         (default_chkp_bound_mode): Likewise.
19257         (default_builtin_chkp_function): Likewise.
19258         (default_chkp_function_value_bounds): Likewise.
19259         (default_chkp_make_bounds_constant): Likewise.
19260         (default_chkp_initialize_bounds): Likewise.
19261         * targhooks.h (default_chkp_bound_type): Likewise.
19262         (default_chkp_bound_mode): Likewise.
19263         (default_builtin_chkp_function): Likewise.
19264         (default_chkp_function_value_bounds): Likewise.
19265         (default_chkp_make_bounds_constant): Likewise.
19266         (default_chkp_initialize_bounds): Likewise.
19267         * toplev.c (compile_file): Likewise.
19268         (process_options): Likewise.
19269         * tree-core.h (DEF_BUILTIN): Likewise.
19270         (DEF_BUILTIN_CHKP): Likewise.
19271         * tree-inline.c (declare_return_variable): Likewise.
19272         (remap_gimple_stmt): Likewise.
19273         (copy_bb): Likewise.
19274         (initialize_inlined_parameters): Likewise.
19275         (expand_call_inline): Likewise.
19276         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
19277         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
19278         (make_pass_ipa_chkp_produce_thunks): Likewise.
19279         (make_pass_chkp): Likewise.
19280         (make_pass_chkp_opt): Likewise.
19281         (make_pass_chkp_instrumentation_passes): Likewise.
19282         * tree-pretty-print.c (dump_generic_node): Likewise.
19283         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
19284         * tree-ssa-dce.c (propagate_necessity): Likewise.
19285         (eliminate_unnecessary_stmts): Likewise.
19286         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
19287         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
19288         * tree-ssa-sccvn.h: Likewise.
19289         * tree-ssa-strlen.c (get_string_length): Likewise.
19290         (valid_builtin_call): Likewise.
19291         (adjust_last_stmt): Likewise.
19292         (handle_builtin_strchr): Likewise.
19293         (handle_builtin_strcpy): Likewise.
19294         (handle_builtin_stxncpy): Likewise.
19295         (handle_builtin_memcpy): Likewise.
19296         (handle_builtin_strcat): Likewise.
19297         (strlen_check_and_optimize_stmt): Likewise.
19298         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
19299         * tree-streamer-in.c: Likewise.
19300         * tree-streamer.c (record_common_node): Likewise.
19301         * tree.c (tree_code_size): Likewise.
19302         (wide_int_to_tree_1): Likewise.
19303         (type_contains_placeholder_1): Likewise.
19304         (build_common_tree_nodes): Likewise.
19305         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
19306         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
19307         (POINTER_BOUNDS_P): Likewise.
19308         (BOUNDED_TYPE_P): Likewise.
19309         (BOUNDED_P): Likewise.
19310         (CALL_WITH_BOUNDS_P): Likewise.
19311         (pointer_bounds_type_node): Likewise.
19312         * value-prof.c (gimple_ic): Likewise.
19313         * var-tracking.c (vt_add_function_parameters): Likewise.
19314         * varasm.c (make_decl_rtl): Likewise.
19315         (assemble_start_function): Likewise.
19316         (output_constant): Likewise.
19317         (maybe_assemble_visibility): Likewise.
19318         * varpool.c (ctor_for_folding): Likewise.
19319         * chkp-builtins.def: Remove.
19320         * ipa-chkp.c: Remove.
19321         * ipa-chkp.h: Remove.
19322         * rtl-chkp.c: Remove.
19323         * rtl-chkp.h: Remove.
19324         * tree-chkp-opt.c: Remove.
19325         * tree-chkp.c: Remove.
19326         * tree-chkp.h: Remove.
19328 2018-06-07  Carl Love  <cel@us.ibm.com>
19330         * config/rs6000/vsx.md (vextract_fp_from_shorth,
19331         vextract_fp_from_shortl): Add BE support.
19333 2018-06-07  Paul Koning  <ni1d@arrl.net>
19335         * compare-elim.c (try_merge_compare): Don't merge compare if
19336         address contains a side effect.
19337         (try_eliminate_compare): Likewise.
19339 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
19341         * config.gcc: Support "tremont".
19342         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
19343         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19344         PROCESSOR_TREMONT.
19345         * config/i386/i386.c (m_TREMONT): Define.
19346         (processor_target_table): Add "tremont".
19347         (PTA_TREMONT): Define.
19348         (ix86_lea_outperforms): Add TARGET_TREMONT.
19349         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
19350         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
19351         and M_INTEL_GOLDMONT_PLUS.
19352         (fold_builtin_cpu): Add "tremont".
19353         (ix86_add_stmt_cost): Add TARGET_TREMONT.
19354         (ix86_option_override_internal): Add "tremont".
19355         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
19356         (processor_type): Add PROCESSOR_TREMONT.
19357         * config/i386/x86-tune.def: Add m_TREMONT.
19358         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
19360 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19362         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
19363         symbol defined for msp430i* devices to be lower case.
19365 2018-06-07  Richard Biener  <rguenther@suse.de>
19367         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
19368         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
19369         Properly wrap signed arithmetic if overflow wraps.
19371 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
19373         PR tree-optimization/69615
19374         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
19375         of a cast from a same precision integral SSA_NAME in a bb dominated
19376         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
19377         cast to utype if rhs2 has already a compatible type.
19379 2018-06-07  Richard Biener  <rguenther@suse.de>
19381         PR tree-optimization/85935
19382         * graphite-scop-detection.c (find_params_in_bb): Analyze
19383         condition operands with respect to the correct loop.  Assert
19384         the analysis doesn't fail.
19386 2018-06-04  Carl Love  <cel@us.ibm.com>
19388         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
19389         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
19390         as it is slightly cheaper.
19391         (first_match_or_eos_index_<mode>):
19392         Calculate index using natural element order.
19393         (first_match_index_<mode>):
19394         Calculate index using natural element order.
19395         (first_match_or_eos_index_<mode>):
19396         Calculate index using natural order.
19397         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
19398         for BE and LE modes.
19399         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
19400         P9V_BUILTIN_VCLZLSBB_V16QI.
19401         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
19402         specific.
19404 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19406         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
19407         indentation and line wrap for many prototypes.  Add missing
19408         @smallexample directives around block of prototypes for vec_xl and
19409         vec_xst.
19411 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
19413         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
19414         track if we pass or return IEEE 128-bit floating point.
19415         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
19416         C++ mangling that is compatible with GCC 8.1.
19417         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
19418         (init_cumulative_args): Note if we pass or return IEEE 128-bit
19419         floating point types.
19420         (rs6000_function_arg_advance_1): Likewise.
19421         (rs6000_mangle_type): Optionally generate mangled names that match
19422         what GCC 8.1 generated for IEEE 128-bit floating point types.
19423         (rs6000_globalize_decl_name): If we have an external function that
19424         passes or returns IEEE 128-bit types, generate a weak reference
19425         from the mangled name used in GCC 8.1 to the current mangled
19426         name.
19427         (rs6000_init_builtins): Make __ibm128 use the long double type if
19428         long double is IBM extended double.  Make __float128 use the long
19429         double type if long double is IEEE 128-bit.
19431         PR target/85657
19432         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
19433         macro for __ibm128 built-in functions.
19434         (PACK_IF): Add __ibm128 pack/unpack functions.
19435         (UNPACK_IF): Likewise.
19436         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
19437         enable long double built-in functions if long double is IEEE
19438         128-bit floating point.
19439         (rs6000_invalid_builtin): Update long double built-in function
19440         error message.
19441         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
19442         functions, adjust the built-in function to use the long double
19443         built-in function if __ibm128 and long double are the same type.
19444         * doc/extend.texi (PowerPC builtins): Update documention for
19445         __builtin_{,un}pack_longdouble.  Add documentation for
19446         __builtin_{,un}pack_ibm128.
19448 2018-06-06  Jim Wilson  <jimw@sifive.com>
19450         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
19451         (struct machine_function): New field interrupt_mode.
19452         (riscv_handle_type_attribute): New function.  Add forward declaration.
19453         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
19454         (riscv_expand_epilogue): Check interrupt_mode field.
19455         (riscv_set_current_function): Check interrupt attribute args and
19456         set interrupt_mode field.
19457         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
19458         (riscv_sret, riscv_uret): New.
19459         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
19460         new arguments to interrupt attribute.
19462 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
19464         PR target/63177
19465         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
19466         Don't handle -mcpu=power8 if -mpower9-vector is also used.
19468 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19470         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
19471         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
19472         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19473         several redundant entries.
19475 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
19477         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
19478         type from "rtx" to "rtx_insn *".
19479         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
19480         for local "call_insn", removing cast.
19481         (ix86_expand_call): Likewise, introducing a "call_insn" local.
19483 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
19485         PR tree-optimization/86066
19486         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
19487         for BIT_INSERT_EXPR stores.
19489 2018-06-06  Richard Biener  <rguenther@suse.de>
19491         PR tree-optimization/86062
19492         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
19493         component refs ontop
19494         of to be offsetted base.
19496 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19498         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
19499         to be static and remove check on interrupt attribute name.
19501 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19503         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
19504         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
19506 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
19508         PR target/79924
19509         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
19510         second argument.
19511         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
19512         Remove second argument, change how error is called.
19513         (aarch64_layout_arg): Remove second argument from
19514         aarch64_err_no_fpadvsimd call.
19515         (aarch64_init_cumulative_args): Ditto.
19516         (aarch64_gimplify_va_arg_expr): Ditto.
19517         * config/aarch64/aarch64.md (mov<mode>): Ditto.
19519 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
19521         * config/i386/i386.md (simple_return_indirect_internal): New expander.
19522         (*simple_return_indirect_internal<mode>): Rename from
19523         simple_return_indirect_internal.  Use W mode iterator.
19524         (rstorssp): New expander.
19525         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
19526         (clrssbsy): New expander.
19527         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
19529 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
19531         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
19532         __typeof__.
19533         (cmse_check_pointed_object): Likewise.
19535 2018-06-05  Martin Liska  <mliska@suse.cz>
19537         PR gcov-profile/47618
19538         * doc/invoke.texi: Document how -fprofile-dir format
19539         is extended.
19541 2018-06-05  Richard Biener  <rguenther@suse.de>
19543         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
19544         removal pretend DOM info isn't available so we do not update
19545         it and only remove edges, not dominated blocks.  Actually free
19546         DOM info in case we removed something.  Remove unreachable blocks.
19547         (mfb_keep_latches): Work with either DOM info or marked backedges.
19548         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
19549         first.  Mark backedges if DOM info isn't available.
19550         (Re-)compute DOM info after cleanup_control_flow_pre.
19552 2018-06-05  Richard Biener  <rguenther@suse.de>
19554         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
19555         (locus_discrim_hasher::hash): Adjust.
19556         (locus_discrim_hasher::equal): Likewise.
19557         (next_discriminator_for_locus): Work on line directly.
19558         (same_line_p): Pass in expanded locus1 as well.
19559         (assign_discriminators): Avoid redundant location expansions.
19561 2018-06-05  Richard Biener  <rguenther@suse.de>
19563         PR tree-optimization/86046
19564         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
19565         if required after clearing TREE_ADDRESSABLE.
19567 2018-06-05  Richard Biener  <rguenther@suse.de>
19569         PR tree-optimization/86047
19570         * tree-ssa-loop.c (for_each_index): Glob handling of all
19571         decls and constants and really handle all of them.
19573 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19575         PR target/81497
19576         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
19577         qualifier_void_pointer and qualifier_const_void_pointer.
19578         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
19579         (arm_init_builtins): Handle the above.
19580         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
19581         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
19582         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
19583         void intrinsics.
19585 2018-06-05  Martin Liska  <mliska@suse.cz>
19587         * auto-profile.c (read_autofdo_file): Do not use
19588         gcov_ctr_summary struct.
19589         (afdo_callsite_hot_enough_for_early_inline): Likewise.
19590         * coverage.c (struct counts_entry): Likewise.
19591         (read_counts_file): Read just single summary entry.
19592         (get_coverage_counts): Use gcov_summary struct.
19593         * coverage.h (get_coverage_counts): Likewise.
19594         * gcov-dump.c (dump_working_sets): Likewise.
19595         (tag_summary): Dump just single summary.
19596         * gcov-io.c (gcov_write_summary): Write just histogram
19597         summary.
19598         (gcov_read_summary): Read just single summary.
19599         (compute_working_sets): Use gcov_summary struct.
19600         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
19601         of GCOV_COUNTERS_SUMMABLE.
19602         (GCOV_COUNTERS_SUMMABLE): Remove.
19603         (GCOV_FIRST_VALUE_COUNTER): Replace with
19604         GCOV_COUNTER_V_INTERVAL.
19605         (struct gcov_ctr_summary): Remove.
19606         (struct gcov_summary): Directly use fields of former
19607         gcov_ctr_summary.
19608         (compute_working_sets): Use gcov_summary struct.
19609         * gcov.c (read_count_file): Do not use ctrs fields.
19610         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
19611         struct.
19612         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
19613         struct.
19614         * profile.c: Likewise.
19615         * profile.h: Likewise.
19617 2018-06-05  Martin Liska  <mliska@suse.cz>
19619         PR gcov-profile/84846
19620         * gcov.c (output_lines): Print working directory only
19621         in intermediate format.
19623 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
19625         * config/s390/s390-builtin-types.def: Add void function type.
19626         * config/s390/s390-builtins.def: Use the function type for the
19627         tbeginc builtin.
19629 2018-06-04  Jim Wilson  <jimw@sifive.com>
19631         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
19632         to int.
19633         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
19634         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
19635         handle EH_RETURN_DATA_REGNO registers properly.
19636         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
19637         (riscv_expand_epilogue): Update comment.  Change argument name and
19638         type.  Update code to use new name and type.  Pass new args to
19639         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
19640         EXCEPTION_RETURN.
19641         * config/riscv/riscv.md (NORMAL_RETURN): New.
19642         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
19643         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
19644         (eh_return): Call gen_eh_return_internal and emit barrier.
19645         (eh_return_internal): Call riscv_expand_epilogue.
19647 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
19649         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
19650         bit_insertion field and declare can_be_merged_into method.
19651         (merged_store_group::can_be_merged_into): New method.
19652         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
19653         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
19654         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
19656 2018-06-04  Richard Biener  <rguenther@suse.de>
19658         PR tree-optimization/85955
19659         * builtins.c (fold_builtin_sincos): Convert pointers to
19660         destination to appropriate type before dereferencing.
19662 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
19664         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
19666 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
19668         * expr.c (expand_expr_real_1): Force the operand into memory if
19669         its TYPE_MODE is BLKmode and if there is no integer mode for
19670         the number of bits being extracted.
19672 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
19674         PR target/85832
19675         PR target/86036
19676         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
19677         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
19679 2018-06-04  Richard Biener  <rguenther@suse.de>
19681         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
19682         (cleanup_tree_cfg_noloop): ... single caller.  Do
19683         start_recording_case_labels later.
19685 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
19687         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
19688         to _IMMINTRIN_H_INCLUDED.
19689         * config/i386/pconfigintrin.h: Ditto.
19690         * config/i386/waitpkgintrin.h: Ditto.
19691         * config/i386/immintrin.h: Add includes for sgxintrin.h,
19692         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
19693         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
19694         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
19695         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
19696         waitpkgintrin.h and cldemoteintrin.h.
19698 2018-06-04  Richard Biener  <rguenther@suse.de>
19700         PR tree-optimization/86038
19701         * tracer.c (find_best_successor): Check probability for
19702         being initialized, bail out if not.
19704 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
19706         PR target/86003
19707         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
19708         of bits to ignore when comparing architectures.
19710 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
19712         PR tree-optimization/69615
19713         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
19714         maximum or minimum of the type, try to merge it also as if
19715         range1 is + [-, x - 1] or + [x + 1, -].
19717         PR c++/86025
19718         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
19720 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
19722         PR tree-optimization/86034
19723         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
19724         the unsigned bitfield type in a bit insertion sequence if it does not
19725         have a larger precision than the bitfield size.
19726         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
19728 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
19730         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
19732 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19734         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
19735         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
19736         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
19737         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
19739 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19741         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
19742         Disable -fdelete-null-pointer-checks for ELF toolchain.
19744 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19745             Kito Cheng  <kito.cheng@gmail.com>
19747         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
19748         (nds32le-*-*, nds32be-*-*): Integrate checking process.
19749         (nds32*-*-*): Add glibc and uclibc conditions.
19750         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
19751         (TARGET_EXCEPT_UNWIND_INFO): Define.
19752         * config/nds32/elf.h: New file.
19753         * config/nds32/linux.h: New file.
19754         * config/nds32/nds32-elf.opt: New file.
19755         * config/nds32/nds32-linux.opt: New file.
19756         * config/nds32/nds32-fp-as-gp.c
19757         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
19758         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
19759         TARGET_LINUX_ABI.
19760         (nds32_asm_file_end): Ditto.
19761         (nds32_print_operand): Ditto.
19762         (nds32_insert_attributes): Ditto.
19763         (nds32_init_libfuncs): New function.
19764         (TARGET_HAVE_TLS): Define.
19765         (TARGET_INIT_LIBFUNCS): Define.
19766         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
19767         spec content.
19768         (TARGET_ELF): Apply different mcmodel setting.
19769         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
19770         been migrated into elf.h and linux.h files.
19771         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
19772         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
19773         (mcmodel): The content has been migrated into nds32-elf.opt and
19774         nds32-linux.opt files.
19775         * config/nds32/t-elf: New file.
19776         * config/nds32/t-linux: New file.
19778 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19779             Shiva Chen  <shiva0217@gmail.com>
19781         * config/nds32/constants.md (unspec_volatile_element): Add
19782         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
19783         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
19784         optimization.
19785         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
19786         (make_pass_nds32_fp_as_gp): Declare.
19787         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
19788         optmization pass.
19789         (nds32_asm_function_end_prologue): Remove unused asm output.
19790         (nds32_asm_function_begin_epilogue): Remove unused asm output.
19791         (nds32_asm_file_start): Output necessary fp_as_gp information.
19792         (nds32_option_override): Adjust register usage.
19793         (nds32_expand_prologue): Consider fp_as_gp situation.
19794         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
19795         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
19796         (epilogue): Ditto.
19797         (return): Ditto.
19798         (simple_return): Ditto.
19799         (omit_fp_begin): Output special directive for fp_as_gp.
19800         (omit_fp_end): Output special directive for fp_as_gp.
19801         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
19802         mforbid-fp-as-gp): New options.
19804 2018-06-01  Mark Wielaard  <mark@klomp.org>
19806         * dwarf2out.c (dwarf2out_finish): Remove generation of
19807         DW_AT_loclists_base.
19809 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
19811         * gimple-ssa-store-merging.c: Include gimple-fold.h.
19812         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
19813         (struct merged_store_group): Add bit_insertion field.
19814         (dump_char_array): Use standard hexadecimal format.
19815         (merged_store_group::merged_store_group): Set bit_insertion to false.
19816         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
19817         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
19818         also print the mask in the dump file.
19819         (pass_store_merging::gate): Minor tweak.
19820         (imm_store_chain_info::coalesce_immediate): Fix wrong association
19821         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
19822         stores with INTEGER_CST stores.
19823         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
19824         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
19825         and use it throughout.  Generate bit insertion sequences if need be.
19826         (pass_store_merging::process_store): Remove redundant condition.
19827         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
19829 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
19831         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
19832         the 128-bit floating point types.  Fix function comment.
19834 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19836         * config/aarch64/aarch64-simd.md
19837         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
19838         mnemonics.
19839         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
19840         mnemonics.
19842 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
19844         PR tree-optimization/85989
19845         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
19846         variable.
19847         (backprop::intersect_uses): Check it when deciding whether this
19848         is a backedge reference.
19849         (backprop::process_block): Add each phi to m_visited_phis
19850         after visiting it, then clear it at the end.
19852 2018-06-01  Richard Biener  <rguenther@suse.de>
19854         * tree-vectorizer.h (vect_dr_stmt): New function.
19855         (vect_get_load_cost): Adjust.
19856         (vect_get_store_cost): Likewise.
19857         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
19858         Use vect_dr_stmt instead of DR_SMTT.
19859         (vect_record_base_alignments): Likewise.
19860         (vect_calculate_target_alignment): Likewise.
19861         (vect_compute_data_ref_alignment): Likewise and make static.
19862         (vect_update_misalignment_for_peel): Likewise.
19863         (vect_verify_datarefs_alignment): Likewise.
19864         (vector_alignment_reachable_p): Likewise.
19865         (vect_get_data_access_cost): Likewise.  Pass down
19866         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
19867         (vect_get_peeling_costs_all_drs): Likewise.
19868         (vect_peeling_hash_get_lowest_cost): Likewise.
19869         (vect_enhance_data_refs_alignment): Likewise.
19870         (vect_find_same_alignment_drs): Likewise.
19871         (vect_analyze_data_refs_alignment): Likewise.
19872         (vect_analyze_group_access_1): Likewise.
19873         (vect_analyze_group_access): Likewise.
19874         (vect_analyze_data_ref_access): Likewise.
19875         (vect_analyze_data_ref_accesses): Likewise.
19876         (vect_vfa_segment_size): Likewise.
19877         (vect_small_gap_p): Likewise.
19878         (vectorizable_with_step_bound_p): Likewise.
19879         (vect_prune_runtime_alias_test_list): Likewise.
19880         (vect_analyze_data_refs): Likewise.
19881         (vect_supportable_dr_alignment): Likewise.
19882         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
19883         (vect_gen_prolog_loop_niters): Likewise.
19884         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
19885         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
19886         modify DR_STMT.
19887         (vect_recog_mask_conversion_pattern): Likewise.
19888         (vect_try_gather_scatter_pattern): Likewise.
19889         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
19890         to vect_get_store_cost.
19891         (vect_get_store_cost): Get stmt_info instead of DR.
19892         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
19893         (vect_get_load_cost): Get stmt_info instead of DR.
19895 2018-06-01  Richard Biener  <rguenther@suse.de>
19897         PR middle-end/86017
19898         * gimple-fold.c (var_decl_component_p): Also allow offsetted
19899         vars wrapped in MEM_REFs.
19901 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
19903         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
19904         Fix subreg tests so that we only return a choice between
19905         GENERAL_REGS and FP_REGS if the original classes included both.
19907 2018-06-01  Richard Biener  <rguenther@suse.de>
19909         PR ipa/85960
19910         * tree-ssa-structalias.c (get_function_part_constraint):
19911         Handle NULL fi->decl.
19912         (find_func_aliases_for_call): Properly handle indirect
19913         fi from direct call.
19914         (find_func_clobbers): Likewise.
19915         (ipa_pta_execute): Likewise.
19916         (create_variable_info_for): For functions that are ifunc_resolver
19917         resolve to a varinfo that contains the result of the resolver call.
19918         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
19919         aliases.
19921 2018-05-31  Michael Collison  <michael.collison@arm.com>
19923         * config/aarch64/aarch64.md:
19924         (*fix_to_zero_extenddfdi2): New pattern.
19925         * gcc.target/aarch64/fix_extend1.c: New testcase.
19927 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
19929         PR middle-end/78809
19930         PR middle-end/83026
19931         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
19932         and BUILT_IN_STRNCMP_EQ.
19933         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
19934         BUILT_IN_STRNCMP_EQ.
19935         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
19936         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
19937         (gimple_fold_builtin): Likewise.
19938         * tree-ssa-strlen.c (compute_string_length): New function.
19939         (determine_min_obsize): New function.
19940         (handle_builtin_string_cmp): New function to handle calls to
19941         string compare functions.
19942         (strlen_optimize_stmt): Add handling to builtin string compare
19943         calls.
19944         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
19945         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
19946         * tree.c (build_common_builtin_nodes): Add new defines of
19947         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
19949 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
19951         PR target/85984
19952         * bb-reorder.c (pass_partition_blocks::gate): Return false for
19953         functions with naked attribute.
19955 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
19957         * config/i386/sse.md (avx_vec_concat<mode>):
19958         Substitute concat_tg_mode mode attribute with xtg_mode.
19959         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
19960         (concat_tg_mode): Remove mode attribute.
19962 2018-05-31  Martin Sebor  <msebor@redhat.com>
19964         PR c/82063
19965         * calls.c (alloc_max_size): Correct a logic error/typo.
19966         Treat excessive arguments as infinite.  Warn for invalid arguments.
19967         * doc/invoke.texi (-Walloc-size-larger-than): Update.
19969 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
19971         PR target/85829
19972         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
19973         and movx for Haswell.
19975 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
19976             Cesar Philippidis  <cesar@codesourcery.com>
19978         PR middle-end/85879
19979         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
19980         when emitting error on private/firstprivate reductions.
19981         * omp-low.c (lower_omp_target): Avoid reference-type processing
19982         on pointers for firstprivate clause.
19984 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
19986         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
19987         (st1x2): Likewise.
19988         (st1x3): Likewise.
19989         * config/aarch64/aarch64-simd.md
19990         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
19991         (aarch64_ld1_x3_<mode>): Likewise
19992         (aarch64_st1x2<VALLDIF:mode>): Likewise
19993         (aarch64_st1_x2_<mode>): Likewise
19994         (aarch64_st1x3<VALLDIF:mode>): Likewise
19995         (aarch64_st1_x3_<mode>): Likewise
19996         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
19997         (vld1_s8_x3): Likewise.
19998         (vld1_u16_x3): Likewise.
19999         (vld1_s16_x3): Likewise.
20000         (vld1_u32_x3): Likewise.
20001         (vld1_s32_x3): Likewise.
20002         (vld1_u64_x3): Likewise.
20003         (vld1_s64_x3): Likewise.
20004         (vld1_f16_x3): Likewise.
20005         (vld1_f32_x3): Likewise.
20006         (vld1_f64_x3): Likewise.
20007         (vld1_p8_x3): Likewise.
20008         (vld1_p16_x3): Likewise.
20009         (vld1_p64_x3): Likewise.
20010         (vld1q_u8_x3): Likewise.
20011         (vld1q_s8_x3): Likewise.
20012         (vld1q_u16_x3): Likewise.
20013         (vld1q_s16_x3): Likewise.
20014         (vld1q_u32_x3): Likewise.
20015         (vld1q_s32_x3): Likewise.
20016         (vld1q_u64_x3): Likewise.
20017         (vld1q_s64_x3): Likewise.
20018         (vld1q_f16_x3): Likewise.
20019         (vld1q_f32_x3): Likewise.
20020         (vld1q_f64_x3): Likewise.
20021         (vld1q_p8_x3): Likewise.
20022         (vld1q_p16_x3): Likewise.
20023         (vld1q_p64_x3): Likewise.
20024         (vst1_s64_x2): Likewise.
20025         (vst1_u64_x2): Likewise.
20026         (vst1_f64_x2): Likewise.
20027         (vst1_s8_x2): Likewise.
20028         (vst1_p8_x2): Likewise.
20029         (vst1_s16_x2): Likewise.
20030         (vst1_p16_x2): Likewise.
20031         (vst1_s32_x2): Likewise.
20032         (vst1_u8_x2): Likewise.
20033         (vst1_u16_x2): Likewise.
20034         (vst1_u32_x2): Likewise.
20035         (vst1_f16_x2): Likewise.
20036         (vst1_f32_x2): Likewise.
20037         (vst1_p64_x2): Likewise.
20038         (vst1q_s8_x2): Likewise.
20039         (vst1q_p8_x2): Likewise.
20040         (vst1q_s16_x2): Likewise.
20041         (vst1q_p16_x2): Likewise.
20042         (vst1q_s32_x2): Likewise.
20043         (vst1q_s64_x2): Likewise.
20044         (vst1q_u8_x2): Likewise.
20045         (vst1q_u16_x2): Likewise.
20046         (vst1q_u32_x2): Likewise.
20047         (vst1q_u64_x2): Likewise.
20048         (vst1q_f16_x2): Likewise.
20049         (vst1q_f32_x2): Likewise.
20050         (vst1q_f64_x2): Likewise.
20051         (vst1q_p64_x2): Likewise.
20052         (vst1_s64_x3): Likewise.
20053         (vst1_u64_x3): Likewise.
20054         (vst1_f64_x3): Likewise.
20055         (vst1_s8_x3): Likewise.
20056         (vst1_p8_x3): Likewise.
20057         (vst1_s16_x3): Likewise.
20058         (vst1_p16_x3): Likewise.
20059         (vst1_s32_x3): Likewise.
20060         (vst1_u8_x3): Likewise.
20061         (vst1_u16_x3): Likewise.
20062         (vst1_u32_x3): Likewise.
20063         (vst1_f16_x3): Likewise.
20064         (vst1_f32_x3): Likewise.
20065         (vst1_p64_x3): Likewise.
20066         (vst1q_s8_x3): Likewise.
20067         (vst1q_p8_x3): Likewise.
20068         (vst1q_s16_x3): Likewise.
20069         (vst1q_p16_x3): Likewise.
20070         (vst1q_s32_x3): Likewise.
20071         (vst1q_s64_x3): Likewise.
20072         (vst1q_u8_x3): Likewise.
20073         (vst1q_u16_x3): Likewise.
20074         (vst1q_u32_x3): Likewise.
20075         (vst1q_u64_x3): Likewise.
20076         (vst1q_f16_x3): Likewise.
20077         (vst1q_f32_x3): Likewise.
20078         (vst1q_f64_x3): Likewise.
20079         (vst1q_p64_x3): Likewise.
20081 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20083         * config/msp430/msp430.c (msp430_output_labelref): Prepend
20084         user_label_prefix to name.
20086         * tree-core.h: Update comment about the format of NAME string
20087         passed to handler in attribute_spec.
20089         * config/msp430/msp430.md: Remove erroneous subreg expression from
20090         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
20091         zero_extend{q,h}isi2.
20093 2018-05-30  Borislav Petkov  <bp@suse.de>
20095         * doc/extend.texi: Document some architecture specific
20096         constraints and sort entries.
20098 2018-05-30  Martin Sebor  <msebor@redhat.com>
20100         PR middle-end/85369
20101         * builtins.c (expand_builtin_stpcpy_1): New function.
20102         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
20103         only if the former succeeds.
20105 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
20107         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
20108         in saphira.
20110 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20112         * doc/invoke.texi (-flinker-output): Document
20114 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20116         * passes.c (ipa_write_summaries): Only modify statements if body
20117         is in memory.
20118         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
20119         incrementally linking.
20120         (ipa_passes): Likewise.
20121         * lto-cgraph.c (lto_output_node): When incrementally linking do not
20122         pass down resolution info.
20123         * common.opt (flag_incremental_link): Update info.
20124         * gcc.c (plugin specs): Turn flinker-output=* to
20125         -plugin-opt=-linker-output-known
20126         * toplev.c (compile_file): Also cut compilation when doing incremental
20127         link.
20128         * flag-types. (enum lto_partition_model): Add
20129         LTO_LINKER_OUTPUT_NOLTOREL.
20130         (invoke.texi): Add -flinker-output docs.
20131         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
20132         link same way as WPA; do not stream in dead initializers.
20134         * dwarf2out.c (dwarf2out_die_ref_for_decl,
20135         darf2out_register_external_decl): Support incremental link.
20137 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20139         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
20141 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20143         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
20144         it down to simple_object_copy_lto_debug_sections.
20145         (run_gcc): Determine incremental LTO link time and configure
20146         lto1 into non-wpa mode, disable renaming of debug sections.
20148 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20150         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20151         descriptions of various incorrectly documented functions.
20153 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
20155         Revert:
20156         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
20157         address check not strict.
20159 2018-05-30  Richard Biener  <rguenther@suse.de>
20161         PR tree-optimization/85964
20162         * tracer.c (better_p): Drop initialized count check, we only
20163         call the function with initialized counts now.
20164         (find_best_successor): Do find a best edge if one
20165         has uninitialized count.
20166         (find_best_predecessor): Likewise.  Do BB frequency check only
20167         if count is initialized.
20169 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
20171         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
20172         (aarch64_ldrstr_offset_compare): New.
20173         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
20174         load/store orderings.
20175         (aarch64_gen_adjusted_ldpstp): Likewise.
20177 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
20179         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
20180         Check for subset of GENERAL_REGS and FP_REGS.
20181         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
20182         r=w alternative.
20184 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
20186         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
20187         and wi::to_poly_offset.  Add the current offset and then check
20188         whether the sum fits, rather than using an unchecked addition of
20189         a checked term.  Check for a shwi rather than a uhwi.
20190         * expr.c (get_bit_range): Use tree_to_poly_uint64.
20191         (store_constructor): Use poly_int_tree_p.
20192         (expand_expr_real_1): Likewise.
20193         * function.c (assign_temp): Likewise.
20194         * fold-const.c (const_binop): Use poly_int_tree_p and
20195         wi::to_poly_offset.
20196         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
20197         division.
20198         * ipa-icf-gimple.c (func_checker::compare_operand): Use
20199         to_poly_offset for MEM offsets.
20200         * ipa-icf.c (sem_variable::equals): Likewise.
20201         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
20202         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20203         wi::to_poly_offset for BIT_FIELD_REF offsets.
20204         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
20205         wi::to_poly_offset.
20206         * var-tracking.c (emit_note_insn_var_location): Use
20207         tree_to_poly_uint64.
20209 2018-05-29  Jim Wilson  <jimw@sifive.com>
20211         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
20213 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
20215         PR target/85950
20216         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
20217         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
20218         sequence.
20219         (sse4_1_round<mode>2): Use nonimmediate_operand
20220         for operand 1 predicate.
20222 2018-05-29  Martin Sebor  <msebor@redhat.com>
20223             Richard Biener  <rguenther@suse.de>
20225         PR testsuite/85888
20226         * calls.c (get_size_range): Call determine_value_range instead
20227         of get_value_range..
20228         * tree-vrp.h (determine_value_range): Declared new function.
20229         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
20231 2018-05-29  Richard Biener  <rguenther@suse.de>
20233         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
20234         sure to use non-pattern stmts for get_earlier_stmt arguments.
20235         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
20236         called on pattern stmts.
20237         (get_later_stmt): Likewise.
20239 2018-05-29  Martin Liska  <mliska@suse.cz>
20241         PR gcov-profile/85759
20242         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
20243         env variables.
20245 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
20247         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
20248         VEC_UNPACK_*_EXPR.
20249         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
20250         VEC_PACK_*_EXPR.
20252         PR target/85918
20253         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
20254         VEC_PACK_FLOAT_EXPR): New tree codes.
20255         * tree-pretty-print.c (op_code_prio): Handle
20256         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
20257         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20258         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
20259         * tree-inline.c (estimate_operator_cost): Likewise.
20260         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
20261         * fold-const.c (const_binop): Likewise.
20262         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
20263         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
20264         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
20265         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
20266         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20267         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
20268         * expr.c (expand_expr_real_2): Likewise.
20269         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
20270         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
20271         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
20272         optabs.
20273         * optabs.c (expand_widen_pattern_expr): For
20274         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
20275         sign from result type rather than operand's type.
20276         (expand_binop_directly): For vec_packu_float_optab and
20277         vec_packs_float_optab allow result type to be different from operand's
20278         type.
20279         * optabs-tree.c (optab_for_tree_code): Handle
20280         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
20281         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
20282         * tree-vect-generic.c (expand_vector_operations_1):  Handle
20283         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
20284         VEC_PACK_FLOAT_EXPR.
20285         * tree-vect-stmts.c (supportable_widening_operation): Handle
20286         FIX_TRUNC_EXPR.
20287         (supportable_narrowing_operation): Handle FLOAT_EXPR.
20288         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
20289         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
20290         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
20291         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
20292         mode attributes.
20293         (vec_pack<floatprefix>_float_<mode>): New expander.
20294         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
20295         attributes.
20296         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
20297         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
20298         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
20299         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
20300         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
20301         Document.
20302         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
20303         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
20304         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
20305         VEC_PACK_FLOAT_EXPR): Document.
20307 2018-05-29  Richard Biener  <rguenther@suse.de>
20309         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
20310         member.
20311         (stmt_vec_info_vec): Make pointer.
20312         (init_stmt_vec_info_vec): Remove.
20313         (free_stmt_vec_info_vec): Likewise.
20314         (set_stmt_vec_info_vec): New function.
20315         (free_stmt_vec_infos): Likewise.
20316         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
20317         (set_vinfo_for_stmt): Likewise.
20318         (get_earlier_stmt): Likewise.
20319         (get_later_stmt): Likewise.
20320         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
20321         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
20322         (vec_info::~vec_info): Free stmt_vec_infos.
20323         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
20324         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
20325         (pass_slp_vectorize::execute): Likewise.
20326         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
20327         (free_stmt_vec_info_vec): Likewise.
20328         (set_stmt_vec_info_vec): New function.
20329         (free_stmt_vec_infos): Likewise.
20330         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
20331         the global stmt_vec_info_vec.
20332         * tree-parloops.c (gather_scalar_reductions): Use
20333         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
20334         vector.
20336 2018-05-29  Richard Biener  <rguenther@suse.de>
20338         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
20340 2018-05-29  Martin Liska  <mliska@suse.cz>
20341             David Malcolm  <dmalcolm@redhat.com>
20343         * vec.c (test_reverse): New.
20344         (vec_c_tests): Add new test.
20345         * vec.h (vl_ptr>::reverse): New function.
20347 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
20349         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
20351         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
20352         and later.
20354 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
20356         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
20358 2018-05-28  Richard Biener  <rguenther@suse.de>
20360         PR tree-optimization/85933
20361         * tree-vect-data-refs.c (vect_record_base_alignments): Only
20362         look at stmts marked as vectorizable.
20364 2018-05-28  Richard Biener  <rguenther@suse.de>
20366         PR tree-optimization/85934
20367         * tree-vect-generic.c (expand_vector_operations_1): Hoist
20368         vector boolean check before scalar optimization.
20370 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
20372         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
20373         for armv5te.
20375 2018-05-28  Mark Wielaard  <mark@klomp.org>
20377         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
20378         if it is an expression containing a minus sign.
20380 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
20382         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
20384 2018-05-27  Paul Koning  <ni1d@arrl.net>
20386         * config/pdp11/pdp11.md (truncsihi2): Remove.
20388 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
20389             Chung-Ju Wu  <jasonwucj@gmail.com>
20391         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
20392         implementation.
20393         (unaligned_store_dw): Ditto.
20394         * config/nds32/nds32-memory-manipulation.c
20395         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
20396         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
20397         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
20398         (emit_setmem_word_loop): Rename to ...
20399         (emit_setmem_doubleword_loop): ... this.
20400         (nds32_gen_dup_4_byte_to_word_value): New function.
20401         (nds32_gen_dup_8_byte_to_double_word_value): New function.
20402         (nds32_expand_setmem_loop): Refine implementation.
20403         (nds32_expand_setmem_loop_v3m): Ditto.
20404         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
20405         pattern.
20407 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20409         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
20411 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20413         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
20414         (nds32_init_machine_status): Initialize machine->attr_naked_p and
20415         machine->attr_no_prologue_p.
20416         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
20417         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
20418         (nds32_expand_epilogue): Consider attr_naked_p.
20419         (nds32_expand_epilogue_v3pop): Likewise.
20420         (nds32_can_use_return_insn): Likewise.
20421         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
20422         attr_no_prologue_p fields.
20423         * config/nds32/nds32.opt (mret-in-naked-func): New option.
20425 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
20427         PR target/85918
20428         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
20429         attributes.
20430         * config/i386/sse.md
20431         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
20432         Rename to ...
20433         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
20434         ... this.
20435         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
20436         Rename to ...
20437         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
20438         ... this.
20439         (*<floatsuffix>floatv2div2sf2): Rename to ...
20440         (*float<floatunssuffix>v2div2sf2): ... this.
20441         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
20442         (float<floatunssuffix>v2div2sf2_mask): ... this.
20443         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
20444         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
20445         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
20446         to ...
20447         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
20448         ... this.
20449         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
20450         Rename to ...
20451         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
20452         ... this.
20453         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
20454         Rename to ...
20455         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
20456         ... this.
20457         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
20458         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
20459         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
20460         gen_ufix_truncv8dfv8si2.
20461         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
20462         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
20463         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
20464         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
20465         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
20466         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
20467         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
20468         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
20470 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
20472         PR target/85900
20473         PR target/85345
20474         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
20476 2018-05-25  Jim Wilson  <jimw@sifive.com>
20478         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
20479         * config/riscv/riscv.c (struct machine_function): Add
20480         interrupt_handler_p and attribute_checked_p fields.
20481         (riscv_attribute_table): Add interrupt.
20482         (riscv_interrupt_type_p): New.
20483         (riscv_save_reg_p): Save extra regs for interrupt handler.
20484         (riscv_use_save_libcall): Return false  for interrupt handler.
20485         (riscv_first_stack_step): Add forward declaration.
20486         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
20487         for interrupt handler with large frame.  Use it for saved reg list.
20488         (riscv_expand_prologue): Move flag_stack_usage_info support to
20489         eliminate duplication.
20490         (riscv_expand_epilogue): Generate mret for interrupt handler.
20491         (riscv_epilogue_uses): New.
20492         (riscv_can_use_return_insn): Return false for interrupt handler.
20493         (riscv_function_ok_for_sibcall): Likewise.
20494         (riscv_set_current_function): Add interrupt handler support.
20495         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
20496         * config/riscv/riscv.md (UNSPECV_MRET): New.
20497         (GP_REGNUM): New.
20498         (riscv_frflags, riscv_fsflags): Use tab after opcode.
20499         (riscv_mret): New.
20500         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
20502 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
20504         PR tree-optimization/85712
20505         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
20506         this candidate has already been replaced in-situ by a copy.
20508 2018-05-25  Jason Merrill  <jason@redhat.com>
20510         PR c++/80485 - inline function non-zero address.
20511         * symtab.c (nonzero_address): Check DECL_COMDAT.
20513 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
20515         PR target/83628
20516         * config/alpha/alpha.md (ashlsi3): New insn pattern.
20517         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
20518         extension of SImode operation.  Use const123_operand predicate.
20519         (*saddsi_1): Remove.
20520         (*saddl_se_1): Ditto.
20521         (*ssubsi_1): Ditto.
20522         (*ssubl_se_1): Ditto.
20523         * config/alpha/predicates.md (const123_operand): New predicate.
20524         * config/alpha/constraints.md (P): Use IN_RANGE.
20526 2018-05-25  Richard Biener  <rguenther@suse.de>
20528         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
20529         defaulted to true.
20530         (ref_maybe_used_by_stmt_p): Likewise.
20531         (stmt_may_clobber_ref_p): Likewise.
20532         (stmt_may_clobber_ref_p_1): Likewise.
20533         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
20534         and pass it along.
20535         (ref_maybe_used_by_stmt_p): Likewise.
20536         (stmt_may_clobber_ref_p): Likewise.
20537         (stmt_may_clobber_ref_p_1): Likewise.
20538         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
20539         the alias oracle to disambiguate DRs with stmts DR analysis
20540         couldn't handle.
20541         (vect_analyze_data_refs): Do not give up on not analyzable
20542         DRs for BB vectorization.  Remove code truncating the dataref
20543         vector.
20545 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
20547         PR target/85832
20548         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
20549         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
20550         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
20552 2018-05-25  Richard Biener  <rguenther@suse.de>
20554         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
20555         function, combining stmt data ref gathering and fatal analysis
20556         parts.
20557         (vect_analyze_data_refs): Remove now redudnant code and simplify.
20558         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
20559         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
20560         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
20561         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
20563 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
20565         PR tree-optimization/85720
20566         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
20567         SCC if all partitions are builtins.
20568         (version_loop_by_alias_check): New parameter.  Generate cancelable
20569         runtime alias check if all partitions are builtins.
20570         (distribute_loop): Update call to above function.
20572 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
20574         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
20575         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
20576         (parm_default_def_partition_arg): Ditto.
20577         (set_parm_default_def_partition): Ditto.
20578         (get_parm_default_def_partitions): Ditto and make it static.
20579         (get_undefined_value_partitions): Ditto and make it static.
20580         (remove_ssa_form): Refactor call to init_var_map here.
20581         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
20582         computation for loop region.
20583         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
20584         (register_default_def): Delete.
20585         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
20586         (parm_default_def_partition_arg): Ditto.
20587         (set_parm_default_def_partition): Ditto.
20588         (get_parm_default_def_partitions): Ditto and make it static.
20589         (get_undefined_value_partitions): Ditto and make it static.
20590         (coalesce_with_default, coalesce_with_default): Update comment.
20591         (create_coalesce_list_for_region): New func factored out from
20592         create_outofssa_var_map.
20593         (populate_coalesce_list_for_outofssa): New func factored out from
20594         create_outofssa_var_map and coalesce_ssa_name.
20595         (create_outofssa_var_map): Delete.
20596         (coalesce_ssa_name): Refactor to support live range computation.
20597         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
20598         (get_parm_default_def_partitions): Delete.
20599         (get_undefined_value_partitions): Ditto.
20600         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
20601         computation for loop region.
20602         (new_tree_live_info, loe_visit_block): Ditto.
20603         (live_worklist, set_var_live_on_entry): Ditto.
20604         (calculate_live_on_exit, verify_live_on_entry): Ditto.
20605         * tree-ssa-live.h (struct _var_map): New fields.
20606         (init_var_map): Change decl.
20607         (region_contains_p): New.
20609 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
20611         * tree-ssa-live.h (live_merge_and_clear): Delete.
20613 2018-05-25  Richard Biener  <rguenther@suse.de>
20615         PR c++/85912
20616         * tree-dump.c (dequeue_and_dump): Remove access to removed
20617         operand 2 of a SWITCH_EXPR.
20619 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20621         * doc/sourcebuild.texi (vect_double_cond_arith): Include
20622         multiplication and division.
20623         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
20624         (cond_udiv@var{m}, cond_umod@var{m}): Document.
20625         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
20626         (cond_udiv_optab, cond_umod_optab): New optabs.
20627         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
20628         (IFN_COND_RDIV): New internal functions.
20629         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
20630         TRUNC_MOD_EXPR and RDIV_EXPR.
20631         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
20632         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
20633         New unspecs.
20634         (SVE_INT_BINARY): Include mult.
20635         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
20636         (optab, sve_int_op): Handle mult.
20637         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
20638         UNSPEC_COND_DIV.
20639         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
20640         for SVE_INT_BINARY_SD.
20642 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20644         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
20645         (optab, sve_int_op): Handle div and udiv.
20646         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
20647         for SVE_INT_BINARY_SD.
20648         (*<optab><mode>3): New insn for the same.
20650 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20652         * tree-vect-patterns.c: Include predict.h.
20653         (vect_recog_divmod_pattern): Restrict check for division support
20654         to when optimizing for size.
20656 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20658         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
20659         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
20660         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
20661         (gimple_match_op::set_op): Likewise.
20662         (gimple_resimplify4): Declare.
20663         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
20664         (expr::gen_transform): Likewise.
20665         (decision_tree::gen): Generate a simplification routine for 4 operands.
20666         * gimple-match-head.c (gimple_simplify): Add an overload for
20667         4 operands.  In the top-level function, handle up to 4 call
20668         arguments and call gimple_resimplify4.
20669         (gimple_resimplify4): New function.
20670         (build_call_internal): Pass a fourth operand.
20671         (maybe_push_to_seq): Likewise.
20672         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
20673         Fold VEC_COND_EXPRs of an operation and a default value into
20674         an IFN_COND_* function if possible.
20675         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
20676         New unspecs.
20677         (SVE_COND_FP_BINARY): Include them.
20678         (optab, sve_fp_op): Handle them.
20679         (SVE_INT_BINARY_REV): New code iterator.
20680         (SVE_COND_FP_BINARY_REV): New int iterator.
20681         (commutative): New int attribute.
20682         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
20683         Declare.
20684         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
20685         function.
20686         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
20687         (*cond_<optab><mode>): New patterns for reversed operands.
20689 2018-05-25  Richard Biener  <rguenther@suse.de>
20691         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
20692         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
20693         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
20694         (STMT_VINFO_GROUPED_ACCESS): Adjust.
20695         * tree-vect-data-refs.c (everywhere): Adjust users.
20696         * tree-vect-loop.c (everywhere): Likewise.
20697         * tree-vect-slp.c (everywhere): Likewise.
20698         * tree-vect-stmts.c (everywhere): Likewise.
20699         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
20701 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20703         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
20704         Rename to...
20705         (gcc_cv_as_section_exclude): ... this.
20706         Try Solaris as #exclude syntax.
20707         * configure: Regenerate.
20708         * config.in: Regenerate.
20709         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
20710         SECTION_EXCLUDE.
20711         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
20712         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
20714         * varasm.c (default_elf_asm_named_section): Don't check if
20715         HAVE_GAS_SECTION_EXCLUDE is defined.
20717 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20719         * doc/md.texi: Update the documentation of the cond_* optabs
20720         to mention the new final operand.  Fix GET_MODE_NUNITS call.
20721         Describe the scalar case too.
20722         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
20723         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
20724         instead of 2.
20725         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
20726         (get_conditional_internal_fn): Update comment.
20727         * tree-vect-loop.c (vectorizable_reduction): Pass the original
20728         accumulator value as a final argument to conditional functions.
20729         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
20730         a define_expand and add an "else" operand.  Assert for now that
20731         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
20732         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
20733         (*cond_<optab><mode>): New patterns.
20734         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
20735         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
20736         (UNSPEC_COND_EOR): Delete.
20737         (optab): Remove associated mappings.
20738         (SVE_INT_BINARY): New code iterator.
20739         (sve_int_op): Remove int attribute and add "minus" to the code
20740         attribute.
20741         (SVE_COND_INT_OP): Delete.
20742         (SVE_COND_FP_OP): Rename to...
20743         (SVE_COND_FP_BINARY): ...this.
20745 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20747         * optabs.c (can_reuse_operands_p): New function.
20748         (maybe_legitimize_operands): Try to reuse the results for
20749         earlier operands.
20751 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
20753         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
20754         Add {q} suffix to insn mnemonic.
20756 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20758         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
20759         (msp430_warn_func_return): New.
20761 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
20763         * fold-const.c (tree_nonzero_bits): New function.
20764         * fold-const.h (tree_nonzero_bits): Likewise.
20765         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
20766         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
20768 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
20770         PR target/85900
20771         PR target/85345
20772         * varasm.c (assemble_alias): Check ifunc_resolver only on
20773         FUNCTION_DECL.
20775 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
20777         PR target/85903
20778         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
20779         when memory input operand is handled.
20781 2018-05-24  Luis Machado  <luis.machado@linaro.org>
20783         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
20784         global.
20785         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
20787 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
20789         * match.pd: Delay FMA folds until after vectorization.
20791 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
20793         PR target/83009
20794         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
20795         address check not strict.
20797 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
20799         * gimple-match.h (gimple_match_op): New class.
20800         (mprts_hook): Replace parameters with a gimple_match_op *.
20801         (maybe_build_generic_op): Likewise.
20802         (gimple_simplified_result_is_gimple_val): Replace parameters with
20803         a const gimple_match_op *.
20804         (gimple_simplify): Replace code_helper * and tree * parameters with
20805         a gimple_match_op * parameter.
20806         (gimple_resimplify1): Replace code_helper *, tree and tree *
20807         parameters with a gimple_match_op * parameter.
20808         (gimple_resimplify2): Likewise.
20809         (gimple_resimplify3): Likewise.
20810         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
20811         parameters with a gimple_match_op * parameter.
20812         * gimple-match-head.c (gimple_simplify): Change prototypes of
20813         auto-generated functions to take a gimple_match_op * instead of
20814         separate code_helper * and tree * parameters.  Make the same
20815         change in the top-level overload and update calls to the
20816         gimple_resimplify routines.  Update calls to the auto-generated
20817         functions and to maybe_push_res_to_seq in the publicly-facing
20818         operation-specific gimple_simplify overloads.
20819         (gimple_match_op::MAX_NUM_OPS): Define.
20820         (gimple_resimplify1): Replace rcode and ops with a single res_op
20821         parameter.  Update call to gimple_simplify.
20822         (gimple_resimplify2): Likewise.
20823         (gimple_resimplify3): Likewise.
20824         (mprts_hook): Replace parameters with a gimple_match_op *.
20825         (maybe_build_generic_op): Likewise.
20826         (build_call_internal): Replace type, nargs and ops with
20827         a gimple_match_op *.
20828         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
20829         with a single gimple_match_op *.  Update calls to mprts_hook,
20830         build_call_internal and gimple_simplified_result_is_gimple_val.
20831         Factor out code that is common to the tree_code and combined_fn cases.
20832         * genmatch.c (expr::gen_transform): Replace tem_code and
20833         tem_ops with a gimple_match_op called tem_op.  Update calls
20834         to the gimple_resimplify functions and maybe_push_res_to_seq.
20835         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
20836         res_ops.  Update call to the gimple_resimplify functions.
20837         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
20838         (decision_tree::gen): Make the functions take a gimple_match_op *
20839         called res_op instead of separate res_code and res_ops parameters.
20840         Update call accordingly.
20841         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
20842         and ops with a single res_op parameter.  Update calls to
20843         maybe_build_generic_op and maybe_push_res_to_seq.
20844         (fold_stmt_1): Update calls to gimple_simplify and
20845         replace_stmt_with_simplification.
20846         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
20847         and gimple_simplified_result_is_gimple_val.
20848         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
20849         gimple_simplify.
20850         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
20851         with a gimple_match_op *.
20852         (vn_nary_build_or_lookup): Likewise.  Update call to
20853         vn_nary_build_or_lookup_1.
20854         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
20855         gimple_match_op *.  Update calls to the gimple_resimplify routines
20856         and to gimple_simplified_result_is_gimple_val.
20857         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
20858         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
20859         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
20860         (visit_nary_op): Likewise.
20861         (visit_reference_op_load): Likewise.
20863 2018-05-23  Luis Machado  <luis.machado@linaro.org>
20865         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
20866         modifier for printing the step amount.
20868 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20870         PR target/78849
20871         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
20872         types.
20874 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
20876         * doc/sourcebuild.texi (Endianness): New subsubsection.
20878 2018-05-23  Luis Machado  <luis.machado@linaro.org>
20880         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20881         <prefetch_dynamic_strides>: New const bool field.
20882         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20883         prefetch_dynamic_strides.
20884         (exynosm1_prefetch_tune): Likewise.
20885         (thunderxt88_prefetch_tune): Likewise.
20886         (thunderx_prefetch_tune): Likewise.
20887         (thunderx2t99_prefetch_tune): Likewise.
20888         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
20889         false.
20890         (aarch64_override_options_internal): Update to set
20891         PARAM_PREFETCH_DYNAMIC_STRIDES.
20892         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20893         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20894         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20895         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20896         prefetch-dynamic-strides setting.
20898 2018-05-23  Luis Machado  <luis.machado@linaro.org>
20900         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20901         <minimum_stride>: New const int field.
20902         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20903         minimum_stride field defaulting to -1.
20904         (exynosm1_prefetch_tune): Likewise.
20905         (thunderxt88_prefetch_tune): Likewise.
20906         (thunderx_prefetch_tune): Likewise.
20907         (thunderx2t99_prefetch_tune): Likewise.
20908         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
20909         <default_opt_level>: Set to 3.
20910         (aarch64_override_options_internal): Update to set
20911         PARAM_PREFETCH_MINIMUM_STRIDE.
20912         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20913         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20914         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20915         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20916         stride is constant and is below the minimum stride threshold.
20918 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20920         * config/arm/arm-cpus.in (mode26): Delete.
20921         (armv4): Delete mode26 reference.
20922         * config/arm/arm.c (arm_configure_build_target): Delete use of
20923         isa_bit_mode26.
20925 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
20927         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
20928         New insn pattern.
20929         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
20930         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
20931         for non-SSE modes.
20932         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
20933         (floatunsdidf2): Ditto.
20935 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
20937         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
20938         (fixuns_trunc<mode>si2_avx512f): Ditto.
20939         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
20940         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
20941         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
20943 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
20945         PR rtl-optimization/79985
20946         * df-scan.c (df_insn_refs_collect): Remove special case for
20947         global registers and asm statements.
20949 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
20951         * extend.texi (Global Register Variables): Rewrite the bullet list.
20952         Note that the register is available for allocation. Note that access
20953         via inline asm must use constraints. Add note about async-signal
20954         handlers. Remove paragraph about automagic register selection.
20956 2018-05-23  Richard Biener  <rguenther@suse.de>
20958         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
20959         of fixed offset from memset VN.
20961 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
20963         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
20964         first_interp field.
20965         (alloc_cand_and_find_basis): Initialize first_interp field.
20966         (slsr_process_mul): Modify first_interp field.
20967         (slsr_process_add): Likewise.
20968         (slsr_process_cast): Modify first_interp field for each new
20969         interpretation.
20970         (slsr_process_copy): Likewise.
20971         (dump_candidate): Dump first_interp field.
20972         (replace_mult_candidate): Process all interpretations, not just
20973         subsequent ones.
20974         (replace_rhs_if_not_dup): Likewise.
20975         (replace_one_candidate): Likewise.
20977 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
20979         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
20980         Add new boolean.
20981         (aarch64_needs_frame_chain): New function.
20982         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
20984 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
20986         PR target/84882
20987         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
20988         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
20989         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
20990         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
20991         as true for strict-align.
20992         (aarch64_can_inline_p): Perform checks even when callee has no
20993         attributes to check for strict alignment.
20994         * doc/extend.texi (AArch64 Function Attributes): Document
20995         no-strict-align.
20996         * doc/invoke.texi: (AArch64 Options): Likewise.
20998 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
21000         PR tree-optimization/85853
21001         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
21002         the handling of the root of the node to...
21003         (vect_slp_analyze_node_operations_1): ...this new function,
21004         and run the whole thing with the child nodes' def types
21005         set according to their SLP node's def type.
21007 2018-05-23  Richard Biener  <rguenther@suse.de>
21009         PR middle-end/85874
21010         * tree-data-ref.c (create_runtime_alias_checks): Defer
21011         and ignore overflow warnings.
21013 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
21015         PR tree-optimization/85822
21016         * tree-vrp.c (is_masked_range_test): Fix handling of negative
21017         constants.
21019 2018-05-23  Richard Biener  <rguenther@suse.de>
21021         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
21022         memset constants via native_interpret_expr.
21024 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
21026         PR target/85345
21027         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
21028         attribute.
21029         (cgraph_node::create_alias): Likewise.
21030         (cgraph_node::get_availability): Check ifunc_resolver instead
21031         of looking up ifunc attribute.
21032         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
21033         * varasm.c (do_assemble_alias): Likewise.
21034         (assemble_alias): Likewise.
21035         (default_binds_local_p_3): Likewise.
21036         * cgraph.h (cgraph_node): Add ifunc_resolver.
21037         (cgraph_node::only_called_directly_or_aliased_p): Return false
21038         for IFUNC resolver.
21039         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
21040         attribute.
21041         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
21042         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
21043         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
21044         instead of looking up ifunc attribute.
21046 2018-05-22  Luis Machado  <luis.machado@linaro.org>
21048         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
21050 2018-05-22  Martin Sebor  <msebor@redhat.com>
21052         PR middle-end/85359
21053         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
21054         only when expasion succeeds.
21055         (expand_builtin_strcmp): Same.
21056         (expand_builtin_strncmp): Same.
21058 2018-05-22  Martin Sebor  <msebor@redhat.com>
21060         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
21062 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
21063             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21065         * config/aarch64/aarch64-ldpstp.md: Replace uses of
21066         aarch64_mem_pair_operand with memory_operand and delete operand swapping
21067         code.
21068         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
21069         Add check for legitimate_address.
21070         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
21071         (aarch64_swap_ldrstr_operands): New.
21072         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
21073         Define prototype.
21075 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
21076             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21078         * config/aarch64/aarch64.md: New patterns to generate stp
21079         and ldp.
21080         (store_pair_sw, store_pair_dw): New patterns to generate stp for
21081         single words and double words.
21082         (load_pair_sw, load_pair_dw): Likewise.
21083         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
21084         Delete.
21085         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
21086         Delete.
21087         * config/aarch64/aarch64-ldpstp.md: Modify peephole
21088         for different mode ldpstp and add peephole for merged zero stores.
21089         Likewise for loads.
21090         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
21091         Add size check.
21092         (aarch64_gen_store_pair): Rename calls to match new patterns.
21093         (aarch64_gen_load_pair): Rename calls to match new patterns.
21094         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
21095         (load_pair<DREG:mode><DREG2:mode>): ... This.
21096         (store_pair<mode>): Rename to...
21097         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
21098         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
21099         New mode iterators.
21100         (V_INT_EQUIV): Handle SImode.
21101         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
21102         New predicate.
21104 2018-05-22  Martin Sebor  <msebor@redhat.com>
21106         PR c/85623
21107         * calls.c (maybe_warn_nonstring_arg): Use string length to set
21108         or ajust the presumed bound on an operation to avoid unnecessary
21109         warnings.
21111 2018-05-22  Martin Sebor  <msebor@redhat.com>
21113         PR tree-optimization/85826
21114         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
21115         assuming that a DECL necesarily has a constant size.
21117 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
21119         PR middle-end/85862
21120         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
21122 2018-05-22  Richard Biener  <rguenther@suse.de>
21124         PR tree-optimization/85834
21125         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
21126         non-constant and non-zero memset arguments.
21128 2018-05-22  Martin Liska  <mliska@suse.cz>
21130         PR ipa/85607
21131         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
21133 2018-05-22  Richard Biener  <rguenther@suse.de>
21135         PR tree-optimization/85863
21136         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
21137         comparisons when vectype is specified.
21138         (vectorizable_condition): Do not specify vectype for
21139         vect_is_simple_cond when SLP vectorizing.
21141 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
21143         PR target/85657
21144         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
21145         define __ibm128 as long double.
21146         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
21147         as a distinct type when IEEE 128-bit support is enabled.
21148         (init_float128_ieee): Fix up conversions between IFmode and IEEE
21149         128-bit types to use the correct functions.
21150         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
21151         convert between 128-bit floating point types that have different
21152         modes but the same representation, instead of using gen_lowpart to
21153         makean alias.
21154         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
21155         KFmode.
21156         (IFKF_reg): New attributes to give the register constraints for
21157         IFmode and KFmode.
21158         (extend<mode>tf2_internal): New insns to mark an explicit
21159         conversion between 128-bit floating point types that have a
21160         different mode but share the same representation.
21162 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
21164         PR tree-optimization/85814
21165         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
21166         a null return from get_strinfo when unsharing the next
21167         strinfo in the chain.
21169 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21171         PR gcc/84923
21172         * varasm.c (weak_finish): Clean up weak_decls.
21174 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21176         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
21177         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
21178         UNSPEC_UADALP values.
21179         * config/aarch64/iterators.md (ABAL): New int iterator.
21180         (ABDL2): Likewise.
21181         (ADALP): Likewise.
21182         (sur): Add mappings for the above.
21183         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
21184         New define_insn.
21185         (aarch64_<sur>abal<mode>_4): Likewise.
21186         (aarch64_<sur>adalp<mode>_3): Likewise.
21187         (<sur>sadv16qi): New define_expand.
21189 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
21191         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
21192         (*movdf_internal): Ditto.
21193         (*rcpsf2_sse): Ditto.
21194         (*rsqrtsf2_sse): Ditto.
21195         (*sqrt<mode>2_sse): Ditto.
21197 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
21199         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
21200         eor3q<mode>4.
21201         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
21202         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
21203         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
21204         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
21205         vbcaxq_s64): New.
21206         * config/aarch64/arm_neon.h: Likewise.
21207         * config/aarch64/iterators.md (VQ_I): New.
21209 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
21211         * config.gcc: Add arc/t-multilib-linux to tmake_file for
21212         arc*-*-linux*.
21213         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
21214         MULTILIB_DIRNAMES
21216 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
21218         * config/nds32/constraints.md (S): New constraint.
21219         * config/nds32/nds32.md (call_internal): Use constraint S.
21220         (call_value_internal): Likewise.
21221         (sibcall_internal): Likewise.
21222         (sibcall_value_internal): Likewise.
21224 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
21225             Chung-Ju Wu  <jasonwucj@gmail.com>
21227         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
21228         into consideration.
21230 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
21231             Chung-Ju Wu  <jasonwucj@gmail.com>
21233         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
21234         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
21235         (nds32_rtx_costs_impl): Simplify.
21236         (nds32_address_cost_impl): Simplify.
21237         (nds32_init_rtx_costs): New function.
21238         (nds32_rtx_costs_speed_prefer): Likewise.
21239         (nds32_rtx_costs_size_prefer): Likewise.
21240         (nds32_address_cost_speed_prefer): Likewise.
21241         (nds32_address_cost_speed_fwprop): Likewise.
21242         (nds32_address_cost_size_prefer): Likewise.
21243         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
21244         * config/nds32/nds32.c (nds32_option_override): Use
21245         nds32_init_rtx_costs function.
21247 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
21249         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
21250         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
21251         (TARGET_PIPELINE_N8): Likewise.
21252         (TARGET_PIPELINE_N10): Likewise.
21253         (TARGET_PIPELINE_N13): Likewise.
21254         (TARGET_PIPELINE_GRAYWOLF): Likewise.
21256 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
21258         * config/nds32/nds32-fpu.md: Update copyright year.
21260 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21262         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
21264 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21266         * config/nds32/nds32.c
21267         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
21268         * config/nds32/nds32.opt (minline-asm-r15): New option.
21270 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21272         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
21273         MASK_HW_ABS.
21274         * config/nds32/nds32.md (abssi2): New pattern.
21276 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
21278         * config/i386/i386.md (rex64namesuffix): New mode attribute.
21279         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
21280         Merge insn pattern from sse_cvtsi2ss<round_name> and
21281         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
21282         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
21283         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
21284         using SWI48 mode iterator.
21285         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
21286         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
21287         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
21288         pattern from sse_cvttss2si<round_saeonly_name>
21289         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
21290         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
21291         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
21292         using SWI48 mode iterator.
21293         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
21294         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
21295         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
21296         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
21297         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
21298         using SWI48 mode iterator.
21299         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
21300         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
21301         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
21302         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
21303         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
21304         SWI48 mode iterator.
21305         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
21306         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
21307         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
21308         pattern from sse_cvttsd2si<round_saeonly_name>
21309         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
21311 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21313         * config/nds32/nds32-md-auxiliary.c
21314         (nds32_valid_smw_lwm_base_p): Refine.
21315         (nds32_output_smw_single_word): Refine.
21316         (nds32_output_smw_double_word): New.
21317         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
21319 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21321         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
21322         (nds32_output_stack_pop): Refine.
21323         (nds32_expand_unaligned_load): Refine.
21324         (nds32_expand_unaligned_store): Refine.
21326 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
21327             Chung-Ju Wu  <jasonwucj@gmail.com>
21329         * config/nds32/constants.md: Add TP_REGNUM constant.
21330         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
21331         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
21332         UNSPEC_ADD32.
21333         * config/nds32/nds32-doubleword.md: Consider flag_pic.
21334         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
21335         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
21336         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
21337         and PIC code generation.
21338         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
21339         code generation.
21340         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
21341         optimization.
21342         * config/nds32/nds32.md: Support TLS and PIC.
21343         * config/nds32/nds32.c: Support TLS and PIC.
21344         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
21345         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
21346         predicate.
21348 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21350         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
21351         mode with E_ prefix.
21353 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
21354             Chung-Ju Wu  <jasonwucj@gmail.com>
21356         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
21357         * config/nds32/nds32-md-auxiliary.c
21358         (symbolic_reference_mentioned_p): New.
21359         (nds32_legitimize_ict_address): New.
21360         (nds32_expand_ict_move): New.
21361         (nds32_indirect_call_referenced_p): New.
21362         (nds32_symbol_binds_local_p): Delete.
21363         (nds32_long_call_p): Modify.
21364         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
21365         * config/nds32/nds32-protos.h
21366         (symbolic_reference_mentioned_p): Declare.
21367         (nds32_legitimize_ict_address): Declare.
21368         (nds32_expand_ict_move): Declare.
21369         (nds32_indirect_call_referenced_p): Declare.
21370         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
21371         (nds32_relax_group): Use nds32_ict_const_p as condition.
21372         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
21373         (nds32_asm_file_start): Output ict_model directive in asm code.
21374         (nds32_legitimate_address_p): Consider indirect call.
21375         (nds32_print_operand): Consider indirect call.
21376         (nds32_print_operand_address): Consider indirect call.
21377         (nds32_insert_attributes): Handle "indirect_call" attribute.
21378         (TARGET_LEGITIMATE_ADDRESS_P): Define.
21379         (TARGET_LEGITIMATE_CONSTANT_P): Define.
21380         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
21381         (TARGET_DELEGITIMIZE_ADDRESS): Define.
21382         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
21383         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
21384         (TARGET_ICT_MODEL_SMALL): Define.
21385         (TARGET_ICT_MODEL_LARGE): Define.
21386         * config/nds32/nds32.md (movsi): Consider ict model.
21387         (call, call_value): Consider ict model.
21388         (sibcall, sibcall_value): Consider ict model.
21389         * config/nds32/nds32.opt (mict-model): New option.
21390         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
21391         model.
21393 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
21394             Monk Chiang  <sh.chiang04@gmail.com>
21395             Jim Wilson <jimw@sifive.com>
21397         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
21398         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
21399         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
21400         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
21401         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
21402         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
21403         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
21404         compute save_libcall_adjustment properly.
21405         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
21406         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
21407         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
21408         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
21409         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
21410         (ABI_SPEC): Handle mabi=ilp32e.
21411         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
21412         (RVE): Add RVE mask.
21413         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
21414         <-march>: Add rv32e as an example.
21416 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
21418         PR c++/82899
21419         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
21420         (intra_create_variable_infos): Handle C++ constructors.
21422 2018-05-18  Martin Liska  <mliska@suse.cz>
21424         * passes.def: Remove a redundant pass.
21426 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
21428         PR bootstrap/85838
21429         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
21431 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21433         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
21434         (ARMv4): Update.
21435         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
21436         (ARMv6m): Update.
21437         (armv2, armv2a, armv3, armv3m): Delete architectures.
21438         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
21439         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
21440         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
21441         Delete cpus.
21442         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
21443         (*mulsidi3adddi): Likewise.
21444         (mulsidi3): Likewise.
21445         (*mulsidi3_nov6): Likewise.
21446         (umulsidi3): Likewise.
21447         (umulsidi3_nov6): Likewise.
21448         (umaddsidi4): Likewise.
21449         (*umulsidi3adddi): Likewise.
21450         (smulsi3_highpart): Likewise.
21451         (*smulsi3_highpart_nov6): Likewise.
21452         (umulsi3_highpart): Likewise.
21453         (*umulsi3_highpart_nov6): Likewise.
21454         * config/arm/arm.h (arm_arch3m): Delete.
21455         * config/arm/arm.c (arm_arch3m): Delete.
21456         (arm_option_override_internal): Update armv3-related comment.
21457         (arm_configure_build_target): Delete use of isa_bit_mode32.
21458         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
21459         (arm_rtx_costs_internal): Delete check of arm_arch3m.
21460         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
21461         (mulsa3): Likewise.
21462         (mulusa3): Likewise.
21463         * config/arm/arm-protos.h (arm_arch3m): Delete.
21464         * config/arm/arm-tables.opt: Regenerate.
21465         * config/arm/arm-tune.md: Likewise.
21466         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
21467         deleted architectures.
21469 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21471         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
21472         (armv5t, armv5te): New features.
21473         (ARMv5, ARMv5e): Delete fgroups.
21474         (ARMv5t, ARMv5te): Adjust for above changes.
21475         (ARMv6m): Likewise.
21476         (armv5, armv5e): Delete arches.
21477         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
21478         arm_arch5.
21479         (*call_reg_arm): Likewise.
21480         (*call_value_reg_armv5): Likewise.
21481         (*call_value_reg_arm): Likewise.
21482         (*call_symbol): Likewise.
21483         (*call_value_symbol): Likewise.
21484         (*sibcall_insn): Likewise.
21485         (*sibcall_value_insn): Likewise.
21486         (clzsi2): Likewise.
21487         (prefetch): Likewise.
21488         (define_split and define_peephole2 dependent on arm_arch5):
21489         Likewise.
21490         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
21491         arm_arch5e.
21492         (TARGET_ARM_QBIT): Likewise.
21493         (TARGET_DSP_MULTIPLY): Likewise.
21494         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
21495         (arm_arch5, arm_arch5e): Delete.
21496         (arm_arch5t, arm_arch5te): Declare.
21497         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
21498         (arm_arch5t): Declare.
21499         (arm_option_reconfigure_globals): Update for the above.
21500         (arm_options_perform_arch_sanity_checks): Update comment, replace
21501         use of arm_arch5 with arm_arch5t.
21502         (use_return_insn): Likewise.
21503         (arm_emit_call_insn): Likewise.
21504         (output_return_instruction): Likewise.
21505         (arm_final_prescan_insn): Likewise.
21506         (arm_coproc_builtin_available): Likewise.
21507         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
21508         arm_arch5e with arm_arch5t and arm_arch5te.
21509         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
21510         (arm_arch5t, arm_arch5te): Declare.
21511         * config/arm/arm-tables.opt: Regenerate.
21512         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
21513         * config/arm/t-multilib: Likewise.
21514         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
21515         instead of arm_arch5.
21516         (*call_reg_thumb1): Likewise.
21517         (*call_value_reg_thumb1_v5): Likewise.
21518         (*call_value_reg_thumb1): Likewise.
21519         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
21520         unreachable path.
21521         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
21523 2018-05-18  Martin Liska  <mliska@suse.cz>
21525         PR gcov-profile/84846
21526         * doc/gcov.texi: Document -t option of gcov tool.
21528 2018-05-18  Martin Liska  <mliska@suse.cz>
21530         PR gcov-profile/84846
21531         * gcov.c (print_usage): Add new -t option.
21532         (process_args): Handle the option.
21533         (generate_results): Use stdout as output when requested by
21534         the option.
21536 2018-05-18  Martin Liska  <mliska@suse.cz>
21538         PR gcov-profile/84846
21539         * coverage.c (coverage_init): Write PWD to .gcno file.
21540         * doc/gcov.texi: Document how working directory is printed.
21541         * gcov-dump.c (dump_gcov_file): Print PWD.
21542         * gcov.c (output_intermediate_file): Likewise.
21543         (read_graph_file): Read PWD string.
21544         (output_lines): Print PWD.
21546 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
21548         PR middle-end/85817
21549         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
21550         for retval and return false if all args to phi are zero.
21552 2018-05-18  Richard Biener  <rguenther@suse.de>
21554         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
21555         method.
21556         (evrp_dom_walker::before_dom_children): Call it.
21558 2018-05-18  Richard Biener  <rguenther@suse.de>
21560         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
21561         results when processing array refs with variable index.
21563 2018-05-18  Toon Moene  <toon@moene.org>
21565         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
21566         directly after that of -floop-interchange. Indicate that both
21567         options are enabled by default when specifying -O3.
21569 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21571         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
21572         iterator.  Delete separate integer-mode vec_set<mode> expander.
21573         (aarch64_simd_vec_setv2di): Delete.
21574         (vec_setv2di): Delete.
21575         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
21576         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
21577         the "w, r" alternative.
21579 2018-05-18  Martin Liska  <mliska@suse.cz>
21581         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
21582         * tree-pass.h (make_pass_lower_switch_O0): New function.
21583         * tree-switch-conversion.c (node_has_low_bound): Remove.
21584         (node_has_high_bound): Likewise.
21585         (node_is_bounded): Likewise.
21586         (class pass_lower_switch): Make it a template type and create
21587         two instances.
21588         (pass_lower_switch::execute): Add template argument.
21589         (make_pass_lower_switch): New function.
21590         (make_pass_lower_switch_O0): New function.
21591         (do_jump_if_equal): Remove.
21592         (emit_case_nodes): Simplify to just handle all 3 cases and leave
21593         all the hard work to tree optimization passes.
21595 2018-05-18  Martin Liska  <mliska@suse.cz>
21597         * dbgcnt.c (limit_low): Renamed from limit.
21598         (limit_high): New variable.
21599         (dbg_cnt_is_enabled): Check for upper limit.
21600         (dbg_cnt): Adjust dumping.
21601         (dbg_cnt_set_limit_by_index): Add new argument for high
21602         value.
21603         (dbg_cnt_set_limit_by_name): Likewise.
21604         (dbg_cnt_process_single_pair): Parse new format.
21605         (dbg_cnt_process_opt): Use strtok.
21606         (dbg_cnt_list_all_counters): Remove 'value' and add
21607         'limit_high'.
21608         * doc/invoke.texi: Document changes.
21610 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
21612         * doc/sourcebuild.texi (scalar_all_fma): Document.
21613         * tree.def (FMA_EXPR): Delete.
21614         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
21615         * internal-fn.c (ternary_direct): New macro.
21616         (expand_ternary_optab_fn): Likewise.
21617         (direct_ternary_optab_supported_p): Likewise.
21618         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
21619         * builtins.c (fold_builtin_fma): Delete.
21620         (fold_builtin_3): Don't call it.
21621         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
21622         * expr.c (expand_expr_real_2): Likewise.
21623         * fold-const.c (operand_equal_p): Likewise.
21624         (fold_ternary_loc): Likewise.
21625         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
21626         * gimple.c (DEFTREECODE): Likewise.
21627         * gimplify.c (gimplify_expr): Likewise.
21628         * optabs-tree.c (optab_for_tree_code): Likewise.
21629         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
21630         * tree-eh.c (operation_could_trap_p): Likewise.
21631         (stmt_could_throw_1_p): Likewise.
21632         * tree-inline.c (estimate_operator_cost): Likewise.
21633         * tree-pretty-print.c (dump_generic_node): Likewise.
21634         (op_code_prio): Likewise.
21635         * tree-ssa-loop-im.c (stmt_cost): Likewise.
21636         * tree-ssa-operands.c (get_expr_operands): Likewise.
21637         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
21638         * fold-const-call.h (fold_fma): Delete.
21639         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
21640         CFN_FNMA and CFN_FNMS.
21641         (fold_fma): Delete.
21642         * genmatch.c (combined_fn): New enum.
21643         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
21644         (commutative_op): New function.
21645         (commutate): Use it.  Handle more than 2 operands.
21646         (dt_operand::gen_gimple_expr): Use commutative_op.
21647         (parser::parse_expr): Allow :c to be used with non-binary
21648         operators if the commutative operand is known.
21649         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
21650         CFN_FMS, CFN_FNMA and CFN_FNMS.
21651         (backprop::process_assign_use): Remove FMA_EXPR handling.
21652         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
21653         (gen_hsa_fma): New function.
21654         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
21655         IFN_FNMA and IFN_FNMS.
21656         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
21657         * gimple-fold.h (follow_all_ssa_edges): Declare.
21658         * gimple-fold.c (follow_all_ssa_edges): New function.
21659         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
21660         gimple_build interface and use follow_all_ssa_edges to fold the result.
21661         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
21662         instead of checking for optabs directly.
21663         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
21664         rather than FMA_EXPRs.
21665         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
21666         call to IFN_FMA instead of an FMA_EXPR.
21668 2018-05-17  Jim Wilson  <jimw@sifive.com>
21670         * expr.c (do_tablejump): When converting index to Pmode, if we have a
21671         sign extended promoted subreg, and the range does not have the sign bit
21672         set, then do a sign extend.
21674         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
21675         test, check for sign extended subreg and/or constant operands, and
21676         do a sign extend in that case.
21678 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
21680         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
21681         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
21682         Add untyped.
21683         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
21684         Change logics_shift_reg to logics_shift_imm.
21685         (thunderx2t99_fp_loadpair_basic): Delete.
21686         (thunderx2t99_fp_storepair_basic): Delete.
21687         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
21688         (thunderx2t99_asimd_polynomial): Delete.
21689         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
21690         and neon_fp_mul_d_scalar_q.
21691         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
21692         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
21693         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
21694         (thunderx2t99_asimd_lut): Add missing tbl types.
21695         (thunderx2t99_asimd_ext): Delete.
21696         (thunderx2t99_asimd_load1_1_mult): Delete.
21697         (thunderx2t99_asimd_load1_2_mult): Delete.
21698         (thunderx2t99_asimd_load1_ldp): New.
21699         (thunderx2t99_asimd_load1): New.
21700         (thunderx2t99_asimd_load2): Add missing *load2* types.
21701         (thunderx2t99_asimd_load3): New.
21702         (thunderx2t99_asimd_load4): New.
21703         (thunderx2t99_asimd_store1_1_mult): Delete.
21704         (thunderx2t99_asimd_store1_2_mult): Delete.
21705         (thunderx2t99_asimd_store2_mult): Delete.
21706         (thunderx2t99_asimd_store2_onelane): Delete.
21707         (thunderx2t99_asimd_store_stp): New.
21708         (thunderx2t99_asimd_store1): New.
21709         (thunderx2t99_asimd_store2): New.
21710         (thunderx2t99_asimd_store3): New.
21711         (thunderx2t99_asimd_store4): New.
21713 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
21715         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
21716         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
21718 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
21719             Segher Boessenkool  <segher@kernel.crashing.org>
21721         PR target/85698
21722         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
21723         operand.
21725 2018-05-17  Richard Biener  <rguenther@suse.de>
21727         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
21728         for pruning loop and prune defs feeding only already visited PHIs.
21730 2018-05-17  Richard Biener  <rguenther@suse.de>
21732         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
21734 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
21735             Richard Biener  <rguenther@suse.de>
21737         PR tree-optimization/85793
21738         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
21739         for VMAT_ELEMENTWISE.
21741 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
21743         * internal-fn.h (lookup_internal_fn): Declare
21744         * internal-fn.c (lookup_internal_fn): New function.
21745         * gimple.c (gimple_build_call_from_tree): Handle calls to
21746         internal functions.
21747         * gimple-pretty-print.c (dump_gimple_call): Print "." before
21748         internal function names.
21749         * tree-pretty-print.c (dump_generic_node): Likewise.
21750         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
21752 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
21754         * gimple-fold.h (gimple_build): Make the function forms take
21755         combined_fn rather than built_in_function.
21756         (gimple_simplify): Likewise.
21757         * gimple-match-head.c (gimple_simplify): Likewise.
21758         * gimple-fold.c (gimple_build): Likewise.
21759         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
21760         rather than gimple_build_call_internal.
21761         (get_initial_defs_for_reduction): Likewise.
21762         (vect_create_epilog_for_reduction): Likewise.
21763         (vectorizable_live_operation): Likewise.
21765 2018-05-17  Martin Liska  <mliska@suse.cz>
21767         * gimple-ssa-sprintf.c (format_directive): Do not use
21768         space in between 'G_' and '('.
21770 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
21772         PR target/85323
21773         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
21774         even if the mask is not all ones.
21776         PR target/85323
21777         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
21778         vector.
21779         (ix86_gimple_fold_builtin): Likewise.
21781         PR target/85323
21782         * config/i386/i386.c: Include tree-vector-builder.h.
21783         (ix86_vector_shift_count): New function.
21784         (ix86_fold_builtin): Fold shift builtins by scalar count.
21785         (ix86_gimple_fold_builtin): Likewise.
21787         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
21788         _mm512_setzero): New intrinsics.
21790 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
21791             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21793         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
21794         code generation for cases where splatting a value is not useful.
21795         * simplify-rtx.c (simplify_ternary_operation): Simplify
21796         vec_merge across a vec_duplicate and a paradoxical subreg forming
21797         a vector mode to a vec_concat.
21799 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
21801         * config.gcc: Support "goldmont-plus".
21802         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
21803         "goldmont-plus".
21804         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21805         PROCESSOR_GOLDMONT_PLUS.
21806         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
21807         (processor_target_table): Add "goldmont-plus".
21808         (PTA_GOLDMONT_PLUS): Define.
21809         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
21810         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
21811         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
21812         (fold_builtin_cpu): Add "goldmont-plus".
21813         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
21814         (ix86_option_override_internal): Add "goldmont-plus".
21815         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
21816         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
21817         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
21818         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
21820 2018-05-17  Richard Biener  <rguenther@suse.de>
21822         PR tree-optimization/85757
21823         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
21824         remove defs that only feed that PHI from further processing.
21826 2018-05-16  Jim Wilson  <jimw@sifive.com>
21828         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
21829         asterisk to name.
21830         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
21832 2018-05-16  Mark Wielaard  <mark@klomp.org>
21834         * dwarf2out.c (count_index_strings): New function.
21835         (output_indirect_strings): Call count_index_strings and generate
21836         header for dwarf_version >= 5.
21838 2018-05-16  Mark Wielaard  <mark@klomp.org>
21840         * dwarf2out.c (dwarf_FORM): New function.
21841         (set_indirect_string): Use dwarf_FORM.
21842         (reset_indirect_string): Likewise.
21843         (size_of_die): Likewise.
21844         (value_format): Likewise.
21845         (output_die): Likewise.
21846         (add_skeleton_AT_string): Likewise.
21847         (output_macinfo_op): Likewise.
21848         (index_string): Likewise.
21849         (output_index_string_offset): Likewise.
21850         (output_index_string): Likewise.
21851         (count_index_strings): Likewise.
21853 2018-05-16  Carl Love  <cel@us.ibm.com>
21855         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
21856         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
21858 2018-05-16  Martin Jambor  <mjambor@suse.cz>
21860         * ipa-prop.c (ipa_free_all_edge_args): Remove.
21861         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
21863 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
21865         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
21866         (fnma<mode>4): Likewise.
21867         (fms<mode>4): Likewise.
21868         (fnms<mode>4): Likewise.
21869         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
21870         (aarch64_fnma<mode>4): Likewise.
21871         (aarch64_fms<mode>4): Likewise.
21872         (aarch64_fnms<mode>4): Likewise.
21873         (aarch64_fnmadd<mode>4): Likewise.
21875 2018-05-16  Jason Merrill  <jason@redhat.com>
21877         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
21879 2018-05-16  Richard Biener  <rguenther@suse.de>
21881         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
21882         (dump_stmt_cost): Declare.
21883         (add_stmt_cost): Dump cost we add.
21884         (add_stmt_costs): New function.
21885         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
21886         No longer exported.
21887         (vect_analyze_stmt): Adjust prototype.
21888         (vectorizable_condition): Likewise.
21889         (vectorizable_live_operation): Likewise.
21890         (vectorizable_reduction): Likewise.
21891         (vectorizable_induction): Likewise.
21892         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
21893         cost vector to pass to vectorizable_ and record afterwards.
21894         (vect_model_reduction_cost): Take cost vector argument and adjust.
21895         (vect_model_induction_cost): Likewise.
21896         (vectorizable_reduction): Likewise.
21897         (vectorizable_induction): Likewise.
21898         (vectorizable_live_operation): Likewise.
21899         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
21900         SLP_TREE_NUMBER_OF_VEC_STMTS.
21901         (vect_analyze_slp_cost_1): Remove.
21902         (vect_analyze_slp_cost): Likewise.
21903         (vect_slp_analyze_node_operations): Take visited args and
21904         a target cost vector.  Avoid processing already visited stmt sets.
21905         (vect_slp_analyze_operations): Use a local cost vector to gather
21906         costs and register those of non-discarded instances.
21907         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
21908         (vect_schedule_slp_instance): Remove copying of
21909         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
21910         zero.
21911         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
21912         adding cost.  Record cost entry location.
21913         (vect_prologue_cost_for_slp_op): Function to compute cost of
21914         a constant or invariant generated for SLP vect in the prologue,
21915         split out from vect_analyze_slp_cost_1.
21916         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
21917         (vect_model_promotion_demotion_cost): Likewise.
21918         (vect_model_store_cost): Likewise, make static.
21919         (vect_model_load_cost): Likewise.
21920         (vectorizable_bswap): Add cost vector arg and adjust.
21921         (vectorizable_call): Likewise.
21922         (vectorizable_simd_clone_call): Likewise.
21923         (vectorizable_conversion): Likewise.
21924         (vectorizable_assignment): Likewise.
21925         (vectorizable_shift): Likewise.
21926         (vectorizable_operation): Likewise.
21927         (vectorizable_store): Likewise.
21928         (vectorizable_load): Likewise.
21929         (vectorizable_condition): Likewise.
21930         (vectorizable_comparison): Likewise.
21931         (can_vectorize_live_stmts): Likewise.
21932         (vect_analyze_stmt): Likewise.
21933         (vect_transform_stmt): Adjust calls to vectorizable_*.
21934         * tree-vectorizer.c: Include gimple-pretty-print.h.
21935         (dump_stmt_cost): New function.
21937 2018-05-16  Richard Biener  <rguenther@suse.de>
21939         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
21940         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
21941         * tree-ssa-dse.c: Include tree-ssa-loop.h.
21942         (check_name): New callback.
21943         (dse_classify_store): Track cycles via a visited bitmap of PHI
21944         defs and simplify handling of in-loop and across loop dead stores
21945         and properly fail for loop-variant refs.  Handle byte-tracking with
21946         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
21947         limiting the walk.
21949 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
21951         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
21952         (vect_get_mask_type_for_stmt): Likewise.
21953         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
21954         split out from...
21955         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
21956         to determine the statement's vector type and the vector type that
21957         should be used for calculating nunits.  Deal with cases in which
21958         the type has to be deferred.
21959         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
21960         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
21961         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
21962         (vect_determine_vf_for_stmt): New functions, split out from...
21963         (vect_determine_vectorization_factor): ...here.
21964         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
21965         (vect_get_mask_type_for_stmt): New functions, split out from
21966         vect_determine_vectorization_factor.
21968 2018-05-16  Richard Biener  <rguenther@suse.de>
21970         * tree-cfg.c (verify_gimple_assign_ternary): Properly
21971         verify the [VEC_]COND_EXPR embedded comparison.
21973 2018-05-15  Martin Sebor  <msebor@redhat.com>
21975         PR tree-optimization/85753
21976         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
21977         RECORD_TYPE in addition to ARRAY_TYPE.
21979 2018-05-15  Martin Sebor  <msebor@redhat.com>
21981         PR middle-end/85643
21982         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
21984 2018-05-15  Richard Biener  <rguenther@suse.de>
21986         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
21987         add by_clobber_p one.  Change algorithm to collect all defs
21988         representing uses we need to walk and try reducing them to
21989         a single one before failing.
21990         (dse_dom_walker::dse_optimize_stmt): Adjust.
21992 2018-05-13  Mark Wielaard  <mark@klomp.org>
21994         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
21995         (size_of_loc_descr): Likewise.
21996         (output_loc_operands): Likewise.
21997         (output_loc_operands_raw): Likewise.
21998         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
21999         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
22000         (hash_loc_operands): Likewise.
22001         (compare_loc_operands): Likewise.
22003 2018-05-14  Mark Wielaard  <mark@klomp.org>
22005         * dwarf2out.c (count_index_addrs): New function.
22006         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
22008 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22010         PR tree-optimization/83648
22011         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
22012         return value as malloc candidate.
22014 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22016         PR ipa/85734
22017         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
22018         param as true in call to suggest_attribute.
22020 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
22022         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
22023         -mreadonly-in-sdata.
22025 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22027         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
22028         New pattern.
22029         (aarch64_crypto_aesd_fused): Likewise.
22031 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
22033         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
22034         (movsi_aarch64): Likewise.
22035         (load_pairsi): Likewise.
22036         (load_pairdi): Likewise.
22037         (store_pairsi): Likewise.
22038         (store_pairdi): Likewise.
22039         (load_pairsf): Likewise.
22040         (load_pairdf): Likewise.
22041         (store_pairsf): Likewise.
22042         (store_pairdf): Likewise.
22043         (zero_extend): Likewise.
22044         (trunc): Swap alternatives.
22045         (fcvt_target): Add '?' to prefer w over r.
22047 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
22049         PR target/85756
22050         * config/i386/i386.md: Disallow non-commutative arithmetics in
22051         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
22052         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
22053         in the peephole2 before it.
22055 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
22057         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
22058         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
22059         (ix86_handle_option): Handle -mcldemote.
22060         * config.gcc: New header.
22061         * config/i386/cldemoteintrin.h: New file.
22062         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
22063         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
22064         -mcldemote.
22065         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22066         OPTION_MASK_ISA_CLDEMOTE.
22067         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
22068         (ix86_valid_target_attribute_inner_p): Ditto.
22069         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
22070         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
22071         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
22072         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
22073         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
22074         (cldemote): New.
22075         * config/i386/i386.opt: Add -mcldemote.
22076         * config/i386/x86intrin.h: New header.
22077         * doc/invoke.texi: Add -mcldemote.
22079 2018-05-14  Richard Biener  <rguenther@suse.de>
22081         * doc/match-and-simplify.texi: Adjust :s documentation.
22083 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
22085         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
22086         intended memcpy size.
22087         (REORDER_45): Likewise.
22089 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
22091         * sort.cc: New file.
22092         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
22093         * vec.c (qsort_chk): Use gcc_qsort.
22094         * Makefile.in (OBJS-libcommon): Add sort.o.
22095         (build/sort.o): New target.  Use it...
22096         (BUILD_RTL): ... here, and...
22097         (build/gencfn-macros): ... here, and...
22098         (build/genmatch): ... here.
22100 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22101             Chung-Ju Wu  <jasonwucj@gmail.com>
22103         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
22104         * config/nds32/nds32-graywolf.md: New file.
22105         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
22106         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
22107         pipeline.
22108         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
22109         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
22110         * config/nds32/nds32.md (pipeline_model): Add graywolf.
22111         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
22112         * config/nds32/pipelines.md: Include n15 settings.
22114 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22115             Chung-Ju Wu  <jasonwucj@gmail.com>
22117         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
22118         * config/nds32/nds32-n13.md: New file.
22119         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
22120         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
22121         pipeline.
22122         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
22123         * config/nds32/nds32.md (pipeline_model): Add n13.
22124         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
22125         * config/nds32/pipelines.md: Include n13 settings.
22127 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22128             Chung-Ju Wu  <jasonwucj@gmail.com>
22130         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
22131         * config/nds32/nds32-n10.md: New file.
22132         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
22133         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
22134         pipeline.
22135         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
22136         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
22137         * config/nds32/nds32.md (pipeline_model): Add n10.
22138         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
22139         * config/nds32/pipelines.md: Include n10 settings.
22141 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
22142             Kito Cheng  <kito.cheng@gmail.com>
22143             Chung-Ju Wu  <jasonwucj@gmail.com>
22145         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
22146         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22147         Add enum values for DSP extension instructions.
22148         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
22149         New constraints.
22150         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
22151         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
22152         New code iterators.
22153         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
22154         * config/nds32/nds32-dspext.md: New file for DSP implementation.
22155         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
22156         * config/nds32/nds32-intrinsic.md: Likewise.
22157         * config/nds32/nds32_intrinsic.h: Likewise.
22158         * config/nds32/nds32-md-auxiliary.c: Likewise.
22159         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
22160         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
22161         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
22162         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
22163         * config/nds32/nds32-protos.h: New declarations for DSP extension.
22164         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
22165         TYPE_DMAC in switch statement.
22166         * config/nds32/nds32.c: New checking and implementation for DSP
22167         extension instructions.
22168         * config/nds32/nds32.h: Likewise.
22169         * config/nds32/nds32.md: Likewise.
22170         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
22171         * config/nds32/predicates.md: Implement new predicates for DSP
22172         extension.
22174 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
22176         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
22177         Reformat alternatives and attributes so it is easier to identify
22178         which constraints/attributes go with which instruction.
22179         (mov<mode>_hardfloat32, FMOVE64): Likewise.
22180         (mov<mode>_softfloat32, FMOVE64): Likewise.
22181         (mov<mode>_hardfloat64, FMOVE64): Likewise.
22182         (mov<mode>_softfloat64, FMOVE64): Likewise.
22184 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22186         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22187         subsection.
22188         (Basic PowerPC Built-in Functions): The new name of the
22189         subsection previously known as "PowerPC Built-in Functions".
22190         (Basic PowerPC Built-in Functions Available on all Configurations):
22191         New subsubsection.
22192         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
22193         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
22194         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
22195         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
22197 2018-05-11  Martin Jambor  <mjambor@suse.cz>
22199         PR ipa/85655
22200         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
22201         single const.
22203 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
22205         PR target/85733
22206         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
22208 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
22210         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
22211         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
22212         (ix86_handle_option): Handle -mwaitpkg.
22213         * config.gcc: New header.
22214         * config/i386/cpuid.h (bit_WAITPKG): New bit.
22215         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
22216         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
22217         function type.
22218         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22219         OPTION_MASK_ISA_WAITPKG.
22220         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
22221         (ix86_option_override_internal): Add PTA_WAITPKG.
22222         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
22223         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
22224         IX86_BUILTIN_TPAUSE.
22225         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
22226         __builtin_ia32_umwait and __builtin_ia32_tpause.
22227         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
22228         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
22229         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
22230         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
22231         UNSPECV_TPAUSE): New.
22232         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
22233         * config/i386/i386.opt: Add -mwaitpkg.
22234         * config/i386/waitpkgintrin.h: New file.
22235         * config/i386/x86intrin.h: New header.
22236         * doc/invoke.texi: Add -mwaitpkg.
22238 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
22240         PR target/85606
22241         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
22242         equivalent.
22243         (cortex-m0): Use armv6s-m isa.
22244         (cortex-m0plus): Likewise.
22245         (cortex-m1): Likewise.
22246         (cortex-m0.small-multiply): Likewise.
22247         (cortex-m0plus.small-multiply): Likewise.
22248         (cortex-m1.small-multiply): Likewise.
22250 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
22251             Jakub Jelinek  <jakub@redhat.com>
22253         PR tree-optimization/85692
22254         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
22255         source permute as well.
22257 2018-05-11  Martin Liska  <mliska@suse.cz>
22259         PR sanitizer/85556
22260         * doc/extend.texi: Document LLVM style format for no_sanitize
22261         attribute.
22263 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
22265         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
22266         mode_supports_vsx_dform_quad to mode_supports_dq_form.
22267         (mode_supports_vsx_dform_quad): Likewise.
22268         (mode_supports_vmx_dform): Move these functions to be next to the
22269         other mode_supports functions.
22270         (mode_supports_dq_form): Likewise.
22271         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
22272         mode_supports_dq_form.
22273         (reg_offset_addressing_ok_p): Likewise.
22274         (offsettable_ok_by_alignment): Likewise.
22275         (rs6000_legitimate_offset_address_p): Likewise.
22276         (legitimate_lo_sum_address_p): Likewise.
22277         (rs6000_legitimize_address): Likewise.
22278         (rs6000_legitimize_reload_address): Likewise.
22279         (rs6000_secondary_reload_inner): Likewise.
22280         (rs6000_preferred_reload_class): Likewise.
22281         (rs6000_output_move_128bit): Likewise.
22283 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
22285         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
22286         Generate SImode target register for null target.
22287         <case IX86_BUILTIN_XGETBV>: Ditto.
22288         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
22289         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
22291 2018-05-10  Carl Love  <cel@us.ibm.com>
22293         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
22294         dcbtt and dcbtstt if operands[2] is 0.
22296 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
22298         PR target/85693
22299         * config/i386/sse.md (usadv64qi): New expander.
22301 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
22303         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
22304         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
22305         -maltivec=be support.
22306         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
22307         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
22308         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
22309         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
22310         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
22311         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
22312         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
22313         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
22314         altivec_vsumsws): Adjust.
22315         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
22316         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
22317         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
22318         support.
22319         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
22320         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
22321         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
22322         (altivec_lve<VI_char>x): Delete expand.
22323         (*altivec_lve<VI_char>x_internal): Rename to...
22324         (altivec_lve<VI_char>x): ... this.
22325         (altivec_lvxl_<mode>): Delete expand.
22326         (*altivec_lvxl_<mode>_internal): Rename to ...
22327         (altivec_lvxl_<mode>): ... this.
22328         (altivec_stvxl_<mode>): Delete expand.
22329         (*altivec_stvxl_<mode>_internal): Rename to ...
22330         (altivec_stvxl_<mode>): ... this.
22331         (altivec_stve<VI_char>x): Delete expand.
22332         (*altivec_stve<VI_char>x_internal): Rename to ...
22333         (altivec_stve<VI_char>x): ... this.
22334         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
22335         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
22336         reduc_plus_scal_<mode>): Adjust.
22337         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
22338         comment.
22339         (rs6000_cpu_cpp_builtins): Adjust.
22340         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
22341         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
22342         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
22343         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
22344         -maltivec=be support.
22345         (rs6000_split_vec_extract_var): Adjust.
22346         (rs6000_split_v4si_init): Adjust.
22347         (swap_selector_for_mode): Delete.
22348         (altivec_expand_lvx_be, altivec_expand_stvx_be,
22349         altivec_expand_stvex_be): Delete.
22350         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
22351         -maltivec=be support.
22352         (rs6000_gimple_fold_builtin): Ditto.
22353         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
22354         Adjust.
22355         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
22356         (TARGET_DIRECT_MOVE_64BIT): Adjust.
22357         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
22358         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
22359         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
22360         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
22361         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
22362         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
22363         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
22364         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
22365         anonymous split): Adjust.
22366         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
22367         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
22369 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
22371         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
22372         when --with-gxx-include-dir is also specified.
22373         * configure: Regenerate.
22375 2018-05-09  Jim Wilson  <jimw@sifive.com>
22377         PR target/84797
22378         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
22379         * config/riscv/t-withmultilib: New.
22380         * config/riscv/withmultilib.h: New.
22381         * doc/install.texi: Document RISC-V --with-multilib-list support.
22383 2018-05-09  Richard Biener  <rguenther@suse.de>
22385         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
22386         vector.
22387         (vect_bb_vectorization_profitable_p): Adjust.  Compute
22388         actual scalar cost using the cost vector and the add_stmt_cost
22389         machinery.
22391 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22393         PR rtl-optimization/85645
22394         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
22395         in the REG_CFA_REGISTER note for LR, don't leave it empty.
22397 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22399         PR rtl-optimization/85645
22400         * shrink-wrap.c (spread_components): Return a boolean saying if
22401         anything was changed.
22402         (try_shrink_wrapping_separate): Iterate spread_components until
22403         nothing changes anymore.
22405 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22407         PR rtl-optimization/85645
22408         * regrename.c (build_def_use): Also kill the chains that include the
22409         destination of a REG_CFA_REGISTER note.
22411 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22413         PR rtl-optimization/85645
22414         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
22415         insn that has a REG_CFA_REGISTER note.
22417 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
22419         * cfgexpand.c (expand_clobber): New function.
22420         (expand_gimple_stmt_1): Use it.
22421         * tree-vect-stmts.c (vect_clobber_variable): New function,
22422         split out from...
22423         (vectorizable_simd_clone_call): ...here.
22424         (vectorizable_store): Emit a clobber either side of an
22425         IFN_STORE_LANES sequence.
22426         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
22428 2018-05-09  Tom de Vries  <tom@codesourcery.com>
22430         PR target/85626
22431         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
22432         (define_insn "trap_if_false"): Add exit after trap.
22434 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
22436         PR rtl-optimization/85638
22437         * bb-reorder.c: Include common/common-target.h.
22438         (create_forwarder_block): New function extracted from...
22439         (fix_up_crossing_landing_pad): ...here.  Rename into...
22440         (dw2_fix_up_crossing_landing_pad): ...this.
22441         (sjlj_fix_up_crossing_landing_pad): New function.
22442         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
22443         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
22444         from both partitions and exit the loop after one iteration.
22446 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22448         Revert:
22449         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22450         subsection.
22451         (Basic PowerPC Built-in Functions): The new name of the
22452         subsection previously known as "PowerPC Built-in Functions".
22453         (Basic PowerPC Built-in Functions Available on all Configurations):
22454         New subsubsection.
22455         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
22456         subsubsection.
22457         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
22458         subsubsection.
22459         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
22460         subsubsection.
22461         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
22462         subsubsection.
22464 2018-05-08  Jim Wilson  <jimw@sifive.com>
22466         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
22467         (LD_EMUL_SUFFIX): New.
22468         (LINK_SPEC): Use it.
22470 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22472         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22473         subsection.
22474         (Basic PowerPC Built-in Functions): The new name of the
22475         subsection previously known as "PowerPC Built-in Functions".
22476         (Basic PowerPC Built-in Functions Available on all Configurations):
22477         New subsubsection.
22478         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
22479         subsubsection.
22480         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
22481         subsubsection.
22482         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
22483         subsubsection.
22484         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
22485         subsubsection.
22487 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
22489         PR target/85683
22490         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
22491         after cmpelim optimization.
22493 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
22495         * config.gcc: Support "goldmont".
22496         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
22497         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22498         PROCESSOR_GOLDMONT.
22499         * config/i386/i386.c (m_GOLDMONT): Define.
22500         (processor_target_table): Add "goldmont".
22501         (PTA_GOLDMONT): Define.
22502         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
22503         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
22504         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
22505         (fold_builtin_cpu): Add "goldmont".
22506         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
22507         (ix86_option_override_internal): Add "goldmont".
22508         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
22509         (processor_type): Add PROCESSOR_GOLDMONT.
22510         * config/i386/i386.md: Add CPU "glm".
22511         * config/i386/glm.md: New file.
22512         * config/i386/x86-tune.def: Add m_GOLDMONT.
22513         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
22515 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
22517         PR target/85572
22518         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
22519         E_V4DImode.
22520         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
22521         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
22522         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
22524         PR target/85317
22525         * config/i386/i386.c (ix86_fold_builtin): Handle
22526         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
22528         PR target/85480
22529         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
22530         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
22532 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
22534         PR target/85658
22535         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
22536         (check_arch): Likewise.
22537         (check_fpu): Return the result rather than printing it.
22538         (end arch): Fix operator precedence.
22539         (end cpu): Likewise.
22540         (END): Print the result from check_fpu.
22542 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
22543             Alan Hayward  <alan.hayward@arm.com>
22544             David Sherwood  <david.sherwood@arm.com>
22546         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
22547         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
22548         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
22549         (*fcmuo<mode>_and): New patterns.
22551 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
22553         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
22554         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
22555         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
22556         (cmp_op, sve_imm_con): New code attributes.
22557         (SVE_COND_INT_CMP, imm_con): Delete.
22558         (cmp_op): Remove above unspecs from int attribute.
22559         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
22560         to...
22561         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
22562         comparison-specific unspecs.
22563         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
22564         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
22565         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
22566         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
22567         (*vec_fcm<cmp_op><mode>): Rename to...
22568         (*fcm<cmp_op><mode>): ...this and adjust likewise.
22569         (*vec_fcmuo<mode>): Rename to...
22570         (*fcmuo<mode>): ...this and adjust likewise.
22571         (*pred_fcm<cmp_op><mode>): New pattern.
22572         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
22573         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
22574         functions.
22575         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
22576         and UNORDERED.
22577         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
22578         (aarch64_emit_sve_predicated_cond): New function.
22579         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
22580         (aarch64_emit_unspec_cond_or): Replace with...
22581         (aarch64_emit_sve_or_conds): ...this new function.  Use
22582         aarch64_emit_sve_ptrue_op for the individual comparisons and
22583         aarch64_emit_binop to OR them together.
22584         (aarch64_emit_inverted_unspec_cond): Replace with...
22585         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
22586         aarch64_emit_sve_ptrue_op for the comparison and
22587         aarch64_emit_unop to invert the result.
22588         (aarch64_expand_sve_vec_cmp_float): Update after the above
22589         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
22591 2018-05-07  Nathan Sidwell  <nathan@acm.org>
22593         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
22594         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
22595         (Backwards Compatibility): Likewise.
22597 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22599         PR bootstrap/85681
22600         Revert:
22601         2018-05-07  Luis Machado  <luis.machado@linaro.org>
22603         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22604         <prefetch_dynamic_strides>: New const bool field.
22605         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22606         prefetch_dynamic_strides.
22607         (exynosm1_prefetch_tune): Likewise.
22608         (thunderxt88_prefetch_tune): Likewise.
22609         (thunderx_prefetch_tune): Likewise.
22610         (thunderx2t99_prefetch_tune): Likewise.
22611         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
22612         to false.
22613         (aarch64_override_options_internal): Update to set
22614         PARAM_PREFETCH_DYNAMIC_STRIDES.
22615         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22616         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22617         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22618         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22619         prefetch-dynamic-strides setting.
22621         2018-05-07  Luis Machado  <luis.machado@linaro.org>
22623         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22624         <minimum_stride>: New const int field.
22625         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22626         minimum_stride field.
22627         (exynosm1_prefetch_tune): Likewise.
22628         (thunderxt88_prefetch_tune): Likewise.
22629         (thunderx_prefetch_tune): Likewise.
22630         (thunderx2t99_prefetch_tune): Likewise.
22631         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
22632         (aarch64_override_options_internal): Update to set
22633         PARAM_PREFETCH_MINIMUM_STRIDE.
22634         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22635         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22636         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22637         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22638         stride is constant and is below the minimum stride threshold.
22640 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22642         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
22643         to 512.
22645 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22647         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22648         <prefetch_dynamic_strides>: New const bool field.
22649         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22650         prefetch_dynamic_strides.
22651         (exynosm1_prefetch_tune): Likewise.
22652         (thunderxt88_prefetch_tune): Likewise.
22653         (thunderx_prefetch_tune): Likewise.
22654         (thunderx2t99_prefetch_tune): Likewise.
22655         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
22656         to false.
22657         (aarch64_override_options_internal): Update to set
22658         PARAM_PREFETCH_DYNAMIC_STRIDES.
22659         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22660         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22661         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22662         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22663         prefetch-dynamic-strides setting.
22665 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22667         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22668         <minimum_stride>: New const int field.
22669         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22670         minimum_stride field.
22671         (exynosm1_prefetch_tune): Likewise.
22672         (thunderxt88_prefetch_tune): Likewise.
22673         (thunderx_prefetch_tune): Likewise.
22674         (thunderx2t99_prefetch_tune): Likewise.
22675         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
22676         (aarch64_override_options_internal): Update to set
22677         PARAM_PREFETCH_MINIMUM_STRIDE.
22678         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22679         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22680         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22681         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22682         stride is constant and is below the minimum stride threshold.
22684 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
22686         PR c++/85659
22687         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
22688         the type is addressable.  Don't force op into register if it has
22689         BLKmode.
22691 2018-05-05  Roland McGrath  <mcgrathr@google.com>
22693         PR other/77609
22694         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
22695         any section for which we don't know a specific type it should have,
22696         regardless of name.  Previously this was done only for the exact
22697         names ".init_array", ".fini_array", and ".preinit_array".
22698         (default_elf_asm_named_section): Add comment about
22699         relationship with default_section_type_flags and SECTION_NOTYPE.
22700         (get_section): Don't consider it a type conflict if one side has
22701         SECTION_NOTYPE and the other doesn't, as long as neither has the
22702         SECTION_BSS et al used in the default_section_type_flags logic.
22704 2018-05-05  Tom de Vries  <tom@codesourcery.com>
22706         PR target/85653
22707         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
22708         (workaround_barsyncs): New function.
22709         (nvptx_reorg): Use workaround_barsyncs.
22710         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22711         (define_expand "nvptx_membar_cta"): New define_expand.
22712         (define_insn "*nvptx_membar_cta"): New insn.
22714 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
22716         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
22717         To improve optimization opportunities.
22718         * builtin-types.def: The new needed builtin types for the above.
22720 2018-05-04  Richard Biener  <rguenther@suse.de>
22722         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
22723         * gimple-ssa-store-merging.c
22724         (imm_store_chain_info::output_merged_store): Remove redundant create,
22725         release split_store vector contents on failure.
22726         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
22727         scalar stmt vector on cache hit.
22729 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
22731         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
22732         Xilinx FP support.
22733         * config.gcc (powerpc-xilinx-eabi*): Remove.
22734         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
22735         support.
22736         (fusion_addis_mem_combo_load): Ditto.
22737         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
22738         FP support.
22739         (rs6000_cpu_cpp_builtins): Ditto.
22740         * config/rs6000/rs6000-linux.c
22741         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
22742         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
22743         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
22744         support.
22745         (rs6000_setup_reg_addr_masks): Ditto.
22746         (rs6000_init_hard_regno_mode_ok): Ditto.
22747         (rs6000_option_override_internal): Ditto.
22748         (legitimate_lo_sum_address_p): Ditto.
22749         (rs6000_legitimize_address): Ditto.
22750         (rs6000_legitimize_reload_address): Ditto.
22751         (rs6000_legitimate_address_p): Ditto.
22752         (abi_v4_pass_in_fpr): Ditto.
22753         (setup_incoming_varargs): Ditto.
22754         (rs6000_gimplify_va_arg): Ditto.
22755         (rs6000_split_multireg_move): Ditto.
22756         (rs6000_savres_strategy): Ditto.
22757         (rs6000_emit_prologue_components): Ditto.
22758         (rs6000_emit_epilogue_components): Ditto.
22759         (rs6000_emit_prologue): Ditto.
22760         (rs6000_emit_epilogue): Ditto.
22761         (rs6000_elf_file_end): Ditto.
22762         (rs6000_function_value): Ditto.
22763         (rs6000_libcall_value): Ditto.
22764         * config/rs6000/rs6000.h: Ditto.
22765         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
22766         (TARGET_MINMAX): ... this.  New.
22767         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
22768         * config/rs6000/rs6000.md: Remove Xilinx FP support.
22769         (*movsi_internal1_single): Delete.
22770         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
22771         mfpu=, mxilinx-fpu): Delete.
22772         * config/rs6000/singlefp.h: Delete.
22773         * config/rs6000/sysv4.h: Remove Xilinx FP support.
22774         * config/rs6000/t-rs6000: Ditto.
22775         * config/rs6000/t-xilinx: Delete.
22776         * config/rs6000/titan.md: Adjust for fp_type removal.
22777         * config/rs6000/vsx.md: Remove Xilinx FP support.
22778         (VStype_simple): Delete.
22779         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
22780         * config/rs6000/xfpu.h: Delete.
22781         * config/rs6000/xfpu.md: Delete.
22782         * config/rs6000/xilinx.h: Delete.
22783         * config/rs6000/xilinx.opt: Delete.
22784         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
22785         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
22787 2018-05-04  Tom de Vries  <tom@codesourcery.com>
22789         PR libgomp/85639
22790         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
22791         if ignore == 0.
22793 2018-05-04  Richard Biener  <rguenther@suse.de>
22795         PR middle-end/85627
22796         * tree-complex.c (update_complex_assignment): We are always in SSA form.
22797         (expand_complex_div_wide): Likewise.
22798         (expand_complex_operations_1): Likewise.
22799         (expand_complex_libcall): Preserve EH info of the original stmt.
22800         (tree_lower_complex): Handle removed blocks.
22801         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
22802         on complex multiplication and division libcall builtins.
22804 2018-05-04  Richard Biener  <rguenther@suse.de>
22806         PR middle-end/85574
22807         * fold-const.c (negate_expr_p): Restrict negation of operand
22808         zero of a division to when we know that can happen without
22809         overflow.
22810         (fold_negate_expr_1): Likewise.
22812 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
22814         PR libstdc++/85466
22815         * real.h (real_nextafter): Declare.
22816         * real.c (real_nextafter): New function.
22817         * fold-const-call.c (fold_const_nextafter): New function.
22818         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
22819         CASE_CFN_NEXTTOWARD.
22820         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
22821         even when arg1_mode is different from arg0_mode.
22823 2018-05-03  Nathan Sidwell  <nathan@acm.org>
22825         * doc/extend.texi (Deprecated Features): Remove
22826         -ffriend-injection.
22827         (Backwards Compatibility): Likewise.
22828         * doc/invoke.texi (C++ Language Options): Likewise.
22829         (C++ Dialect Options): Likewise.
22831 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
22833         PR target/85530
22834         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
22835         _mm512_mask_mullox_epi64): New intrinsics.
22837 2018-05-03  Tom de Vries  <tom@codesourcery.com>
22839         PR testsuite/85106
22840         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22841         dump files): Add offload-tree.
22843 2018-05-03  Richard Biener  <rguenther@suse.de>
22845         PR tree-optimization/85615
22846         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
22847         to loops not nested in BBs loop father to avoid creating multi-entry
22848         loops.
22850 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22852         PR tree-optimization/70291
22853         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
22854         arguments.  Change return type to tree.  Emit libcall as a new
22855         statement rather than replacing existing one when inplace_p is true.
22856         (expand_complex_multiplication_components): New function.
22857         (expand_complex_multiplication): Expand floating-point complex
22858         multiplication using the above.
22859         (expand_complex_division): Rename inner_type parameter to type.
22860         Update expand_complex_libcall call-site.
22861         (expand_complex_operations_1): Update expand_complex_multiplication
22862         and expand_complex_division call-sites.
22864 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
22866         PR target/85582
22867         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
22868         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
22869         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
22870         the highest significant bit of the shift count mask is clear.  In
22871         check whether and[sq]i3 is needed verify that all significant bits
22872         of the shift count other than the highest are set.
22874 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22876         PR libgomp/82428
22877         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
22878         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
22879         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
22880         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
22881         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
22882         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
22883         __builtin_goacc_parlevel_size.
22885 2018-05-02  Richard Biener  <rguenther@suse.de>
22887         PR tree-optimization/85597
22888         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
22889         do not use split vect_get_vec_defs call but call vect_get_slp_defs
22890         directly.
22892 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22894         PR testsuite/85106
22895         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22896         dump files): Add ltrans-tree.
22898 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22900         PR testsuite/85106
22901         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22902         dump files): Add wpa-ipa.
22904 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
22906         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
22907         powerpc*-*-linux*paired* target.
22908         * config/rs6000/750cl.h: Delete.
22909         * config/rs6000/paired.h: Delete.
22910         * config/rs6000/paired.md: Delete.
22911         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
22912         float support.
22913         * config/rs6000/rs6000-builtin.def: Remove paired float support.
22914         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
22915         comment.  Remove paired float support.
22916         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
22917         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
22918         VECTOR_PAIRED.
22919         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
22920         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
22921         declarations.
22922         * config/rs6000/rs6000.c: Remove paired float support.
22923         (paired_expand_vector_init, paired_expand_vector_move,
22924         paired_emit_vector_compare, paired_emit_vector_cond_expr,
22925         (paired_expand_lv_builtin, paired_expand_stv_builtin,
22926         paired_expand_builtin, paired_expand_predicate_builtin,
22927         paired_init_builtins): Delete.
22928         * config/rs6000/rs6000.h: Remove paired float support.
22929         * config/rs6000/rs6000.md: Remove paired float support.
22930         (move_from_CR_ov_bit): Delete.
22931         * config/rs6000/rs6000.opt (mpaired): Delete.
22932         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
22933         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
22935 2018-05-02  Richard Biener  <rguenther@suse.de>
22937         PR middle-end/85567
22938         * gimplify.c (gimplify_save_expr): When in SSA form allow
22939         SAVE_EXPRs to compute to SSA vars.
22941 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
22943         PR target/85582
22944         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
22945         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
22946         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
22947         clobber operands[2], instead use a new pseudo.  Formatting fixes.
22949 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
22951         PR tree-optimization/85586
22952         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
22953         exit early for statements in the same group if the accesses are
22954         not strided.
22956 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22958         PR lto/85451
22959         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
22960         error message.
22962 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
22964         PR tree-optimization/85143
22965         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
22967 2018-05-01  Tom de Vries  <tom@codesourcery.com>
22969         PR lto/85451
22970         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
22971         not found" error message.
22973 2018-05-01  Tom de Vries  <tom@codesourcery.com>
22975         PR other/83786
22976         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
22977         * vec.c (test_ordered_remove_if): New function.
22978         (vec_c_tests): Call test_ordered_remove_if.
22979         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
22980         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
22981         * tree-vect-patterns.c (vect_pattern_recog_1): Use
22982         VEC_ORDERED_REMOVE_IF.
22984 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22986         PR tree-optimization/82665
22987         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
22988         pointer subtraction where arguments come from a memchr call.
22990 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
22992         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
22993         --push-state --as-needed and --pop-state instead of --as-needed and
22994         --no-as-needed if ld supports it.
22995         * configure: Regenerated.
22997         PR web/85578
22998         * doc/install.texi2html: Replace _002d with - and _002a with * in
22999         generated html files using sed.
23001 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23003         PR c++/85523
23004         * gcc-rich-location.c (blank_line_before_p): New function.
23005         (use_new_line): New function.
23006         (gcc_rich_location::add_fixit_insert_formatted): New function.
23007         * gcc-rich-location.h
23008         (gcc_rich_location::add_fixit_insert_formatted): New function.
23010 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23012         * selftest.c (assert_streq): Rename "expected" and "actual" to
23013         "val1" and "val2".  Extend NULL-handling to cover both inputs
23014         symmetrically, while still requiring both to be non-NULL for a pass.
23015         * selftest.h (assert_streq): Rename "expected" and "actual" to
23016         "val1" and "val2".
23017         (ASSERT_EQ): Likewise.
23018         (ASSERT_EQ_AT): Likewise.
23019         (ASSERT_KNOWN_EQ): Likewise.
23020         (ASSERT_KNOWN_EQ_AT): Likewise.
23021         (ASSERT_NE): Likewise.
23022         (ASSERT_MAYBE_NE): Likewise.
23023         (ASSERT_MAYBE_NE_AT): Likewise.
23024         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
23025         the assertion to pass.
23026         (ASSERT_STREQ_AT): Likewise.
23028 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
23030         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
23031         interaction with -pie.
23033 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23035         * selftest.h: Fix alphabetization of per-source-file selftest
23036         declarations.
23038 2018-04-30  Jason Merrill  <jason@redhat.com>
23040         PR c++/61982 - dead stores to destroyed objects.
23041         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
23042         of clobber.
23044 2018-04-30  Jason Merrill  <jason@redhat.com>
23046         * tree.c (build_clobber): New.
23047         * tree.h: Declare it.
23048         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
23050 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23052         * diagnostic-show-locus.c (layout::layout): Update for
23053         location_get_source_line returning a char_span.
23054         (struct char_span): Move to input.h.
23055         (struct correction): Update for fields in char_span becoming
23056         private.
23057         (struct source_line): Update for location_get_source_line
23058         returning a char_span.
23059         (layout::print_line): Likewise.
23060         * edit-context.c (edited_file::print_content): Likewise.
23061         (edited_file::print_diff_hunk): Likewise.
23062         (edited_file::print_run_of_changed_lines): Likewise.
23063         (edited_file::get_num_lines): Likewise.
23064         (edited_line::edited_line): Likewise.
23065         * final.c (asm_show_source): Likewise.
23066         * input.c (location_get_source_line): Convert return type
23067         from const char * to char_span, losing the final "line_len"
23068         param.
23069         (dump_location_info): Update for the above.
23070         (get_substring_ranges_for_loc): Likewise.  Use a char_span
23071         when handling the literal within the line.
23072         (test_reading_source_line): Update for location_get_source_line
23073         returning a char_span.
23074         * input.h (class char_span): Move here from
23075         diagnostic-show-locus.c, converting from a struct to a class.
23076         Make data members private.
23077         (char_span::operator bool): New.
23078         (char_span::length): New.
23079         (char_span::get_buffer): New.
23080         (char_span::operator[]): New.
23081         (char_span::subspan): Make const.
23082         (char_span::xstrdup): New.
23083         (location_get_source_line): Convert return type from const char *
23084         to char_span, losing the final "line_size" param.
23086 2018-04-30  Jan Hubicka  <jh@suse.cz>
23088         * lto-wrapper.c (ltrans_priorities): New static var.
23089         (cmp_priority): New.
23090         (run_gcc): Read priorities and if doing parallel build order
23091         the Makefile by them.
23093 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23095         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
23097 2018-04-30  Richard Biener  <rguenther@suse.de>
23099         * tree-cfg.c (verify_address): Remove base argument, add
23100         flag whether to check TREE_ADDRESSABLE and do that.
23101         (verify_expr): Remove.
23102         (verify_types_in_gimple_reference): Add pieces from verify_expr.
23103         (verify_gimple_assign_single): Likewise.
23104         (verify_gimple_switch): Likewise.
23105         (verify_expr_location_1): Dereference tp once.  Add (disabled)
23106         piece from verify_expr.
23107         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
23109 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23111         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
23113 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23115         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
23116         (small_data_pattern): Likewise.
23117         (arc_rewrite_small_data): Likewise.
23118         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
23119         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
23120         (get_symbol_alignment): New function.
23121         (legitimate_small_data_address_p): Likewise.
23122         (legitimate_scaled_address): Update, call
23123         legitimate_small_data_address_p.
23124         (output_sdata): New static variable.
23125         (arc_print_operand): Update how we handle small data operands.
23126         (arc_print_operand_address): Likewise.
23127         (arc_legitimate_address_p): Update, use
23128         legitimate_small_data_address_p.
23129         (arc_rewrite_small_data_p): Remove.
23130         (arc_rewrite_small_data_1): Likewise.
23131         (arc_rewrite_small_data): Likewise.
23132         (small_data_pattern): Likewise.
23133         (compact_sda_memory_operand): Update to use
23134         legitimate_small_data_address_p and get_symbol_alignment.
23135         (prepare_move_operands): Don't rewite sdata pattern.
23136         (prepare_extend_operands): Remove.
23137         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
23138         pattern.
23139         (zero_extendqisi2): Likewise.
23140         (zero_extendhisi2): Likewise.
23141         (extendqihi2): Likewise.
23142         (extendqisi2): Likewise.
23143         (extendhisi2): Likewise.
23144         (addsi3): Likewise.
23145         (subsi3): Likewise.
23146         (andsi3): Likewise.
23147         * config/arc/constraints.md (Usd): Change it to memory constraint.
23149 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23151         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
23152         as source of std instructions.
23153         * config/arc/arc.md (movsi_insn): Update pattern predicate to
23154         allow 6-bit constants as source for store instructions.
23155         (movdi_insn): Update instruction pattern to allow 6-bit constants
23156         as source for store instructions.
23158 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
23160         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
23162 2018-04-30  Nathan Sidwell  <nathan@acm.org>
23163             Sandra Loosemore <sandra@codesourcery.com>
23165         * dumpfile.c (dump_open): Allow '-' for stdout.
23166         * doc/invoke.texi (Developer Options): Document dump filename
23167         determination early.  Document stdin/stdout selection.
23169 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
23171         Microblaze Target: PIC data text relative
23173         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
23174         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
23175         Add declaration.
23176         * config/microblaze/microblaze.h (microblaze_constant_address_p):
23177         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
23178         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
23179         New addressing mode for data-text relative position indepenedent code.
23180         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
23181         'ADDRESS_SYMBOLIC_TXT_REL'.
23182         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
23183         (microblaze_legitimate_pic_operand): Exclude function calls from
23184         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
23185         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
23186         addresses cases.
23187         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
23188         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
23189         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
23190         for 'address + offset'.
23191         (microblaze_expand_prologue): Add new function prologue call for
23192         'r20' assignation.
23193         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
23194         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
23195         table in case of TARGET_PIC_DATA_TEXT_REL.
23196         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
23197         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
23198         Add new macros 'UNSPEC_TEXT',
23199         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
23200         + exclude function calls from 'UNSPEC_PLT' in case of data text
23201         relative mode.
23202         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
23203         new target hook for generating address diff vector tables in case of
23204         flag_pic.
23205         * doc/tm.texi : Regenerate.
23206         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
23207         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
23208         of addr diff vector generation.
23209         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
23210         target hook definition.
23211         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
23212         Add default function for generate_pic_addr_diff_vec -> flag_pic.
23213         * doc/invoke.texi (Add new pic option): Add new microblaze pic
23214         option for data text relative.
23216 2018-04-30  Richard Biener  <rguenther@suse.de>
23218         * tree-chrec.h (evolution_function_is_constant_p): Remove
23219         redundant check.
23220         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
23222 2018-04-30  Richard Biener  <rguenther@suse.de>
23224         PR bootstrap/85571
23225         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
23227 2018-04-30  Richard Biener  <rguenther@suse.de>
23229         PR tree-optimization/28364
23230         PR tree-optimization/85275
23231         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
23232         copying first exit test.
23234 2018-04-28  Mark Wielaard  <mark@klomp.org>
23236         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
23237         dwarf_version >= 5.
23238         (dwarf_AT): Handle DW_AT_addr_base.
23239         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
23241 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
23243         PR target/84431
23244         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
23245         (*ashl<dwi>3_doubleword_mask_1): Ditto.
23246         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
23247         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
23249 2018-04-28  Richard Biener  <rguenther@suse.de>
23251         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
23252         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
23253         to reflect use.  Only add interesting stmts.
23255 2018-04-27  Martin Jambor  <mjambor@suse.cz>
23257         PR ipa/85549
23258         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
23259         the jump function allows for passing through aggregate values.
23261 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
23263         * input.h (in_system_header_at): Convert from macro to inline
23264         function.
23265         (from_macro_expansion_at): Likewise.
23266         (from_macro_definition_at): Likewise.
23268 2018-04-27  Jeff Law  <law@redhat.com>
23270         * config.gcc: Mark tile* targets as deprecated/obsolete.
23272 2018-04-27  Richard Biener  <rguenther@suse.de>
23274         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
23275         fix for ILP32.
23277 2018-04-27  Richard Biener  <rguenther@suse.de>
23279         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
23281 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
23283         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
23284         with Yd constraint. Set "preferred_for_speed" attribute from
23285         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
23286         with Yd constraint.
23287         (*movdi_internal): Ditto.
23288         (movti_interunit splitters): Remove
23289         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
23290         (movdi_interunit splitters): Ditto.
23291         * config/i386/constraints.md (Ye): Remove.
23292         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
23294 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23296         PR target/85512
23297         * config/aarch64/constraints.md (Usg): Limit to 31.
23298         (Usj): Limit to 63.
23300 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
23302         PR tree-optimization/85529
23303         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
23304         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
23305         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
23306         zero extension or masking of the MSB bit.
23307         (optimize_range_tests): Add FIRST_BB argument, pass it through
23308         to optimize_range_tests_var_bound.
23309         (maybe_optimize_range_tests, reassociate_bb): Adjust
23310         optimize_range_tests callers.
23312 2018-04-26  Richard Biener  <rguenther@suse.de>
23313             Jakub Jelinek  <jakub@redhat.com>
23315         * cgraph.h (symbol_table): Just declare debug method here.
23316         * symtab.c (symbol_table::debug): Define.
23318 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
23320         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
23322 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
23324         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
23325         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
23326         (*movdi_internal): Substitute Yi and Yj constraint with x
23327         and Ym and Yn constraint with y constraint.  Update "isa"
23328         attribute and set "preferred_for_speed" attribute from
23329         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
23330         (*movsi_internal): Ditto.
23331         (*movdf_internal): Ditto.
23332         (*movsf_internal): Ditto.
23333         (*zero_extendsidi2): Ditto.
23334         * config/i386/sse.md (vec_set<mode>_0): Ditto.
23335         (sse2_loadld): Ditto.
23336         (*vec_extract<ssevecmodelower>_0): Ditto.
23337         (*vec_extractv4si_0_zext_sse4): Ditto.
23338         (vec_concatv2di): Ditto.
23339         (*vec_dup<mode>): Ditto.
23340         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
23341         * config/i386/constraints.md (Yi): Remove.
23342         (Yj): Remove.
23343         (Ym): Remove.
23344         (Yn): Remove.
23346 2018-04-26  Nathan Sidwell  <nathan@acm.org>
23348         * dumpfile.c (dump_open): New.
23349         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
23350         (dump_finish): Detect stdio/stderr by value not name.
23352 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
23354         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
23356 2018-04-26  Tom de Vries  <tom@codesourcery.com>
23358         PR target/84952
23359         * config/nvptx/nvptx.c (verify_neutering_jumps)
23360         (verify_neutering_labels): New function
23361         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
23363 2018-04-26  Tom de Vries  <tom@codesourcery.com>
23365         PR target/84025
23366         * config/nvptx/nvptx.c (needs_neutering_p): New function.
23367         (nvptx_single): Use needs_neutering_p to skip over insns that do not
23368         need neutering.
23370 2018-04-26  Richard Biener <rguenther@suse.de>
23371             Tom de Vries  <tom@codesourcery.com>
23373         PR lto/85422
23374         * lto-streamer-out.c (output_function): Fixup loops if required to match
23375         discovery done in the reader.
23377 2018-04-26  Richard Biener  <rguenther@suse.de>
23379         PR tree-optimization/85116
23380         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
23381         have a loop exit from the single latch predecessor.  Remove
23382         case of header with just condition.
23383         (ch_base::copy_headers): Exclude infinite loops from any
23384         processing.
23385         (pass_ch::execute): Record exits.
23387 2018-04-26  Richard Biener  <rguenther@suse.de>
23389         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
23390         prologue cost vector and pass it to vect_get_load_cost.
23391         (vect_get_peeling_costs_all_drs): Likewise.
23392         (vect_peeling_hash_get_lowest_cost): Likewise.
23393         (vect_enhance_data_refs_alignment): Likewise.
23395 2018-04-26  Richard Biener  <rguenther@suse.de>
23397         PR middle-end/85450
23398         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
23399         checking of integer<->pointer conversions.
23400         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
23401         sign-/zero-extending pointer types.
23402         (expand_omp_for_static_chunk): Likewise.
23404 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
23405             Jean Lee  <xiaoyur347@gmail.com>
23407         * config/mips/mips.c (mips_asan_shadow_offset): New function.
23408         (TARGET_ASAN_SHADOW_OFFSET): Define.
23409         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
23410         true for -fsanitize=address.
23412 2018-04-25  Mark Wielaard  <mark@klomp.org>
23414         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
23415         shorter ones.
23417 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
23419         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
23420         than "alu", remove explicit "memory" and "imm_disp" attributes.
23421         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
23423         PR middle-end/85414
23424         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
23425         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
23426         gen_lowpart_no_emit.
23428 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
23430         PR target/85473
23431         * config/i386/i386.c (ix86_expand_builtin): Change memory
23432         operand to XI, extend p0 to Pmode.
23433         * config/i386/i386.md: Change unspec volatile and operand
23434         1 mode to XI, change operand 0 mode to P.
23436 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23438         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
23439         GET_MODE_MASK before any checking.
23440         (nds32_can_use_bset_p): Likewise.
23441         (nds32_can_use_btgl_p): Likewise.
23443 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23445         * config/nds32/nds32-doubleword.md: New define_split pattern for
23446         illegal register number.
23448 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23450         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
23452 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23454         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
23456 2018-04-25  Richard Biener  <rguenther@suse.de>
23458         * lto-streamer.h (LTO_major_version): Bump to 8.
23460 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
23462         * BASE-VER: Set to 9.0.0.
23464 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
23466         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
23467         in __abskf2 and __powikf2.
23469 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23471         PR target/85512
23472         * config/aarch64/constraints.md (Usg, Usj): New constraints.
23473         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
23474         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
23475         Use the above on operand 2.  Reindent.
23476         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
23478 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
23480         PR target/85485
23481         * common/config/i386/i386-common.c (ix86_handle_option): Don't
23482         handle OPT_mcet.
23483         * config/i386/i386.opt (mcet): Removed.
23484         * doc/install.texi: Remove -mcet documentation.
23485         * doc/invoke.texi: Likewise.
23487 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
23489         PR target/85485
23490         * doc/install.texi: Remove -mcet from bootstrap-cet.
23492 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
23494         PR target/85511
23495         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
23496         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
23497         if TARGET_64BIT.
23499         PR target/85503
23500         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
23501         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
23502         containing a CONST_VECTOR.
23504 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
23506         * doc/install.texi: Update newlib dependency for nvptx.
23508 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
23510         PR target/85508
23511         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
23512         instead of INTVAL when shifting x left.
23514 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
23516         PR tree-optimization/85478
23517         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
23518         vect_grouped_store_supported for single element vectors.
23520 2018-04-24  Richard Biener  <rguenther@suse.de>
23522         PR target/85491
23523         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
23524         load cost increase to the case of non-constant step.
23526 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
23528         PR target/84828
23529         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
23530         destination if any_malformed_asm.
23532 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
23534         PR middle-end/85496
23535         * expr.c (store_field): In the bitfield case, if the value comes from
23536         a function call and is returned in registers by means of a PARALLEL,
23537         do not change the mode of the temporary unless BLKmode and VOIDmode.
23539 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
23541         PR rtl-optimization/85423
23542         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
23543         dependencies to debug insns when the previous insn is non-debug.
23545 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
23547         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
23548         enums into a single definition.
23549         (fls): Fix predicates and printing.
23550         (seti): Likewise.
23552 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
23554         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
23555         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
23556         and short u6 immediate.
23557         (check_if_valid_sleep_operand): Remove.
23558         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
23560 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23562         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
23563         flag_always_save_lp condition.
23564         * config/nds32/nds32.opt (malways-save-lp): New option.
23566 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
23568         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
23569         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
23570         * config/nds32/nds32.h
23571         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
23572         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
23574 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
23576         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
23577         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
23579 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
23580             Chung-Ju Wu  <jasonwucj@gmail.com>
23582         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
23583         Declare.
23584         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
23585         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
23587 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23589         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
23591 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23593         * config/nds32/nds32-protos.h (nds32_data_alignment,
23594         nds32_local_alignment): Declare.
23595         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
23596         nds32_local_alignment): New functions.
23597         (TARGET_CONSTANT_ALIGNMENT): Define.
23598         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
23600 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23602         * config/nds32/nds32.c
23603         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
23604         (TARGET_MODES_TIEABLE_P): Likewise.
23606 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23608         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
23609         level Ofast and Og.
23611 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
23612             Chung-Ju Wu  <jasonwucj@gmail.com>
23614         * config/nds32/constants.md (unspec_volatile_element): Add enum values
23615         for unaligned access.
23616         * config/nds32/nds32-intrinsic.c: Implementation of expanding
23617         unaligned access.
23618         * config/nds32/nds32-intrinsic.md: Likewise.
23619         * config/nds32/nds32_intrinsic.h: Likewise.
23620         * config/nds32/nds32.h (nds32_builtins): Likewise.
23621         * config/nds32/nds32.opt (munaligned-access): New option.
23622         * config/nds32/nds32.c (nds32_asm_file_start): Display
23623         flag_unaligned_access status.
23625 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
23627         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
23628         -mno-relax is present.
23629         * config/riscv/linux.h (LINK_SPEC): Ditto.
23631 2018-04-20  Martin Sebor  <msebor@redhat.com>
23633         PR c/85365
23634         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
23635         for null pointers.
23636         (gimple_fold_builtin_stxcpy_chk): Same.
23637         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
23639 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
23641         PR target/85456
23642         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
23643         __powikf2 when long double is IEEE 128-bit.
23645 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
23647         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
23648         step to make sure stack always aligned.
23650 2018-04-20  Carl Love  <cel@us.ibm.com>
23652         PR target/83402
23653         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
23654         size check for arg0.
23656 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
23657             Tom de Vries  <tom@codesourcery.com>
23659         PR target/85445
23660         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
23661         Emit insns for calls too.
23662         (nvptx_find_par): Always look for worker-level predecessor insn.
23663         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
23664         calls.
23665         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
23666         (nvptx_process_pars): Propagate frames for calls.
23668 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
23670         PR target/85469
23671         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
23672         Removed.
23673         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
23674         (ix86_handle_option): Don't handle OPT_mibt.
23675         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
23676         __SHSTK__.
23677         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
23678         has_ibt and ibt.
23679         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
23680         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
23681         (ix86_target_macros): Define __CET__ with flag_cf_protection
23682         for -fcf-protection.
23683         * config/i386/i386.c (isa2_opts): Remove -mibt.
23684         * config/i386/i386.h (TARGET_IBT): Removed.
23685         (TARGET_IBT_P): Likewise.
23686         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
23687         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
23688         * config/i386/i386.opt (mcet): Update help message.
23689         (mshstk): Likewise.
23690         (mibt): Removed.
23691         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
23692         -mcet as an alias for -mshstk.
23694 2018-04-20  Richard Biener <rguenther@suse.de>
23696         PR middle-end/85475
23697         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
23698         complexity by forcing a single use of the multiply operand.
23700 2018-04-20  Martin Jambor  <mjambor@suse.cz>
23702         ipa/85449
23703         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
23704         recursion dependency to only apply to non-clones.
23706 2018-04-20  Martin Jambor  <mjambor@suse.cz>
23708         ipa/85447
23709         * ipa-cp.c (create_specialized_node): Check that clones of
23710         self-recursive edges exist during IPA-CP.
23712 2018-04-19  Toon Moene  <toon@moene.org>
23714         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
23715         by -O3.
23717 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
23719         PR tree-optimization/85467
23720         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
23721         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
23722         VECTOR_CST element to type.
23724 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
23726         PR target/85397
23727         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
23728         * config/i386/i386.md (builtin_setjmp_setup): Removed.
23729         (builtin_longjmp): Likewise.
23730         (save_stack_nonlocal): New pattern.
23731         (restore_stack_nonlocal): Likewise.
23733 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
23735         PR target/85404
23736         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
23737         Replace ASM_OUTPUT_LABEL with fprintf.
23739 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
23741         PR target/85417
23742         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
23743         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
23744         * config/i386/i386-c.c (ix86_target_macros_internal): Also
23745         define __IBT__ and __SHSTK__ for -fcf-protection.
23746         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
23747         TARGET_IBT.
23748         (ix86_trampoline_init): Likewise.
23749         (x86_output_mi_thunk): Likewise.
23750         (ix86_notrack_prefixed_insn_p): Likewise.
23751         (ix86_option_override_internal): Don't disallow -fcf-protection.
23752         * config/i386/i386.md (rdssp<mode>): Also enable for
23753         -fcf-protection.
23754         (incssp<mode>): Likewise.
23755         (nop_endbr): Likewise.
23756         * config/i386/i386.opt (mcet): Change help message to built-in
23757         functions only.
23758         (mibt): Likewise.
23759         (mshstk): Likewise.
23760         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
23761         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
23762         enable CET built-in functions.
23764 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
23766         * common/config/i386/i386-common.c
23767         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
23768         OPTION_MASK_ISA_MOVDIRI_UNSET,
23769         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
23770         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
23771         * config.gcc (movdirintrin.h): New header.
23772         * config/i386/cpuid.h (bit_MOVDIRI,
23773         bit_MOVDIR64B): New bits.
23774         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
23775         and -mmvodir64b.
23776         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
23777         (VOID, PVOID, PCVOID)): New function types.
23778         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
23779         __builtin_ia32_directstoreu_u64,
23780         __builtin_ia32_movdir64b): New builtins.
23781         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
23782         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
23783         and -mmovdiri.
23784         (ix86_valid_target_attribute_inner_p): Ditto.
23785         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
23786         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
23787         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
23788         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
23789         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
23790         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
23791         (movdiri<mode>, movdir64b_<mode>): New.
23792         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
23793         * config/i386/immintrin.h: Include movdirintrin.h.
23794         * config/i386/movdirintrin.h: New file.
23795         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
23797 2018-04-19  Richard Biener  <rguenther@suse.de>
23799         PR middle-end/85455
23800         * cfg.c (clear_bb_flags): When loop state says we have
23801         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
23803 2018-04-19  Richard Biener  <rguenther@suse.de>
23805         PR tree-optimization/84737
23806         * tree-vect-data-refs.c (vect_copy_ref_info): New function
23807         copying restrict info.
23808         (vect_setup_realignment): Use it.
23809         * tree-vectorizer.h (vect_copy_ref_info): Declare.
23810         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
23811         the first DR to all generated stores.
23812         (vectorizable_load): Likewise for loads.
23814 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
23816         PR tree-optimization/85446
23817         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
23818         the integral and pointer types to have the same precision.
23820         * doc/install.texi: Document --disable-cet being the default and
23821         --enable-cet=auto.
23823 2018-04-18  Martin Liska  <mliska@suse.cz>
23825         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
23826         style.
23828 2018-04-18  Martin Liska  <mliska@suse.cz>
23830         Revert
23831         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
23833         PR ipa/83983
23834         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23835         arguments if they are comparable.
23837 2018-04-18  Martin Liska  <mliska@suse.cz>
23839         Revert
23840         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23842         PR lto/84805
23843         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23844         incomplete types.
23846 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
23848         PR target/85388
23849         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
23850         ENDBR after calling __morestack.
23852 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
23854         PR jit/85384
23855         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
23856         by using gcc_base_ver to generate a gcc_driver_version, and use
23857         it when generating GCC_DRIVER_NAME.
23858         * configure: Regenerate.
23860 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
23862         PR target/81084
23863         * config.gcc: Obsolete powerpc*-*-*spe*.
23865 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23867         PR debug/84637
23868         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
23869         (stabstr_D): Change type of unum from unsigned int to
23870         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
23871         type.
23873 2018-04-17  Jim Wilson  <jimw@sifive.com>
23875         PR 84856
23876         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
23877         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
23878         Set arg_pointer_offset after using pretend_args_size.
23880 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23882         PR rtl-optimization/85431
23883         * dse.c (record_store): Ignore zero width stores.
23885         PR sanitizer/85230
23886         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
23887         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
23888         __builtin_stack_restore rather than after it.
23889         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
23890         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
23891         argument instead of virtual_dynamic_stack_rtx.
23893 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23895         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
23896         New prototype.
23897         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23898         Add note to error message to explain internal mapping of overloaded
23899         built-in function name to non-overloaded built-in function name.
23900         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
23901         function.
23903 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
23905         PR target/85424
23906         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
23907         where the inputs overlap with the output.
23909 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23911         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
23912         (=v, v) alternative and explicit "memory" attribute.
23913         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
23914         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23915         attributes.
23916         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
23917         "sselog1" type instead of "sselog".
23918         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
23919         "sselog".  Remove explicit "memory" attribute.
23920         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
23921         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23922         attributes.
23923         (vec_extract_hi_v32hi): Merge all alternatives into one, use
23924         "sselog1" type instead of "sselog".  Remove explicit "memory"
23925         attribute.
23926         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
23927         use "sselog1" type instead of "sselog".  Remove explicit "memory"
23928         attribute.
23929         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
23930         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23931         attributes.
23932         (vec_extract_hi_v64qi): Merge all alternatives into one, use
23933         "sselog1" type instead of "sselog".  Remove explicit "memory"
23934         attribute.
23935         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
23936         use "sselog1" type instead of "sselog".  Remove explicit "memory"
23937         attribute.
23939         PR target/85430
23940         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
23942         PR middle-end/85414
23943         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
23944         on a SUBREG.
23946 2018-04-17  Martin Jambor  <mjambor@suse.cz>
23948         PR ipa/85421
23949         * ipa-cp.c (create_specialized_node): Call
23950         expand_all_artificial_thunks if necessary.
23952 2018-04-17  Martin Liska  <mliska@suse.cz>
23954         PR lto/85405
23955         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
23956         in message, remote space in between '_G' and '('.
23958 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23960         PR target/85281
23961         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
23962         avx512f_vmcmp<mode>3<round_saeonly_name>,
23963         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
23964         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
23965         avx512f_rndscale<mode><round_saeonly_name>,
23966         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
23967         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
23968         Use %<iptr>2 instead of %2 for -masm=intel.
23969         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
23970         avx512f_vcvttss2usi<round_saeonly_name>,
23971         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
23972         -masm=intel.
23973         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
23974         avx512f_vcvttsd2usi<round_saeonly_name>,
23975         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
23976         Use %q1 instead of %1 for -masm=intel.
23977         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
23978         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
23979         of %3 for -masm=intel.
23980         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
23981         -masm=intel.
23982         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
23983         -masm=intel.
23984         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
23985         -masm=intel.
23986         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
23987         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
23988         %g1.
23989         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
23990         -masm=intel.
23991         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
23992         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
23993         %g1 and one with %0 and %1.
23994         (avx512er_vmrcp28<mode><round_saeonly_name>,
23995         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
23996         %1 for -masm=intel.
23997         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
23998         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
23999         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
24000         of %0 and %{%4%} for -masm=intel.
24001         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
24002         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
24003         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
24004         order of %0 and %{%5%}%{z%} for -masm=intel.
24006 2018-04-17  Jan Hubicka  <jh@suse.cz>
24008         PR lto/85405
24009         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
24011 2018-04-17  Martin Liska  <mliska@suse.cz>
24013         PR ipa/85329
24014         * multiple_target.c (create_dispatcher_calls): Set apostrophes
24015         for target_clone error message.  Make default implementation
24016         clone to be a local declaration.
24017         (separate_attrs): Add new argument and check for an empty
24018         string.
24019         (expand_target_clones): Handle it.
24020         (ipa_target_clone): Make redirection just for target_clones
24021         functions.
24023 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
24024             Tom de Vries  <tom@codesourcery.com>
24026         PR middle-end/84955
24027         * omp-expand.c (expand_oacc_for): Add dummy false branch for
24028         tiled basic blocks without omp continue statements.
24030 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
24032         PR target/83660
24033         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
24034         vec_extract expression as having side effects to make sure it gets
24035         a cleanup point.
24037 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
24039         PR target/85403
24040         * config/i386/i386.c (get_builtin_code_for_version): Check
24041         error_mark_node.
24043 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
24045         PR target/84331
24046         * config.gcc: Support "skylake".
24047         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
24048         PROCESSOR_SKYLAKE.
24049         * config/i386/i386.c (m_SKYLAKE): Define.
24050         (processor_target_table): Add "skylake".
24051         (ix86_option_override_internal): Add "skylake".
24052         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
24053         PROCESSOR_CANNONLAKE.
24054         (get_builtin_code_for_version): Fix priority for
24055         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
24056         PROCESSOR_SKYLAKE-AVX512.
24057         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
24058         (processor_type): Add PROCESSOR_SKYLAKE.
24060 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
24061             Jason Merrill  <jason@redhat.com>
24063         PR c++/85112
24064         * convert.c (convert_to_integer_1): Use direct recursion for
24065         enumeral types and types with a precision less than the number
24066         of bits in their mode.
24068 2018-04-16  Julia Koval  <julia.koval@intel.com>
24070         PR target/84413
24071         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
24072         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
24074 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
24076         PR target/85293
24077         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
24078         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
24079         and -mno-direct-move.
24081 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
24083         PR target/83402
24084         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
24085         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
24086         Ensure negative shifts result in {0}.
24088 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
24090         PR rtl-optimization/79916
24091         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
24092         regs (if any) to define how to gnerate SD moves when LRA is in
24093         progress.
24095 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
24097         PR rtl-optimization/85393
24098         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
24099         * except.c (expand_dw2_landing_pad_for_region): Make static.
24100         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
24101         a label and unconditional jump to old_bb, rather than
24102         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
24103         basic block.
24105         PR rtl-optimization/85376
24106         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
24107         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
24108         instead of a specific value.
24110 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
24111             Bin Cheng  <bin.cheng@arm.com>
24113         PR tree-optimization/82965
24114         PR tree-optimization/83991
24115         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
24116         by_profile_only parameter.
24117         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
24118         information if the loop was predicted to iterate too many times.
24119         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
24121 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
24123         PR lto/71991
24124         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
24125         always inline.
24127 2018-04-13  Martin Liska  <mliska@suse.cz>
24128             Jakub Jelinek  <jakub@redhat.com>
24130         PR middle-end/81657
24131         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
24132         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
24133         * builtins.c (expand_builtin_memory_copy_args): Use
24134         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
24135         handle dest_addr == pc_rtx.
24137 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
24139         PR target/85291
24140         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
24141         asked to not generate direct moves.
24142         (fix_trunc<mode>si2_stfiwx): Similar.
24143         (fix_trunc<mode>si2_internal): Similar.
24145 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24147         PR debug/83157
24148         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
24149         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
24150         lookup if dest in some wider mode is known to be const0_rtx and
24151         if so, record permanent equivalence for it to be ZERO_EXTEND of
24152         the narrower mode destination.
24154 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
24156         * lto-streamer-out.c (output_function): Revert 259346.
24157         * omp-expand.c (expand_oacc_for): Likewise.
24159 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
24161         PR rtl-optimization/85354
24162         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
24163         * sel-sched.c (sel_global_init): ... here.
24165 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
24167         PR target/85238
24168         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
24169         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
24170         mode for PE-COFF targets.
24171         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
24172         (i386_pe_asm_lto_end): Likewise.
24173         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
24174         (TARGET_ASM_LTO_END): Likewise.
24175         * config/i386/winnt.c (saved_debug_info_level): New static variable.
24176         (i386_pe_asm_lto_start): New function.
24177         (i386_pe_asm_lto_end): Likewise.
24179 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
24180             Richard Biener  <rguenther@suse.de>
24182         PR middle-end/84955
24183         * lto-streamer-out.c (output_function): Fix CFG loop state before
24184         streaming out.
24185         * omp-expand.c (expand_oacc_for): Handle calls to internal
24186         functions like regular functions.
24188 2018-04-12  Richard Biener  <rguenther@suse.de>
24190         PR lto/85371
24191         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
24192         for the early LTO debug to properly generate references to it
24193         during DIE emission.  Do not re-use that for the skeleton for
24194         split-dwarf.
24195         (dwarf2out_early_finish): Likewise.
24197 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24199         PR target/85328
24200         * config/i386/sse.md
24201         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
24202         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
24203         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
24204         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
24205         and output is a reg, avoid creating invalid lowpart subreg, but
24206         instead split into a 512-bit move.  Don't split if not AVX512VL,
24207         input is xmm16+ reg and output is a mem.
24208         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
24209         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
24210         xmm16+ reg and output is a mem.
24212 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24214         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
24215         also for flag_dwarf2_cfi_asm.
24217 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24219         PR rtl-optimization/85342
24220         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
24221         a bool scalar var inside of the loop instead.  Don't try to update
24222         recog_data.operand after failed apply_change_group.
24224 2018-04-12  Tom de Vries  <tom@codesourcery.com>
24226         PR target/85296
24227         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
24228         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
24229         array with flexible array member as array without given dimension.
24230         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
24231         argument for undefined param to true.
24233 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
24235         PR target/85321
24236         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
24237         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
24238         from PowerPC section.
24239         * config/rs6000/sysv4.opt (mcall-): Improve help text.
24240         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
24241         help text that is too long.
24242         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
24243         help text that is too long.
24244         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
24245         help text that is too long.
24247 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
24249         * config/alpha/alpha.md (stack_probe_internal): Rename
24250         from "probe_stack".  Update all callers.
24252 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24254         PR rtl-optimization/84566
24255         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
24256         sched_macro_fuse_insns.
24258 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24260         PR target/84301
24261         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
24262         (compute_block_dependences): ... from here.
24264 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24266         PR tree-optimization/85331
24267         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
24268         from int to HOST_WIDE_INT.
24270 2018-04-11  Martin Jambor  <mjambor@suse.cz>
24272         PR ipa/84149
24273         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
24274         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
24275         not the same as the source val.
24276         (cgraph_edge_brings_value_p): New parameter.
24277         (gather_edges_for_value): Pass destination value to
24278         cgraph_edge_brings_value_p.
24279         (perhaps_add_new_callers): Likewise.
24280         (get_info_about_necessary_edges): Likewise and exclude values brought
24281         only by self-recursive edges.
24282         (create_specialized_node): Redirect only clones of self-calling edges.
24283         (+self_recursive_pass_through_p): New function.
24284         (find_more_scalar_values_for_callers_subset): Use it.
24285         (find_aggregate_values_for_callers_subset): Likewise.
24286         (known_aggs_to_agg_replacement_list): Removed.
24287         (decide_whether_version_node): Re-calculate known constants for all
24288         remaining context clones.
24290 2018-04-11  Richard Biener  <rguenther@suse.de>
24292         PR lto/85339
24293         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
24294         from early DWARF output.
24295         (dwarf2out_early_finish): Output line info unconditionally into
24296         early DWARF and add reference to it.
24298 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24300         PR target/85281
24301         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
24302         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
24303         other than V2DFmode using iptr mode attribute.
24304         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
24306 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24308         PR rtl-optimization/84659
24309         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
24311 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24313         PR debug/85302
24314         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
24315         SIZEP is NULL.
24316         (output_loc_list): Pass address of a dummy size variable even in the
24317         locview handling loop.
24318         (index_location_lists): Add comment on why skip_loc_list_entry can't
24319         call size_of_locs.
24321 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24323         PR target/85261
24324         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
24325         into register.
24327 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
24329         PR target/85321
24330         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
24331         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
24332         and -mstring-compare-inline-limit.
24334 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24336         PR target/85287
24337         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
24338         for stack clash protection in a register whenever we need it to be in
24339         a register.
24341 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24343         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
24344         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
24346 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24348         PR target/85321
24349         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
24350         the help text.
24351         (mlong-double-): Ditto.
24352         * config/rs6000/sysv4.opt (msdata=): Ditto.
24353         (mtls-size=): Ditto.
24355 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24357         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
24358         erroneous entries for
24359         "vector int vec_ldl (int, long int *)", and
24360         "vector unsigned int vec_ldl (int, unsigned long int *)".
24361         Add comments and entries for
24362         "vector bool char vec_ldl (int, bool char *)",
24363         "vector bool short vec_ldl (int, bool short *)",
24364         "vector bool int vec_ldl (int, bool int *)",
24365         "vector bool long long vec_ldl (int, bool long long *)",
24366         "vector pixel vec_ldl (int, pixel *)",
24367         "vector long long vec_ldl (int, long long *)",
24368         "vector unsigned long long vec_ldl (int, unsigned long long *)".
24369         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
24370         type tree bool_long_long_type_node and correct definition of
24371         bool_V2DI_type_node to make reference to this new type tree.
24372         (rs6000_mangle_type): Replace erroneous reference to
24373         bool_long_type_node with bool_long_long_type_node.
24374         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
24375         comments to emphasize sign distinctions for char and int types and
24376         replace RS6000_BTI_bool_long constant with
24377         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
24378         use of RS6000_BTI_pixel.
24379         (bool_long_type_node): Remove this macro definition.
24380         (bool_long_long_type_node): New macro definition
24382 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
24384         PR rtl-optimization/85300
24385         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
24386         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
24387         simplify_unary_operation fails.
24389 2018-04-10  Martin Liska  <mliska@suse.cz>
24391         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
24392         cgraph_edge and ipa_ref.
24394 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
24396         PR target/85177
24397         PR target/85255
24398         * config/i386/sse.md
24399         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
24400         computation of the VEC_MERGE selector from mask.
24401         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
24402         Fix decoding of the VEC_MERGE selector into mask.
24404 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
24406         PR tree-optimization/85286
24407         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
24409 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
24411         * final.c (final_1): Set insn_last_address as well as
24412         insn_current_address.
24414 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24416         PR target/85173
24417         * explow.c (emit_stack_probe): Call validize_mem on memory location
24418         before passing it to gen_probe_stack.  Create address operand and
24419         legitimize it for the probe_stack_address case.
24421 2018-04-09  Jan Hubicka  <jh@suse.cz>
24423         PR lto/85078
24424         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
24425         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
24426         * tree.c (free_lang_data_in_type): Fix handling of binfos;
24427         walk basetypes.
24428         (free_lang_data): Rebuild type inheritance graph.
24430 2018-04-09  Martin Sebor  <msebor@redhat.com>
24432         * invoke.texi (-finline-small-functions): Mention other optimization
24433         options.
24434         (-findirect-inlining, -fpartial-inlining): Same.
24435         (-finline-functions-called-once): Same.
24436         (-freorder-blocks-and-partition): Same.
24438 2018-04-09  Jan Hubicka  <jh@suse.cz>
24440         PR rtl/84058
24441         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
24442         jumps; choose last target that matches the criteria (i.e.
24443         no partition changes for non-crossing jumps).
24444         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
24445         support for redirecting crossing jumps to non-crossing.
24447 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
24449         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
24450         also for naked functions.
24452 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
24454         * config/arc/arc.md (add_shift): New pattern.
24455         (add_shift2): Likewise.
24456         (sub_shift): Likewise.
24457         (sub_shift_cmp0_noout): Likewise.
24458         (compare_si_ashiftsi): Likewise.
24459         (xbfu_cmp0_noout): New combine pattern.
24460         (xbfu_cmp0"): Likewise.
24461         (movsi_set_cc_insn): Place the predicable variant first.
24462         (commutative_binary_cmp0_noout): Remove clobber.
24463         (commutative_binary_cmp0): New pattern.
24464         (noncommutative_binary_cmp0): Likewise.
24465         (noncommutative_binary_cmp0_noout): Likewise.
24466         (noncommutative_binary_comparison_result_used): Removed.
24467         (rsub_cmp0): New pattern.
24468         (rsub_cmp0_noout): Likewise.
24469         (extzvsi): Changed, keep only meaningful variants.
24470         (SQH, SEZ): New iterators.
24471         (SQH_postfix): New mode attribute.
24472         (SEZ_prefix): New code attribute.
24473         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
24474         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
24475         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
24476         of numerical value.
24477         (noncommutative_operator): Check the availability of barrel
24478         shifter option.
24480 2018-04-09  Richard Biener  <rguenther@suse.de>
24482         PR tree-optimization/85284
24483         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
24484         Only use the niter constraining form of simple_iv when the exit
24485         is always executed.
24487 2018-04-09  Tom de Vries  <tom@codesourcery.com>
24489         PR target/84041
24490         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
24491         (define_expand "*memory_barrier"): New define_expand.
24492         (define_insn "memory_barrier"): New insn.
24494 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24496         PR rtl-optimization/80463
24497         PR rtl-optimization/83972
24498         PR rtl-optimization/83480
24500         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
24501         correct producer for the insn.
24502         (tidy_control_flow): Fixup seqnos in case of debug insns.
24504 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24506         PR rtl-optimization/83913
24508         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
24509         different sched-times when merging exprs.
24511 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24513         PR rtl-optimization/83962
24515         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
24516         tidy_fallthru_edge and tidy_control_flow.
24518 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24520         PR rtl-optimization/83530
24522         * sel-sched.c (force_next_insn): New global variable.
24523         (remove_insn_for_debug): When force_next_insn is true, also leave only
24524         next insn in the ready list.
24525         (sel_sched_region): When the region wasn't scheduled, make another pass
24526         over it with force_next_insn set to 1.
24528 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
24530         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
24531         into tm_file.
24532         * config/nds32/constants.md (unspec_volatile_element): Add enum values
24533         for interrupt control.
24534         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
24535         functions for interrupt control.
24536         * config/nds32/nds32-intrinsic.md: Likewise.
24537         * config/nds32/nds32_intrinsic.h: Likewise.
24538         * config/nds32/nds32.h (nds32_builtins): Likewise.
24540 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
24542         * config/nds32/nds32.c (nds32_init_machine_status,
24543         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
24544         strict_aligned_p field.
24545         (nds32_expand_to_rtl_hook): New function.
24546         (TARGET_EXPAND_TO_RTL_HOOK): Define.
24547         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
24549 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24550             Chung-Ju Wu  <jasonwucj@gmail.com>
24552         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
24553         * config/nds32/nds32-n7.md: New file.
24554         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
24555         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
24556         pipeline.
24557         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
24558         * config/nds32/nds32.md (pipeline_model): Add n7.
24559         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
24560         * config/nds32/pipelines.md: Include n7 settings.
24562 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24563             Chung-Ju Wu  <jasonwucj@gmail.com>
24565         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
24566         * config/nds32/nds32-e8.md: New file.
24567         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
24568         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
24569         pipeline.
24570         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
24571         * config/nds32/nds32.md (pipeline_model): Add e8.
24572         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
24573         * config/nds32/pipelines.md: Include e8 settings.
24575 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24576             Chung-Ju Wu  <jasonwucj@gmail.com>
24578         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
24579         * config/nds32/nds32-n8.md: New file.
24580         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
24581         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
24582         pipeline.
24583         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
24584         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
24585         * config/nds32/nds32.md (pipeline_model): Add n8.
24586         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
24587         * config/nds32/pipelines.md: Include n8 settings.
24589 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24590             Chung-Ju Wu  <jasonwucj@gmail.com>
24592         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
24593         * config/nds32/nds32-n9-2r1w.md: New file.
24594         * config/nds32/nds32-n9-3r2w.md: New file.
24595         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
24596         nds32_register_ports): New or modify for cpu n9.
24597         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
24598         pipeline.
24599         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
24600         * config/nds32/nds32-utils.c: New file.
24601         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
24602         TARGET_MUL_SLOW): Define.
24603         * config/nds32/nds32.md (pipeline_model): New attribute.
24604         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
24605         New options that support cpu n9.
24606         * config/nds32/pipelines.md: Include n9 settings.
24607         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
24609 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
24611         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
24612         information if necessary.
24613         (output_cond_branch_compare_zero): Likewise.
24614         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
24615         (nds32_target_alignment): Refine for alignment.
24616         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
24617         (FUNCTION_BOUNDARY): Modify.
24618         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
24619         align case.
24620         * config/nds32/nds32.opt (malways-align, malign-functions): New.
24622 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
24624         * config/nds32/constants.md (unspec_volatile_element): Add values for
24625         TLB operation and data prefetch.
24626         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
24627         functions for TLB operation and data prefetch.
24628         * config/nds32/nds32-intrinsic.md: Likewise.
24629         * config/nds32/nds32_intrinsic.h: Likewise.
24630         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
24631         (nds32_print_operand): Likewise.
24632         * config/nds32/nds32.h (nds32_builtins): Likewise.
24634 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
24635         Andrew Pinski <pinsika@gcc.gnu.org>
24637         PR middle-end/82976
24638         * match.pd: Use constant_boolean_node of correct type instead of
24639         boolean_true_node or boolean_false_node for simplifying
24640         pointer comparisons to zero.
24642 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
24644         PR tree-optimization/80021
24645         * tree.c (verify_type_variant): Make error call in verify_variant_match
24646         translatable and remove final full stop.
24648 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24650         * config/nds32/constants.md (unspec_volatile_element): Add
24651         UNSPEC_VOLATILE_EH_RETURN.
24652         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
24653         nds32_output_stack_pop): Support dwarf exception handling process.
24654         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
24655         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
24656         exception handling process.
24657         (nds32_compute_stack_frame): Likewise.
24658         (nds32_return_addr_rtx): Likewise.
24659         (nds32_initial_elimination_offset): Likewise.
24660         (nds32_expand_prologue): Likewise.
24661         (nds32_expand_epilogue): Likewise.
24662         (nds32_dynamic_chain_address): New function.
24663         * config/nds32/nds32.h (machine_function): Add fields for dwarf
24664         exception handling.
24665         (DYNAMIC_CHAIN_ADDRESS): Define.
24666         (EH_RETURN_DATA_REGNO): Define.
24667         (EH_RETURN_STACKADJ_RTX): Define.
24668         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
24669         patterns for dwarf exception handling.
24671 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24673         * config/nds32/nds32.h: Clean up obsolete macros.
24675 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24677         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
24678         Add enum values for particular instructions.
24679         * config/nds32/nds32-intrinsic.c: Implementation of expanding
24680         particular intrinsic functions.
24681         * config/nds32/nds32-intrinsic.md: Likewise.
24682         * config/nds32/nds32_intrinsic.h: Likewise.
24683         * config/nds32/nds32.h (nds32_builtins): Likewise.
24684         * config/nds32/nds32.md (type): Add pbsad and pbsada.
24685         (btst, ave): New patterns for particular instructions.
24687 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24689         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
24690         Add enum values for atomic load/store and memory sync.
24691         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
24692         and memory sync.
24693         * config/nds32/nds32-intrinsic.md: Likewise.
24694         * config/nds32/nds32_intrinsic.h: Likewise.
24695         * config/nds32/nds32.h (nds32_builtins): Likewise.
24697 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
24699         PR tree-optimization/85257
24700         * fold-const.c (native_encode_vector): If not all elts could fit
24701         and off is -1, return 0 rather than offset.
24702         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
24703         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
24704         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
24705         adjust buffer in native_interpret_expr call.
24707 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24709         * config/nds32/constants.md (unspec_volatile_element): Add cache
24710         control enum values.
24711         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
24712         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
24713         * config/nds32/nds32.c (nds32_cctl_names): New.
24714         (nds32_print_operand): Handle cache control register names.
24715         * config/nds32/nds32.h (nds32_builtins): New enum values.
24716         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
24717         macros.
24718         * config/nds32/nds32.md (type): Add mmu.
24719         * config/nds32/pipelines.md (simple_insn): Add mmu.
24721 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24723         * config/nds32/nds32.md (type): Remove call.
24724         * config/nds32/pipelines.md (simple_insn): Likewise.
24726 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24728         * config/nds32/constants.md (unspec_volatile_element): Add
24729         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
24730         UNSPEC_VOLATILE_FMFCFG.
24731         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
24732         description for fmfcfg and fmfcsr.
24733         (bdesc_1arg): Add fmtcsr.
24734         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
24735         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
24736         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
24737         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
24738         unspec_fmfcfg): New patterns.
24739         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
24740         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
24741         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
24742         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
24743         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
24744         __nds32__fmfcfg): Define.
24746 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24748         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
24749         intrinsic register names.
24750         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
24751         intrinsic register enum values and macros.
24753 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24755         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
24756         for load/store addressing form.
24757         (nds32_print_operand_address): Likewise.
24759 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
24761         PR target/85196
24762         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
24763         based on LABEL_REF.  Remove useless assertion.
24764         (pic_address_needs_scratch): Fix formatting.
24765         (sparc_legitimize_pic_address): Minor tweaks.
24766         (sparc_delegitimize_address): Adjust assertion accordingly.
24767         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
24768         into symbolic_operand.
24769         (movsi_high_pic_label_ref): Likewise.
24770         (movsi_lo_sum_pic_label_ref): Likewise.
24771         (movdi_pic_label_ref): Likewise.
24772         (movdi_high_pic_label_ref): Likewise.
24773         (movdi_lo_sum_pic_label_ref): Likewise.
24775 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
24777         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
24778         custom LIB_SPEC setup.
24780 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
24781             Kito Cheng  <kito.cheng@gmail.com>
24783         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
24784         * config/riscv/freebsd.h: New.
24786 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
24788         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
24789         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
24790         file.
24792 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
24793             Kito Cheng  <kito.cheng@gmail.com>
24795         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
24796         nds32_output_call, nds32_symbol_binds_local_p): New functions.
24797         * config/nds32/nds32-protos.h (nds32_output_call,
24798         nds32_output_return): Declare.
24799         * config/nds32/nds32.md: Refine all the call and return patterns.
24801 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
24803         PR debug/85252
24804         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
24805         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
24807         PR rtl-optimization/84872
24808         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
24809         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
24810         EDGE_CROSSING edge.
24812 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
24814         * expr.c (copy_blkmode_to_reg): Revert 254862.
24815         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
24817 2018-04-06  Richard Biener  <rguenther@suse.de>
24819         PR middle-end/85244
24820         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
24821         after seeing a component reference with an adjacent field.  Treat
24822         refs to arrays at struct end of external decls similar to
24823         refs to unconstrained commons.
24825 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
24827         PR sanitizer/85213
24828         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
24829         look through SAVE_EXPRs with non-side-effects argument.  Adjust
24830         recursive calls.
24831         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
24832         save_p here.
24834 2018-04-06  Richard Biener  <rguenther@suse.de>
24836         PR middle-end/85180
24837         * alias.c (find_base_term): New wrapper around find_base_term
24838         unwinding CSELIB_VAL_PTR changes.
24839         (find_base_term): Do not restore CSELIB_VAL_PTR during the
24840         recursion.
24842 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24844         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
24845         instructions.
24846         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
24847         constant definitions.
24848         ("nop"): lr 0,0 -> nopr r0
24849         ("nop_lr0", "nop_lr1"): New insn definitions.
24851 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
24853         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
24854         NDS32_V3PUSH_AVAILABLE_P macro.
24856 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
24857             Chung-Ju Wu  <jasonwucj@gmail.com>
24859         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
24860         (nds32*-*-*): Add float and fpu_config into supported_defaults.
24861         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
24862         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
24863         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
24864         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
24865         * config/nds32/constraints.md: New constraints and checking for hard
24866         float configuration.
24867         * config/nds32/iterators.md: New mode iterator and attribute for hard
24868         float configuration.
24869         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
24870         patterns.
24871         * config/nds32/nds32-fpu.md: New file.
24872         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
24873         deal with hard float code generation.
24874         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
24875         ARCH_V3S.
24876         (abi_type, float_reg_number): New enum type.
24877         * config/nds32/nds32-predicates.c: New predicates for hard float.
24878         * config/nds32/nds32-protos.h: Declare functions for hard float.
24879         * config/nds32/nds32.c: Implementation for hard float configuration.
24880         * config/nds32/nds32.h: Definitions for hard float configuration.
24881         * config/nds32/nds32.md: Include hard float machine description and
24882         modify patterns for hard float configuration.
24883         * config/nds32/nds32.opt: New options for hard float configuration.
24884         * config/nds32/predicates.md: New predicates for hard float
24885         configuration.
24887 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
24889         * common/config/nds32/nds32-common.c
24890         (nds32_option_optimization_table): Enable -mreleax-hint by default.
24892 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
24894         PR middle-end/85195
24895         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
24896         CONSTRUCTOR_ELT (ctor, ...)->value.
24898 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
24900         PR target/85193
24901         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
24903 2018-04-05  Tom de Vries  <tom@codesourcery.com>
24905         PR target/85204
24906         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
24907         cond jump.
24909 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
24910             Kito Cheng  <kito.cheng@gmail.com>
24912         * config/nds32/constraints.md (U33): Fine-tune checking condition.
24913         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
24914         * config/nds32/nds32.h (nds32_16bit_address_type): Add
24915         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
24917 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
24918             Kito Cheng  <kito.cheng@gmail.com>
24920         * config/nds32/constraints.md (Ufe): New memory constraint.
24921         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
24922         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
24923         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
24924         operands.
24925         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
24926         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
24928 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24930         * config/nds32/nds32.md: Use optimize_size in the condition for
24931         alu-shift instructions.
24933 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24935         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
24937 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24939         * config/nds32/nds32.md (negsi2): Refine pattern.
24941 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
24942             Chung-Ju Wu  <jasonwucj@gmail.com>
24944         * config/nds32/iterators.md (shift_rotate): New code iterator.
24945         (shift): New code attribute.
24946         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
24947         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
24948         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
24949         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
24950         bit-wise operations.
24951         (andsi3, *andsi3): Ditto.
24952         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
24953         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
24954         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
24955         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
24956         nds32_ior_operand, nds32_xor_operand): New predicates.
24958 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24960         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
24961         (addsi3, subsi3): ... this.
24963 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24965         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
24967 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24969         * config/nds32/nds32.md: Adjust indention.
24971 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
24973         * config/nds32/nds32.md (feature): New attribute.
24975 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24977         * config/nds32/nds32.md (subtype): New attribute.
24979 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24981         PR target/85203
24982         * config/arm/arm-builtins.c (arm_expand_builtin): Change
24983         expansion to perform a bitwise AND of the argument followed by a
24984         boolean negation of the result.
24986 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
24988         PR rtl-optimization/84878
24989         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
24990         the basic block.  Assert the use reference is not artificial and that
24991         it has an associated insn.
24993 2018-04-04  Michael Matz  <matz@suse.de>
24995         * builtins.c (compute_objsize): Pass correct operand
24996         to array_at_struct_end_p.
24998 2018-04-04  Richard Biener  <rguenther@suse.de>
25000         PR lto/85176
25001         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
25002         from contexts for DINFO_LEVEL_TERSE and below.
25004 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25006         * config/nds32/nds32-doubleword.md (move_<mode>): Require
25007         resiter_operand condition.
25008         * config/nds32/nds32.md (*move<mode>): Ditto.
25010 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25011             Monk Chiang  <sh.chiang04@gmail.com>
25013         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
25015 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25017         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
25019 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25020             Kito Cheng  <kito.cheng@gmail.com>
25022         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
25023         nds32_cond_code_str, output_cond_branch,
25024         output_cond_branch_compare_zero, nds32_expand_cbranch,
25025         nds32_expand_cstore, nds32_expand_movcc,
25026         nds32_output_cbranchsi4_equality_zero,
25027         nds32_output_cbranchsi4_equality_reg,
25028         nds32_output_cbranchsi4_equality_reg_or_const_int,
25029         nds32_output_cbranchsi4_greater_less_zero: New functions.
25030         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
25031         nds32_expand_cstore, nds32_expand_movcc,
25032         nds32_output_cbranchsi4_equality_zero,
25033         nds32_output_cbranchsi4_equality_reg,
25034         nds32_output_cbranchsi4_equality_reg_or_const_int,
25035         nds32_output_cbranchsi4_greater_less_zero): Declare.
25036         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
25037         nds32_rimm11s_operand): New predicates.
25038         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
25039         * config/nds32/nds32.md: Rewrite all the branch and conditional move
25040         patterns.
25042 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25044         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
25045         * config/nds32/nds32.md: Ditto.
25046         * config/nds32/pipelines.md: Ditto.
25048 2018-04-04  Richard Biener  <rguenther@suse.de>
25050         PR tree-optimization/85168
25051         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
25052         propagating abnormals.
25054 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25056         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
25058 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25059             Kito Cheng  <kito.cheng@gmail.com>
25061         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
25062         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
25063         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
25064         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
25065         * config/nds32/nds32.md (sibcall_internal): New.
25066         (sibcall_register): Remove.
25067         (sibcall_immediate): Remove.
25068         (sibcall_value_internal): New.
25069         (sibcall_value_register): Remove.
25070         (sibcall_value_immediate): Remove.
25071         * config/nds32/predicates.md (nds32_general_register_operand): New.
25072         (nds32_call_address_operand): New.
25074 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
25076         PR rtl-optimization/85167
25077         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
25078         bb_defs if *split_p, instead preinitialize it to NULL.
25080         PR tree-optimization/85156
25081         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
25082         evaluating the argument multiple times.
25084 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
25086         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
25087         than vector.
25088         (_mm_cvtpd_ps): Likewise.
25089         (_mm_cvttpd_epi32): Likewise.
25090         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
25091         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
25092         vector, pixel, and bool following altivec.h include.
25094 2018-04-03  Martin Sebor  <msebor@redhat.com>
25096         * doc/extend.texi (Common Function Attributes): Clarify.
25097         (const attribute): Likewise.
25098         (pure attribute): Likewise.
25100 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
25102         PR target/85169
25103         * config/i386/i386.c (ix86_expand_vector_set): Use
25104         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
25106 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
25108         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
25109         instructions when changing rounding bits to preserve precision bits
25110         in the x87 control word.
25112 2018-04-03  Martin Liska  <mliska@suse.cz>
25114         PR tree-optimization/82491
25115         * rtl.h (strip_offset_and_add): Replace += suboffset with
25116         poly_uint64 () + suboffset.
25118 2018-03-29  Martin Liska  <mliska@suse.cz>
25119             Martin Jambor  <mjambor@suse.cz>
25121         PR ipa/84947
25122         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
25123         param_type is not an integral or pointer type.
25125 2018-04-03  Richard Biener  <rguenther@suse.de>
25127         * sese.h (recompute_all_dominators): Remove.
25129 2018-04-02  Martin Sebor  <msebor@redhat.com>
25131         * doc/invoke.texi (-Wrestrict): Fix typos.
25133 2018-04-02  Jim Wilson  <jimw@sifive.com>
25135         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
25136         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
25137         (<optab>di3, <optab>si3_extend): Likewise.
25138         (<optab>si3_mask, <optab>si3_mask_1): New.
25139         (<optab>di3_mask, <optab>di3_mask_1): New.
25140         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
25141         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
25142         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
25144 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
25146         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
25147         example.
25149 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
25151         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
25152         (nds32_canonicalize_comparison): New function.
25154 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25155             Kito Cheng  <kito.cheng@gmail.com>
25156             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
25158         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
25159         * config/nds32/constants.md (unspec_volatile_element): Add
25160         UNSPEC_VOLATILE_RELAX_GROUP.
25161         * config/nds32/nds32-relax-opt.c: New file.
25162         * config/nds32/nds32-predicates.c
25163         (nds32_symbol_load_store_p): New function.
25164         * config/nds32/nds32-protos.h
25165         (nds32_symbol_load_store_p): Declare function.
25166         (make_pass_nds32_relax_opt): Declare new rtl pass function.
25167         * config/nds32/nds32.c
25168         (nds32_register_pass): New function to register pass.
25169         (nds32_register_passes): New function to register passes.
25170         * config/nds32/nds32.md (relax_group): New pattern.
25171         * config/nds32/nds32.opt (mrelax-hint): New option.
25172         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
25174 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
25176         * config/nds32/t-nds32: Modify files dependency.
25178 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25180         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
25181         (PROFILE_HOOK): Define its implementation.
25183 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25185         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
25186         type and 32-bit size.
25188 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
25190         PR middle-end/85090
25191         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
25192         (V_128_256): New mode iterator.
25193         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
25194         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
25195         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
25196         of V.
25197         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
25198         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
25200 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
25202         PR target/83315
25203         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
25204         NaN inputs correctly.
25206 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
25208         PR target/80546
25209         * config/rs6000/vsx.md (??r): New mode attribute.
25210         (*vsx_mov<mode>_64bit): Use it.
25211         (*vsx_mov<mode>_32bit): Likewise.
25213 2018-03-30  Martin Sebor  <msebor@redhat.com>
25215         PR tree-optimization/84818
25216         * builtins.c (check_access): Use warning_n.
25218 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25220         PR target/83822
25221         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
25222         condition.
25223         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
25224         condition.
25226 2018-03-30  Julia Koval  <julia.koval@intel.com>
25228         PR target/84413
25229         * x86-tune.def (movx, partial_reg_dependency): Enable for
25230         m_SKYLAKE_AVX512.
25232 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
25234         PR inline-asm/84985
25235         * lra-constraints.c (process_alt_operands): Move setting
25236         this_alternative_matches below.
25238 2018-03-29  Martin Liska  <mliska@suse.cz>
25240         PR lto/84995.
25241         * doc/invoke.texi: Document how LTO works with debug info.
25242         Describe auto-load support of binutils.  Mention 'x86-64'
25243         as valid option value of -march option.
25245 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
25247         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
25249         PR c/85094
25250         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
25251         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
25252         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
25253         checking.
25255 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
25257         PR target/84912
25258         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
25259         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
25260         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
25261         for RS6000_BTM_POWERPC64.
25262         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
25263         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
25264         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
25265         definition.
25266         (DIVDE): Use it.
25267         (DIVDEU): Likewise.
25269 2018-03-28  Carl Love  <cel@us.ibm.com>
25271         Revert
25272         2017-09-27  Carl Love  <cel@us.ibm.com>
25274         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
25275         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
25276         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
25277         fctiw instruction.
25279 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25281         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
25282         instead of __vector bool.
25283         (_mm_max_pu8): Likewise.
25284         (_mm_min_pi16): Likewise.
25286 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
25288         PR target/84912
25289         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
25290         (DIVWEUO): Likewise.
25291         (DIVDEO): Likewise.
25292         (DIVDEUO): Likewise.
25293         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
25294         DIVWEUO and DIVDEUO.
25295         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
25296         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
25297         (div_extend): Likewise.
25298         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
25299         builtin function.
25300         (__builtin_divweuo): Likewise.
25301         (__builtin_divdeo): Likewise.
25302         (__builtin_divdeuo): Likewise.
25304 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
25306         PR target/85095
25307         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
25308         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
25310         PR tree-optimization/82004
25311         * gimple-match-head.c (optimize_pow_to_exp): New function.
25312         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
25313         Don't fold to exp if optimize_pow_to_exp is false.
25315 2018-03-28  Martin Liska  <mliska@suse.cz>
25317         PR other/84819
25318         * calls.c (initialize_argument_information): Fix trailing space.
25319         * common.opt: Fix typo and provide better explanation for
25320         -fsanitize-coverage option.
25321         * config/i386/i386.opt: Fix typo.
25323 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
25324             Martin Liska  <mliska@suse.cz>
25326         PR sanitizer/85081
25327         * gimplify.c (asan_poison_variable): Don't do the check for
25328         gimplify_omp_ctxp here.
25329         (gimplify_decl_expr): Do it here.
25330         (gimplify_target_expr): Likewise.
25332 2018-03-28  Martin Liska  <mliska@suse.cz>
25334         PR target/84988
25335         * config/i386/i386.c (ix86_function_arg_advance): Do not call
25336         chkp_type_bounds_count if MPX is not enabled.
25338 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
25340         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
25342 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
25344         PR target/84914
25345         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
25346         function to create the function decl for complex long double
25347         multiply and divide for -mabi=ieeelongdouble.
25348         (init_float128_ieee): Call it.
25350 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
25352         PR target/85044
25353         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
25354         -fcf-protection=branch -mibt.
25355         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
25357 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25359         PR target/81863
25360         * config/arm/arm.c (arm_valid_symbolic_address): Handle
25361         arm_word_relocations.
25363 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
25365         PR target/85056
25366         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
25367         extern array declarations.
25369 2018-03-27  Richard Biener  <rguenther@suse.de>
25371         PR middle-end/84067
25372         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
25373         explicit single_use checks.
25375 2018-03-27  Richard Biener  <rguenther@suse.de>
25377         PR tree-optimization/85082
25378         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
25379         Valueize the VUSE.
25381 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25383         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
25384         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
25385         Turn on fasynchronous-unwind-tables and funwind-tables.
25387 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
25389         PR target/85073
25390         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
25391         (*bmi_blsr_<mode>_ccz): Ditto.
25393 2018-03-26  Tom de Vries  <tom@codesourcery.com>
25395         PR tree-optimization/85063
25396         * omp-general.c (offloading_function_p): New function.  Factor out
25397         of ...
25398         * omp-offload.c (pass_omp_target_link::gate): ... here.
25399         * omp-general.h (offloading_function_p): Declare.
25400         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
25401         with attribute omp declare target for offloading functions.
25403 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
25405         PR tree-optimization/84005
25406         * tree-data-ref.h (get_base_for_alignment): Declare.
25407         * tree-data-ref.c (get_base_for_alignment_1): New function.
25408         (get_base_for_alignment): Likewise.
25409         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
25410         get_base_for_alignment to find a suitable base object, instead
25411         of always using drb->base_address.
25413 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
25415         PR inline-asm/85022
25416         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
25417         known size by default.
25419 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
25421         PR inline-asm/85030
25422         * lra-constraints.c (process_alt_operands): Don't match BLKmode
25423         and non BLKmode operands.
25425 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25427         PR target/85026
25428         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
25429         Clean up attributes.
25431 2018-03-23  Richard Biener  <rguenther@suse.de>
25433         PR debug/85020
25434         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
25435         we are going to emit early debug for LTO.
25437 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
25439         PR inline-asm/85034
25440         * function.c (match_asm_constraints_1): Don't optimize if input
25441         doesn't satisfy general_operand predicate for output's mode.
25443         PR inline-asm/85022
25444         * alias.c (write_dependence_p): Don't require for x_canonicalized
25445         non-VOIDmode if x has VOIDmode.
25447         PR sanitizer/85029
25448         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
25449         just don't try to optimize it rather than assert it never happens.
25451 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25453         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
25454         macro expansions for definition of ST_INTERNAL_<mode> and
25455         LD_INTERNAL_<mode> builtins.
25456         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
25457         Remove prototype.
25458         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
25459         function.
25460         (altivec_expand_st_builtin): Likewise.
25461         (altivec_expand_builtin): Remove calls to deleted functions.
25462         (rs6000_address_for_altivec): Delete this function.
25463         * config/rs6000/vector.md: Remove expands for
25464         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
25466 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
25468         PR target/84826
25469         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
25470         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
25471         re-computing once computed.
25472         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
25473         (arm_init_machine_status): Initialize
25474         machine->static_chain_stack_bytes.
25476 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25478         PR target/84760
25479         * doc/extend.texi: Add four new prototypes for vec_ld.
25480         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
25481         definitions for more logical presentation.
25482         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
25483         entries for V1TI variants of __builtin_altivec_ld builtin.
25484         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
25485         handling of V1TI variant of LVX icode pattern.
25486         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
25487         (rs6000_gimple_fold_builtin): Likewise.
25488         (altivec_init_builtins): Add code to define
25489         __builtin_altivec_lvx_v1ti function.
25491 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
25493         PR inline-asm/84941
25494         * function.c (match_asm_constraints_1): Don't do the optimization
25495         if input isn't a REG, SUBREG, MEM or constant.
25497 2018-03-22  Tom de Vries  <tom@codesourcery.com>
25499         PR tree-optimization/84956
25500         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
25501         bb_has_abnormal_pred.
25503 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
25505         PR sanitizer/85018
25506         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
25507         DECL_INITIAL (decl) to decl at the end.
25508         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
25509         adjust the comment.
25511 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
25513         * doc/extend.texi (__builtin_tgmath): Document when complex
25514         integer types are treated as _Complex _Float64.
25516 2018-03-21  Tom de Vries  <tom@codesourcery.com>
25518         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
25520 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
25522         PR tree-optimization/84960
25523         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
25524         if it is ENTRY block, move them into single succ of ENTRY in that case.
25526 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
25528         PR tree-optimization/84811
25529         * poly-int.h (poly_span_traits): Remove the T3 parameter and
25530         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
25531         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
25532         (known_subrange_p): Update accordingly.  Cast each value involved
25533         in the size comparison, rather than casting the result of the
25534         subtraction.
25536 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
25538         PR tree-optimization/84982
25539         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
25540         by flipping the least significant bit rather than all bits from
25541         bitpos to bitpos + bitsize - 1.
25543 2018-03-21  Nathan Sidwell  <nathan@acm.org>
25545         * doc/extend.texi (Deprecated Features): Remove mention of
25546         long-deleted deprecations.
25548 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25550         PR jit/84288
25551         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
25552         * configure: Regenerate.
25554 2018-03-21  Tom de Vries  <tom@codesourcery.com>
25556         PR tree-optimization/83126
25557         * tree-parloops.c (num_phis): New function.
25558         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
25560 2018-03-21  Nathan Sidwell  <nathan@acm.org>
25562         * doc/extend.texi (Deprecated Features): Update deprecated flags,
25563         mention anon-struct/union members and trailing attributes.
25565 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
25567         PR tree-optimization/84969
25568         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
25569         builtin memset partitions if they set different rhs values.
25571 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
25573         PR rtl-optimization/84989
25574         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
25575         VEC_DUPLICATE with scalar result mode.
25577 2018-03-21  Martin Liska  <mliska@suse.cz>
25579         PR ipa/84963
25580         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
25581         not intended return statement.
25583 2018-03-21  Martin Liska  <mliska@suse.cz>
25585         PR target/84988
25586         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
25587         (chkp_find_bound_slots_1): Limit number of iterations.
25589 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
25591         PR target/84838
25592         * Minor grammar fixes for x86 options.
25594 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
25596         PR debug/84875
25597         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
25598         holding REG_CFA_RESTORE notes, instead turn them into a USE.
25600 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
25602         PR target/83789
25603         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
25604         (altivec_lvx_<mode>_1op): Likewise.
25605         (altivec_stvx_<mode>_2op): Likewise.
25606         (altivec_stvx_<mode>_1op): Likewise.
25607         (altivec_lvx_<VM2:mode>): New define_expand.
25608         (altivec_stvx_<VM2:mode>): Likewise.
25609         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
25610         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
25611         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
25612         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
25613         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
25614         (rs6000_gen_lvx): Likewise.
25615         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
25616         (altivec_expand_stv_builtin): Likewise.
25617         (altivec_expand_builtin): Likewise.
25618         * config/rs6000/vector.md: Likewise.
25620 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25622         PR target/82518
25623         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
25624         BYTES_BIG_ENDIAN.
25626 2018-03-20  Richard Biener  <rguenther@suse.de>
25628         PR target/84986
25629         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
25630         sign-conversions as zero, fall back to standard scalar_stmt
25631         cost for the rest.
25633 2018-03-20  Martin Liska  <mliska@suse.cz>
25635         PR ipa/84825
25636         * predict.c (rebuild_frequencies): Handle case when we have
25637         PROFILE_ABSENT, but flag_guess_branch_prob is false.
25639 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
25641         PR target/84990
25642         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
25643         flag_section_anchors.
25644         * varasm.c (use_blocks_for_decl_p): Remove hack for
25645         dw2_force_const_mem.
25647         PR target/84845
25648         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
25649         to ...
25650         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
25651         be created, use lowpart_subreg of operands[0] rather than operands[0]
25652         itself.
25653         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
25654         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
25655         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
25656         and n constraint instead of aarch64_shift_imm_di and Usd.
25657         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
25658         (*aarch64_<optab>_reg_minus<mode>3): ... this.
25660 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
25662         PR target/82989
25663         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
25664         to favor GPR over NEON registers.
25665         (<shift>di3_neon): Likewise.
25667 2018-03-20  Tom de Vries  <tom@codesourcery.com>
25669         PR target/84952
25670         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
25671         (nvptx_process_pars): Emit bar.sync asap and alap.
25673 2018-03-20  Tom de Vries  <tom@codesourcery.com>
25675         PR target/84954
25676         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
25677         seen_label if seen_label is already set.
25679 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
25681         PR target/84945
25682         * config/i386/i386.c (fold_builtin_cpu): For features above 31
25683         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
25684         Use 1U instead of 1.  Formatting fixes.
25686         PR c/84953
25687         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
25688         instead of TREE_TYPE (s1) for the return value.
25690 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
25692         PR tree-optimization/84946
25693         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
25694         bitsize + bitsize in poly_uint64 rather than poly_int64.
25696         PR sanitizer/78651
25697         * dwarf2asm.c: Include fold-const.c.
25698         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
25699         of decl rather than decl itself.
25701         PR rtl-optimization/84643
25702         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
25704 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
25706         PR sanitizer/78651
25707         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
25708         calling assemble_variable.
25710 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
25712         PR target/81647
25713         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
25714         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
25716 2018-03-19  Jim Wilson  <jimw@sifive.com>
25718         PR bootstrap/84856
25719         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
25720         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
25721         (riscv_first_stack_step): Likewise.
25722         (riscv_option_override): Use STACK_BOUNDARY instead of
25723         MIN_STACK_BOUNDARY.
25724         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
25725         MIN_STACK_BOUNDARY.
25726         (BIGGEST_ALIGNMENT): Set to 128.
25727         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
25728         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
25729         STACK_BOUNDARY.
25731 2018-03-19  Richard Biener  <rguenther@suse.de>
25733         PR tree-optimization/84933
25734         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
25735         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
25737 2018-03-19  Richard Biener  <rguenther@suse.de>
25739         PR tree-optimization/84859
25740         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
25741         (cond_if_else_store_replacement): Perform sinking operation on
25742         single-store BBs regardless of MAX_STORES_TO_SINK setting.
25743         Generalize what a BB with a single eligible store is.
25745 2018-03-19  Richard Biener  <rguenther@suse.de>
25747         PR tree-optimization/84929
25748         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
25749         chrec_is_positive against non-chrec arg.
25751 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
25753         PR target/84711
25754         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
25756 2018-03-18  Martin Liska  <mliska@suse.cz>
25758         PR rtl-optimization/84635
25759         * regrename.c (build_def_use): Use matches_mode only when
25760         matches >= 0.
25762 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
25764         PR tree-optimization/84913
25765         * tree-vect-loop.c (vectorizable_reduction): Don't try to
25766         vectorize chains of COND_EXPRs.
25768 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
25770         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
25772 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
25774         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
25776 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
25778         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
25780 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
25781             Kito Cheng  <kito.cheng@gmail.com>
25783         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
25784         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
25785         (nds32_adjust_reg_alloc_order): New function.
25786         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
25788 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
25790         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
25791         nds32_print_operand, nds32_print_operand_address): Use
25792         HOST_WIDE_INT_PRINT_DEC instead.
25794 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
25796         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
25798 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
25800         PR target/84902
25801         * config/i386/i386.c (initial_ix86_tune_features,
25802         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
25803         unsigned long long.
25804         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
25805         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
25806         rather than 1u << ix86_tune.  Formatting fix.
25807         (ix86_option_override_internal): Change ix86_arch_mask from
25808         unsigned int to unsigned HOST_WIDE_INT, initialize to
25809         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
25810         (ix86_function_specific_restore): Likewise.
25812 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
25814         PR target/84899
25815         * postreload.c (reload_combine_recognize_pattern): Perform
25816         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
25817         truncate_int_for_mode the result for the destination's mode.
25819         PR c/84909
25820         * hsa-gen.c (mem_type_for_type): Fix comment typo.
25821         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
25822         Likewise.
25823         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
25824         Likewise.
25826 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
25828         PR target/84876
25829         * lra-assigns.c (lra_split_hard_reg_for): Don't use
25830         regno_allocno_class_array and sorted_pseudos.
25831         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
25832         insns where regno is used.
25834 2018-03-16  Martin Liska  <mliska@suse.cz>
25836         PR ipa/84833
25837         * multiple_target.c (create_dispatcher_calls): Redirect
25838         reference in the symbol table.
25840 2018-03-16  Martin Liska  <mliska@suse.cz>
25842         PR ipa/84722
25843         * multiple_target.c (create_dispatcher_calls): Redirect also
25844         an alias.
25846 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
25848         PR c++/79937
25849         PR c++/82410
25850         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
25851         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
25852         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
25854 2018-03-16  Julia Koval  <julia.koval@intel.com>
25856         * doc/invoke.texi (Skylake Server): Add CLWB.
25857         Cannonlake): Remove CLWB.
25859 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
25861         PR tree-optimization/84841
25862         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
25863         1 << 3.
25864         (FLOAT_ONE_CONST_TYPE): Define.
25865         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
25866         (sort_by_operand_rank): Put entries with higher constant_type last
25867         rather than first to match comments.
25869 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
25871         * config/nios2/nios2.md (movsi_internal): Fix thinko in
25872         split predicate.
25874 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
25876         PR c++/79085
25877         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
25878         check and use address of target always.
25880 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
25882         PR target/84574
25883         * config/i386/i386.c (indirect_thunk_needed): Update comments.
25884         (indirect_thunk_bnd_needed): Likewise.
25885         (indirect_thunks_used): Likewise.
25886         (indirect_thunks_bnd_used): Likewise.
25887         (indirect_return_needed): New.
25888         (indirect_return_bnd_needed): Likewise.
25889         (output_indirect_thunk_function): Add a bool argument for
25890         function return.
25891         (output_indirect_thunk_function): Don't generate alias for
25892         function return thunk.
25893         (ix86_code_end): Call output_indirect_thunk_function to generate
25894         function return thunks.
25895         (ix86_output_function_return): Set indirect_return_bnd_needed
25896         and indirect_return_needed instead of indirect_thunk_bnd_needed
25897         and indirect_thunk_needed.
25899 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
25901         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
25902         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
25903         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
25905 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
25906             Paul Hua <paul.hua.gm@gmail.com>
25908         PR c/84852
25909         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
25911 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
25913         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
25914         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
25915         resp. SFmode cases.
25917 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
25919         PR target/84711
25920         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
25921         instead of GET_MODE_SIZE when comparing Units.
25923 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
25925         PR target/68256
25926         * varasm.c (hash_section): Return an unchangeble hash value
25927         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
25928         Return !aarch64_can_use_per_function_literal_pools_p ().
25930 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
25932         PR target/84860
25933         * optabs.c (emit_conditional_move): Pass address of cmode's copy
25934         rather than address of cmode as last argument to prepare_cmp_insn.
25936 2018-03-15  Julia Koval  <julia.koval@intel.com>
25938         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
25939         F_AVX512VNNI, F_AVX512BITALG): New.
25941 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
25943         PR target/83451
25944         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
25945         insn for floating-point loads and stores.
25947 2018-03-14  Carl Love  <cel@us.ibm.com>
25949         * config/rs6000/rs6000-c.c: Add macro definitions for
25950         ALTIVEC_BUILTIN_VEC_PERMXOR.
25951         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
25952         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
25953         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
25954         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
25955         UNSPEC_VPERMXOR.
25956         * config/doc/extend.texi: Add prototypes for vec_permxor.
25958 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
25960         PR c/84852
25961         * diagnostic-show-locus.c (class layout_point): Convert m_line
25962         from int to linenum_type.
25963         (line_span::comparator): Use linenum "compare" function when
25964         comparing line numbers.
25965         (test_line_span): New function.
25966         (layout_range::contains_point): Convert param "row" from int to
25967         linenum_type.
25968         (layout_range::intersects_line_p): Likewise.
25969         (layout::will_show_line_p): Likewise.
25970         (layout::print_source_line): Likewise.
25971         (layout::should_print_annotation_line_p): Likewise.
25972         (layout::print_annotation_line): Likewise.
25973         (layout::print_leading_fixits): Likewise.
25974         (layout::annotation_line_showed_range_p): Likewise.
25975         (struct line_corrections): Likewise for field m_row.
25976         (line_corrections::line_corrections): Likewise for param "row".
25977         (layout::print_trailing_fixits): Likewise.
25978         (layout::get_state_at_point): Likewise.
25979         (layout::get_x_bound_for_row): Likewise.
25980         (layout::print_line): Likewise.
25981         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
25982         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
25983         * input.c (selftest::test_linenum_comparisons): New function.
25984         (selftest::input_c_tests): Call it.
25985         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
25986         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
25987         * selftest.h (ASSERT_GT): New macro.
25988         (ASSERT_GT_AT): New macro.
25989         (ASSERT_LT): New macro.
25990         (ASSERT_LT_AT): New macro.
25992 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
25994         PR rtl-optimization/84780
25995         * combine.c (distribute_links): Don't make a link based on pc_rtx.
25997 2018-03-14  Martin Liska  <mliska@suse.cz>
25999         * tree.c (record_node_allocation_statistics): Use
26000         get_stats_node_kind.
26001         (get_stats_node_kind): New function extracted from
26002         record_node_allocation_statistics.
26003         (free_node): Use get_stats_node_kind.
26005 2018-03-14  Richard Biener  <rguenther@suse.de>
26007         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
26008         that the value-set of ANTIC_IN doesn't grow.
26010         Revert
26011         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
26012         member.
26013         (BB_VISITED_WITH_VISITED_SUCCS): New define.
26014         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
26016 2018-03-14  Julia Koval  <julia.koval@intel.com>
26018         * config.gcc (icelake-client, icelake-server): New.
26019         (icelake): Remove.
26020         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
26021         (initial_ix86_arch_features): Ditto.
26022         (PTA_SKYLAKE): Add SGX.
26023         (PTA_ICELAKE): Remove.
26024         (PTA_ICELAKE_CLIENT): New.
26025         (PTA_ICELAKE_SERVER): New.
26026         (ix86_option_override_internal): Split up icelake on icelake client and
26027         icelake server.
26028         (get_builtin_code_for_version): Ditto.
26029         (fold_builtin_cpu): Ditto.
26030         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
26031         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
26032         * config/i386/i386.h (processor_type): Ditto.
26033         * doc/invoke.texi: Ditto.
26035 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
26037         PR sanitizer/83392
26038         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
26039         INTEGER_CST offset, add it together with bitpos / 8 and
26040         sign extend based on POINTER_SIZE.
26042         PR target/84844
26043         Revert
26044         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
26046         PR target/78090
26047         * config/i386/constraints.md (Yc): New register constraint.
26048         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
26049         Use Yc constraint for alternative 2 of operand 0.  Remove
26050         preferred_for_speed attribute.
26052 2018-03-14  Richard Biener  <rguenther@suse.de>
26054         PR tree-optimization/84830
26055         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
26056         with the old one to avoid oscillations.
26058 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
26060         PR target/83712
26061         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
26062         pseudos.
26063         (assign_by_spills): Return a flag of reload assignment failure.
26064         Do not process the reload assignment failures.  Do not spill other
26065         reload pseudos if they has the same reg class.  Update n if
26066         necessary.
26067         (lra_assign): Add a return arg.  Set up from the result of
26068         assign_by_spills call.
26069         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26070         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26071         usage_insns if it is not NULL.
26072         (spill_hard_reg_in_range): New function.
26073         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26074         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26075         function prototypes.
26076         (lra_assign): Change prototype.
26077         * lra.c (lra): Add code to deal with fails by splitting hard reg
26078         live ranges.
26080 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
26082         * config/riscv/riscv.opt (mrelax): New option.
26083         * config/riscv/riscv.c (riscv_file_start): Emit ".option
26084         "norelax" when riscv_mrelax is disabled.
26085         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
26087 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26089         PR target/84743
26090         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
26091         reassociation for int modes.
26093 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26095         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
26096         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
26097         for big-endian.
26098         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
26099         * config/aarch64/aarch64-sve.md
26100         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
26101         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
26102         (*extend<mode><Vwide>2): Rename to...
26103         (aarch64_sve_extend<mode><Vwide>2): ...this.
26104         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
26105         renaming the old pattern to...
26106         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
26107         unsigned packs.
26108         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
26109         define_expand, renaming the old pattern to...
26110         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
26111         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
26112         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
26113         account when deciding which SVE instruction the optab should use.
26114         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
26116 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26118         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
26119         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
26120         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
26121         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
26122         (tlsdesc_small_<mode>): Turn a define_expand and use
26123         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
26124         (tlsdesc_small_advsimd_<mode>): ...this.
26125         (tlsdesc_small_sve_<mode>): New pattern.
26127 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26129         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
26130         (UNSPEC_UMUL_HIGHPART): New constants.
26131         (MUL_HIGHPART): New int iteraor.
26132         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
26133         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
26134         define_expand.
26135         (*<su>mul<mode>3_highpart): New define_insn.
26137 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
26139         PR lto/84805
26140         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
26141         incomplete types.
26143 2018-03-13  Martin Liska  <mliska@suse.cz>
26145         PR ipa/84658.
26146         * (sem_item_optimizer::sem_item_optimizer): Initialize new
26147         vector.
26148         (sem_item_optimizer::~sem_item_optimizer): Release it.
26149         (sem_item_optimizer::merge_classes): Register variable aliases.
26150         (sem_item_optimizer::fixup_pt_set): New function.
26151         (sem_item_optimizer::fixup_points_to_sets): Likewise.
26152         * ipa-icf.h: Declare new variables and functions.
26154 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
26156         PR middle-end/84834
26157         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
26158         integer_pow2p@2 and test integer_pow2p in condition.
26159         (A < 0 ? C : 0): Similarly for @1.
26161         PR middle-end/84831
26162         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
26163         characters starting at p contain '\0' character, don't look beyond
26164         that.
26166         PR target/84827
26167         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
26168         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
26170         PR target/84828
26171         * reg-stack.c (change_stack): Change update_end var from int to
26172         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
26173         also call set_block_for_insn on the newly added insns and rescan.
26175         PR target/84786
26176         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
26177         on the last operand.
26179         PR c++/84704
26180         * tree.c (stabilize_reference_1): Return save_expr (e) for
26181         STATEMENT_LIST even if it doesn't have side-effects.
26183 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
26185         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
26187 2018-03-12  Renlin Li  <renlin.li@arm.com>
26189         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
26190         aarch64_output_scalar_simd_mov_immediate.
26192 2018-03-12  Martin Sebor  <msebor@redhat.com>
26194         PR tree-optimization/83456
26195         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
26196         for perfectly overlapping calls to memcpy.
26197         (gimple_fold_builtin_memory_chk): Same.
26198         (gimple_fold_builtin_strcpy): Handle no-warning.
26199         (gimple_fold_builtin_stxcpy_chk): Same.
26200         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
26202 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
26204         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
26205         parameter.  Use it for SFmode.
26206         (rs6000_function_arg_advance_1): Adjust.
26207         (rs6000_function_arg): Adjust.
26208         (rs6000_gimplify_va_arg): Pass false for that new parameter.
26210 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
26212         PR rtl-optimization/84169
26213         PR rtl-optimization/84780
26214         * combine.c (can_combine_p): Check for a 2-insn combination whether
26215         the destination register is used between the two insns, too.
26217 2018-03-12  Richard Biener  <rguenther@suse.de>
26219         PR tree-optimization/84803
26220         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
26221         for refs DR analysis didn't process.
26223 2018-03-12  Richard Biener  <rguenther@suse.de>
26225         PR tree-optimization/84777
26226         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
26227         force-vectorize loops ignore whether we are optimizing for size.
26229 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26231         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
26232         (TARGET_MD_ASM_ADJUST): Define.
26234 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
26235             Kito Cheng  <kito.cheng@gmail.com>
26236             Chung-Ju Wu  <jasonwucj@gmail.com>
26238         * config/nds32/nds32.c (nds32_compute_stack_frame,
26239         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
26240         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
26241         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
26242         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
26243         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
26244         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
26245         * config/nds32/nds32.md (prologue, epilogue): Use macro
26246         NDS32_V3PUSH_AVAILABLE_P to do checking.
26248 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
26250         PR debug/58150
26251         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
26252         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
26253         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
26254         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
26255         addition of most attributes on !orig_type_die or the attribute not
26256         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
26258 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26259             Chung-Ju Wu  <jasonwucj@gmail.com>
26261         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
26262         __NDS32_VH__ macro.
26263         * config/nds32/nds32.opt (mvh): New option.
26265 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26266             Chung-Ju Wu  <jasonwucj@gmail.com>
26268         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
26269         function.
26270         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
26271         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
26272         definition.
26274 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26275             Chung-Ju Wu  <jasonwucj@gmail.com>
26277         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
26278         function.
26279         * config/nds32/nds32-multiple.md (strlensi): New pattern.
26280         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
26282 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
26283             Kito Cheng  <kito.cheng@gmail.com>
26284             Chung-Ju Wu  <jasonwucj@gmail.com>
26286         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
26287         UNSPEC_FFMISM and UNSPEC_FLMISM.
26288         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
26289         for ffb, ffmism and flmism.
26290         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
26291         (unspec_ffmism): Ditto.
26292         (unspec_flmism): Ditto.
26293         (nds32_expand_builtin_impl): Check if string extension is available.
26294         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
26295         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
26297 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
26299         Reverting patch:
26300         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
26302         PR target/83712
26303         * lra-assigns.c (assign_by_spills): Return a flag of reload
26304         assignment failure.  Do not process the reload assignment
26305         failures.  Do not spill other reload pseudos if they has the same
26306         reg class.
26307         (lra_assign): Add a return arg.  Set up from the result of
26308         assign_by_spills call.
26309         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26310         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26311         usage_insns if it is not NULL.
26312         (spill_hard_reg_in_range): New function.
26313         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26314         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26315         function prototypes.
26316         (lra_assign): Change prototype.
26317         * lra.c (lra): Add code to deal with fails by splitting hard reg
26318         live ranges.
26320 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
26322         PR target/84807
26323         * config/i386/i386.opt: Replace Enforcment with Enforcement.
26325 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
26327         PR debug/84620
26328         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
26329         (dw_val_node): Add val_symbolic_view.
26330         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
26331         (symview_upper_bound): New.
26332         (new_line_info_table): Initialize symviews_since_reset.
26333         (dwarf2out_source_line): Count symviews_since_reset and set
26334         symview_upper_bound.
26335         (dw_val_equal_p): Handle symview.
26336         (add_AT_symview): New.
26337         (print_dw_val): Handle symview.
26338         (attr_checksum, attr_checksum_ordered): Likewise.
26339         (same_dw_val_p, size_of_die): Likewise.
26340         (value_format, output_die): Likewise.
26341         (add_high_low_attributes): Use add_AT_symview for entry_view.
26342         (dwarf2out_finish): Reset symview_upper_bound, clear
26343         zero_view_p.
26345 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
26347         PR target/83969
26348         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
26349         Add strict argument and use it.
26350         (rs6000_split_multireg_move): Update for new strict argument.
26351         (mem_operand_gpr): Disallow all non-offsettable addresses.
26352         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
26354 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
26356         PR target/84772
26357         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
26358         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
26359         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
26361         PR c++/84767
26362         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
26363         decl, use remap_type if we want to use the type.
26365 2018-03-09  Martin Sebor  <msebor@redhat.com>
26367         PR tree-optimization/84526
26368         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
26369         Remove dead code.
26370         (builtin_access::generic_overlap): Be prepared to handle non-array
26371         base objects.
26373 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
26375         PR rtl-optimization/84682
26376         * lra-constraints.c (process_address_1): Check is_address flag
26377         for address constraints.
26378         (process_alt_operands): Likewise.
26379         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
26380         preprocess_constraints.
26381         * recog.h (preprocess_constraints): Add oploc parameter.
26382         Adjust callers.
26383         * recog.c (preprocess_constraints): Test address_operand for
26384         CT_ADDRESS constraints.
26386 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
26388         PR target/83712
26389         * lra-assigns.c (assign_by_spills): Return a flag of reload
26390         assignment failure.  Do not process the reload assignment
26391         failures.  Do not spill other reload pseudos if they has the same
26392         reg class.
26393         (lra_assign): Add a return arg.  Set up from the result of
26394         assign_by_spills call.
26395         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26396         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26397         usage_insns if it is not NULL.
26398         (spill_hard_reg_in_range): New function.
26399         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26400         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26401         function prototypes.
26402         (lra_assign): Change prototype.
26403         * lra.c (lra): Add code to deal with fails by splitting hard reg
26404         live ranges.
26406 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26408         PR target/83193
26409         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
26410         Accept complain bool parameter.  Only emit errors if it is true.
26411         (arm_parse_cpu_option_name): Likewise.
26412         (arm_target_thumb_only): Adjust callers of the above.
26413         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
26414         prototype to take a default true bool parameter.
26415         (arm_parse_arch_option_name): Likewise.
26417 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
26418             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
26420         PR jit/64089
26421         PR jit/84288
26422         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
26423         * configure: Regenerate.
26424         * configure.ac ("linker --version-script option"): New.
26425         ("linker soname option"): New.
26427 2018-03-09  Richard Biener  <rguenther@suse.de>
26429         PR tree-optimization/84775
26430         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
26431         immediate uses of predicate stmts and mark them modified.
26433         Revert
26434         PR tree-optimization/84178
26435         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
26436         to caller.
26437         (version_loop_for_if_conversion): Delay update_ssa call.
26438         (tree_if_conversion): Delay update_ssa until after predicate
26439         insertion.
26441 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
26443         PR target/84763
26444         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
26445         when the function accesses prior frames.
26447 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
26449         PR debug/84456
26450         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
26451         gen_llsym, otherwise call maybe_gen_llsym.
26453         PR inline-asm/84742
26454         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
26455         has ',' character inside of it.
26457 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26459         PR target/84748
26460         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
26461         as clobbering CC_REGNUM.
26463 2018-03-08  Richard Biener  <rguenther@suse.de>
26465         PR middle-end/84552
26466         * tree-scalar-evolution.c: Include tree-into-ssa.h.
26467         (follow_copies_to_constant): Do not follow SSA names registered
26468         for update.
26470 2018-03-08  Richard Biener  <rguenther@suse.de>
26472         PR tree-optimization/84178
26473         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
26474         to caller.
26475         (version_loop_for_if_conversion): Delay update_ssa call.
26476         (tree_if_conversion): Delay update_ssa until after predicate
26477         insertion.
26479 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
26481         PR tree-optimization/84178
26482         * tree-if-conv.c (release_bb_predicate): Remove the
26483         the assertion that the stmts have NULL use_ops.
26484         Discard the statements, asserting that they haven't
26485         yet been added to a BB.
26487 2018-03-08  Richard Biener  <rguenther@suse.de>
26489         PR tree-optimization/84746
26490         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
26491         (phi_translate): Pass in destination ANTIC_OUT set.
26492         (phi_translate_1): Likewise.  For a simplified result lookup
26493         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
26494         (phi_translate_set): Adjust.
26495         (do_pre_regular_insertion): Likewise.
26496         (do_pre_partial_partial_insertion): Likewise.
26498 2018-03-08  Martin Liska  <mliska@suse.cz>
26500         PR gcov-profile/84735
26501         * doc/gcov.texi: Document usage of profile files.
26502         * gcov-io.h: Document changes in the format.
26504 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
26506         PR debug/84404
26507         PR debug/84408
26508         * dwarf2out.c (struct dw_line_info_table): Update comments for
26509         view == -1.
26510         (FORCE_RESET_NEXT_VIEW): New.
26511         (FORCE_RESETTING_VIEW_P): New.
26512         (RESETTING_VIEW_P): Check for -1 too.
26513         (ZERO_VIEW_P): Likewise.
26514         (new_line_info_table): Force-reset next view.
26515         (dwarf2out_begin_function): Likewise.
26516         (dwarf2out_source_line): Simplify zero_view_p initialization.
26517         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
26518         view directly.  Omit view when omitting .loc at line 0.
26520 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
26522         PR tree-optimization/84740
26523         * tree-switch-conversion.c (process_switch): Call build_constructors
26524         only if info.phi_count is non-zero.
26526         PR tree-optimization/84739
26527         * tree-tailcall.c (find_tail_calls): Check call arguments against
26528         DECL_ARGUMENTS (current_function_decl) rather than
26529         DECL_ARGUMENTS (func) when checking for tail recursion.
26531 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
26533         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
26534         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
26535         Volker Reichelt's entry and add entries for people that perform
26536         GCC fuzzy testing and report numerous bugs.
26538 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
26540         PR target/82411
26541         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
26542         readonly data in sdata, if that is disabled.
26543         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
26544         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
26545         -mreadonly-in-sdata option.
26547 2018-03-07  Martin Sebor  <msebor@redhat.com>
26549         PR tree-optimization/84468
26550         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
26551         basic block when looking for nul assignment.
26553 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
26555         PR target/84277
26556         * except.h (output_function_exception_table): Adjust prototype.
26557         * except.c (output_function_exception_table): Remove FNNAME parameter
26558         and add SECTION parameter.  Ouput one part of the table at a time.
26559         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
26560         the first part of the exception table and emit unwind directives.
26561         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
26562         (i386_pe_seh_cold_init): Likewise.
26563         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
26564         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
26565         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
26566         (ix86_output_call_insn): Emit a nop in one more case for SEH.
26567         * config/i386/winnt.c: Include except.h.
26568         (struct seh_frame_state): Add reg_offset, after_prologue and
26569         in_cold_section fields.
26570         (i386_pe_seh_end_prologue): Set seh->after_prologue.
26571         (i386_pe_seh_cold_init): New function.
26572         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
26573         to seh->in_cold_section.
26574         (seh_emit_push): Record the offset of the push.
26575         (seh_emit_save): Record the offet of the save.
26576         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
26577         Test seh->after_prologue to disregard the epilogue.
26578         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
26579         (i386_pe_end_cold_function): New function.
26581 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
26583         PR fortran/84565
26584         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
26585         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
26587         PR c++/84704
26588         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
26589         on tmp_var.
26590         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
26591         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
26593         PR middle-end/84723
26594         * multiple_target.c: Include tree-inline.h and intl.h.
26595         (expand_target_clones): Diagnose and fail if node->definition and
26596         !tree_versionable_function_p (node->decl).
26598 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
26600         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
26601         sprint_ul.
26602         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
26603         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
26604         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
26606 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
26608         PR target/84710
26609         * combine.c (try_combine): Use reg_or_subregno instead of handling
26610         just paradoxical SUBREGs and REGs.
26612 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
26614         * config/arc/arc.c (arc_finalize_pic): Remove function.
26615         (arc_must_save_register): We use single base PIC register, remove
26616         checks to save/restore the PIC register.
26617         (arc_expand_prologue): Likewise.
26618         * config/arc/arc-protos.h (arc_set_default_type_attributes):
26619         Remove.
26620         (arc_verify_short): Likewise.
26621         (arc_attr_type): Likewise.
26622         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
26623         (walk_stores): Likewise.
26624         (arc_address_cost): Make it static.
26625         (arc_verify_short): Likewise.
26626         (branch_dest): Likewise.
26627         (arc_attr_type): Likewise.
26628         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
26629         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
26630         (arc_final_prescan_insn): Remove inserting the nops due to
26631         hardware hazards.  It is done in reorg step.
26632         (insn_length_variant_t): Remove.
26633         (insn_length_parameters_t): Likewise.
26634         (arc_insn_length_parameters): Likewise.
26635         (arc_get_insn_variants): Likewise.
26636         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
26638 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
26640         PR inline-asm/84683
26641         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
26642         assertion failure.
26644         PR tree-optimization/84687
26645         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
26646         on new_node->decl.
26647         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
26649 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26651         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
26652         Rename to ppc_speculation_barrier.
26653         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
26654         __builtin_ppc_speculation_barrier.
26656 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
26658         PR target/84700
26659         * combine.c (combine_simplify_rtx): Don't try to simplify if
26660         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
26661         are equal to x.
26663 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
26665         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
26666         to 32 bytes when compiling for POWER9.
26668 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
26670         PR target/84564
26671         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
26672         regparm >= 3 with no arg reg available also for calls with
26673         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
26675         PR target/84524
26676         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
26677         orig,vex.
26678         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
26680 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
26682         PR target/84264
26683         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
26685 2018-03-05  Richard Biener  <rguenther@suse.de>
26687         PR tree-optimization/84486
26688         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
26689         When inserting a __builtin_assume_aligned call set the LHS
26690         SSA name alignment info accordingly.
26692 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
26694         PR tree-optimization/84114
26695         * config/aarch64/aarch64.c (aarch64_reassociation_width)
26696         Avoid reassociation of FLOAT_MODE addition.
26698 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
26700         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
26701         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
26702         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
26703         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
26704         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
26705         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
26706         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
26707         and -mwbnoinvd.
26708         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
26709         __builtin_ia32_wbinvd): New builtins.
26710         (SPECIAL_ARGS2): New.
26711         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
26712         (SPECIAL_ARGS2): New.
26713         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
26714         (ix86_valid_target_attribute_inner_p): Ditto.
26715         (ix86_init_mmx_sse_builtins): Add special_args2.
26716         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
26717         TARGET_WBNOINVD_P): New.
26718         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
26719         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
26720         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
26721         * config/i386/immintrin.h (_wbinvd): New intrinsic.
26722         * config/i386/pconfigintrin.h: New file.
26723         * config/i386/wbnoinvdintrin.h: Ditto.
26724         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
26725         wbnoinvdintrin.h.
26726         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
26728 2018-03-05  Richard Biener  <rguenther@suse.de>
26730         PR tree-optimization/84670
26731         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
26732         member.
26733         (BB_VISITED_WITH_VISITED_SUCCS): New define.
26734         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
26735         (compute_antic_aux): Only assert the number of values in ANTIC_IN
26736         doesn't grow if all successors (recursively) were visited at least
26737         once.
26739 2018-03-05  Richard Biener  <rguenther@suse.de>
26741         PR tree-optimization/84650
26742         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
26743         if executed in the loop pipeline.
26745 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
26747         * doc/configfiles.texi (Configuration Files): Move info about
26748         conditionalizing $target-protos.h to...
26749         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
26750         differs from $target-protos.h.
26752 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
26753             Chung-Ju Wu  <jasonwucj@gmail.com>
26755         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
26756         * config/nds32/nds32-multiple.md (setmemsi): Define.
26757         * config/nds32/nds32-memory-manipulation.c
26758         (nds32_gen_dup_4_byte_to_word_value): New.
26759         (emit_setmem_word_loop): New.
26760         (emit_setmem_byte_loop): New.
26761         (nds32_expand_setmem_loop): New.
26762         (nds32_expand_setmem_loop_v3m): New.
26763         (nds32_expand_setmem_unroll): New.
26764         (nds32_expand_setmem): New.
26766 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
26767             Chung-Ju Wu  <jasonwucj@gmail.com>
26769         * config/nds32/nds32-memory-manipulation.c
26770         (nds32_emit_load_store): New.
26771         (nds32_emit_post_inc_load_store): New.
26772         (nds32_emit_mem_move): New.
26773         (nds32_emit_mem_move_block): New.
26774         (nds32_expand_movmemsi_loop_unknown_size): New.
26775         (nds32_expand_movmemsi_loop_known_size): New.
26776         (nds32_expand_movmemsi_loop): New.
26777         (nds32_expand_movmemsi_unroll): New.
26778         (nds32_expand_movmemqi): Rename ...
26779         (nds32_expand_movmemsi): ... to this.
26780         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
26781         (movmemsi): ... to this.
26782         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
26783         (nds32_expand_movmemsi): ... to this.
26785 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
26786             Monk Chiang  <sh.chiang04@gmail.com>
26787             Chung-Ju Wu  <jasonwucj@gmail.com>
26789         * config/nds32/nds32-protos.h
26790         (nds32_expand_load_multiple): New arguments.
26791         (nds32_expand_store_multiple): Ditto.
26792         (nds32_valid_multiple_load_store): Rename ...
26793         (nds32_valid_multiple_load_store_p): ... to this.
26794         * config/nds32/nds32-memory-manipulation.c
26795         (nds32_expand_load_multiple): Refine implementation.
26796         (nds32_expand_store_multiple): Ditto.
26797         * config/nds32/nds32-multiple.md
26798         (load_multiple): Update nds32_expand_load_multiple interface.
26799         (store_multiple): Update nds32_expand_store_multiple interface.
26800         * config/nds32/nds32-predicates.c
26801         (nds32_valid_multiple_load_store): Rename ...
26802         (nds32_valid_multiple_load_store_p): ... to this and refine
26803         implementation.
26804         * config/nds32/predicates.md
26805         (nds32_load_multiple_and_update_address_operation): New predicate.
26806         (nds32_store_multiple_and_update_address_operation): New predicate.
26808 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
26809             Chung-Ju Wu  <jasonwucj@gmail.com>
26811         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
26812         (combo): New attribute.
26813         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
26815 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
26817         * config/nds32/nds32.opt: Change -mcmodel= default value.
26819 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
26820             Monk Chiang  <sh.chiang04@gmail.com>
26821             Chung-Ju Wu  <jasonwucj@gmail.com>
26823         * config/nds32/constants.md (unspec_element): New enum.
26824         * config/nds32/constraints.md (Umw): New constraint.
26825         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
26826         * config/nds32/nds32-intrinsic.md: Likewise.
26827         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
26828         (nds32_valid_smw_lwm_base_p): New.
26829         (nds32_output_smw_single_word): New.
26830         (nds32_output_lmw_single_word): New.
26831         (nds32_expand_unaligned_load): New.
26832         (nds32_expand_unaligned_store): New.
26833         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
26834         (nds32_output_smw_single_word): Declare.
26835         (nds32_output_lmw_single_word): Declare.
26836         (nds32_expand_unaligned_load): Declare.
26837         (nds32_expand_unaligned_store): Declare.
26838         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
26839         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
26840         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
26841         NDS32_BUILTIN_UASTORE_DW.
26842         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
26843         predicate.
26845 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
26846             Kito Cheng  <kito.cheng@gmail.com>
26847             Chung-Ju Wu  <jasonwucj@gmail.com>
26849         * config/nds32/nds32-intrinsic.c
26850         (nds32_expand_builtin_null_ftype_reg): Delete.
26851         (nds32_expand_builtin_reg_ftype_imm): Ditto.
26852         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
26853         (nds32_read_argument): New.
26854         (nds32_legitimize_target): Ditto.
26855         (nds32_legitimize_argument): Ditto.
26856         (nds32_check_constant_argument): Ditto.
26857         (nds32_expand_unop_builtin): Ditto.
26858         (nds32_expand_unopimm_builtin): Ditto.
26859         (nds32_expand_binop_builtin): Ditto.
26860         (nds32_builtin_decl_impl): Ditto.
26861         (builtin_description): Ditto.
26862         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
26863         (nds32_init_builtins_impl): Ditto.
26864         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
26865         (nds32_builtin_decl): New.
26866         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
26867         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
26869 2018-03-02  Jeff Law  <law@redhat.com>
26871         * reorg.c (stop_search_p): Handle DEBUG_INSN.
26872         (redundant_insn, fill_simple_delay_slots): Likewise.
26873         (fill_slots_from_thread): Likewise.
26874         * resource.c (mark_referenced_resources): Likewise.
26875         (mark_set_resources, find_dead_or_set_registers): Likewise.
26877 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26879         * substring-locations.h (format_warning_va): Formatting fix for
26880         ATTRIBUTE_GCC_DIAG.
26881         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
26882         argument.
26883         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
26884         * substring-locations.c: Include intl.h.
26885         (format_warning_va): Turned into small wrapper around
26886         format_warning_n_va, renamed to ...
26887         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
26888         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
26889         use ngettext.
26890         (format_warning_at_substring_n): New function.
26891         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
26892         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
26893         format_warning_at_substring with just a shorter name instead of
26894         const function pointer.
26895         (fmtwarn_n): New function.
26896         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
26897         appropriate, get rid of all the fmtstr temporaries, move conditionals
26898         with G_() wrapped string literals directly into fmtwarn arguments,
26899         cast dir.len to (int), formatting fixes.
26901 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
26903         * doc/invoke.texi: Remove "Cilk Plus" references.
26905 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26906             Richard Biener  <rguenther@suse.de>
26908         PR ipa/84628
26909         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
26910         for error or warning attributes if CALL_FROM_THUNK_P is set.
26911         Formatting fixes.
26913 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26915         PR target/56540
26916         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
26917         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
26919         PR target/56540
26920         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
26921         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
26923         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
26924         instead of -1U in last predictors element's probability member.
26926 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
26928         PR ipa/83983
26929         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
26930         arguments if they are comparable.
26932 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
26934         PR tree-optimization/84634
26935         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
26936         masks and masked_loop_p with a single loop_masks, making sure it's
26937         null for bb vectorization.
26939 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
26941         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
26942         (vect_analyze_data_ref_access): Use loop->safe_len rather than
26943         loop->force_vectorize to check whether there is no alias.
26945 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26947         PR target/84614
26948         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
26949         prototypes.
26950         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
26951         comments.
26952         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
26953         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
26954         instead of a loop around prev_real_insn.
26955         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
26956         prev_real_insn.
26958         PR inline-asm/84625
26959         * config/i386/i386.c (ix86_print_operand): Use conditional
26960         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
26961         zero vector.
26963 2018-03-02  Richard Biener  <rguenther@suse.de>
26965         PR tree-optimization/84427
26966         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
26967         (bitmap_set_subtract_values): Rewrite to handle multiple
26968         exprs per value.
26969         (clean): Likewise.
26970         (prune_clobbered_mems): Likewise.
26971         (phi_translate): Take edge instead of pred/phiblock.
26972         (phi_translate_1): Likewise.
26973         (phi_translate_set): Likewise.  Insert all translated
26974         exprs for a value into the set, keeping possibly multiple
26975         expressions per value.
26976         (compute_antic_aux): Adjust for phi_translate changes.
26977         When intersecting union the expressions and prune those
26978         not in the final value set, keeping possibly multiple
26979         expressions per value.  Do not use value-insertion
26980         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
26981         all expressions.  Add verification that the value-sets
26982         only shrink during iteration.
26983         (compute_partial_antic_aux): Adjust for the phi_translate changes.
26984         (do_pre_regular_insertion): Likewise.
26985         (do_pre_partial_partial_insertion): Likewise.
26987 2018-03-02  Richard Biener  <rguenther@suse.de>
26989         PR target/82005
26990         * config/darwin.c (saved_debug_info_level): New static global.
26991         (darwin_asm_lto_start): Disable debug info generation for LTO out.
26992         (darwin_asm_lto_end): Restore debug info generation settings.
26994 2018-03-01  Martin Liska  <mliska@suse.cz>
26996         PR sanitizer/82484
26997         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
26998         volatile arguments.
27000 2018-03-01  Richard Biener  <rguenther@suse.de>
27002         PR debug/84645
27003         * dwarf2out.c (gen_variable_die): Properly handle late VLA
27004         type annotation with LTO when debug was disabled at compile-time.
27006 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
27008         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
27009         XINT with INTVAL.
27010         (mips_final_postscan_insn): Likewise.
27012 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
27014         PR rtl-optimization/84528
27015         * alias.c (init_alias_target): Add commentary.
27016         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
27017         a unique base value if the frame pointer is not eliminated
27018         to the stack pointer.
27020 2018-03-01  Tom de Vries  <tom@codesourcery.com>
27022         PR rtl-optimization/83327
27023         * lra-int.h (hard_regs_spilled_into): Declare.
27024         * lra.c (hard_regs_spilled_into): Define.
27025         (init_reg_info): Init hard_regs_spilled_into.
27026         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
27027         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
27028         (process_bb_lives): Handle hard_regs_spilled_into.
27029         (lra_create_live_ranges_1): Before doing liveness propagation, clear
27030         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
27032 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
27034         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
27035         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
27036         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
27037         * config/rs6000/aix72.h: New file.
27039 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
27041         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
27042         instead of warning_at with conditional singular and plural messages
27043         where possible.
27045         PR target/52991
27046         * stor-layout.c (update_alignment_for_field): For
27047         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
27048         && !DECL_PACKED (field), do the alignment update, just use
27049         only desired_align instead of MAX (type_align, desired_align)
27050         as the alignment.
27051         (place_field): Don't do known_align < desired_align handling
27052         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
27053         is non-NULL, instead do it after rli->prev_field handling and
27054         only if not within a bitfield word.  For DECL_PACKED (field)
27055         use type_align of BITS_PER_UNIT.
27057 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
27059         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
27060         superfluous parentheses and trailing spaces.
27062 2018-02-28  Richard Biener  <rguenther@suse.de>
27064         PR tree-optimization/84584
27065         * graphite-scop-detection.c (scop_detection::add_scop): Discard
27066         SCoPs with fake exit edge.
27068 2018-02-28  Martin Liska  <mliska@suse.cz>
27070         PR testsuite/84597
27071         * timevar.c (timer::print): Fix format to properly print 100%
27072         values.
27074 2018-02-28  Richard Biener  <rguenther@suse.de>
27076         PR middle-end/84607
27077         * genmatch.c (capture_info::walk_match): Do not mark
27078         captured expressions without operands as expr_p given
27079         they act more like predicates and should be subject to
27080         "lost tail" side-effect preserving.
27082 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
27084         PR rtl-optimization/81611
27085         * auto-inc-dec.c (attempt_change): Move dead note from
27086         mem_insn if it's the next use of regno
27087         (find_address): Take address use of reg holding
27088         non-incremented value.  Add parm to limit search to the named
27089         reg only.
27090         (merge_in_block): Attempt to use a mem insn that is the next
27091         use of the original regno.
27093 2018-02-27  Martin Sebor  <msebor@redhat.com>
27095         PR c++/83871
27096         * doc/invoke.texi (-Wmissing-attributes): New option.
27097         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
27099 2018-02-27  Martin Sebor  <msebor@redhat.com>
27101         PR translation/84207
27102         * diagnostic-core.h (warning_n, error_n, inform_n): Change
27103         n argument to unsigned HOST_WIDE_INT.
27104         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
27105         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
27106         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
27107         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
27109 2018-02-27  Richard Biener  <rguenther@suse.de>
27111         PR tree-optimization/84512
27112         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
27113         Do not use the estimate returned from record_stmt_cost for
27114         the scalar iteration cost but sum properly using add_stmt_cost.
27116 2018-02-27  Richard Biener  <rguenther@suse.de>
27118         PR tree-optimization/84466
27119         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
27120         Adjust last change to less strictly validate use operands.
27122 2018-02-27  Martin Liska  <mliska@suse.cz>
27124         PR gcov-profile/84548
27125         * gcov.c (process_file): Allow partial overlap and consider it
27126         also as group functions.
27127         (output_lines): Properly calculate range of lines for a group.
27129 2018-02-27  Martin Liska  <mliska@suse.cz>
27131         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
27132         'ggc' suffixes.  Change first column width.
27133         (timer::print): Fix formatting of the column.
27135 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
27137         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
27138         preserve inline entry blocks for the sake of debug inline
27139         entry point markers alone.
27140         (remove_unused_locals): Suggest in comments a better place to
27141         force the preservation of inline entry blocks that are
27142         otherwise unused, but do not preserve them.
27144 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27146         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
27148 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27150         PR target/84039
27151         * config/i386/constraints.md (Bs): Replace
27152         ix86_indirect_branch_register with
27153         TARGET_INDIRECT_BRANCH_REGISTER.
27154         (Bw): Likewise.
27155         * config/i386/i386.md (indirect_jump): Likewise.
27156         (tablejump): Likewise.
27157         (*sibcall_memory): Likewise.
27158         (*sibcall_value_memory): Likewise.
27159         Peepholes of indirect call and jump via memory: Likewise.
27160         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
27161         (*sibcall_value_GOT_32): Likewise.
27162         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27163         (GOT_memory_operand): Likewise.
27164         (call_insn_operand): Likewise.
27165         (sibcall_insn_operand): Likewise.
27166         (GOT32_symbol_operand): Likewise.
27167         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
27169 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
27171         PR rtl-optimization/83496
27172         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
27173         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
27174         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
27175         redundant insn, if any.
27176         (relax_delay_slots): Likewise.
27177         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
27179 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
27181         PR tree-optimization/83965
27182         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
27183         that grouped statements are part of a reduction chain.  Return
27184         true if the statement is not marked as a reduction itself but
27185         is part of a group.
27186         (vect_recog_dot_prod_pattern): Don't check whether the statement
27187         is part of a group here.
27188         (vect_recog_sad_pattern): Likewise.
27189         (vect_recog_widen_sum_pattern): Likewise.
27191 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
27193         PR debug/84545
27194         * final.c (rest_of_clean_state): Also look for calls inside sequences.
27196 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27198         PR target/84530
27199         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
27200         the bool argument.
27201         (ix86_output_indirect_function_return): New prototype.
27202         (ix86_split_simple_return_pop_internal): Likewise.
27203         * config/i386/i386.c (indirect_return_via_cx): New.
27204         (indirect_return_via_cx_bnd): Likewise.
27205         (indirect_thunk_name): Handle return va CX_REG.
27206         (output_indirect_thunk_function): Create alias for
27207         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
27208         (ix86_output_indirect_jmp): Remove the bool argument.
27209         (ix86_output_indirect_function_return): New function.
27210         (ix86_split_simple_return_pop_internal): Likewise.
27211         * config/i386/i386.md (*indirect_jump): Don't pass false
27212         to ix86_output_indirect_jmp.
27213         (*tablejump_1): Likewise.
27214         (simple_return_pop_internal): Change it to define_insn_and_split.
27215         Call ix86_split_simple_return_pop_internal to split it for
27216         -mfunction-return=.
27217         (simple_return_indirect_internal): Call
27218         ix86_output_indirect_function_return instead of
27219         ix86_output_indirect_jmp.
27221 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
27223         PR bootstrap/84405
27224         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
27225         memset and value initialization afterwards.
27227 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
27229         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
27231 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27233         PR target/84521
27234         * common/config/aarch64/aarch64-common.c
27235         (aarch_option_optimization_table[]): Switch
27236         off fomit-frame-pointer
27238 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
27239             Chung-Ju Wu  <jasonwucj@gmail.com>
27241         * config/nds32/nds32-multiple.md (load_multiple): Disallow
27242         volatile memory.
27243         (store_multiple): Ditto.
27245 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
27247         * config.gcc: Add --with-cpu support for nds32 target.
27248         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
27249         * config/nds32/nds32.opt: Add -mcpu= option.
27251 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
27253         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
27254         isel=yes): Warn for these deprecated options.
27256 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
27258         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
27259         ISA_2_5_MASKS_EMBEDDED.
27261 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
27263         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
27264         p->max as pointers rather than using iterative_hash_expr.
27266 2018-02-23  Carl Love  <cel@us.ibm.com>
27268         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
27269         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
27270         BU_P8V_OVERLOAD_2.
27271         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
27272         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
27273         P8V_BUILTIN_VEC_VUNSIGNED2.
27275 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
27277         PR target/81572
27278         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
27279         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
27280         LRA_UNKNOWN_ALT.
27281         * lra-constraints.c (curr_insn_transform): Set up
27282         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
27283         LRA_UNKNOWN_ALT.
27284         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
27285         * lra-eliminations.c (spill_pseudos): Ditto.
27286         (process_insn_for_elimination): Ditto.
27287         * lra-lives.c (reg_early_clobber_p): Use the new macros.
27288         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
27289         LRA_NON_CLOBBERED_ALT.
27291 2018-02-22  Martin Sebor  <msebor@redhat.com>
27293         PR tree-optimization/84480
27294         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
27295         to maybe_diag_stxncpy_trunc.  Call it.
27296         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
27297         from gimple_fold_builtin_strcpy.  Print inlining stack.
27298         (handle_builtin_stxncpy): Print inlining stack.
27299         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
27301 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
27303         PR target/84176
27304         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
27305         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
27306         and -fcheck-pointer-bounds are used together.
27307         (indirect_thunk_prefix): New enum.
27308         (indirect_thunk_need_prefix): New function.
27309         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
27310         "_nt" instead of "_bnd" for NOTRACK prefix.
27311         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
27312         (output_indirect_thunk_function): Likewise.
27313         (): Likewise.
27314         (ix86_code_end): Update output_indirect_thunk_function calls.
27315         (ix86_output_indirect_branch_via_reg): Replace
27316         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
27317         (ix86_output_indirect_branch_via_push): Likewise.
27318         (ix86_output_function_return): Likewise.
27319         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
27320         incompatible with -fcf-protection=branch and
27321         -fcheck-pointer-bounds.
27323 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
27325         PR target/83335
27326         * config/aarch64/aarch64.c (aarch64_print_address_internal):
27327         Change gcc_assert call to output_operand_lossage.
27329 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
27331         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
27333 2018-02-22  DJ Delorie  <dj@redhat.com>
27334             Sebastian Perta  <sebastian.perta@renesas.com>
27335             Oleg Endo  <olegendo@gcc.gnu.org>
27337         * config/rx/rx.c (rx_rtx_costs): New function.
27338         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
27340 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
27342         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
27344 2018-02-22  Martin Liska  <mliska@suse.cz>
27346         PR driver/83193
27347         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
27348         Add "native" as a possible value.
27350 2018-02-22  Martin Liska  <mliska@suse.cz>
27352         PR driver/83193
27353         * config/i386/i386.c (ix86_option_override_internal):
27354         Add "native" as a possible value for -march and -mtune.
27356 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
27358         PR target/84502
27359         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
27360         to all type variants.
27362         PR tree-optimization/84503
27363         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
27364         width as info->bitpos + info->bitsize - start.
27365         (merged_store_group::merge_overlapping): Simplify width computation.
27366         (check_no_overlap): New function.
27367         (imm_store_chain_info::try_coalesce_bswap): Compute expected
27368         start + width and last_order of the group, fail if check_no_overlap
27369         fails.
27370         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
27371         to group if check_no_overlap fails.
27373 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
27375         * config/rs6000/altivec.md: Delete contraint arguments to
27376         define_expand, define_split, and define_peephole2, and in
27377         define_insn_and_split if always unused.
27378         * config/rs6000/darwin.md: Ditto.
27379         * config/rs6000/dfp.md: Ditto.
27380         * config/rs6000/rs6000.md: Ditto.
27381         * config/rs6000/sync.md: Ditto.
27382         * config/rs6000/vector.md: Ditto.
27383         * config/rs6000/vsx.md: Ditto.
27385 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
27387         * config/rs6000/altivec.md: Write output control strings as braced
27388         blocks instead of double-quoted strings.
27389         * config/rs6000/darwin.md: Ditto.
27390         * config/rs6000/rs6000.md: Ditto.
27391         * config/rs6000/vector.md: Ditto.
27392         * config/rs6000/vsx.md: Ditto.
27394 2018-02-21  Jason Merrill  <jason@redhat.com>
27396         PR c++/84314 - ICE with templates and fastcall attribute.
27397         * attribs.c (build_type_attribute_qual_variant): Remove assert.
27399 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
27401         * ipa-cp.c (determine_versionability): Fix comment typos.
27403 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
27405         PR c/84229
27406         * ipa-cp.c (determine_versionability): Do not version functions caling
27407         va_arg_pack.
27409 2018-02-21  Martin Liska  <mliska@suse.cz>
27411         PR driver/83193
27412         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
27413         Add "native" as a possible value.
27414         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
27415         the macro when native cpu detection is available.
27417 2018-02-21  Martin Liska  <mliska@suse.cz>
27419         PR driver/83193
27420         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
27421         Add "native" as a possible value.
27422         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
27423         when native cpu detection is available.
27425 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
27426             Martin Sebor  <msebor@redhat.com>
27428         PR tree-optimization/84478
27429         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
27430         false.
27431         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
27432         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
27433         support which is conservatively correct, for 2 only stay conservative
27434         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
27435         argument to the 2 argument get_range_strlen, adjust 6 arg
27436         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
27437         false.
27438         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
27439         (gimple_fold_builtin_strlen): Pass true as last argument to
27440         get_range_strlen.
27442 2018-02-20  Martin Sebor  <msebor@redhat.com>
27444         PR middle-end/84095
27445         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
27446         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
27447         (builtin_memref::builtin_memref): Factor out parts into
27448         set_base_and_offset and call it.
27450 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
27452         PR middle-end/84406
27453         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
27454         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
27455         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
27456         search at the associated MODE_INT.
27458 2018-02-20  Jeff Law  <law@redhat.com>
27460         PR middle-end/82123
27461         PR tree-optimization/81592
27462         PR middle-end/79257
27463         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
27464         for range data rather than using global data.
27465         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
27466         range data rather than using global data.
27467         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
27468         pass it to children as needed.
27469         (struct directive::fmtresult): Similarly.
27470         (struct directive::set_width): Similarly.
27471         (struct directive::set_precision): Similarly.
27472         (format_integer, format_directive, parse_directive): Similarly.
27473         (format_none): Accept unnamed vr_values parameter.
27474         (format_percent, format_floating, format_character): Similarly.
27475         (format_string, format_plain): Similarly.
27476         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
27477         the EVRP range analyzer for range data rather than using global data.
27478         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
27479         gimple-ssa-evrp-analyze.h
27480         (class sprintf_dom_walker): Add after_dom_children member function.
27481         Add evrp_range_analyzer member.
27482         (sprintf_dom_walker::before_dom_children): Call into the EVRP
27483         range analyzer as needed.
27484         (sprintf_dom_walker::after_dom_children): New member function.
27485         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
27486         if not optimizing.
27487         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
27488         (evrp_range_analyzer::pop_to_marker): Likewise.
27490 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
27492         PR tree-optimization/84419
27493         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
27494         with the required type if its current type is compatible but
27495         different.
27497 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
27499         PR middle-end/82004
27500         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
27501         after vectorization.
27503 2018-02-20  Martin Liska  <mliska@suse.cz>
27505         PR driver/83193
27506         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
27507         possible values if we don't have a hint.
27509 2018-02-20  Martin Liska  <mliska@suse.cz>
27511         PR c/84310
27512         PR target/79747
27513         * final.c (shorten_branches): Build align_tab array with one
27514         more element.
27515         * opts.c (finish_options): Add alignment option limit check.
27516         (MAX_CODE_ALIGN): Likewise.
27517         (MAX_CODE_ALIGN_VALUE): Likewise.
27518         * doc/invoke.texi: Document maximum allowed option value for
27519         all -falign-* options.
27521 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
27523         PR target/84146
27524         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
27525         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
27526         * var-tracking.c (emit_note_insn_var_location): Remove all references
27527         to NOTE_INSN_CALL_ARG_LOCATION.
27528         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
27529         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
27530         Use copy_rtx_if_shared.
27531         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
27532         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
27533         (dwarf2out_var_location): Remove handling of
27534         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
27535         on call_insn.
27536         * final.c (final_scan_insn): Remove all references to
27537         NOTE_INSN_CALL_ARG_LOCATION.
27538         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
27539         before dumping final insns.
27540         * except.c (emit_note_eh_region_end): Remove all references to
27541         NOTE_INSN_CALL_ARG_LOCATION.
27542         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
27543         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
27544         * config/arc/arc.c (hwloop_optimize): Likewise.
27545         * config/arm/arm.c (create_fix_barrier): Likewise.
27546         * config/s390/s390.c (s390_chunkify_start): Likewise.
27547         * config/sh/sh.c (find_barrier): Likewise.
27548         * config/i386/i386.c (rest_of_insert_endbranch,
27549         ix86_seh_fixup_eh_fallthru): Likewise.
27550         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
27551         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
27552         * config/frv/frv.c (frv_function_prologue): Likewise.
27553         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
27554         reg note.
27555         (note_outside_basic_block_p): Remove all references to
27556         NOTE_INSN_CALL_ARG_LOCATION.
27557         * gengtype.c (adjust_field_rtx_def): Likewise.
27558         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
27559         Likewise.
27560         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
27561         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
27563         PR c++/84444
27564         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
27565         is ADDR_EXPR.
27567         PR tree-optimization/84452
27568         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
27569         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
27570         is NULL.
27572 2018-02-19  Martin Liska  <mliska@suse.cz>
27574         PR sanitizer/82183
27575         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
27577 2018-02-19  Martin Liska  <mliska@suse.cz>
27578             Richard Sandiford  <richard.sandiford@linaro.org>
27580         PR tree-optimization/82491
27581         * gimple-fold.c (get_base_constructor): Make earlier bail out
27582         to prevent ubsan.
27584 2018-02-19  Carl Love  <cel@us.ibm.com>
27586         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
27587         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
27588         BU_P8V_OVERLOAD_1.
27589         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
27590         P8V_BUILTIN_VEC_NEG.
27592 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
27594         * config/rl78/rl78.md (movdf): New define expand.
27596 2018-02-19  Martin Liska  <mliska@suse.cz>
27598         PR other/80589
27599         * doc/invoke.texi: Fix typo.
27600         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
27602 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
27604         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
27605         handle rs6000_single_float and rs6000_double_float specially for
27606         e500 family CPUs.
27608 2018-02-16  Jeff Law  <law@redhat.com>
27610         * config/rx/rx.c (add_pop_cfi_notes): New function.;
27611         (pop_regs): Use it.
27613 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
27615         PR ipa/84425
27616         * ipa-inline.c (inline_small_functions): Fix a typo.
27618 2018-02-16  Nathan Sidwell  <nathan@acm.org>
27620         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
27622 2018-02-16  Carl Love  <cel@us.ibm.com>
27624         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
27625         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
27626         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
27627         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
27628         expansion to P8V_BUILTIN_VEC_FLOAT2.
27630 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
27632         PR rtl-optimization/70023
27633         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
27634         src_regno into account.
27636 2018-02-16  Carl Love  <cel@us.ibm.com>
27638         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
27639         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
27640         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
27641         * config/rs6000/rs6000.c: Remove case statements for
27642         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
27643         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
27644         and P9V_BUILTIN_VEC_VINSERT4B.
27645         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
27646         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
27647         * config/rs6000/vsx.md:
27648         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
27649         vec_insert4b.
27651 2018-02-16  Carl Love  <cel@us.ibm.com>
27653         * config/rs6000/altivec.h: Add builtin names vec_extract4b
27654         vec_insert4b.
27655         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
27656         definitions.
27657         * config/rs6000/rs6000-c.c: Add the definitions for
27658         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
27659         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
27660         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
27661         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
27662         definition for insert4b and define insn *insert3b_internal.
27663         * doc/extend.texi: Add documentation for vec_extract4b.
27665 2018-02-16  Nathan Sidwell  <nathan@acm.org>
27667         * doc/extend.texi (Backwards Compatibility): Mention friend
27668         injection.  Note for-scope is deprecated.
27669         * doc/invoke.texi (-ffriend-injection): Deprecate.
27671 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
27673         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
27674         that moved to I2, also allow destinations that are a paradoxical
27675         subreg (instead of a normal reg).
27677 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
27679         PR target/83831
27680         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
27681         to QImode.
27683 2018-02-16  Richard Biener  <rguenther@suse.de>
27685         PR tree-optimization/84037
27686         PR tree-optimization/84016
27687         PR target/82862
27688         * config/i386/i386.c (ix86_builtin_vectorization_cost):
27689         Adjust vec_construct for the fact we need additional higher latency
27690         128bit inserts for AVX256 and AVX512 vector builds.
27691         (ix86_add_stmt_cost): Scale vector construction cost for
27692         elementwise loads.
27694 2018-02-16  Richard Biener  <rguenther@suse.de>
27696         PR tree-optimization/84417
27697         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
27698         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
27699         (non_rewritable_lvalue_p): Likewise, use poly-ints.
27701 2018-02-16  Martin Liska  <mliska@suse.cz>
27703         PR sanitizer/84307
27704         * internal-fn.def (ASAN_CHECK): Set proper flags.
27705         (ASAN_MARK): Likewise.
27707 2018-02-16  Julia Koval  <julia.koval@intel.com>
27709         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
27710         from PTA_CANNONLAKE.
27712 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
27714         PR target/84272
27715         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
27716         Use ++iter rather than iter++ for std::list iterators.
27717         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
27718         defer deleting them until all nodes in the forest are processed.  Do
27719         free even leaf nodes.  Change to_process into auto_vec.
27721         PR bootstrap/84405
27722         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
27723         * vec.h (vec_default_construct): Use memset instead of placement new
27724         if BROKEN_VALUE_INITIALIZATION is defined.
27725         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
27726         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
27727         is defined.
27729         PR rtl-optimization/83723
27730         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
27731         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
27732         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
27733         recursive calls.
27734         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
27735         callers.
27736         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
27738 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
27740         PR rtl-optimization/81443
27741         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
27742         from inner REGs to paradoxical SUBREGs.
27744 2018-02-16  Richard Biener  <rguenther@suse.de>
27746         PR tree-optimization/84399
27747         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
27748         For operands we can analyze at their definition make sure we can
27749         analyze them at each use as well.
27751 2018-02-16  Richard Biener  <rguenther@suse.de>
27753         PR tree-optimization/84190
27754         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
27755         volatile accesses if the decl isn't volatile.
27757 2018-02-15  Jason Merrill  <jason@redhat.com>
27759         PR c++/84314 - ICE with templates and fastcall attribute.
27760         * attribs.c (build_type_attribute_qual_variant): Don't clobber
27761         TYPE_CANONICAL on an existing type.
27763 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
27765         PR tree-optimization/84383
27766         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
27767         dstoff nor call operand_equal_p if dstbase is NULL.
27769         PR tree-optimization/84334
27770         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
27771         also a CONSTANT_CLASS_P, punt.
27773 2018-02-14  Jim Wilson  <jimw@sifive.com>
27775         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
27776         first SMALL_OPERAND check.  New local min_second_step.  Move assert
27777         to where locals are set.  Add TARGET_RVC support.
27778         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
27780 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
27782         * doc/invoke.texi: Correct -Wformat-overflow code sample.
27784 2018-02-14  Martin Sebor  <msebor@redhat.com>
27786         PR tree-optimization/83698
27787         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
27788         arrays constrain the offset range to their bounds.
27789         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
27790         (builtin_access::overlap): Avoid setting the size of overlap if it's
27791         already been set.
27792         (maybe_diag_overlap): Also consider arrays when deciding what values
27793         of offsets to include in diagnostics.
27795 2018-02-14  Martin Sebor  <msebor@redhat.com>
27797         PR c/84108
27798         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
27799         that correspond to the kind of a declaration.
27801 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
27803         PR target/83984
27804         * config/pa/pa.md: Load address of PIC label using the linkage table
27805         if the label is nonlocal.
27807 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27809         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
27810         warning message if user requests -maltivec=be.
27811         * doc/invoke.texi: Document deprecation of -maltivec=be.
27813 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
27815         PR target/84220
27816         * config/rs6000/rs6000-c.c: Update definitions for
27817         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
27818         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
27820 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
27822         PR target/84239
27823         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
27824         add _get_ssp intrinsics. Remove argument from
27825         __builtin_ia32_rdssp[d|q].
27826         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
27827         * config/i386/i386-builtin.def: Remove argument from
27828         __builtin_ia32_rdssp[d|q].
27829         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
27830         ix86_expand_special_args_builtin for _rdssp[d|q].
27831         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
27832         Clear register before usage.
27833         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
27834         Add documentation for new _get_ssp and _inc_ssp intrinsics.
27836 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
27838         PR tree-optimization/84357
27839         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
27840         operand 1 of an ARRAY_REF too.
27842 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
27844         PR target/83831
27845         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
27846         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
27847         declarations.
27848         (set_of_reg): New struct.
27849         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
27850         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
27851         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
27852         functions.
27853         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
27854         Split into bitclr, bitset, bitinvert patterns if appropriate.
27855         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
27856         use rx_fuse_in_memory_bitop.
27857         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
27858         to named insn, correct maximum insn length.
27860 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
27862         PR target/79242
27863         * machmode.def: Define a complex mode for PARTIAL_INT.
27864         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
27865         MODE_PARTIAL_INT.
27866         * doc/rtl.texi: Document CSPImode.
27867         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
27868         handling.
27869         (msp430_hard_regno_nregs_with_padding): Likewise.
27871 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
27873         PR target/84279
27874         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
27876 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
27878         PR rtl-optimization/84169
27879         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
27880         we generated a parallel as new i3 and we split that to new i2 and i3
27881         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
27882         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
27883         those to i2, not i1.  Partially rewrite this scan code.
27885 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
27887         PR c/82210
27888         * stor-layout.c (place_field): For variable length fields, adjust
27889         offset_align afterwards not just based on the field's alignment,
27890         but also on the size.
27892         PR middle-end/84309
27893         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
27894         of exps and logs in the use_exp2 case.
27896 2018-02-13  Jeff Law  <law@redhat.com>
27898         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
27899         entry for "vector".
27901         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
27902         ARGS as unused.
27904 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
27906         PR debug/84342
27907         PR debug/84319
27908         * common.opt (gas-loc-support, gas-locview-support): New.
27909         (ginline-points, ginternal-reset-location-views): New.
27910         * doc/invoke.texi: Document them.  Use @itemx where intended.
27911         (gvariable-location-views): Adjust.
27912         * target.def (reset_location_view): New.
27913         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
27914         (TARGET_RESET_LOCATION_VIEW): New.
27915         * doc/tm.texi: Rebuilt.
27916         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
27917         (dwarf2out_default_as_locview_support): New.
27918         (output_asm_line_debug_info): Use option variables.
27919         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
27920         (output_loc_list): Likewise.
27921         (add_high_low_attributes): Check option variables.
27922         Don't output entry view attribute in strict mode.
27923         (gen_inlined_subroutine_die): Check option variables.
27924         (dwarf2out_inline_entry): Likewise.
27925         (init_sections_and_labels): Likewise.
27926         (dwarf2out_early_finish): Likewise.
27927         (maybe_reset_location_view): New, from...
27928         (dwarf2out_var_location): ... here.  Call it.
27929         * debug.h (dwarf2out_default_as_loc_support): Declare.
27930         (dwarf2out_default_as_locview_support): Declare.
27931         * hooks.c (hook_int_rtx_insn_0): New.
27932         * hooks.h (hook_int_rtx_insn_0): Declare.
27933         * toplev.c (process_options): Take -gas-loc-support and
27934         -gas-locview-support from dwarf2out.  Enable
27935         -gvariable-location-views by default only with locview
27936         assembler support.  Enable -ginternal-reset-location-views by
27937         default only if the target defines the corresponding hook.
27938         Enable -ginline-points by default if location views are
27939         enabled; force it disabled if statement frontiers are
27940         disabled.
27941         * tree-inline.c (expand_call_inline): Check option variables.
27942         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
27944 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
27946         PR tree-optimization/84321
27947         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
27948         handling.  Also check whether the anti-range contains any values
27949         that satisfy the mask; switch to a VR_RANGE if not.
27951 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
27953         PR sanitizer/84340
27954         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
27956 2018-02-13  Martin Jambor  <mjambor@suse.cz>
27958         PR c++/83990
27959         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
27960         of call statements, also set location of a load to a temporary.
27962 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
27964         * config/rl78/rl78.c (add_vector_labels): New function.
27965         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
27966         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
27967         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
27968         which checks that no arguments are passed.
27969         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
27970         * doc/extend.texi: Documentation for the new attribute.
27972 2018-02-13  Andreas Schwab  <schwab@suse.de>
27974         * config/riscv/linux.h (CPP_SPEC): Define.
27976 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
27978         PR target/84335
27979         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
27980         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
27981         OPTION_MASK_ISA_AES as first argument to def_builtin_const
27982         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
27983         instead of OPTION_MASK_ISA_PCLMUL as first argument to
27984         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
27985         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
27986         temporarily for AES and PCLMUL builtins.
27988         PR tree-optimization/84339
27989         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
27990         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
27991         Formatting fixes.
27993         PR middle-end/84309
27994         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
27995         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
27996         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
27997         inline function.
27998         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
27999         inline function.
28000         * omp-simd-clone.h: New file.
28001         * omp-simd-clone.c: Include omp-simd-clone.h.
28002         (expand_simd_clones): No longer static.
28003         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
28004         cgraph.h and omp-simd-clone.h.
28005         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
28006         (vect_recog_widen_shift_pattern): Formatting fix.
28007         (vect_pattern_recog_1): Don't check optab for calls.
28009         PR target/84336
28010         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
28011         operands[2] into a REG before using gen_lowpart on it.
28013 2018-02-12  Jeff Law  <law@redhat.com>
28015         PR target/83760
28016         * config/sh/sh.c (find_barrier): Consider a sibling call
28017         a barrier as well.
28019         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
28020         successfully back substituting a reg.
28022 2018-02-12  Richard Biener  <rguenther@suse.de>
28024         PR tree-optimization/84037
28025         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
28026         parameter, move visited init to caller.
28027         (vect_slp_analyze_operations): Separate cost from validity
28028         check, initialize visited once for all instances.
28029         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
28030         for all instances.
28031         * tree-vect-stmts.c (vect_model_simple_cost): Make early
28032         out an assert.
28033         (vect_model_promotion_demotion_cost): Likewise.
28034         (vectorizable_bswap): Guard cost modeling with !slp_node
28035         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
28036         SLP stmts.
28037         (vectorizable_call): Likewise.
28038         (vectorizable_conversion): Likewise.
28039         (vectorizable_assignment): Likewise.
28040         (vectorizable_shift): Likewise.
28041         (vectorizable_operation): Likewise.
28042         (vectorizable_store): Likewise.
28043         (vectorizable_load): Likewise.
28044         (vectorizable_condition): Likewise.
28045         (vectorizable_comparison): Likewise.
28047 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
28049         PR sanitizer/84307
28050         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
28051         (ASAN_MARK): Fix fnspec to account for return value, change pointer
28052         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
28054 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
28056         PR middle-end/83665
28057         * params.def (inline-min-speedup): Increase from 8 to 15.
28058         (max-inline-insns-auto): Decrease from 40 to 30.
28059         * ipa-split.c (consider_split): Add some buffer for function to
28060         be considered inlining candidate.
28061         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
28062         default values.
28064 2018-02-12  Richard Biener  <rguenther@suse.de>
28066         PR tree-optimization/84037
28067         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
28068         matched stmts if we cannot swap the non-matched ones.
28070 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
28072         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
28073         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
28074         _mm_maskz_scalef_round_ss): New intrinsics.
28075         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
28076         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
28077         __builtin_ia32_scalefss_round): Remove.
28078         (__builtin_ia32_scalefsd_mask_round,
28079         __builtin_ia32_scalefss_mask_round): New intrinsics.
28080         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
28081         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
28082         ((match_operand:VF_128 2 "<round_nimm_predicate>"
28083         "<round_constraint>")): Changed to ...
28084         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
28085         "<round_scalar_constraint>")): ... this.
28086         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
28087         %0, %1, %2<round_op3>}"): Changed to ...
28088         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
28089         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
28090         %2<round_scalar_mask_op3>}"): ... this.
28091         * config/i386/subst.md (round_scalar_nimm_predicate): New.
28093 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
28095         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
28096         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
28097         (_mm_maskz_sqrt_round_ss): New intrinsics.
28098         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
28099         (__builtin_ia32_sqrtsd_mask_round)
28100         (__builtin_ia32_sqrtss_mask_round): New builtins.
28101         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
28102         (__builtin_ia32_sqrtss_round): Remove.
28103         (__builtin_ia32_sqrtsd_mask_round)
28104         (__builtin_ia32_sqrtss_mask_round): New builtins.
28105         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
28106         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
28107         ((match_operand:VF_128 1 "vector_operand"
28108         "xBm,<round_constraint>")): Changed to ...
28109         ((match_operand:VF_128 1 "vector_operand"
28110         "xBm,<round_scalar_constraint>")): ... this.
28111         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
28112         %0, %2, %<iptr>1<round_op3>}): Changed to ...
28113         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
28114         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
28115         %<iptr>1<round_scalar_mask_op3>}): ... this.
28116         ((set_attr "prefix" "<round_prefix>")): Changed to ...
28117         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
28119 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
28121         PR target/84266
28122         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
28123         Cast vec_cmpeq result to correct type.
28124         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
28125         Cast vec_cmpgt result to correct type.
28127 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
28129         * final.c (final_scan_insn_1): Renamed from...
28130         (final_scan_insn): ... this.  New wrapper, to recover
28131         seen from the outermost call in recursive ones.
28132         * config/sparc/sparc.c (output_return): Drop seen from call.
28133         (output_sibcall): Likewise.
28134         * config/visium/visium.c (output_branch): Likewise.
28136 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
28138         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
28139         function label.
28141 2018-02-10  Alan Modra  <amodra@gmail.com>
28143         PR target/84300
28144         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
28145         Specify LR as an input.
28147 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
28149         PR sanitizer/83987
28150         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
28151         remove_member_access_dummy_vars): New functions.
28152         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
28153         lower_omp_1, execute_lower_omp): Use them.
28155         PR rtl-optimization/84308
28156         * shrink-wrap.c (spread_components): Release todo vector.
28158 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
28160         PR rtl-optimization/57193
28161         * ira-color.c (struct allocno_color_data): Add member
28162         conflict_allocno_hard_prefs.
28163         (update_conflict_allocno_hard_prefs): New.
28164         (bucket_allocno_compare_func): Add a preference based on
28165         conflict_allocno_hard_prefs.
28166         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
28167         (color_allocnos): Remove a dead code.  Initiate
28168         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
28170 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
28172         PR target/84226
28173         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
28174         constraint from =wa to wa.  Avoid a subreg on the output operand,
28175         instead use a pseudo and subreg it in a move.
28176         (p9_xxbrd_<mode>): Changed to ...
28177         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
28178         (p9_xxbrd_v2df): New expander.
28179         (p9_xxbrw_<mode>): Changed to ...
28180         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
28181         (p9_xxbrw_v4sf): New expander.
28183 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
28185         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
28187 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
28189         PR target/83926
28190         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
28191         multiply in 32-bit mode.
28192         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
28193         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
28194         mode.
28196 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
28198         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
28199         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
28200         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
28201         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
28203 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
28205         PR lto/84213
28206         * dwarf2out.c (is_trivial_indirect_ref): New function.
28207         (dwarf2out_late_global_decl): Do not generate a location
28208         attribute for variables that have a non-trivial DECL_VALUE_EXPR
28209         and that are not defined in the current unit.
28211 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
28213         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
28214         instead of a libcall for UNORDERED.
28216 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
28218         PR target/82641
28219         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
28220         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
28222 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28224         PR target/PR84295
28225         * config/s390/s390.c (s390_set_current_function): Invoke
28226         s390_indirect_branch_settings also if fndecl didn't change.
28228 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
28230         * config/rs6000/rs6000.md (blockage): Set length to zero.
28232 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
28234         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
28236 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
28238         PR sanitizer/84285
28239         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
28240         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
28241         -static-lib*san.
28243         PR debug/84252
28244         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
28245         PARALLEL incoming that failed vt_get_decl_and_offset check.
28247         PR middle-end/84237
28248         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
28249         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
28250         TREE_READONLY bit.
28251         (get_variable_section): For decls in named .bss* sections pass true as
28252         second argument to bss_initializer_p.
28254 2018-02-09  Marek Polacek  <polacek@redhat.com>
28255             Jakub Jelinek  <jakub@redhat.com>
28257         PR c++/83659
28258         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
28259         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
28260         Sync some changes from cxx_fold_indirect_ref.
28262 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
28264         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
28265         markers.
28266         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
28267         (BLOCK_INLINE_ENTRY_LABEL): New.
28268         (dwarf2out_var_location): Disregard inline entry markers.
28269         (inline_entry_data): New struct.
28270         (inline_entry_data_hasher): New hashtable type.
28271         (inline_entry_data_hasher::hash): New.
28272         (inline_entry_data_hasher::equal): New.
28273         (inline_entry_data_table): New variable.
28274         (add_high_low_attributes): Add DW_AT_entry_pc and
28275         DW_AT_GNU_entry_view attributes if a pending entry is found
28276         in inline_entry_data_table.  Add old entry_pc attribute only
28277         if debug nonbinding markers are disabled.
28278         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
28279         markers are enabled.
28280         (block_within_block_p, dwarf2out_inline_entry): New.
28281         (dwarf2out_finish): Check that no entries remained in
28282         inline_entry_data_table.
28283         * final.c (reemit_insn_block_notes): Handle inline entry notes.
28284         (final_scan_insn, notice_source_line): Likewise.
28285         (rest_of_clean_state): Skip inline entry markers.
28286         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
28287         markers.
28288         * gimple.c (gimple_build_debug_inline_entry): New.
28289         * gimple.h (enum gimple_debug_subcode): Add
28290         GIMPLE_DEBUG_INLINE_ENTRY.
28291         (gimple_build_debug_inline_entry): Declare.
28292         (gimple_debug_inline_entry_p): New.
28293         (gimple_debug_nonbind_marker_p): Adjust.
28294         * insn-notes.def (INLINE_ENTRY): New.
28295         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
28296         inline entry marker notes.
28297         (print_insn): Likewise.
28298         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
28299         (INSN_DEBUG_MARKER_KIND): Likewise.
28300         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
28301         * tree-inline.c (expand_call_inline): Build and insert
28302         debug_inline_entry stmt.
28303         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
28304         inline entry blocks early, if nonbind markers are enabled.
28305         (dump_scope_block): Dump fragment info.
28306         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
28307         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
28308         (gimple_build_debug_inline_entry): New.
28309         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
28310         Enable/disable inline entry points too.
28311         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
28312         (DEBUG_INSN): Describe inline entry markers.
28314         * common.opt (gvariable-location-views): New.
28315         (gvariable-location-views=incompat5): New.
28316         * config.in: Rebuilt.
28317         * configure: Rebuilt.
28318         * configure.ac: Test assembler for view support.
28319         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
28320         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
28321         * dwarf2out.c (var_loc_view): New typedef.
28322         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
28323         (dwarf2out_locviews_in_attribute): New.
28324         (dwarf2out_locviews_in_loclist): New.
28325         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
28326         (enum dw_line_info_opcode): Add LI_adv_address.
28327         (struct dw_line_info_table): Add view.
28328         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
28329         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
28330         (zero_view_p): New variable.
28331         (ZERO_VIEW_P): New macro.
28332         (output_asm_line_debug_info): New.
28333         (struct var_loc_node): Add view.
28334         (add_AT_view_list, AT_loc_list): New.
28335         (add_var_loc_to_decl): Add view param.  Test it against last.
28336         (new_loc_list): Add view params.  Record them.
28337         (AT_loc_list_ptr): Handle loc and view lists.
28338         (view_list_to_loc_list_val_node): New.
28339         (print_dw_val): Handle dw_val_class_view_list.
28340         (size_of_die): Likewise.
28341         (value_format): Likewise.
28342         (loc_list_has_views): New.
28343         (gen_llsym): Set vl_symbol too.
28344         (maybe_gen_llsym, skip_loc_list_entry): New.
28345         (dwarf2out_maybe_output_loclist_view_pair): New.
28346         (output_loc_list): Output view list or entries too.
28347         (output_view_list_offset): New.
28348         (output_die): Handle dw_val_class_view_list.
28349         (output_dwarf_version): New.
28350         (output_compilation_unit_header): Use it.
28351         (output_skeleton_debug_sections): Likewise.
28352         (output_rnglists, output_line_info): Likewise.
28353         (output_pubnames, output_aranges): Update version comments.
28354         (output_one_line_info_table): Output view numbers in asm comments.
28355         (dw_loc_list): Determine current endview, pass it to new_loc_list.
28356         Call maybe_gen_llsym.
28357         (loc_list_from_tree_1): Adjust.
28358         (add_AT_location_description): Create view list attribute if
28359         needed, check it's absent otherwise.
28360         (convert_cfa_to_fb_loc_list): Adjust.
28361         (maybe_emit_file): Call output_asm_line_debug_info for test.
28362         (dwarf2out_var_location): Reset views as needed.  Precompute
28363         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
28364         attribute.  Set view.
28365         (new_line_info_table): Reset next view.
28366         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
28367         (dwarf2out_source_line): Likewise.  Output view resets and labels to
28368         the assembler, or select appropriate line info opcodes.
28369         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
28370         (optimize_string_length): Catch it.  Adjust.
28371         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
28372         dw_val_class_view_list, and remove it if no longer needed.
28373         (hash_loc_list): Hash view numbers.
28374         (loc_list_hasher::equal): Compare them.
28375         (optimize_location_lists): Check whether a view list symbol is
28376         needed, and whether the locview attribute is present, and
28377         whether they match.  Remove the locview attribute if no longer
28378         needed.
28379         (index_location_lists): Call skip_loc_list_entry for test.
28380         (dwarf2out_finish): Call output_asm_line_debug_info for test.
28381         Use output_dwarf_version.
28382         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
28383         (struct dw_val_node): Add val_view_list.
28384         * final.c (SEEN_NEXT_VIEW): New.
28385         (set_next_view_needed): New.
28386         (clear_next_view_needed): New.
28387         (maybe_output_next_view): New.
28388         (final_start_function): Rename to...
28389         (final_start_function_1): ... this.  Take pointer to FIRST,
28390         add SEEN parameter.  Emit param bindings in the initial view.
28391         (final_start_function): Reintroduce SEEN-less interface.
28392         (final): Rename to...
28393         (final_1): ... this.  Take SEEN parameter.  Output final pending
28394         next view at the end.
28395         (final): Reintroduce seen-less interface.
28396         (final_scan_insn): Output pending next view before switching
28397         sections or ending a block.  Mark the next view as needed when
28398         outputting variable locations.  Notify debug backend of section
28399         changes, and of location view changes.
28400         (rest_of_handle_final): Adjust.
28401         * toplev.c (process_options): Autodetect value for debug variable
28402         location views option.  Warn on incompat5 without -gdwarf-5.
28403         * doc/invoke.texi (gvariable-location-views): New.
28404         (gvariable-location-views=incompat5): New.
28405         (gno-variable-location-views): New.
28407 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
28409         PR tree-optimization/84136
28410         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
28411         that the result of find_edge is non-NULL.
28413 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
28415         PR target/83008
28416         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
28417         storing integer register in SImode.  Fix cost of 256 and 512
28418         byte aligned SSE register store.
28420 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
28422         * config/i386/i386.c (ix86_multiplication_cost): Fix
28423         multiplication cost for TARGET_AVX512DQ.
28425 2018-02-08  Marek Polacek  <polacek@redhat.com>
28427         PR tree-optimization/84238
28428         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
28429         get_range_strlen.
28431 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
28433         PR tree-optimization/84265
28434         * tree-vect-stmts.c (vectorizable_store): Don't treat
28435         VMAT_CONTIGUOUS accesses as grouped.
28436         (vectorizable_load): Likewise.
28438 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
28440         PR tree-optimization/81635
28441         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
28442         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
28443         (test_round_for_mask): New functions.
28444         (wide_int_cc_tests): Call test_round_for_mask.
28445         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
28446         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
28447         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
28448         range returned by get_range_info.
28450 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
28452         PR ipa/81360
28453         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
28454         * symtab.c: Include builtins.h
28455         (symtab_node::output_to_lto_symbol_table_p): Move here
28456         from lto-streamer-out.c:output_symbol_p.
28457         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
28458         (output_symbol_p): Move all logic to symtab.c
28459         (produce_symtab): Update.
28461 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28463         * config/s390/s390-opts.h (enum indirect_branch): Define.
28464         * config/s390/s390-protos.h (s390_return_addr_from_memory)
28465         (s390_indirect_branch_via_thunk)
28466         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
28467         (enum s390_indirect_branch_type): Define.
28468         * config/s390/s390.c (struct s390_frame_layout, struct
28469         machine_function): Remove.
28470         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
28471         (indirect_branch_table_label_no, indirect_branch_table_name):
28472         Define variables.
28473         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
28474         (enum s390_indirect_branch_option): Define.
28475         (s390_return_addr_from_memory): New function.
28476         (s390_handle_string_attribute): New function.
28477         (s390_attribute_table): Add new attribute handler.
28478         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
28479         (s390_indirect_branch_via_thunk): New function.
28480         (s390_indirect_branch_via_inline_thunk): New function.
28481         (s390_function_ok_for_sibcall): When jumping via thunk disallow
28482         sibling call optimization for non z10 compiles.
28483         (s390_emit_call): Force indirect branch target to be a single
28484         register.  Add r1 clobber for non-z10 compiles.
28485         (s390_emit_epilogue): Emit return jump via return_use expander.
28486         (s390_reorg): Handle JUMP_INSNs as execute targets.
28487         (s390_option_override_internal): Perform validity checks for the
28488         new command line options.
28489         (s390_indirect_branch_attrvalue): New function.
28490         (s390_indirect_branch_settings): New function.
28491         (s390_set_current_function): Invoke s390_indirect_branch_settings.
28492         (s390_output_indirect_thunk_function):  New function.
28493         (s390_code_end): Implement target hook.
28494         (s390_case_values_threshold): Implement target hook.
28495         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
28496         macros.
28497         * config/s390/s390.h (struct s390_frame_layout)
28498         (struct machine_function): Move here from s390.c.
28499         (TARGET_INDIRECT_BRANCH_NOBP_RET)
28500         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
28501         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
28502         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
28503         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
28504         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
28505         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
28506         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
28507         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
28508         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
28509         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
28510         (mnemonic attribute): Add values which aren't recognized
28511         automatically.
28512         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
28513         pattern for branch conversion.  Fix mnemonic attribute.
28514         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
28515         indirect branch via thunk if requested.
28516         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
28517         ("*indirect_jump"): Disable for branch conversion using out of
28518         line thunks.
28519         ("indirect_jump_via_thunk<mode>_z10")
28520         ("indirect_jump_via_thunk<mode>")
28521         ("indirect_jump_via_inlinethunk<mode>_z10")
28522         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
28523         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
28524         ("casesi_jump_via_inlinethunk<mode>_z10")
28525         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
28526         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
28527         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
28528         ("*indirect2_jump"): Disable for branch conversion.
28529         ("casesi_jump"): Turn into expander and expand patterns for branch
28530         conversion.
28531         ("return_use"): New expander.
28532         ("*return"): Emit return via thunk and rename it to ...
28533         ("*return<mode>"): ... this one.
28534         * config/s390/s390.opt: Add new options and and enum for the
28535         option values.
28537 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
28539         * lra-constraints.c (match_reload): Unconditionally use
28540         gen_lowpart_SUBREG, rather than selecting between that
28541         and equivalent gen_rtx_SUBREG code.
28543 2018-02-08  Richard Biener  <rguenther@suse.de>
28545         PR tree-optimization/84233
28546         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
28547         changed flag instead of boguously re-using phi_inserted.
28549 2018-02-08  Martin Jambor  <mjambor@suse.cz>
28551         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
28552         static local variables.
28554 2018-02-08  Richard Biener  <rguenther@suse.de>
28556         PR tree-optimization/84278
28557         * tree-vect-stmts.c (vectorizable_store): When looking for
28558         smaller vector types to perform grouped strided loads/stores
28559         make sure the mode is supported by the target.
28560         (vectorizable_load): Likewise.
28562 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
28564         * config/aarch64/aarch64.c (aarch64_components_for_bb):
28565         Increase LDP/STP opportunities by adding adjacent callee-saves.
28567 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
28569         PR rtl-optimization/84068
28570         PR rtl-optimization/83459
28571         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
28573 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
28575         PR tree-optimization/84224
28576         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
28577         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
28578         non-zero arguments.
28580 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
28582         PR target/84113
28583         * config/rs6000/altivec.md (*restore_world): Remove LR use.
28584         * config/rs6000/predicates.md (restore_world_operation): Adjust op
28585         count, remove one USE.
28587 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
28589         * doc/install.texi (Configuration): Document the
28590         --with-long-double-format={ibm,ieee} PowerPC configuration
28591         options.
28593         PR target/84154
28594         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
28595         Convert from define_expand to be define_insn_and_split.  Rework
28596         float/double/_Float128 conversions to QI/HI/SImode to work with
28597         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
28598         conversions to QI/HImode types did a store and then a load to
28599         truncate the value.  For conversions to VSX registers, don't split
28600         the insn, instead emit the code directly.  Use the code iterator
28601         any_fix to combine signed and unsigned conversions.
28602         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
28603         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
28604         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
28605         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
28606         (fix_<mode>di2_hw): Likewise.
28607         (fixuns_<mode>di2_hw): Likewise.
28608         (fix_<mode>si2_hw): Likewise.
28609         (fixuns_<mode>si2_hw): Likewise.
28610         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
28611         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
28612         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
28613         fix<uns>_trunc<SFDF:mode>si2_p8.
28614         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
28615         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
28616         (fix<uns>_<mode>_mem): Likewise.
28617         (fctiw<u>z_<mode>_mem): Likewise.
28618         (fix<uns>_<mode>_mem): Likewise.
28619         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
28620         the register allocator from doing a direct move to the GPRs to do
28621         a store, and instead use the ISA 3.0 store byte/half-word from
28622         vector register instruction.  For IEEE 128-bit floating point,
28623         also optimize stores of 32-bit ints.
28624         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
28626 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
28628         * genextract.c (push_pathstr_operand): New function to support
28629         [a-zA-Z].
28630         (walk_rtx): Call push_pathstr_operand.
28631         (print_path): Support [a-zA-Z].
28633 2018-02-07  Richard Biener  <rguenther@suse.de>
28635         PR tree-optimization/84037
28636         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
28637         (cse_and_gimplify_to_preheader): Declare.
28638         (vect_get_place_in_interleaving_chain): Likewise.
28639         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28640         ivexpr_map.
28641         (_loop_vec_info::~_loop_vec_info): Delete it.
28642         (cse_and_gimplify_to_preheader): New function.
28643         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
28644         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
28645         (vectorizable_load): Likewise.  For grouped stores always base
28646         the IV on the first element.
28647         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
28648         condition before gimplifying.
28650 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
28652         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
28653         *DIV_EXPR and *MOD_EXPR.
28655 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
28657         PR target/84248
28658         * config/i386/i386.c (ix86_option_override_internal): Mask out
28659         the CF_SET bit when checking -fcf-protection.
28661 2018-02-07  Tom de Vries  <tom@codesourcery.com>
28663         PR libgomp/84217
28664         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
28665         enough.
28667 2018-02-07  Richard Biener  <rguenther@suse.de>
28669         PR tree-optimization/84204
28670         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
28671         this place.
28673         PR tree-optimization/84205
28674         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
28675         special-case isl_ast_op_zdiv_r.
28677         PR tree-optimization/84223
28678         * graphite-scop-detection.c (gather_bbs::before_dom_children):
28679         Only add conditions from within the region.
28680         (gather_bbs::after_dom_children): Adjust.
28682 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
28684         PR target/84209
28685         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
28686         * config/avr/avr.md: Only post-reload split REG-REG moves if
28687         either register is GENERAL_REG_P.
28689 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
28691         PR tree-optimization/84235
28692         * tree-ssa-scopedtables.c
28693         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
28694         if the subtraction is performed in floating point type where NaNs are
28695         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
28696         build 1.  Formatting fix.
28698 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
28700         PR target/84146
28701         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
28702         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
28703         and skip it regardless of bb boundaries.  Use CALL_P macro,
28704         don't test INSN_P (insn) together with CALL_P or JUMP_P check
28705         unnecessarily, formatting fix.
28707 2018-02-06  Michael Collison  <michael.collison@arm.com>
28709         * config/arm/thumb2.md:
28710         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
28711         (*thumb_mov_notscc): Ditto.
28713 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
28715         PR target/84154
28716         * config/rs6000/rs6000.md (su code attribute): Use "u" for
28717         unsigned_fix, not "s".
28719 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28721         * configure.ac (gcc_fn_eh_frame_ro): New function.
28722         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
28723         correct .eh_frame permissions.
28724         * configure: Regenerate.
28726 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
28728         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
28729         irrelevant options.
28731 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28733         * config/rs6000/rs6000.c (rs6000_option_override_internal):
28734         Display warning message for -mno-speculate-indirect-jumps.
28736 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
28738         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
28739         Undocumented.
28740         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
28742 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
28744         PR tree-optimization/84225
28745         * tree-eh.c (find_trapping_overflow): Only call
28746         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
28748 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
28750         PR target/84145
28751         * config/i386/i386.c: Reimplement the check of possible options
28752         -mibt/-mshstk conbination. Change error messages.
28753         * doc/invoke.texi: Fix a typo: remove extra '='.
28755 2018-02-06  Marek Polacek  <polacek@redhat.com>
28757         PR tree-optimization/84228
28758         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
28760 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
28762         PR target/82641
28763         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
28764         emitted arch directives.
28765         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
28766         __ARM_FEATURE_COPROC before changing architectures.
28768 2018-02-06  Richard Biener  <rguenther@suse.de>
28770         * config/i386/i386.c (print_reg): Fix typo.
28771         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
28773 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
28775         * configure: Regenerate.
28777 2018-02-05  Martin Sebor  <msebor@redhat.com>
28779         PR tree-optimization/83369
28780         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
28781         inlining context.
28783 2018-02-05  Martin Liska  <mliska@suse.cz>
28785         * doc/invoke.texi: Cherry-pick upstream r323995.
28787 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
28789         * ira.c (ira_init_register_move_cost): Adjust comment.
28791 2018-02-05  Martin Liska  <mliska@suse.cz>
28793         PR gcov-profile/84137
28794         * doc/gcov.texi: Fix typo in documentation.
28796 2018-02-05  Martin Liska  <mliska@suse.cz>
28798         PR gcov-profile/83879
28799         * doc/gcov.texi: Document necessity of --dynamic-list-data when
28800         using dlopen functionality.
28802 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
28804         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
28805         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
28806         _mm_maskz_range_ss, _mm_mask_range_round_ss,
28807         _mm_maskz_range_round_ss): New intrinsics.
28808         (__builtin_ia32_rangesd128_round)
28809         (__builtin_ia32_rangess128_round): Remove.
28810         (__builtin_ia32_rangesd128_mask_round,
28811         __builtin_ia32_rangess128_mask_round): New builtins.
28812         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
28813         __builtin_ia32_rangess128_round): Remove.
28814         (__builtin_ia32_rangesd128_mask_round,
28815         __builtin_ia32_rangess128_mask_round): New builtins.
28816         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
28817         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
28818         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
28819         "<round_saeonly_constraint>")): Changed to ...
28820         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
28821         "<round_saeonly_scalar_constraint>")): ... this.
28822         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
28823         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
28824         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
28825         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
28826         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
28828 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
28830         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
28831         options.
28832         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
28833         Remove all values except native, 8540 and 8548.
28835 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
28837         * config/i386/i386.c (ix86_output_function_return): Pass
28838         INVALID_REGNUM, instead of -1, as invalid register number to
28839         indirect_thunk_name and output_indirect_thunk.
28841 2018-02-02  Julia Koval  <julia.koval@intel.com>
28843         * config.gcc: Add -march=icelake.
28844         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
28845         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
28846         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
28847         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
28848         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
28849         (processor_target_table): Add icelake.
28850         (ix86_option_override_internal): Handle new PTAs.
28851         (get_builtin_code_for_version): Handle icelake.
28852         (M_INTEL_COREI7_ICELAKE): New.
28853         (fold_builtin_cpu): Handle icelake.
28854         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
28855         * doc/invoke.texi: Add -march=icelake.
28857 2018-02-02  Julia Koval  <julia.koval@intel.com>
28859         * config/i386/i386.c (ix86_option_override_internal): Change flags type
28860         to wide_int_bitmask.
28861         * wide-int-bitmask.h: New.
28863 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
28865         PR target/84066
28866         * config/i386/i386.md: Replace Pmode with word_mode in
28867         builtin_setjmp_setup and builtin_longjmp to support x32.
28869 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
28871         PR target/56010
28872         PR target/83743
28873         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
28874         #include "opts.h".
28875         (rs6000_supported_cpu_names): New static variable.
28876         (linux_cpu_translation_table): Likewise.
28877         (elf_platform) <cpu>: Define new static variable and use it.
28878         Translate kernel AT_PLATFORM name to canonical name if needed.
28879         Error if platform name is unknown.
28881 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
28883         PR target/84089
28884         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
28886 2018-02-01  Jeff Law  <law@redhat.com>
28888         PR target/84128
28889         * config/i386/i386.c (release_scratch_register_on_entry): Add new
28890         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
28891         the scratch if RELEASE_VIA_POP is false.
28892         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
28893         If we have to save a temporary register, decrement SIZE appropriately.
28894         Pass new arguments to release_scratch_register_on_entry.
28895         (ix86_adjust_stack_and_probe): Likewise.
28896         (ix86_emit_probe_stack_range): Pass new arguments to
28897         release_scratch_register_on_entry.
28899 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
28901         PR rtl-optimization/84157
28902         * combine.c (change_zero_ext): Use REG_P predicate in
28903         front of HARD_REGISTER_P predicate.
28905 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
28907         * config/avr/avr.c (avr_option_override): Move disabling of
28908         -fdelete-null-pointer-checks to...
28909         * common/config/avr/avr-common.c (avr_option_optimization_table):
28910         ...here.
28912 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28914         PR tree-optimization/81635
28915         * tree-data-ref.c (split_constant_offset_1): For types that
28916         wrap on overflow, try to use range info to prove that wrapping
28917         cannot occur.
28919 2018-02-01  Renlin Li  <renlin.li@arm.com>
28921         PR target/83370
28922         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
28923         TAILCALL_ADDR_REGS.
28924         (aarch64_register_move_cost): Likewise.
28925         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
28926         TAILCALL_ADDR_REGS.
28927         (REG_CLASS_NAMES): Likewise.
28928         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
28929         TAILCALL_ADDR_REGS. Remove IP registers.
28930         * config/aarch64/aarch64.md (Ucs): Update register constraint.
28932 2018-02-01  Richard Biener  <rguenther@suse.de>
28934         * domwalk.h (dom_walker::dom_walker): Add additional constructor
28935         for specifying RPO order and allow NULL for that.
28936         * domwalk.c (dom_walker::dom_walker): Likewise.
28937         (dom_walker::walk): Handle NULL RPO order.
28938         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
28939         in RPO order.
28940         (rewrite_update_dom_walker): Likewise.
28941         (mark_def_dom_walker): Likewise.
28943 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28945         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
28946         (aarch64_maybe_expand_sve_subreg_move): Declare.
28947         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
28948         * config/aarch64/predicates.md (aarch64_any_register_operand): New
28949         predicate.
28950         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
28951         that are semantically a reverse operation.
28952         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
28953         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
28954         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
28955         functions.
28956         (aarch64_can_change_mode_class): For big-endian, forbid changes
28957         between two SVE modes if they have different element sizes.
28959 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28961         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
28962         the TImode handling for big-endian targets.
28964 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28966         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
28967         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
28968         not just bytes.
28969         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
28970         Remove BSWAP handing for big-endian targets and use the form of
28971         LD1RQ appropariate for the mode.
28973 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28975         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
28976         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
28977         duplicated element.
28979 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28981         PR tearget/83845
28982         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
28983         check for operands that need to go through aarch64_sve_reload_be.
28985 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
28987         PR tree-optimization/81661
28988         PR tree-optimization/84117
28989         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
28990         * tree-eh.c: Include gimplify.h.
28991         (find_trapping_overflow, replace_trapping_overflow,
28992         rewrite_to_non_trapping_overflow): New functions.
28993         * tree-vect-loop.c: Include tree-eh.h.
28994         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
28995         * tree-data-ref.c: Include tree-eh.h.
28996         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
28998 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
29000         PR rtl-optimization/84123
29001         * combine.c (change_zero_ext): Check if hard register satisfies
29002         can_change_dest_mode before calling gen_lowpart_SUBREG.
29004 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
29006         PR target/82444
29007         * ira.c (ira_init_register_move_cost): Remove assert.
29009 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29011         PR rtl-optimization/84071
29012         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
29013         * doc/tm.texi: Regenerate.
29015 2018-01-31  Richard Biener  <rguenther@suse.de>
29017         PR tree-optimization/84132
29018         * tree-data-ref.c (analyze_miv_subscript): Properly
29019         check whether evolution_function_is_affine_multivariate_p
29020         before calling gcd_of_steps_may_divide_p.
29022 2018-01-31  Julia Koval  <julia.koval@intel.com>
29024         PR target/83618
29025         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
29026         * config/i386/i386.md (rdpid_rex64) New.
29027         (rdpid): Make 32bit only.
29029 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
29031         PR lto/84105
29032         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
29033         an IDENTIFIER_NODE for FUNCTION_TYPE's.
29035 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29037         Revert
29038         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29040         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29042 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29044         PR rtl-optimization/84071
29045         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
29046         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
29048 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
29050         * config/arc/arc.c (arc_handle_aux_attribute): New function.
29051         (arc_attribute_table): Add 'aux' attribute.
29052         (arc_in_small_data_p): Consider aux like variables.
29053         (arc_is_aux_reg_p): New function.
29054         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
29055         (arc_get_aux_arg): New function.
29056         (prepare_move_operands): Handle aux-register access.
29057         (arc_handle_aux_attribute): New function.
29058         * doc/extend.texi (ARC Variable attributes): Add subsection.
29060 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
29062         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
29063         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
29064         (arc_attribute_table): Add 'uncached' attribute.
29065         (arc_print_operand): Print '.di' flag for uncached memory
29066         accesses.
29067         (arc_in_small_data_p): Do not consider for small data the uncached
29068         types.
29069         (arc_is_uncached_mem_p): New function.
29070         * config/arc/predicates.md (compact_store_memory_operand): Check
29071         for uncached memory accesses.
29072         (nonvol_nonimm_operand): Likewise.
29073         * doc/extend.texi (ARC Type Attribute): New subsection.
29075 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
29077         PR c/84100
29078         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
29079         falign-loops=): Add Optimization flag.
29081 2018-01-30  Jeff Law  <law@redhat.com>
29083         PR target/84064
29084         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
29085         INT_REGISTERS_SAVED.  Check it prior to calling
29086         get_scratch_register_on_entry.
29087         (ix86_adjust_stack_and_probe): Similarly.
29088         (ix86_emit_probe_stack_range): Similarly.
29089         (ix86_expand_prologue): Corresponding changes.
29091 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29093         PR target/40411
29094         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
29095         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
29097 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
29099         PR target/84112
29100         * lra-constraints.c (curr_insn_transform): Process AND in the
29101         address.
29103 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
29105         PR rtl-optimization/83986
29106         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
29107         dependence against last_pending_memory_flush in addition to
29108         pending_jump_insns.
29110 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
29112         PR tree-optimization/81611
29113         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
29114         copies.
29116 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29118         PR target/83758
29119         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
29120         a reg rtx.
29122 2018-01-30  Richard Biener  <rguenther@suse.de>
29123             Jakub Jelinek  <jakub@redhat.com>
29125         PR tree-optimization/84111
29126         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
29127         inner loops added during recursion, as they don't have up-to-date
29128         SSA form.
29130 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
29132         PR ipa/81360
29133         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
29134         (can_inline_edge_by_limits_p): ... here.
29135         (can_early_inline_edge_p, check_callers,
29136         update_caller_keys, update_callee_keys, recursive_inlining,
29137         add_new_edges_to_heap, speculation_useful_p,
29138         inline_small_functions,
29139         inline_small_functions, flatten_function,
29140         inline_to_all_callers_1): Update.
29142 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
29144         * profile-count.c (profile_count::combine_with_ipa_count): Handle
29145         zeros correctly.
29147 2018-01-30  Richard Biener  <rguenther@suse.de>
29149         PR tree-optimization/83008
29150         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
29151         invariant and constant vector uses in stmts when they need
29152         more than one stmt.
29154 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29156         PR bootstrap/84017
29157         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
29158         * configure: Regenerate.
29160 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29162         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
29163         pattern.
29164         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
29165         Use gen_rtx_REG rather than gen_lowpart.
29167 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29169         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
29170         rather than 0 when creating partial subregs.
29172 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29174         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
29175         of usage.
29177 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
29179         PR target/81550
29180         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
29181         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
29182         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
29183         flags.  This restores the settings used before the 2017-07-24.
29184         Turning off pre increment/decrement/modify allows IVOPTS to
29185         optimize DF/SF loops where the index is an int.
29187 2018-01-29  Richard Biener  <rguenther@suse.de>
29188             Kelvin Nilsen  <kelvin@gcc.gnu.org>
29190         PR bootstrap/80867
29191         * tree-vect-stmts.c (vectorizable_call): Don't call
29192         targetm.vectorize_builtin_md_vectorized_function if callee is
29193         NULL.
29195 2018-01-22  Carl Love  <cel@us.ibm.com>
29197         * doc/extend.tex: Fix typo in second arg in
29198         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
29200 2018-01-29  Richard Biener  <rguenther@suse.de>
29202         PR tree-optimization/84086
29203         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
29204         (flush_ssaname_freelist): When SSA names were released reset
29205         the SCEV hash table.
29207 2018-01-29  Richard Biener  <rguenther@suse.de>
29209         PR tree-optimization/84057
29210         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
29211         removed paths when removing edges.
29213 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
29215         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
29216         -mfunction-return=@var{choice}.
29218 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29220         PR diagnostic/84034
29221         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
29222         Handle CR like TAB.
29223         (layout::print_source_line): Likewise.
29224         (test_get_line_width_without_trailing_whitespace): Add test cases.
29226 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
29228         PR middle-end/84040
29229         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
29230         debug insns.
29232 2018-01-26  Jim Wilson  <jimw@sifive.com>
29234         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
29236         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
29237         specified.
29239 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29241         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
29242         and CMP + SUB-immediate -> SUBS.
29244 2018-01-26  Martin Sebor  <msebor@redhat.com>
29246         PR tree-optimization/83896
29247         * tree-ssa-strlen.c (get_string_len): Rename...
29248         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
29249         Avoid assuming length is constant.
29250         (handle_char_store): Use HOST_WIDE_INT for string length.
29252 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
29254         PR target/81763
29255         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
29256         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
29258 2018-01-26  Richard Biener  <rguenther@suse.de>
29260         PR rtl-optimization/84003
29261         * dse.c (record_store): Only record redundant stores when
29262         the earlier store aliases at least all accesses the later one does.
29264 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
29266         PR rtl-optimization/83985
29267         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
29268         REG_CFA_RESTORE insns.
29269         (delete_unmarked_insns): Don't ignore separate shrink wrapping
29270         REG_CFA_RESTORE insns here.
29272         PR c/83989
29273         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
29274         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
29276 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29278         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
29279         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
29280         (arc_init): Likewise.
29281         (arc_override_options): Likewise.
29282         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
29283         value.
29284         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
29285         support.
29286         * config/arc/arc.h (TARGET_DBNZ): Define.
29287         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
29288         properly set the tune attribute.
29289         (dbnz): Use TARGET_DBNZ guard.
29290         * config/arc/arc.opt (mtune): Add core3 option.
29292 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29294         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
29295         recognize new pic like addresses.
29296         (arc_delegitimize_address): Clean up.
29298 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29300         * config/arc/arc-arches.def: Option mrf16 valid for all
29301         architectures.
29302         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
29303         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
29304         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
29305         * config/arc/arc-tables.opt: Regenerate.
29306         * config/arc/arc.c (arc_conditional_register_usage): Handle
29307         reduced register file case.
29308         (arc_file_start): Set must have build attributes.
29309         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
29310         mrf16 option value.
29311         * config/arc/arc.opt (mrf16): Add new option.
29312         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
29313         * config/arc/genmultilib.awk: Handle new mrf16 option.
29314         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
29315         * config/arc/t-multilib: Regenerate.
29316         * doc/invoke.texi (ARC Options): Document mrf16 option.
29318 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29320         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
29321         * config/arc/arc.c (arc_handle_secure_attribute): New function.
29322         (arc_attribute_table): Add 'secure_call' attribute.
29323         (arc_print_operand): Print secure call operand.
29324         (arc_function_ok_for_sibcall): Don't optimize tail calls when
29325         secure.
29326         (arc_is_secure_call_p): New function.  * config/arc/arc.md
29327         (call_i): Add support for sjli instruction.
29328         (call_value_i): Likewise.
29329         * config/arc/constraints.md (Csc): New constraint.
29331 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29332             John Eric Martin  <John.Martin@emmicro-us.com>
29334         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
29335         * config/arc/arc.c (_arc_jli_section): New struct.
29336         (arc_jli_section): New type.
29337         (rc_jli_sections): New static variable.
29338         (arc_handle_jli_attribute): New function.
29339         (arc_attribute_table): Add jli_always and jli_fixed attribute.
29340         (arc_file_end): New function.
29341         (TARGET_ASM_FILE_END): Define.
29342         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
29343         (arc_add_jli_section): New function.
29344         (jli_call_scan): Likewise.
29345         (arc_reorg): Call jli_call_scan.
29346         (arc_output_addsi): Remove 'S' from printing asm operand.
29347         (arc_is_jli_call_p): New function.
29348         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
29349         operand.
29350         (movhi_insn): Likewise.
29351         (movsi_insn): Likewise.
29352         (movsi_set_cc_insn): Likewise.
29353         (loadqi_update): Likewise.
29354         (load_zeroextendqisi_update): Likewise.
29355         (load_signextendqisi_update): Likewise.
29356         (loadhi_update): Likewise.
29357         (load_zeroextendhisi_update): Likewise.
29358         (load_signextendhisi_update): Likewise.
29359         (loadsi_update): Likewise.
29360         (loadsf_update): Likewise.
29361         (movsicc_insn): Likewise.
29362         (bset_insn): Likewise.
29363         (bxor_insn): Likewise.
29364         (bclr_insn): Likewise.
29365         (bmsk_insn): Likewise.
29366         (bicsi3_insn): Likewise.
29367         (cmpsi_cc_c_insn): Likewise.
29368         (movsi_ne): Likewise.
29369         (movsi_cond_exec): Likewise.
29370         (clrsbsi2): Likewise.
29371         (norm_f): Likewise.
29372         (normw): Likewise.
29373         (swap): Likewise.
29374         (divaw): Likewise.
29375         (flag): Likewise.
29376         (sr): Likewise.
29377         (kflag): Likewise.
29378         (ffs): Likewise.
29379         (ffs_f): Likewise.
29380         (fls): Likewise.
29381         (call_i): Remove 'S' asm letter, add jli instruction.
29382         (call_value_i): Likewise.
29383         * config/arc/arc.op (mjli-always): New option.
29384         * config/arc/constraints.md (Cji): New constraint.
29385         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
29386         operand.
29387         (subsf3_fpx): Likewise.
29388         (mulsf3_fpx): Likewise.
29389         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
29390         asm operand.
29391         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
29392         function attrbutes.
29393         * doc/invoke.texi (ARC): Document mjli-always option.
29395 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
29397         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
29398         avoid addition with 0 and use incw and decw where possible.
29400 2018-01-26  Richard Biener  <rguenther@suse.de>
29402         PR tree-optimization/81082
29403         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
29404         association if it requires casting to unsigned.
29405         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
29406         from fold_plusminus_mult_expr to catch important cases late when
29407         range info is available.
29409 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29411         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
29412         * configure.ac (hidden_linkonce): New test.
29413         * configure: Regenerate.
29414         * config.in: Regenerate.
29416 2018-01-26  Julia Koval  <julia.koval@intel.com>
29418         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
29419         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
29420         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
29421         _mm_mask_bitshuffle_epi64_mask): Fix type.
29422         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
29423         USI_FTYPE_V4DI_V4DI_USI): Remove.
29424         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
29425         __builtin_ia32_vpshufbitqmb256_mask,
29426         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
29427         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
29428         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
29430 2018-01-26  Alan Modra  <amodra@gmail.com>
29432         PR target/84033
29433         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
29434         UNSPEC_VBPERMQ.  Sort other unspecs.
29436 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
29438         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
29440 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
29442         PR middle-end/83055
29443         * predict.c (drop_profile): Do not push/pop cfun; update also
29444         node->count.
29445         (handle_missing_profiles): Fix logic looking for zero profiles.
29447 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
29449         PR middle-end/83977
29450         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
29451         on functions with #pragma omp declare simd or functions with simd
29452         attribute.
29453         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
29454         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
29455         Remove trailing \n from warning_at calls.
29457 2018-01-25  Tom de Vries  <tom@codesourcery.com>
29459         PR target/84028
29460         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
29461         for neutered workers.
29463 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
29465         PR target/68467
29466         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
29467         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
29469 2018-01-24  Jeff Law  <law@redhat.com>
29471         PR target/83994
29472         * i386.c (get_probe_interval): Move to earlier point.
29473         (ix86_compute_frame_layout): If -fstack-clash-protection and
29474         the frame is larger than the probe interval, then use pushes
29475         to save registers rather than reg->mem moves.
29476         (ix86_expand_prologue): Remove conditional for int_registers_saved
29477         assertion.
29479 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
29481         PR target/84014
29482         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
29483         min/max for never referenced object.
29485 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
29487         PR middle-end/83977
29488         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
29489         here.
29490         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
29491         attributes from DECL_ATTRIBUTES (decl) without affecting
29492         DECL_ATTRIBUTES (current_function_decl).
29493         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
29494         functions with non-NULL DECL_ABSTRACT_ORIGIN.
29496 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
29498         PR tree-optimization/83979
29499         * fold-const.c (fold_comparison): Use constant_boolean_node
29500         instead of boolean_{true,false}_node.
29502 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
29504         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
29505         with zero counts.
29507 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29509         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
29510         Simplify the clause that sets the length attribute.
29511         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
29512         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
29513         clause that sets the length attribute.
29514         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
29516 2018-01-24  Tom de Vries  <tom@codesourcery.com>
29518         PR target/83589
29519         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
29520         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
29521         Add strict parameter.
29522         (prevent_branch_around_nothing): Insert dummy insn between branch to
29523         label and label with no ptx insn inbetween.
29524         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
29526 2018-01-24  Tom de Vries  <tom@codesourcery.com>
29528         PR target/81352
29529         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
29530         for neutered threads in warp.
29531         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
29533 2018-01-24  Richard Biener  <rguenther@suse.de>
29535         PR tree-optimization/83176
29536         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
29537         operands.
29539 2018-01-24  Richard Biener  <rguenther@suse.de>
29541         PR tree-optimization/82819
29542         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
29543         code generating pluses that are no-ops in the target precision.
29545 2018-01-24  Richard Biener  <rguenther@suse.de>
29547         PR middle-end/84000
29548         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
29550 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
29552         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
29553         to merge probabilities.
29554         * predict.c (probably_never_executed): Also mark as cold functions
29555         with global 0 profile and guessed local profile.
29556         * profile-count.c (profile_probability::combine_with_count): New
29557         member function.
29558         * profile-count.h (profile_probability::operator*,
29559         profile_probability::operator*=, profile_probability::operator/,
29560         profile_probability::operator/=): Reduce precision to adjusted
29561         and set value to guessed on contradictory divisions.
29562         (profile_probability::combine_with_freq): Remove.
29563         (profile_probability::combine_wiht_count): Declare.
29564         (profile_count::force_nonzero):: Set to adjusted.
29565         (profile_count::probability_in):: Set quality to adjusted.
29566         * tree-ssa-tail-merge.c (replace_block_by): Use
29567         combine_with_count.
29569 2018-01-23  Andrew Waterman  <andrew@sifive.com>
29570             Jim Wilson  <jimw@sifive.com>
29572         * config/riscv/riscv.c (riscv_stack_boundary): New.
29573         (riscv_option_override): Set riscv_stack_boundary.  Handle
29574         riscv_preferred_stack_boundary_arg.
29575         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
29576         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
29577         (STACK_BOUNDARY): Set to riscv_stack_boundary.
29578         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
29579         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
29580         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
29582 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
29584         PR target/83905
29585         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
29586         of struct ix86_frame.
29587         (ix86_expand_epilogue): Likewise.  Add a local variable for
29588         the reg_save_offset field in struct ix86_frame.
29590 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
29592         PR tree-optimization/82604
29593         * tree-loop-distribution.c (enum partition_kind): New enum item
29594         PKIND_PARTIAL_MEMSET.
29595         (partition_builtin_p): Support above new enum item.
29596         (generate_code_for_partition): Ditto.
29597         (compute_access_range): Differentiate cases that equality can be
29598         proven at all loops, the innermost loops or no loops.
29599         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
29600         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
29601         (finalize_partitions, distribute_loop): Don't fuse partition of
29602         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
29603         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
29604         parloop is enabled.
29606 2018-01-23  Martin Liska  <mliska@suse.cz>
29608         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
29609         order to ignore the predictor.
29610         (PRED_POLYMORPHIC_CALL): Likewise.
29611         (PRED_RECURSIVE_CALL): Likewise.
29613 2018-01-23  Martin Liska  <mliska@suse.cz>
29615         * tree-profile.c (tree_profiling): Print function header to
29616         aware reader which function we are working on.
29617         * value-prof.c (gimple_find_values_to_profile): Do not print
29618         not interesting value histograms.
29620 2018-01-23  Martin Liska  <mliska@suse.cz>
29622         * profile-count.h (enum profile_quality): Add
29623         profile_uninitialized as the first value. Do not number values
29624         as they are zero based.
29625         (profile_count::verify): Update sanity check.
29626         (profile_probability::verify): Likewise.
29628 2018-01-23  Nathan Sidwell  <nathan@acm.org>
29630         * doc/invoke.texi (ffor-scope): Deprecate.
29632 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
29634         PR tree-optimization/83510
29635         * domwalk.c (set_all_edges_as_executable): New function.
29636         (dom_walker::dom_walker): Convert bool param
29637         "skip_unreachable_blocks" to enum reachability.  Move setup of
29638         edge flags to set_all_edges_as_executable and only do it when
29639         reachability is REACHABLE_BLOCKS.
29640         * domwalk.h (enum dom_walker::reachability): New enum.
29641         (dom_walker::dom_walker): Convert bool param
29642         "skip_unreachable_blocks" to enum reachability.
29643         (set_all_edges_as_executable): New decl.
29644         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
29645         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
29646         "reachability".
29647         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
29648         but converting true to REACHABLE_BLOCKS.
29649         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
29650         * tree-vrp.c
29651         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
29652         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
29653         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
29654         REACHABLE_BLOCKS.
29655         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
29656         if check_all_array_refs will be called.
29658 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
29660         * tree.c (selftest::test_location_wrappers): Add more test
29661         coverage.
29663 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
29665         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
29666         (selftest::test_bit_in_range): Likewise.
29668 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
29670         PR testsuite/83888
29671         * doc/sourcebuild.texi (vect_float): Say that the selector
29672         only describes the situation when -funsafe-math-optimizations is on.
29673         (vect_float_strict): Document.
29675 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
29677         PR tree-optimization/83965
29678         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
29679         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
29680         instead of checking only for a reduction.
29681         (vect_recog_widen_sum_pattern): Likewise.
29683 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
29685         * predict.c (probably_never_executed): Only use precise profile info.
29686         (compute_function_frequency): Skip after inlining hack since we now
29687         have quality checking.
29689 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
29691         * profile-count.h (profile_probability::very_unlikely,
29692         profile_probability::unlikely, profile_probability::even): Set
29693         precision to guessed.
29695 2018-01-23  Richard Biener  <rguenther@suse.de>
29697         PR tree-optimization/83963
29698         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
29699         Properly terminate dominator walk when crossing the exit edge not
29700         when visiting its source block.
29702 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
29704         PR c++/83918
29705         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
29706         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
29708 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
29710         PR tree-optimization/83957
29711         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
29712         semicolon after for body surrounded by braces.
29714         PR tree-optimization/83081
29715         * profile-count.h (profile_probability::split): New method.
29716         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
29717         Use profile_probability::split.
29718         (do_compare_rtx_and_jump): Fix adjustment of probabilities
29719         when splitting a single conditional jump into 2.
29721 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
29723         PR tree-optimization/69452
29724         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
29725         decl.
29727 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29729         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
29730         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
29731         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
29733 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29735         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
29736         declaration.
29737         * config/rl78/rl78.md (movdi): New define_expand.
29738         * config/rl78/rl78.c (rl78_split_movdi): New function.
29740 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
29742         PR target/83862
29743         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
29744         no longer used.
29745         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
29746         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
29747         128-bit to produce an UNSPEC move to get the double word with the
29748         signbit and then a shift directly to do signbit.
29749         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
29750         implementation with a new version that just does either a direct
29751         move or a regular move.  Move memory interface to separate insns.
29752         Move insns so they are next to the expander.
29753         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
29754         with signbit move.  Split big and little endian case.
29755         (signbit<mode>2_dm_mem_le): Likewise.
29756         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
29757         (signbit<mode>2_dm2): Likewise.
29759 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29761         * config/rl78/rl78.md (anddi3): New define_expand.
29763 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29765         * config/rl78/rl78.md (umindi3): New define_expand.
29767 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29769         * config/rl78/rl78.md (smindi3): New define_expand.
29771 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29773         * config/rl78/rl78.md (smaxdi3): New define_expand.
29775 2018-01-22  Carl Love  <cel@us.ibm.com>
29777         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
29778         LVX_V1TI): Add macro expansion.
29779         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
29780         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
29781         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
29782         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
29783         Change check to determine if the instruction is a byte reversing
29784         entry.  Fix typo in comment.
29785         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
29786         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
29787         Add def_builtin calls for new builtins.
29788         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
29789         Add define_insn expansion.
29791 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29793         * config/rl78/rl78.md (umaxdi3): New define_expand.
29795 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29797         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
29798         for non-QImode registers.
29800 2018-01-22  Richard Biener  <rguenther@suse.de>
29802         PR tree-optimization/83963
29803         * graphite-scop-detection.c (scop_detection::get_sese): Delay
29804         including the loop exit block.
29805         (scop_detection::merge_sese): Likewise.
29806         (scop_detection::add_scop): Do it here instead.
29808 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29810         * doc/sourcebuild.texi (arm_softfloat): Document.
29812 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
29814         PR gcc/77734
29815         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
29816         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
29817         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
29819 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29820             David Edelsohn  <dje.gcc@gmail.com>
29822         PR target/83946
29823         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
29824         Change "crset eq" to "crset 2".
29825         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
29826         (*call_indirect_aix<mode>_nospec): Likewise.
29827         (*call_value_indirect_aix<mode>_nospec): Likewise.
29828         (*call_indirect_elfv2<mode>_nospec): Likewise.
29829         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
29830         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
29831         change assembly output from . to $.
29832         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
29833         (indirect_jump<mode>_nospec): Change assembly output from . to $.
29834         (*tablejump<mode>_internal1_nospec): Likewise.
29836 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
29838         PR target/80870
29839         * config/sh/sh_optimize_sett_clrt.cc:
29840         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
29842 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
29844         PR tree-optimization/83940
29845         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
29846         offset_dt to vect_constant_def rather than vect_unknown_def_type.
29847         (vect_check_load_store_mask): Add a mask_dt_out parameter and
29848         use it to pass back the definition type.
29849         (vect_check_store_rhs): Likewise rhs_dt_out.
29850         (vect_build_gather_load_calls): Add a mask_dt argument and use
29851         it instead of a call to vect_is_simple_use.
29852         (vectorizable_store): Update calls to vect_check_load_store_mask
29853         and vect_check_store_rhs.  Use the dt returned by the latter instead
29854         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
29855         instead of calls to vect_is_simple_use.  Pass the scalar rather
29856         than the vector operand to vect_is_simple_use when handling
29857         second and subsequent copies of an rhs value.
29858         (vectorizable_load): Update calls to vect_check_load_store_mask
29859         and vect_build_gather_load_calls.  Use the cached mask_dt and
29860         gs_info.offset_dt instead of calls to vect_is_simple_use.
29862 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
29864         PR middle-end/83945
29865         * tree-emutls.c: Include gimplify.h.
29866         (lower_emutls_2): New function.
29867         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
29868         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
29869         it before further processing.
29871         PR target/83930
29872         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
29873         UINTVAL (trueop1) instead of INTVAL (op1).
29875 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
29877         PR debug/81570
29878         PR debug/83728
29879         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
29880         INCOMING_FRAME_SP_OFFSET if not defined.
29881         (scan_trace): Add ENTRY argument.  If true and
29882         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
29883         emit a note to adjust the CFA offset.
29884         (create_cfi_notes): Adjust scan_trace callers.
29885         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
29886         INCOMING_FRAME_SP_OFFSET in the CIE.
29887         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
29888         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
29889         Likewise.
29890         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
29891         * doc/tm.texi: Regenerated.
29893 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29895         PR rtl-optimization/83147
29896         * lra-constraints.c (remove_inheritance_pseudos): Use
29897         lra_substitute_pseudo_within_insn.
29899 2018-01-19  Tom de Vries  <tom@codesourcery.com>
29900             Cesar Philippidis  <cesar@codesourcery.com>
29902         PR target/83920
29903         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
29905 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
29907         PR target/83790
29908         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
29909         spaces for function labels.
29911 2018-01-19  Martin Liska  <mliska@suse.cz>
29913         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
29914         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
29915         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
29916         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
29917         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
29918         (PRED_CONST_RETURN): Change from 69 to 65.
29919         (PRED_NULL_RETURN): Change from 91 to 71.
29920         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
29921         (PRED_LOOP_GUARD): Change from 66 to 73.
29923 2018-01-19  Martin Liska  <mliska@suse.cz>
29925         * predict.c (predict_insn_def): Add new assert.
29926         (struct branch_predictor): Change type to signed integer.
29927         (test_prediction_value_range): Amend test to cover
29928         PROB_UNINITIALIZED.
29929         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
29930         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
29931         (PRED_LOOP_ITERATIONS_MAX): Likewise.
29932         (PRED_LOOP_IV_COMPARE): Likewise.
29933         * predict.h (PROB_UNINITIALIZED): Define new constant.
29935 2018-01-19  Martin Liska  <mliska@suse.cz>
29937         * predict.c (dump_prediction): Add new format for
29938         analyze_brprob.py script which is enabled with -details
29939         suboption.
29940         * profile-count.h (precise_p): New function.
29942 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
29944         PR tree-optimization/83922
29945         * tree-vect-loop.c (vect_verify_full_masking): Return false if
29946         there are no statements that need masking.
29947         (vect_active_double_reduction_p): New function.
29948         (vect_analyze_loop_operations): Use it when handling phis that
29949         are not in the loop header.
29951 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
29953         PR tree-optimization/83914
29954         * tree-vect-loop.c (vectorizable_induction): Don't convert
29955         init_expr or apply the peeling adjustment for inductions
29956         that are nested within the vectorized loop.
29958 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29960         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
29961         instead of NEG.
29963 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
29965         PR sanitizer/81715
29966         PR testsuite/83882
29967         * function.h (gimplify_parameters): Add gimple_seq * argument.
29968         * function.c: Include gimple.h and options.h.
29969         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
29970         for the added local temporaries if needed.
29971         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
29972         if there are any parameter cleanups, wrap whole body into a
29973         try/finally with the cleanups.
29975 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
29977         PR target/82964
29978         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
29979         Use GET_MODE_CLASS for scalar floating point.
29981 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
29983         PR ipa/82256
29984         patch by PaX Team
29985         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
29986         Fix call of call_cgraph_insertion_hooks.
29988 2018-01-18  Martin Sebor  <msebor@redhat.com>
29990         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
29992 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
29994         PR ipa/83619
29995         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
29996         frequencies.
29998 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
30000         PR other/70268
30001         * common.opt: (-ffile-prefix-map): New option.
30002         * opts.c (common_handle_option): Defer it.
30003         * opts-global.c (handle_common_deferred_options): Handle it.
30004         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
30005         * file-prefix-map.h: New file.
30006         (remap_debug_filename, add_debug_prefix_map): ...here.
30007         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
30008         * final.c (debug_prefix_map, add_debug_prefix_map
30009         remap_debug_filename): Move to...
30010         * file-prefix-map.c: New file.
30011         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
30012         generalize, get rid of alloca(), use strrchr() instead of strchr().
30013         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
30014         Implement in terms of add_prefix_map().
30015         (remap_macro_filename, remap_debug_filename): Implement in term of
30016         remap_filename().
30017         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
30018         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
30019         * dbxout.c: Include file-prefix-map.h.
30020         * varasm.c: Likewise.
30021         * vmsdbgout.c: Likewise.
30022         * xcoffout.c: Likewise.
30023         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
30024         * doc/cppopts.texi (-fmacro-prefix-map): Document.
30025         * doc/invoke.texi (-ffile-prefix-map): Document.
30026         (-fdebug-prefix-map): Update description.
30028 2018-01-18  Martin Liska  <mliska@suse.cz>
30030         * config/i386/i386.c (indirect_thunk_name): Document that also
30031         lfence is emitted.
30032         (output_indirect_thunk): Document why both instructions
30033         (pause and lfence) are generated.
30035 2018-01-18  Richard Biener  <rguenther@suse.de>
30037         PR tree-optimization/83887
30038         * graphite-scop-detection.c
30039         (scop_detection::get_nearest_dom_with_single_entry): Remove.
30040         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
30041         (scop_detection::merge_sese): Re-implement with a flood-fill
30042         algorithm that properly finds a SESE region if it exists.
30044 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
30046         PR c/61240
30047         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
30048         pointer_diff optimizations use view_convert instead of convert.
30050 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30052         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30053         Generate different code for -mno-speculate-indirect-jumps.
30054         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30055         (*call_indirect_aix<mode>): Disable for
30056         -mno-speculate-indirect-jumps.
30057         (*call_indirect_aix<mode>_nospec): New define_insn.
30058         (*call_value_indirect_aix<mode>): Disable for
30059         -mno-speculate-indirect-jumps.
30060         (*call_value_indirect_aix<mode>_nospec): New define_insn.
30061         (*sibcall_nonlocal_sysv<mode>): Generate different code for
30062         -mno-speculate-indirect-jumps.
30063         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30065 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
30067         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
30068         long double type, set the flags for noting the default long double
30069         type, even if we don't pass or return a long double type.
30071 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
30073         PR ipa/83051
30074         * ipa-inline.c (flatten_function): Do not overwrite final inlining
30075         failure.
30077 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
30079         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
30080         support for merge[hl].
30081         (fold_mergehl_helper): New helper function.
30082         (tree-vector-builder.h): New #include for tree_vector_builder usage.
30083         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
30084         (altivec_vmrglw_direct): Add xxmrglw insn.
30086 2018-01-17  Andrew Waterman  <andrew@sifive.com>
30088         * config/riscv/riscv.c (riscv_conditional_register_usage): If
30089         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
30091 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
30093         PR lto/83121
30094         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
30095         call the lto_location_cache before reading the
30096         DECL_SOURCE_LOCATION of the types.
30098 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
30099             Richard Sandiford  <richard.sandiford@linaro.org>
30101         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
30102         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
30103         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
30104         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
30105         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
30106         Add declaration.
30107         * config/aarch64/constraints.md (aarch64_movti_operand):
30108         Limit immediates.
30109         * config/aarch64/predicates.md (Uti): Add new constraint.
30111 2018-01-17  Carl Love  <cel@us.ibm.com>
30113         * config/rs6000/vsx.md (define_expand xl_len_r,
30114         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
30115         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
30116         lxvll.
30117         (define_expand, define_insn): Move the shift left from  the
30118         define_insn to the define_expand for lxvl and stxvl instructions.
30119         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
30120         and XL_LEN_R definitions to PURE.
30122 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
30124         * config/i386/i386.c (indirect_thunk_name): Declare regno
30125         as unsigned int.  Compare regno with INVALID_REGNUM.
30126         (output_indirect_thunk): Ditto.
30127         (output_indirect_thunk_function): Ditto.
30128         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
30129         in the call to output_indirect_thunk_function.
30131 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
30133         PR middle-end/83884
30134         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
30135         rather than the size of inner_type to determine the stack slot size
30136         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
30138 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
30140         PR target/83546
30141         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
30142         to PTA_SILVERMONT.
30144 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
30146         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
30147         endian Linux systems to optionally enable multilibs for selecting
30148         the long double type if the user configured an explicit type.
30149         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
30150         have no long double multilibs if not defined.
30151         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
30152         warn if the user used -mabi={ieee,ibm}longdouble and we built
30153         multilibs for long double.
30154         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
30155         appropriate multilib option.
30156         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
30157         multilib options.
30158         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
30159         for building long double multilibs.
30160         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
30162 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
30164         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
30165         copies.
30167         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
30168         64 bits.
30169         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
30170         128 bits.
30172         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
30173         variables.
30175         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
30176         return value.
30178 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
30180         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
30181         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
30183 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
30185         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
30186         different rtl trees depending on TARGET_64BIT.
30187         (rs6000_gen_lvx): Likewise.
30189 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
30191         * config/visium/visium.md (nop): Tweak comment.
30192         (hazard_nop): Likewise.
30194 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30196         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
30197         -mspeculate-indirect-jumps.
30198         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
30199         for -mno-speculate-indirect-jumps.
30200         (*call_indirect_elfv2<mode>_nospec): New define_insn.
30201         (*call_value_indirect_elfv2<mode>): Disable for
30202         -mno-speculate-indirect-jumps.
30203         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
30204         (indirect_jump): Emit different RTL for
30205         -mno-speculate-indirect-jumps.
30206         (*indirect_jump<mode>): Disable for
30207         -mno-speculate-indirect-jumps.
30208         (*indirect_jump<mode>_nospec): New define_insn.
30209         (tablejump): Emit different RTL for
30210         -mno-speculate-indirect-jumps.
30211         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
30212         (tablejumpsi_nospec): New define_expand.
30213         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
30214         (tablejumpdi_nospec): New define_expand.
30215         (*tablejump<mode>_internal1): Disable for
30216         -mno-speculate-indirect-jumps.
30217         (*tablejump<mode>_internal1_nospec): New define_insn.
30218         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
30219         option.
30221 2018-01-16  Artyom Skrobov tyomitch@gmail.com
30223         * caller-save.c (insert_save): Drop unnecessary parameter.  All
30224         callers updated.
30226 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30227             Richard Biener  <rguenth@suse.de>
30229         PR libgomp/83590
30230         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
30231         return early, inline manually is_gimple_sizepos.  Make sure if we
30232         call gimplify_expr we don't end up with a gimple constant.
30233         * tree.c (variably_modified_type_p): Don't return true for
30234         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
30235         * gimplify.h (is_gimple_sizepos): Remove.
30237 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30239         PR tree-optimization/83857
30240         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
30241         vectorizable_live_operation for pure SLP statements.
30242         (vectorizable_live_operation): Handle PHIs.
30244 2018-01-16  Richard Biener  <rguenther@suse.de>
30246         PR tree-optimization/83867
30247         * tree-vect-stmts.c (vect_transform_stmt): Precompute
30248         nested_in_vect_loop_p since the scalar stmt may get invalidated.
30250 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30252         PR c/83844
30253         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
30254         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
30255         If off is not INTEGER_CST, issue a may not be aligned warning
30256         rather than isn't aligned.  Use isn%'t rather than isn't.
30257         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
30258         into MULT_EXPR.
30259         <case MULT_EXPR>: Improve the case when bottom and one of the
30260         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
30261         operand, in that case check if the other operand is multiple of
30262         bottom divided by the INTEGER_CST operand.
30264 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30266         PR target/83858
30267         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
30268         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
30269         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
30270         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
30271         * config/pa/pa.c (pa_function_arg_advance): Likewise.
30272         (pa_function_arg, pa_arg_partial_bytes): Likewise.
30273         (pa_function_arg_size): New function.
30275 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30277         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
30278         in a separate statement.
30280 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30282         PR tree-optimization/83847
30283         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
30284         group gathers and scatters.
30286 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30288         PR rtl-optimization/86620
30289         * params.def (max-sched-ready-insns): Bump minimum value to 1.
30291         PR rtl-optimization/83213
30292         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
30293         to last if both are JUMP_INSNs.
30295         PR tree-optimization/83843
30296         * gimple-ssa-store-merging.c
30297         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
30298         store_immediate_info for bswap/nop orig_stores.
30300 2018-01-15  Andrew Waterman  <andrew@sifive.com>
30302         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
30303         !TARGET_MUL.
30304         <UDIV>: Increase cost if !TARGET_DIV.
30306 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
30308         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
30309         (define_attr "cr_logical_3op"): New.
30310         (cceq_ior_compare): Adjust.
30311         (cceq_ior_compare_complement): Adjust.
30312         (*cceq_rev_compare): Adjust.
30313         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
30314         (is_cracked_insn): Adjust.
30315         (insn_must_be_first_in_group): Adjust.
30316         * config/rs6000/40x.md: Adjust.
30317         * config/rs6000/440.md: Adjust.
30318         * config/rs6000/476.md: Adjust.
30319         * config/rs6000/601.md: Adjust.
30320         * config/rs6000/603.md: Adjust.
30321         * config/rs6000/6xx.md: Adjust.
30322         * config/rs6000/7450.md: Adjust.
30323         * config/rs6000/7xx.md: Adjust.
30324         * config/rs6000/8540.md: Adjust.
30325         * config/rs6000/cell.md: Adjust.
30326         * config/rs6000/e300c2c3.md: Adjust.
30327         * config/rs6000/e500mc.md: Adjust.
30328         * config/rs6000/e500mc64.md: Adjust.
30329         * config/rs6000/e5500.md: Adjust.
30330         * config/rs6000/e6500.md: Adjust.
30331         * config/rs6000/mpc.md: Adjust.
30332         * config/rs6000/power4.md: Adjust.
30333         * config/rs6000/power5.md: Adjust.
30334         * config/rs6000/power6.md: Adjust.
30335         * config/rs6000/power7.md: Adjust.
30336         * config/rs6000/power8.md: Adjust.
30337         * config/rs6000/power9.md: Adjust.
30338         * config/rs6000/rs64.md: Adjust.
30339         * config/rs6000/titan.md: Adjust.
30341 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30343         * config/i386/predicates.md (indirect_branch_operand): Rewrite
30344         ix86_indirect_branch_register logic.
30346 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30348         * config/i386/constraints.md (Bs): Update
30349         ix86_indirect_branch_register check.  Don't check
30350         ix86_indirect_branch_register with GOT_memory_operand.
30351         (Bw): Likewise.
30352         * config/i386/predicates.md (GOT_memory_operand): Don't check
30353         ix86_indirect_branch_register here.
30354         (GOT32_symbol_operand): Likewise.
30356 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30358         * config/i386/predicates.md (constant_call_address_operand):
30359         Rewrite ix86_indirect_branch_register logic.
30360         (sibcall_insn_operand): Likewise.
30362 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30364         * config/i386/constraints.md (Bs): Replace
30365         ix86_indirect_branch_thunk_register with
30366         ix86_indirect_branch_register.
30367         (Bw): Likewise.
30368         * config/i386/i386.md (indirect_jump): Likewise.
30369         (tablejump): Likewise.
30370         (*sibcall_memory): Likewise.
30371         (*sibcall_value_memory): Likewise.
30372         Peepholes of indirect call and jump via memory: Likewise.
30373         * config/i386/i386.opt: Likewise.
30374         * config/i386/predicates.md (indirect_branch_operand): Likewise.
30375         (GOT_memory_operand): Likewise.
30376         (call_insn_operand): Likewise.
30377         (sibcall_insn_operand): Likewise.
30378         (GOT32_symbol_operand): Likewise.
30380 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
30382         PR middle-end/83837
30383         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
30384         type rather than type addr's type points to.
30385         (expand_omp_atomic_mutex): Likewise.
30386         (expand_omp_atomic): Likewise.
30388 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30390         PR target/83839
30391         * config/i386/i386.c (output_indirect_thunk_function): Use
30392         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
30393         for  __x86_return_thunk.
30395 2018-01-15  Richard Biener  <rguenther@suse.de>
30397         PR middle-end/83850
30398         * expmed.c (extract_bit_field_1): Fix typo.
30400 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30402         PR target/83687
30403         * config/arm/iterators.md (VF): New mode iterator.
30404         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
30405         Remove integer-related logic from pattern.
30406         (neon_vabd<mode>_3): Likewise.
30408 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
30410         PR middle-end/82694
30411         * common.opt (fstrict-overflow): No longer an alias.
30412         (fwrapv-pointer): New option.
30413         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
30414         also for pointer types based on flag_wrapv_pointer.
30415         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
30416         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
30417         opts->x_flag_wrapv got set.
30418         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
30419         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
30420         POINTER_TYPE_OVERFLOW_UNDEFINED.
30421         * match.pd: Likewise in address comparison pattern.
30422         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
30424 2018-01-15  Richard Biener  <rguenther@suse.de>
30426         PR lto/83804
30427         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
30428         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
30429         Reset type names to their identifier if their TYPE_DECL doesn't
30430         have linkage (and thus is used for ODR and devirt).
30431         (save_debug_info_for_decl): Remove.
30432         (save_debug_info_for_type): Likewise.
30433         (add_tree_to_fld_list): Adjust.
30434         * tree-pretty-print.c (dump_generic_node): Make dumping of
30435         type names more robust.
30437 2018-01-15  Richard Biener  <rguenther@suse.de>
30439         * BASE-VER: Bump to 8.0.1.
30441 2018-01-14  Martin Sebor  <msebor@redhat.com>
30443         PR other/83508
30444         * builtins.c (check_access): Avoid warning when the no-warning bit
30445         is set.
30447 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
30449         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
30450         * ira-color (allocno_hard_regs_compare): Likewise.
30452 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
30454         PR target/83013
30455         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
30456         Use .pushsection/.popsection.
30458 2018-01-14  Martin Sebor  <msebor@redhat.com>
30460         PR c++/81327
30461         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
30463 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
30465         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
30466         entry from extra_headers.
30467         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
30468         extra_headers, make the list bitwise identical to the i?86-*-* one.
30470 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30472         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
30473         -mcmodel=large with -mindirect-branch=thunk,
30474         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
30475         -mfunction-return=thunk-extern.
30476         * doc/invoke.texi: Document -mcmodel=large is incompatible with
30477         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
30478         -mfunction-return=thunk and -mfunction-return=thunk-extern.
30480 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30482         * config/i386/i386.c (print_reg): Print the name of the full
30483         integer register without '%'.
30484         (ix86_print_operand): Handle 'V'.
30485         * doc/extend.texi: Document 'V' modifier.
30487 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30489         * config/i386/constraints.md (Bs): Disallow memory operand for
30490         -mindirect-branch-register.
30491         (Bw): Likewise.
30492         * config/i386/predicates.md (indirect_branch_operand): Likewise.
30493         (GOT_memory_operand): Likewise.
30494         (call_insn_operand): Likewise.
30495         (sibcall_insn_operand): Likewise.
30496         (GOT32_symbol_operand): Likewise.
30497         * config/i386/i386.md (indirect_jump): Call convert_memory_address
30498         for -mindirect-branch-register.
30499         (tablejump): Likewise.
30500         (*sibcall_memory): Likewise.
30501         (*sibcall_value_memory): Likewise.
30502         Disallow peepholes of indirect call and jump via memory for
30503         -mindirect-branch-register.
30504         (*call_pop): Replace m with Bw.
30505         (*call_value_pop): Likewise.
30506         (*sibcall_pop_memory): Replace m with Bs.
30507         * config/i386/i386.opt (mindirect-branch-register): New option.
30508         * doc/invoke.texi: Document -mindirect-branch-register option.
30510 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30512         * config/i386/i386-protos.h (ix86_output_function_return): New.
30513         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
30514         set function_return_type.
30515         (indirect_thunk_name): Add ret_p to indicate thunk for function
30516         return.
30517         (output_indirect_thunk_function): Pass false to
30518         indirect_thunk_name.
30519         (ix86_output_indirect_branch_via_reg): Likewise.
30520         (ix86_output_indirect_branch_via_push): Likewise.
30521         (output_indirect_thunk_function): Create alias for function
30522         return thunk if regno < 0.
30523         (ix86_output_function_return): New function.
30524         (ix86_handle_fndecl_attribute): Handle function_return.
30525         (ix86_attribute_table): Add function_return.
30526         * config/i386/i386.h (machine_function): Add
30527         function_return_type.
30528         * config/i386/i386.md (simple_return_internal): Use
30529         ix86_output_function_return.
30530         (simple_return_internal_long): Likewise.
30531         * config/i386/i386.opt (mfunction-return=): New option.
30532         (indirect_branch): Mention -mfunction-return=.
30533         * doc/extend.texi: Document function_return function attribute.
30534         * doc/invoke.texi: Document -mfunction-return= option.
30536 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30538         * config/i386/i386-opts.h (indirect_branch): New.
30539         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
30540         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
30541         with local indirect jump when converting indirect call and jump.
30542         (ix86_set_indirect_branch_type): New.
30543         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
30544         (indirectlabelno): New.
30545         (indirect_thunk_needed): Likewise.
30546         (indirect_thunk_bnd_needed): Likewise.
30547         (indirect_thunks_used): Likewise.
30548         (indirect_thunks_bnd_used): Likewise.
30549         (INDIRECT_LABEL): Likewise.
30550         (indirect_thunk_name): Likewise.
30551         (output_indirect_thunk): Likewise.
30552         (output_indirect_thunk_function): Likewise.
30553         (ix86_output_indirect_branch_via_reg): Likewise.
30554         (ix86_output_indirect_branch_via_push): Likewise.
30555         (ix86_output_indirect_branch): Likewise.
30556         (ix86_output_indirect_jmp): Likewise.
30557         (ix86_code_end): Call output_indirect_thunk_function if needed.
30558         (ix86_output_call_insn): Call ix86_output_indirect_branch if
30559         needed.
30560         (ix86_handle_fndecl_attribute): Handle indirect_branch.
30561         (ix86_attribute_table): Add indirect_branch.
30562         * config/i386/i386.h (machine_function): Add indirect_branch_type
30563         and has_local_indirect_jump.
30564         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
30565         to true.
30566         (tablejump): Likewise.
30567         (*indirect_jump): Use ix86_output_indirect_jmp.
30568         (*tablejump_1): Likewise.
30569         (simple_return_indirect_internal): Likewise.
30570         * config/i386/i386.opt (mindirect-branch=): New option.
30571         (indirect_branch): New.
30572         (keep): Likewise.
30573         (thunk): Likewise.
30574         (thunk-inline): Likewise.
30575         (thunk-extern): Likewise.
30576         * doc/extend.texi: Document indirect_branch function attribute.
30577         * doc/invoke.texi: Document -mindirect-branch= option.
30579 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
30581         PR ipa/83051
30582         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
30584 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
30586         * ipa-inline.c (want_inline_small_function_p): Return false if
30587         inlining has already failed with CIF_FINAL_ERROR.
30588         (update_caller_keys): Call want_inline_small_function_p before
30589         can_inline_edge_p.
30590         (update_callee_keys): Likewise.
30592 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
30594         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
30595         New function.
30596         (rs6000_quadword_masked_address_p): Likewise.
30597         (quad_aligned_load_p): Likewise.
30598         (quad_aligned_store_p): Likewise.
30599         (const_load_sequence_p): Add comment to describe the outer-most loop.
30600         (mimic_memory_attributes_and_flags): New function.
30601         (rs6000_gen_stvx): Likewise.
30602         (replace_swapped_aligned_store): Likewise.
30603         (rs6000_gen_lvx): Likewise.
30604         (replace_swapped_aligned_load): Likewise.
30605         (replace_swapped_load_constant): Capitalize argument name in
30606         comment describing this function.
30607         (rs6000_analyze_swaps): Add a third pass to search for vector loads
30608         and stores that access quad-word aligned addresses and replace
30609         with stvx or lvx instructions when appropriate.
30610         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
30611         New function prototype.
30612         (rs6000_quadword_masked_address_p): Likewise.
30613         (rs6000_gen_lvx): Likewise.
30614         (rs6000_gen_stvx): Likewise.
30615         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
30616         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
30617         when memory address is aligned.
30618         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
30619         this split to select lvx instruction when memory address is aligned.
30620         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
30621         instruction when memory address is aligned.
30622         (*vsx_le_perm_load_v16qi): Likewise.
30623         (four unnamed splitters): Modify to select the stvx instruction
30624         when memory is aligned.
30626 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
30628         * predict.c (determine_unlikely_bbs): Handle correctly BBs
30629         which appears in the queue multiple times.
30631 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30632             Alan Hayward  <alan.hayward@arm.com>
30633             David Sherwood  <david.sherwood@arm.com>
30635         * tree-vectorizer.h (vec_lower_bound): New structure.
30636         (_loop_vec_info): Add check_nonzero and lower_bounds.
30637         (LOOP_VINFO_CHECK_NONZERO): New macro.
30638         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
30639         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
30640         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
30641         fields.  Make seg_len the distance travelled, not including the
30642         access size.
30643         (dr_direction_indicator): Declare.
30644         (dr_zero_step_indicator): Likewise.
30645         (dr_known_forward_stride_p): Likewise.
30646         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
30647         tree-ssanames.h.
30648         (runtime_alias_check_p): Allow runtime alias checks with
30649         variable strides.
30650         (operator ==): Compare access_size and align.
30651         (prune_runtime_alias_test_list): Rework for new distinction between
30652         the access_size and seg_len.
30653         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
30654         segment lengths.
30655         (get_segment_min_max): New function.
30656         (create_intersect_range_checks): Use it.
30657         (dr_step_indicator): New function.
30658         (dr_direction_indicator): Likewise.
30659         (dr_zero_step_indicator): Likewise.
30660         (dr_known_forward_stride_p): Likewise.
30661         * tree-loop-distribution.c (data_ref_segment_size): Return
30662         DR_STEP * (niters - 1).
30663         (compute_alias_check_pairs): Update call to the dr_with_seg_len
30664         constructor.
30665         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
30666         (vect_preserves_scalar_order_p): New function, split out from...
30667         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
30668         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
30669         (vect_vfa_access_size): New function.
30670         (vect_vfa_align): Likewise.
30671         (vect_compile_time_alias): Take access_size_a and access_b arguments.
30672         (dump_lower_bound): New function.
30673         (vect_check_lower_bound): Likewise.
30674         (vect_small_gap_p): Likewise.
30675         (vectorizable_with_step_bound_p): Likewise.
30676         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
30677         depencies if the vectorization factor is 1.  Convert the checks
30678         for nonzero steps into checks on the bounds of DR_STEP.  Try using
30679         a bunds check for variable steps if the minimum required step is
30680         relatively small. Update calls to the dr_with_seg_len
30681         constructor and to vect_compile_time_alias.
30682         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
30683         function.
30684         (vect_loop_versioning): Call it.
30685         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
30686         when retrying.
30687         (vect_estimate_min_profitable_iters): Account for any bounds checks.
30689 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30690             Alan Hayward  <alan.hayward@arm.com>
30691             David Sherwood  <david.sherwood@arm.com>
30693         * doc/sourcebuild.texi (vect_scatter_store): Document.
30694         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
30695         optabs.
30696         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
30697         Document.
30698         * genopinit.c (main): Add supports_vec_scatter_store and
30699         supports_vec_scatter_store_cached to target_optabs.
30700         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
30701         IFN_MASK_SCATTER_STORE.
30702         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
30703         functions.
30704         * internal-fn.h (internal_store_fn_p): Declare.
30705         (internal_fn_stored_value_index): Likewise.
30706         * internal-fn.c (scatter_store_direct): New macro.
30707         (expand_scatter_store_optab_fn): New function.
30708         (direct_scatter_store_optab_supported_p): New macro.
30709         (internal_store_fn_p): New function.
30710         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
30711         IFN_MASK_SCATTER_STORE.
30712         (internal_fn_mask_index): Likewise.
30713         (internal_fn_stored_value_index): New function.
30714         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
30715         for scatter stores.
30716         * optabs-query.h (supports_vec_scatter_store_p): Declare.
30717         * optabs-query.c (supports_vec_scatter_store_p): New function.
30718         * tree-vectorizer.h (vect_get_store_rhs): Declare.
30719         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
30720         true for scatter stores.
30721         (vect_gather_scatter_fn_p): Handle scatter stores too.
30722         (vect_check_gather_scatter): Consider using scatter stores if
30723         supports_vec_scatter_store_p.
30724         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
30725         scatter stores too.
30726         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
30727         internal_fn_stored_value_index.
30728         (check_load_store_masking): Handle scatter stores too.
30729         (vect_get_store_rhs): Make public.
30730         (vectorizable_call): Use internal_store_fn_p.
30731         (vectorizable_store): Handle scatter store internal functions.
30732         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
30733         when deciding whether the end of the group has been reached.
30734         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
30735         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
30736         (mask_scatter_store<mode>): New insns.
30738 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30739             Alan Hayward  <alan.hayward@arm.com>
30740             David Sherwood  <david.sherwood@arm.com>
30742         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
30743         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
30744         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
30745         function.
30746         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
30747         Use vect_truncate_gather_scatter_offset if we can't treat the
30748         operation as a normal gather load or scatter store.
30749         (get_group_load_store_type): Take the gather_scatter_info
30750         as argument.  Try using a gather load or scatter store for
30751         single-element groups.
30752         (get_load_store_type): Update calls to get_group_load_store_type
30753         and vect_use_strided_gather_scatters_p.
30755 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30756             Alan Hayward  <alan.hayward@arm.com>
30757             David Sherwood  <david.sherwood@arm.com>
30759         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
30760         optional tree argument.
30761         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
30762         null target hooks.
30763         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
30764         but continue to use the current value as a fallback.
30765         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
30766         to compare the updates.
30767         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
30768         (get_load_store_type): Use it when handling a strided access.
30769         (vect_get_strided_load_store_ops): New function.
30770         (vect_get_data_ptr_increment): Likewise.
30771         (vectorizable_load): Handle strided gather loads.  Always pass
30772         a step to vect_create_data_ref_ptr and bump_vector_ptr.
30774 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30775             Alan Hayward  <alan.hayward@arm.com>
30776             David Sherwood  <david.sherwood@arm.com>
30778         * doc/md.texi (gather_load@var{m}): Document.
30779         (mask_gather_load@var{m}): Likewise.
30780         * genopinit.c (main): Add supports_vec_gather_load and
30781         supports_vec_gather_load_cached to target_optabs.
30782         * optabs-tree.c (init_tree_optimization_optabs): Use
30783         ggc_cleared_alloc to allocate target_optabs.
30784         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
30785         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
30786         functions.
30787         * internal-fn.h (internal_load_fn_p): Declare.
30788         (internal_gather_scatter_fn_p): Likewise.
30789         (internal_fn_mask_index): Likewise.
30790         (internal_gather_scatter_fn_supported_p): Likewise.
30791         * internal-fn.c (gather_load_direct): New macro.
30792         (expand_gather_load_optab_fn): New function.
30793         (direct_gather_load_optab_supported_p): New macro.
30794         (direct_internal_fn_optab): New function.
30795         (internal_load_fn_p): Likewise.
30796         (internal_gather_scatter_fn_p): Likewise.
30797         (internal_fn_mask_index): Likewise.
30798         (internal_gather_scatter_fn_supported_p): Likewise.
30799         * optabs-query.c (supports_at_least_one_mode_p): New function.
30800         (supports_vec_gather_load_p): Likewise.
30801         * optabs-query.h (supports_vec_gather_load_p): Declare.
30802         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
30803         and memory_type field.
30804         (NUM_PATTERNS): Bump to 15.
30805         * tree-vect-data-refs.c: Include internal-fn.h.
30806         (vect_gather_scatter_fn_p): New function.
30807         (vect_describe_gather_scatter_call): Likewise.
30808         (vect_check_gather_scatter): Try using internal functions for
30809         gather loads.  Recognize existing calls to a gather load function.
30810         (vect_analyze_data_refs): Consider using gather loads if
30811         supports_vec_gather_load_p.
30812         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
30813         (vect_get_gather_scatter_offset_type): Likewise.
30814         (vect_convert_mask_for_vectype): Likewise.
30815         (vect_add_conversion_to_patterm): Likewise.
30816         (vect_try_gather_scatter_pattern): Likewise.
30817         (vect_recog_gather_scatter_pattern): New pattern recognizer.
30818         (vect_vect_recog_func_ptrs): Add it.
30819         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
30820         internal_fn_mask_index and internal_gather_scatter_fn_p.
30821         (check_load_store_masking): Take the gather_scatter_info as an
30822         argument and handle gather loads.
30823         (vect_get_gather_scatter_ops): New function.
30824         (vectorizable_call): Check internal_load_fn_p.
30825         (vectorizable_load): Likewise.  Handle gather load internal
30826         functions.
30827         (vectorizable_store): Update call to check_load_store_masking.
30828         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
30829         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
30830         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
30831         (aarch64_gather_scale_operand_d): New predicates.
30832         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
30833         (mask_gather_load<mode>): New insns.
30835 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30836             Alan Hayward  <alan.hayward@arm.com>
30837             David Sherwood  <david.sherwood@arm.com>
30839         * optabs.def (fold_left_plus_optab): New optab.
30840         * doc/md.texi (fold_left_plus_@var{m}): Document.
30841         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
30842         * internal-fn.c (fold_left_direct): Define.
30843         (expand_fold_left_optab_fn): Likewise.
30844         (direct_fold_left_optab_supported_p): Likewise.
30845         * fold-const-call.c (fold_const_fold_left): New function.
30846         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
30847         * tree-parloops.c (valid_reduction_p): New function.
30848         (gather_scalar_reductions): Use it.
30849         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
30850         (vect_finish_replace_stmt): Declare.
30851         * tree-vect-loop.c (fold_left_reduction_fn): New function.
30852         (needs_fold_left_reduction_p): New function, split out from...
30853         (vect_is_simple_reduction): ...here.  Accept reductions that
30854         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
30855         (vect_force_simple_reduction): Also store the reduction type in
30856         the assignment's STMT_VINFO_REDUC_TYPE.
30857         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
30858         (merge_with_identity): New function.
30859         (vect_expand_fold_left): Likewise.
30860         (vectorize_fold_left_reduction): Likewise.
30861         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
30862         scalar phi in place for it.  Check for target support and reject
30863         cases that would reassociate the operation.  Defer the transform
30864         phase to vectorize_fold_left_reduction.
30865         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
30866         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
30867         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
30869 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30871         * tree-if-conv.c (predicate_mem_writes): Remove redundant
30872         call to ifc_temp_var.
30874 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30875             Alan Hayward  <alan.hayward@arm.com>
30876             David Sherwood  <david.sherwood@arm.com>
30878         * target.def (legitimize_address_displacement): Take the original
30879         offset as a poly_int.
30880         * targhooks.h (default_legitimize_address_displacement): Update
30881         accordingly.
30882         * targhooks.c (default_legitimize_address_displacement): Likewise.
30883         * doc/tm.texi: Regenerate.
30884         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
30885         as an argument, moving assert of ad->disp == ad->disp_term to...
30886         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
30887         Try calling targetm.legitimize_address_displacement before expanding
30888         the address rather than afterwards, and adjust for the new interface.
30889         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
30890         Match the new hook interface.  Handle SVE addresses.
30891         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
30892         new hook interface.
30894 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30896         * Makefile.in (OBJS): Add early-remat.o.
30897         * target.def (select_early_remat_modes): New hook.
30898         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
30899         * doc/tm.texi: Regenerate.
30900         * targhooks.h (default_select_early_remat_modes): Declare.
30901         * targhooks.c (default_select_early_remat_modes): New function.
30902         * timevar.def (TV_EARLY_REMAT): New timevar.
30903         * passes.def (pass_early_remat): New pass.
30904         * tree-pass.h (make_pass_early_remat): Declare.
30905         * early-remat.c: New file.
30906         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
30907         function.
30908         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
30910 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30911             Alan Hayward  <alan.hayward@arm.com>
30912             David Sherwood  <david.sherwood@arm.com>
30914         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
30915         vfm1 with a bound_epilog parameter.
30916         (vect_do_peeling): Update calls accordingly, and move the prologue
30917         call earlier in the function.  Treat the base bound_epilog as 0 for
30918         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
30919         this base when peeling for gaps.
30920         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
30921         with fully-masked loops.
30922         (vect_estimate_min_profitable_iters): Handle the single peeled
30923         iteration in that case.
30925 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30926             Alan Hayward  <alan.hayward@arm.com>
30927             David Sherwood  <david.sherwood@arm.com>
30929         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
30930         single-element interleaving even if the size is not a power of 2.
30931         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
30932         accesses for single-element interleaving if the group size is
30933         not a power of 2.
30935 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30936             Alan Hayward  <alan.hayward@arm.com>
30937             David Sherwood  <david.sherwood@arm.com>
30939         * doc/md.texi (fold_extract_last_@var{m}): Document.
30940         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
30941         * optabs.def (fold_extract_last_optab): New optab.
30942         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
30943         * internal-fn.c (fold_extract_direct): New macro.
30944         (expand_fold_extract_optab_fn): Likewise.
30945         (direct_fold_extract_optab_supported_p): Likewise.
30946         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
30947         * tree-vect-loop.c (vect_model_reduction_cost): Handle
30948         EXTRACT_LAST_REDUCTION.
30949         (get_initial_def_for_reduction): Do not create an initial vector
30950         for EXTRACT_LAST_REDUCTION reductions.
30951         (vectorizable_reduction): Leave the scalar phi in place for
30952         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
30953         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
30954         epilogue code for EXTRACT_LAST_REDUCTION and defer the
30955         transform phase to vectorizable_condition.
30956         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
30957         split out from...
30958         (vect_finish_stmt_generation): ...here.
30959         (vect_finish_replace_stmt): New function.
30960         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
30961         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
30962         pattern.
30963         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
30965 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30966             Alan Hayward  <alan.hayward@arm.com>
30967             David Sherwood  <david.sherwood@arm.com>
30969         * doc/md.texi (extract_last_@var{m}): Document.
30970         * optabs.def (extract_last_optab): New optab.
30971         * internal-fn.def (EXTRACT_LAST): New internal function.
30972         * internal-fn.c (cond_unary_direct): New macro.
30973         (expand_cond_unary_optab_fn): Likewise.
30974         (direct_cond_unary_optab_supported_p): Likewise.
30975         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
30976         loops using EXTRACT_LAST.
30977         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
30978         (extract_last_<mode>): ...this optab.
30979         (vec_extract<mode><Vel>): Update accordingly.
30981 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30982             Alan Hayward  <alan.hayward@arm.com>
30983             David Sherwood  <david.sherwood@arm.com>
30985         * target.def (empty_mask_is_expensive): New hook.
30986         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
30987         * doc/tm.texi: Regenerate.
30988         * targhooks.h (default_empty_mask_is_expensive): Declare.
30989         * targhooks.c (default_empty_mask_is_expensive): New function.
30990         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
30991         if the target says that empty masks are expensive.
30992         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
30993         New function.
30994         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
30996 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30997             Alan Hayward  <alan.hayward@arm.com>
30998             David Sherwood  <david.sherwood@arm.com>
31000         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
31001         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
31002         (vect_use_loop_mask_for_alignment_p): New function.
31003         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
31004         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
31005         niters_skip argument.  Make sure that the first niters_skip elements
31006         of the first iteration are inactive.
31007         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
31008         Update call to vect_set_loop_masks_directly.
31009         (get_misalign_in_elems): New function, split out from...
31010         (vect_gen_prolog_loop_niters): ...here.
31011         (vect_update_init_of_dr): Take a code argument that specifies whether
31012         the adjustment should be added or subtracted.
31013         (vect_update_init_of_drs): Likewise.
31014         (vect_prepare_for_masked_peels): New function.
31015         (vect_do_peeling): Skip prologue peeling if we're using a mask
31016         instead.  Update call to vect_update_inits_of_drs.
31017         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
31018         mask_skip_niters.
31019         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
31020         alignment.  Do not include the number of peeled iterations in
31021         the minimum threshold in that case.
31022         (vectorizable_induction): Adjust the start value down by
31023         LOOP_VINFO_MASK_SKIP_NITERS iterations.
31024         (vect_transform_loop): Call vect_prepare_for_masked_peels.
31025         Take the number of skipped iterations into account when calculating
31026         the loop bounds.
31027         * tree-vect-stmts.c (vect_gen_while_not): New function.
31029 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31030             Alan Hayward  <alan.hayward@arm.com>
31031             David Sherwood  <david.sherwood@arm.com>
31033         * doc/sourcebuild.texi (vect_fully_masked): Document.
31034         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
31035         default value to 0.
31036         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
31037         split out from...
31038         (vect_analyze_loop_2): ...here. Don't check the vectorization
31039         factor against the number of loop iterations if the loop is
31040         fully-masked.
31042 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31043             Alan Hayward  <alan.hayward@arm.com>
31044             David Sherwood  <david.sherwood@arm.com>
31046         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
31047         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
31048         (dump_groups): Update accordingly.
31049         (iv_use::mem_type): New member variable.
31050         (address_p): New function.
31051         (record_use): Add a mem_type argument and initialize the new
31052         mem_type field.
31053         (record_group_use): Add a mem_type argument.  Use address_p.
31054         Remove obsolete null checks of base_object.  Update call to record_use.
31055         (find_interesting_uses_op): Update call to record_group_use.
31056         (find_interesting_uses_cond): Likewise.
31057         (find_interesting_uses_address): Likewise.
31058         (get_mem_type_for_internal_fn): New function.
31059         (find_address_like_use): Likewise.
31060         (find_interesting_uses_stmt): Try find_address_like_use before
31061         calling find_interesting_uses_op.
31062         (addr_offset_valid_p): Use the iv mem_type field as the type
31063         of the addressed memory.
31064         (add_autoinc_candidates): Likewise.
31065         (get_address_cost): Likewise.
31066         (split_small_address_groups_p): Use address_p.
31067         (split_address_groups): Likewise.
31068         (add_iv_candidate_for_use): Likewise.
31069         (autoinc_possible_for_pair): Likewise.
31070         (rewrite_groups): Likewise.
31071         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
31072         (determine_group_iv_cost): Update after split of USE_ADDRESS.
31073         (get_alias_ptr_type_for_ptr_address): New function.
31074         (rewrite_use_address): Rewrite address uses in calls that were
31075         identified by find_address_like_use.
31077 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31078             Alan Hayward  <alan.hayward@arm.com>
31079             David Sherwood  <david.sherwood@arm.com>
31081         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
31082         TARGET_MEM_REFs.
31083         * gimple-expr.h (is_gimple_addressable: Likewise.
31084         * gimple-expr.c (is_gimple_address): Likewise.
31085         * internal-fn.c (expand_call_mem_ref): New function.
31086         (expand_mask_load_optab_fn): Use it.
31087         (expand_mask_store_optab_fn): Likewise.
31089 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31090             Alan Hayward  <alan.hayward@arm.com>
31091             David Sherwood  <david.sherwood@arm.com>
31093         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
31094         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
31095         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
31096         (cond_umax@var{mode}): Document.
31097         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
31098         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
31099         (cond_umin_optab, cond_umax_optab): New optabs.
31100         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
31101         (COND_IOR, COND_XOR): New internal functions.
31102         * internal-fn.h (get_conditional_internal_fn): Declare.
31103         * internal-fn.c (cond_binary_direct): New macro.
31104         (expand_cond_binary_optab_fn): Likewise.
31105         (direct_cond_binary_optab_supported_p): Likewise.
31106         (get_conditional_internal_fn): New function.
31107         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
31108         Cope with reduction statements that are vectorized as calls rather
31109         than assignments.
31110         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
31111         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
31112         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
31113         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
31114         (UNSPEC_COND_EOR): New unspecs.
31115         (optab): Add mappings for them.
31116         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
31117         (sve_int_op, sve_fp_op): New int attributes.
31119 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31120             Alan Hayward  <alan.hayward@arm.com>
31121             David Sherwood  <david.sherwood@arm.com>
31123         * optabs.def (while_ult_optab): New optab.
31124         * doc/md.texi (while_ult@var{m}@var{n}): Document.
31125         * internal-fn.def (WHILE_ULT): New internal function.
31126         * internal-fn.h (direct_internal_fn_supported_p): New override
31127         that takes two types as argument.
31128         * internal-fn.c (while_direct): New macro.
31129         (expand_while_optab_fn): New function.
31130         (convert_optab_supported_p): Likewise.
31131         (direct_while_optab_supported_p): New macro.
31132         * wide-int.h (wi::udiv_ceil): New function.
31133         * tree-vectorizer.h (rgroup_masks): New structure.
31134         (vec_loop_masks): New typedef.
31135         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
31136         and fully_masked_p.
31137         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
31138         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
31139         (vect_max_vf): New function.
31140         (slpeel_make_loop_iterate_ntimes): Delete.
31141         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
31142         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
31143         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
31144         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
31145         internal-fn.h, stor-layout.h and optabs-query.h.
31146         (vect_set_loop_mask): New function.
31147         (add_preheader_seq): Likewise.
31148         (add_header_seq): Likewise.
31149         (interleave_supported_p): Likewise.
31150         (vect_maybe_permute_loop_masks): Likewise.
31151         (vect_set_loop_masks_directly): Likewise.
31152         (vect_set_loop_condition_masked): Likewise.
31153         (vect_set_loop_condition_unmasked): New function, split out from
31154         slpeel_make_loop_iterate_ntimes.
31155         (slpeel_make_loop_iterate_ntimes): Rename to..
31156         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
31157         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
31158         (vect_do_peeling): Update call accordingly.
31159         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
31160         loops.
31161         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
31162         mask_compare_type, can_fully_mask_p and fully_masked_p.
31163         (release_vec_loop_masks): New function.
31164         (_loop_vec_info): Use it to free the loop masks.
31165         (can_produce_all_loop_masks_p): New function.
31166         (vect_get_max_nscalars_per_iter): Likewise.
31167         (vect_verify_full_masking): Likewise.
31168         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
31169         retries, and free the mask rgroups before retrying.  Check loop-wide
31170         reasons for disallowing fully-masked loops.  Make the final decision
31171         about whether use a fully-masked loop or not.
31172         (vect_estimate_min_profitable_iters): Do not assume that peeling
31173         for the number of iterations will be needed for fully-masked loops.
31174         (vectorizable_reduction): Disable fully-masked loops.
31175         (vectorizable_live_operation): Likewise.
31176         (vect_halve_mask_nunits): New function.
31177         (vect_double_mask_nunits): Likewise.
31178         (vect_record_loop_mask): Likewise.
31179         (vect_get_loop_mask): Likewise.
31180         (vect_transform_loop): Handle the case in which the final loop
31181         iteration might handle a partial vector.  Call vect_set_loop_condition
31182         instead of slpeel_make_loop_iterate_ntimes.
31183         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
31184         (check_load_store_masking): New function.
31185         (prepare_load_store_mask): Likewise.
31186         (vectorizable_store): Handle fully-masked loops.
31187         (vectorizable_load): Likewise.
31188         (supportable_widening_operation): Use vect_halve_mask_nunits for
31189         booleans.
31190         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
31191         (vect_gen_while): New function.
31192         * config/aarch64/aarch64.md (umax<mode>3): New expander.
31193         (aarch64_uqdec<mode>): New insn.
31195 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31196             Alan Hayward  <alan.hayward@arm.com>
31197             David Sherwood  <david.sherwood@arm.com>
31199         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
31200         (reduc_xor_scal_optab): New optabs.
31201         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
31202         (reduc_xor_scal_@var{m}): Document.
31203         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
31204         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
31205         internal functions.
31206         * fold-const-call.c (fold_const_call): Handle them.
31207         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
31208         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
31209         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
31210         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
31211         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
31212         (UNSPEC_XORV): New unspecs.
31213         (optab): Add entries for them.
31214         (BITWISEV): New int iterator.
31215         (bit_reduc_op): New int attributes.
31217 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31218             Alan Hayward  <alan.hayward@arm.com>
31219             David Sherwood  <david.sherwood@arm.com>
31221         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
31222         * internal-fn.def (VEC_SHL_INSERT): New internal function.
31223         * optabs.def (vec_shl_insert_optab): New optab.
31224         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
31225         (duplicate_and_interleave): Likewise.
31226         * tree-vect-loop.c: Include internal-fn.h.
31227         (neutral_op_for_slp_reduction): New function, split out from
31228         get_initial_defs_for_reduction.
31229         (get_initial_def_for_reduction): Handle option 2 for variable-length
31230         vectors by loading the neutral value into a vector and then shifting
31231         the initial value into element 0.
31232         (get_initial_defs_for_reduction): Replace the code argument with
31233         the neutral value calculated by neutral_op_for_slp_reduction.
31234         Use gimple_build_vector for constant-length vectors.
31235         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
31236         but the first group_size elements have a neutral value.
31237         Use duplicate_and_interleave otherwise.
31238         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
31239         Update call to get_initial_defs_for_reduction.  Handle SLP
31240         reductions for variable-length vectors by creating one vector
31241         result for each scalar result, with the elements associated
31242         with other scalar results stubbed out with the neutral value.
31243         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
31244         Require IFN_VEC_SHL_INSERT for double reductions on
31245         variable-length vectors, or SLP reductions that have
31246         a neutral value.  Require can_duplicate_and_interleave_p
31247         support for variable-length unchained SLP reductions if there
31248         is no neutral value, such as for MIN/MAX reductions.  Also require
31249         the number of vector elements to be a multiple of the number of
31250         SLP statements when doing variable-length unchained SLP reductions.
31251         Update call to vect_create_epilog_for_reduction.
31252         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
31253         and remove initial values.
31254         (duplicate_and_interleave): Make public.
31255         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
31256         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
31258 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31259             Alan Hayward  <alan.hayward@arm.com>
31260             David Sherwood  <david.sherwood@arm.com>
31262         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
31263         (can_duplicate_and_interleave_p): New function.
31264         (vect_get_and_check_slp_defs): Take the vector of statements
31265         rather than just the current one.  Remove excess parentheses.
31266         Restriction rejectinon of vect_constant_def and vect_external_def
31267         for variable-length vectors to boolean types, or types for which
31268         can_duplicate_and_interleave_p is false.
31269         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
31270         (duplicate_and_interleave): New function.
31271         (vect_get_constant_vectors): Use gimple_build_vector for
31272         constant-length vectors and suitable variable-length constant
31273         vectors.  Use duplicate_and_interleave for other variable-length
31274         vectors.  Don't defer the update when inserting new statements.
31276 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31277             Alan Hayward  <alan.hayward@arm.com>
31278             David Sherwood  <david.sherwood@arm.com>
31280         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
31281         min_profitable_iters doesn't go negative.
31283 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31284             Alan Hayward  <alan.hayward@arm.com>
31285             David Sherwood  <david.sherwood@arm.com>
31287         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
31288         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
31289         * optabs.def (vec_mask_load_lanes_optab): New optab.
31290         (vec_mask_store_lanes_optab): Likewise.
31291         * internal-fn.def (MASK_LOAD_LANES): New internal function.
31292         (MASK_STORE_LANES): Likewise.
31293         * internal-fn.c (mask_load_lanes_direct): New macro.
31294         (mask_store_lanes_direct): Likewise.
31295         (expand_mask_load_optab_fn): Handle masked operations.
31296         (expand_mask_load_lanes_optab_fn): New macro.
31297         (expand_mask_store_optab_fn): Handle masked operations.
31298         (expand_mask_store_lanes_optab_fn): New macro.
31299         (direct_mask_load_lanes_optab_supported_p): Likewise.
31300         (direct_mask_store_lanes_optab_supported_p): Likewise.
31301         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
31302         parameter.
31303         (vect_load_lanes_supported): Likewise.
31304         * tree-vect-data-refs.c (strip_conversion): New function.
31305         (can_group_stmts_p): Likewise.
31306         (vect_analyze_data_ref_accesses): Use it instead of checking
31307         for a pair of assignments.
31308         (vect_store_lanes_supported): Take a masked_p parameter.
31309         (vect_load_lanes_supported): Likewise.
31310         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
31311         vect_store_lanes_supported and vect_load_lanes_supported.
31312         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
31313         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
31314         parameter.  Don't allow gaps for masked accesses.
31315         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
31316         and vect_load_lanes_supported.
31317         (get_load_store_type): Take a masked_p parameter and update
31318         call to get_group_load_store_type.
31319         (vectorizable_store): Update call to get_load_store_type.
31320         Handle IFN_MASK_STORE_LANES.
31321         (vectorizable_load): Update call to get_load_store_type.
31322         Handle IFN_MASK_LOAD_LANES.
31324 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31325             Alan Hayward  <alan.hayward@arm.com>
31326             David Sherwood  <david.sherwood@arm.com>
31328         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
31329         modes for SVE.
31330         * config/aarch64/aarch64-protos.h
31331         (aarch64_sve_struct_memory_operand_p): Declare.
31332         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
31333         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
31334         (VPRED, vpred): Handle SVE structure modes.
31335         * config/aarch64/constraints.md (Utx): New constraint.
31336         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
31337         (aarch64_sve_struct_nonimmediate_operand): New predicates.
31338         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
31339         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
31340         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
31341         structure modes.  Split into pieces after RA.
31342         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
31343         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
31344         New patterns.
31345         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
31346         SVE structure modes.
31347         (aarch64_classify_address): Likewise.
31348         (sizetochar): Move earlier in file.
31349         (aarch64_print_operand): Handle SVE register lists.
31350         (aarch64_array_mode): New function.
31351         (aarch64_sve_struct_memory_operand_p): Likewise.
31352         (TARGET_ARRAY_MODE): Redefine.
31354 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31355             Alan Hayward  <alan.hayward@arm.com>
31356             David Sherwood  <david.sherwood@arm.com>
31358         * target.def (array_mode): New target hook.
31359         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
31360         * doc/tm.texi: Regenerate.
31361         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
31362         * hooks.c (hook_optmode_mode_uhwi_none): New function.
31363         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
31364         targetm.array_mode.
31365         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
31366         type sizes.
31368 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31369             Alan Hayward  <alan.hayward@arm.com>
31370             David Sherwood  <david.sherwood@arm.com>
31372         * fold-const.c (fold_binary_loc): Check the argument types
31373         rather than the result type when testing for a vector operation.
31375 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31377         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
31378         * doc/tm.texi: Regenerate.
31380 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31381             Alan Hayward  <alan.hayward@arm.com>
31382             David Sherwood  <david.sherwood@arm.com>
31384         * doc/invoke.texi (-msve-vector-bits=): Document new option.
31385         (sve): Document new AArch64 extension.
31386         * doc/md.texi (w): Extend the description of the AArch64
31387         constraint to include SVE vectors.
31388         (Upl, Upa): Document new AArch64 predicate constraints.
31389         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
31390         enum.
31391         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
31392         (msve-vector-bits=): New option.
31393         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
31394         SVE when these are disabled.
31395         (sve): New extension.
31396         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
31397         modes.  Adjust their number of units based on aarch64_sve_vg.
31398         (MAX_BITSIZE_MODE_ANY_MODE): Define.
31399         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
31400         aarch64_addr_query_type.
31401         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
31402         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
31403         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
31404         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
31405         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
31406         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
31407         (aarch64_simd_imm_zero_p): Delete.
31408         (aarch64_check_zero_based_sve_index_immediate): Declare.
31409         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
31410         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
31411         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
31412         (aarch64_sve_float_mul_immediate_p): Likewise.
31413         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
31414         rather than an rtx.
31415         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
31416         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
31417         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
31418         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
31419         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
31420         (aarch64_regmode_natural_size): Likewise.
31421         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
31422         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
31423         left one place.
31424         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
31425         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
31426         for VG and the SVE predicate registers.
31427         (V_ALIASES): Add a "z"-prefixed alias.
31428         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
31429         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
31430         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
31431         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
31432         (REG_CLASS_NAMES): Add entries for them.
31433         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
31434         and the predicate registers.
31435         (aarch64_sve_vg): Declare.
31436         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
31437         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
31438         (REGMODE_NATURAL_SIZE): Define.
31439         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
31440         SVE macros.
31441         * config/aarch64/aarch64.c: Include cfgrtl.h.
31442         (simd_immediate_info): Add a constructor for series vectors,
31443         and an associated step field.
31444         (aarch64_sve_vg): New variable.
31445         (aarch64_dbx_register_number): Handle VG and the predicate registers.
31446         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
31447         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
31448         (VEC_ANY_DATA, VEC_STRUCT): New constants.
31449         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
31450         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
31451         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
31452         (aarch64_get_mask_mode): New functions.
31453         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
31454         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
31455         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
31456         predicate modes and predicate registers.  Explicitly restrict
31457         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
31458         to store a vector mode if it is recognized by
31459         aarch64_classify_vector_mode.
31460         (aarch64_regmode_natural_size): New function.
31461         (aarch64_hard_regno_caller_save_mode): Return the original mode
31462         for predicates.
31463         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
31464         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
31465         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
31466         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
31467         functions.
31468         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
31469         does not overlap dest if the function is frame-related.  Handle
31470         SVE constants.
31471         (aarch64_split_add_offset): New function.
31472         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
31473         them aarch64_add_offset.
31474         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
31475         and update call to aarch64_sub_sp.
31476         (aarch64_add_cfa_expression): New function.
31477         (aarch64_expand_prologue): Pass extra temporary registers to the
31478         functions above.  Handle the case in which we need to emit new
31479         DW_CFA_expressions for registers that were originally saved
31480         relative to the stack pointer, but now have to be expressed
31481         relative to the frame pointer.
31482         (aarch64_output_mi_thunk): Pass extra temporary registers to the
31483         functions above.
31484         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
31485         IP0 and IP1 values for SVE frames.
31486         (aarch64_expand_vec_series): New function.
31487         (aarch64_expand_sve_widened_duplicate): Likewise.
31488         (aarch64_expand_sve_const_vector): Likewise.
31489         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
31490         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
31491         into the register, rather than emitting a SET directly.
31492         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
31493         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
31494         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
31495         (offset_9bit_signed_scaled_p): New functions.
31496         (aarch64_replicate_bitmask_imm): New function.
31497         (aarch64_bitmask_imm): Use it.
31498         (aarch64_cannot_force_const_mem): Reject expressions involving
31499         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
31500         (aarch64_classify_index): Handle SVE indices, by requiring
31501         a plain register index with a scale that matches the element size.
31502         (aarch64_classify_address): Handle SVE addresses.  Assert that
31503         the mode of the address is VOIDmode or an integer mode.
31504         Update call to aarch64_classify_symbol.
31505         (aarch64_classify_symbolic_expression): Update call to
31506         aarch64_classify_symbol.
31507         (aarch64_const_vec_all_in_range_p): New function.
31508         (aarch64_print_vector_float_operand): Likewise.
31509         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
31510         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
31511         and the FP immediates 1.0 and 0.5.
31512         (aarch64_print_address_internal): Handle SVE addresses.
31513         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
31514         (aarch64_regno_regclass): Handle predicate registers.
31515         (aarch64_secondary_reload): Handle big-endian reloads of SVE
31516         data modes.
31517         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
31518         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
31519         (aarch64_convert_sve_vector_bits): New function.
31520         (aarch64_override_options): Use it to handle -msve-vector-bits=.
31521         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
31522         rather than an rtx.
31523         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
31524         Handle SVE vector and predicate modes.  Accept VL-based constants
31525         that need only one temporary register, and VL offsets that require
31526         no temporary registers.
31527         (aarch64_conditional_register_usage): Mark the predicate registers
31528         as fixed if SVE isn't available.
31529         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
31530         Return true for SVE vector and predicate modes.
31531         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
31532         rather than an unsigned int.  Handle SVE modes.
31533         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
31534         SVE modes.
31535         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
31536         if SVE is enabled.
31537         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
31538         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
31539         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
31540         (aarch64_sve_float_mul_immediate_p): New functions.
31541         (aarch64_sve_valid_immediate): New function.
31542         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
31543         Explicitly reject structure modes.  Check for INDEX constants.
31544         Handle PTRUE and PFALSE constants.
31545         (aarch64_check_zero_based_sve_index_immediate): New function.
31546         (aarch64_simd_imm_zero_p): Delete.
31547         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
31548         vector modes.  Accept constants in the range of CNT[BHWD].
31549         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
31550         ask for an Advanced SIMD mode.
31551         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
31552         (aarch64_simd_vector_alignment): Handle SVE predicates.
31553         (aarch64_vectorize_preferred_vector_alignment): New function.
31554         (aarch64_simd_vector_alignment_reachable): Use it instead of
31555         the vector size.
31556         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
31557         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
31558         functions.
31559         (MAX_VECT_LEN): Delete.
31560         (expand_vec_perm_d): Add a vec_flags field.
31561         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
31562         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
31563         (aarch64_evpc_ext): Don't apply a big-endian lane correction
31564         for SVE modes.
31565         (aarch64_evpc_rev): Rename to...
31566         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
31567         (aarch64_evpc_rev_global): New function.
31568         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
31569         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
31570         MAX_VECT_LEN.
31571         (aarch64_evpc_sve_tbl): New function.
31572         (aarch64_expand_vec_perm_const_1): Update after rename of
31573         aarch64_evpc_rev.  Handle SVE permutes too, trying
31574         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
31575         than aarch64_evpc_tbl.
31576         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
31577         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
31578         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
31579         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
31580         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
31581         (aarch64_expand_sve_vcond): New functions.
31582         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
31583         of aarch64_vector_mode_p.
31584         (aarch64_dwarf_poly_indeterminate_value): New function.
31585         (aarch64_compute_pressure_classes): Likewise.
31586         (aarch64_can_change_mode_class): Likewise.
31587         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
31588         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
31589         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
31590         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
31591         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
31592         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
31593         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
31594         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
31595         constraints.
31596         (Dn, Dl, Dr): Accept const as well as const_vector.
31597         (Dz): Likewise.  Compare against CONST0_RTX.
31598         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
31599         of "vector" where appropriate.
31600         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
31601         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
31602         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
31603         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
31604         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
31605         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
31606         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
31607         (v_int_equiv): Extend to SVE modes.
31608         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
31609         mode attributes.
31610         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
31611         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
31612         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
31613         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
31614         (SVE_COND_FP_CMP): New int iterators.
31615         (perm_hilo): Handle the new unpack unspecs.
31616         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
31617         attributes.
31618         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
31619         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
31620         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
31621         (aarch64_equality_operator, aarch64_constant_vector_operand)
31622         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
31623         (aarch64_sve_nonimmediate_operand): Likewise.
31624         (aarch64_sve_general_operand): Likewise.
31625         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
31626         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
31627         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
31628         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
31629         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
31630         (aarch64_sve_float_arith_immediate): Likewise.
31631         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
31632         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
31633         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
31634         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
31635         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
31636         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
31637         (aarch64_sve_float_arith_operand): Likewise.
31638         (aarch64_sve_float_arith_with_sub_operand): Likewise.
31639         (aarch64_sve_float_mul_operand): Likewise.
31640         (aarch64_sve_vec_perm_operand): Likewise.
31641         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
31642         (aarch64_mov_operand): Accept const_poly_int and const_vector.
31643         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
31644         as well as const_vector.
31645         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
31646         in file.  Use CONST0_RTX and CONSTM1_RTX.
31647         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
31648         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
31649         Use aarch64_simd_imm_zero.
31650         * config/aarch64/aarch64-sve.md: New file.
31651         * config/aarch64/aarch64.md: Include it.
31652         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
31653         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
31654         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
31655         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
31656         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
31657         (sve): New attribute.
31658         (enabled): Disable instructions with the sve attribute unless
31659         TARGET_SVE.
31660         (movqi, movhi): Pass CONST_POLY_INT operaneds through
31661         aarch64_expand_mov_immediate.
31662         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
31663         CNT[BHSD] immediates.
31664         (movti): Split CONST_POLY_INT moves into two halves.
31665         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
31666         Split additions that need a temporary here if the destination
31667         is the stack pointer.
31668         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
31669         (*add<mode>3_poly_1): New instruction.
31670         (set_clobber_cc): New expander.
31672 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31674         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
31675         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
31676         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
31677         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
31678         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
31679         Change innermode from fixed_mode_size to machine_mode.
31680         (simplify_subreg): Update call accordingly.  Handle a constant-sized
31681         subreg of a variable-length CONST_VECTOR.
31683 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31684             Alan Hayward  <alan.hayward@arm.com>
31685             David Sherwood  <david.sherwood@arm.com>
31687         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
31688         (add_offset_to_base): New function, split out from...
31689         (create_mem_ref): ...here.  When handling a scale other than 1,
31690         check first whether the address is valid without the offset.
31691         Add it into the base if so, leaving the index and scale as-is.
31693 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
31695         PR c++/83778
31696         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
31697         fold_for_warn before checking if arg2 is INTEGER_CST.
31699 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
31701         * config/rs6000/predicates.md (load_multiple_operation): Delete.
31702         (store_multiple_operation): Delete.
31703         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
31704         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
31705         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
31706         guarded by TARGET_STRING.
31707         (rs6000_output_load_multiple): Delete.
31708         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
31709         OPTION_MASK_STRING / TARGET_STRING handling.
31710         (print_operand) <'N', 'O'>: Add comment that these are unused now.
31711         (const rs6000_opt_masks) <"string">: Change mask to 0.
31712         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
31713         (MASK_STRING): Delete.
31714         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
31715         parts.  Simplify.
31716         (load_multiple): Delete.
31717         (*ldmsi8): Delete.
31718         (*ldmsi7): Delete.
31719         (*ldmsi6): Delete.
31720         (*ldmsi5): Delete.
31721         (*ldmsi4): Delete.
31722         (*ldmsi3): Delete.
31723         (store_multiple): Delete.
31724         (*stmsi8): Delete.
31725         (*stmsi7): Delete.
31726         (*stmsi6): Delete.
31727         (*stmsi5): Delete.
31728         (*stmsi4): Delete.
31729         (*stmsi3): Delete.
31730         (movmemsi_8reg): Delete.
31731         (corresponding unnamed define_insn): Delete.
31732         (movmemsi_6reg): Delete.
31733         (corresponding unnamed define_insn): Delete.
31734         (movmemsi_4reg): Delete.
31735         (corresponding unnamed define_insn): Delete.
31736         (movmemsi_2reg): Delete.
31737         (corresponding unnamed define_insn): Delete.
31738         (movmemsi_1reg): Delete.
31739         (corresponding unnamed define_insn): Delete.
31740         * config/rs6000/rs6000.opt (mno-string): New.
31741         (mstring): Replace by deprecation warning stub.
31742         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
31744 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
31746         * regrename.c (regrename_do_replace): If replacing the same
31747         reg multiple times, try to reuse last created gen_raw_REG.
31749         PR debug/81155
31750         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
31751         main to workaround a bug in GDB.
31753 2018-01-12  Tom de Vries  <tom@codesourcery.com>
31755         PR target/83737
31756         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
31758 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
31760         PR rtl-optimization/80481
31761         * ira-color.c (get_cap_member): New function.
31762         (allocnos_conflict_by_live_ranges_p): Use it.
31763         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
31764         (setup_slot_coalesced_allocno_live_ranges): Ditto.
31766 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
31768         PR target/83628
31769         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
31770         (*saddl_se_1): Ditto.
31771         (*ssubsi_1): Ditto.
31772         (*ssubl_se_1): Ditto.
31774 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
31776         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
31777         rather than wi::to_widest for DR_INITs.
31778         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
31779         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
31780         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
31781         INTEGER_CSTs.
31782         (vect_analyze_group_access_1): Note that here.
31784 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
31786         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
31787         polynomial type sizes.
31789 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
31791         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
31792         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
31793         (gimple_add_tmp_var): Likewise.
31795 2018-01-12  Martin Liska  <mliska@suse.cz>
31797         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
31798         (gimple_alloc_sizes): Likewise.
31799         (dump_gimple_statistics): Use PRIu64 in printf format.
31800         * gimple.h: Change uint64_t to int.
31802 2018-01-12  Martin Liska  <mliska@suse.cz>
31804         * tree-core.h: Use uint64_t instead of int.
31805         * tree.c (tree_node_counts): Likewise.
31806         (tree_node_sizes): Likewise.
31807         (dump_tree_statistics): Use PRIu64 in printf format.
31809 2018-01-12  Martin Liska  <mliska@suse.cz>
31811         * Makefile.in: As qsort_chk is implemented in vec.c, add
31812         vec.o to linkage of gencfn-macros.
31813         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
31814         passing the info to record_node_allocation_statistics.
31815         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
31816         and pass the info.
31817         * ggc-common.c (struct ggc_usage): Add operator== and use
31818         it in operator< and compare function.
31819         * mem-stats.h (struct mem_usage): Likewise.
31820         * vec.c (struct vec_usage): Remove operator< and compare
31821         function. Can be simply inherited.
31823 2018-01-12  Martin Jambor  <mjambor@suse.cz>
31825         PR target/81616
31826         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
31827         * tree-ssa-math-opts.c: Include domwalk.h.
31828         (convert_mult_to_fma_1): New function.
31829         (fma_transformation_info): New type.
31830         (fma_deferring_state): Likewise.
31831         (cancel_fma_deferring): New function.
31832         (result_of_phi): Likewise.
31833         (last_fma_candidate_feeds_initial_phi): Likewise.
31834         (convert_mult_to_fma): Added deferring logic, split actual
31835         transformation to convert_mult_to_fma_1.
31836         (math_opts_dom_walker): New type.
31837         (math_opts_dom_walker::after_dom_children): New method, body moved
31838         here from pass_optimize_widening_mul::execute, added deferring logic
31839         bits.
31840         (pass_optimize_widening_mul::execute): Moved most of code to
31841         math_opts_dom_walker::after_dom_children.
31842         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
31843         * config/i386/i386.c (ix86_option_override_internal): Added
31844         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
31846 2018-01-12  Richard Biener  <rguenther@suse.de>
31848         PR debug/83157
31849         * dwarf2out.c (gen_variable_die): Do not reset old_die for
31850         inline instance vars.
31852 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
31854         PR target/81819
31855         * config/rx/rx.c (rx_is_restricted_memory_address):
31856         Handle SUBREG case.
31858 2018-01-12  Richard Biener  <rguenther@suse.de>
31860         PR tree-optimization/80846
31861         * target.def (split_reduction): New target hook.
31862         * targhooks.c (default_split_reduction): New function.
31863         * targhooks.h (default_split_reduction): Declare.
31864         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
31865         target requests first reduce vectors by combining low and high
31866         parts.
31867         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
31868         (get_vectype_for_scalar_type_and_size): Export.
31869         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
31870         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
31871         * doc/tm.texi: Regenerate.
31872         * config/i386/i386.c (ix86_split_reduction): Implement
31873         TARGET_VECTORIZE_SPLIT_REDUCTION.
31875 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
31877         PR target/83368
31878         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
31879         in PIC mode except for TARGET_VXWORKS_RTP.
31880         * config/sparc/sparc.c: Include cfgrtl.h.
31881         (TARGET_INIT_PIC_REG): Define.
31882         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
31883         (sparc_pic_register_p): New predicate.
31884         (sparc_legitimate_address_p): Use it.
31885         (sparc_legitimize_pic_address): Likewise.
31886         (sparc_delegitimize_address): Likewise.
31887         (sparc_mode_dependent_address_p): Likewise.
31888         (gen_load_pcrel_sym): Remove 4th parameter.
31889         (load_got_register): Adjust call to above.  Remove obsolete stuff.
31890         (sparc_expand_prologue): Do not call load_got_register here.
31891         (sparc_flat_expand_prologue): Likewise.
31892         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
31893         (sparc_use_pseudo_pic_reg): New function.
31894         (sparc_init_pic_reg): Likewise.
31895         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
31896         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
31898 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
31900         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
31901         Add item for branch_cost.
31903 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
31905         PR rtl-optimization/83565
31906         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
31907         not extend the result to a larger mode for rotate operations.
31908         (num_sign_bit_copies1): Likewise.
31910 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
31912         PR target/40411
31913         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
31914         -symbolic.
31915         Use values-Xc.o for -pedantic.
31916         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
31918 2018-01-12  Martin Liska  <mliska@suse.cz>
31920         PR ipa/83054
31921         * ipa-devirt.c (final_warning_record::grow_type_warnings):
31922         New function.
31923         (possible_polymorphic_call_targets): Use it.
31924         (ipa_devirt): Likewise.
31926 2018-01-12  Martin Liska  <mliska@suse.cz>
31928         * profile-count.h (enum profile_quality): Use 0 as invalid
31929         enum value of profile_quality.
31931 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
31933         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
31934         -mext-string options.
31936 2018-01-12  Richard Biener  <rguenther@suse.de>
31938         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
31939         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
31940         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
31941         Likewise.
31942         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
31944 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31946         * configure.ac (--with-long-double-format): Add support for the
31947         configuration option to change the default long double format on
31948         PowerPC systems.
31949         * config.gcc (powerpc*-linux*-*): Likewise.
31950         * configure: Regenerate.
31951         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
31952         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
31953         used without modification.
31955 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31957         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
31958         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
31959         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
31960         MISC_BUILTIN_SPEC_BARRIER.
31961         (rs6000_init_builtins): Likewise.
31962         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
31963         enum value.
31964         (speculation_barrier): New define_insn.
31965         * doc/extend.texi: Document __builtin_speculation_barrier.
31967 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
31969         PR target/83203
31970         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
31971         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
31972         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
31973         iterators.
31974         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
31975         integral modes instead of "ss" and "sd".
31976         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
31977         vectors with 32-bit and 64-bit elements.
31978         (vecdupssescalarmodesuffix): New mode attribute.
31979         (vec_dup<mode>): Use it.
31981 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
31983         PR target/83330
31984         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
31985         frame if argument is passed on stack.
31987 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
31989         PR target/82682
31990         * ree.c (combine_reaching_defs): Optimize also
31991         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
31992         reg2=any_extend(exp); reg1=reg2;, formatting fix.
31994 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
31996         PR middle-end/83189
31997         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
31999 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
32001         PR middle-end/83718
32002         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
32003         after they are computed.
32005 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
32007         PR tree-optimization/83695
32008         * gimple-loop-linterchange.cc
32009         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
32010         reset cached scev information after interchange.
32011         (pass_linterchange::execute): Remove call to scev_reset_htab.
32013 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32015         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
32016         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
32017         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
32018         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
32019         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
32020         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
32021         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
32022         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
32023         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
32024         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
32025         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
32026         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
32027         (V_lane_reg): Likewise.
32028         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
32029         New define_expand.
32030         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
32031         (vfmal_lane_low<mode>_intrinsic,
32032         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
32033         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
32034         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
32035         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
32036         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
32037         vfmsl_lane_high<mode>_intrinsic): New define_insns.
32039 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32041         * config/arm/arm-cpus.in (fp16fml): New feature.
32042         (ALL_SIMD): Add fp16fml.
32043         (armv8.2-a): Add fp16fml as an option.
32044         (armv8.3-a): Likewise.
32045         (armv8.4-a): Add fp16fml as part of fp16.
32046         * config/arm/arm.h (TARGET_FP16FML): Define.
32047         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
32048         when appropriate.
32049         * config/arm/arm-modes.def (V2HF): Define.
32050         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
32051         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
32052         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
32053         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
32054         vfmsl_low, vfmsl_high): New set of builtins.
32055         * config/arm/iterators.md (PLUSMINUS): New code iterator.
32056         (vfml_op): New code attribute.
32057         (VFMLHALVES): New int iterator.
32058         (VFML, VFMLSEL): New mode attributes.
32059         (V_reg): Define mapping for V2HF.
32060         (V_hi, V_lo): New mode attributes.
32061         (VF_constraint): Likewise.
32062         (vfml_half, vfml_half_selector): New int attributes.
32063         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
32064         define_expand.
32065         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
32066         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
32067         New define_insn.
32068         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
32069         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
32070         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
32071         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
32072         documentation.
32073         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
32074         Document new effective target and option set.
32076 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32078         * config/arm/arm-cpus.in (armv8_4): New feature.
32079         (ARMv8_4a): New fgroup.
32080         (armv8.4-a): New arch.
32081         * config/arm/arm-tables.opt: Regenerate.
32082         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
32083         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
32084         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
32085         Add matching rules for -march=armv8.4-a and extensions.
32086         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
32088 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
32090         PR target/81821
32091         * config/rx/rx.md (BW): New mode attribute.
32092         (sync_lock_test_and_setsi): Add mode suffix to insn output.
32094 2018-01-11  Richard Biener  <rguenther@suse.de>
32096         PR tree-optimization/83435
32097         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
32098         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
32099         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
32101 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32102             Alan Hayward  <alan.hayward@arm.com>
32103             David Sherwood  <david.sherwood@arm.com>
32105         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
32106         field.
32107         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
32108         (aarch64_print_address_internal): Use it to check for a zero offset.
32110 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32111             Alan Hayward  <alan.hayward@arm.com>
32112             David Sherwood  <david.sherwood@arm.com>
32114         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
32115         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
32116         Return a poly_int64 rather than a HOST_WIDE_INT.
32117         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
32118         rather than a HOST_WIDE_INT.
32119         * config/aarch64/aarch64.h (aarch64_frame): Protect with
32120         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
32121         hard_fp_offset, frame_size, initial_adjust, callee_offset and
32122         final_offset from HOST_WIDE_INT to poly_int64.
32123         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
32124         to_constant when getting the number of units in an Advanced SIMD
32125         mode.
32126         (aarch64_builtin_vectorized_function): Check for a constant number
32127         of units.
32128         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
32129         GET_MODE_SIZE.
32130         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
32131         attribute instead of GET_MODE_NUNITS.
32132         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
32133         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
32134         GET_MODE_SIZE for fixed-size registers.
32135         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
32136         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
32137         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
32138         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
32139         (aarch64_print_operand, aarch64_print_address_internal)
32140         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
32141         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
32142         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
32143         Handle polynomial GET_MODE_SIZE.
32144         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
32145         wider than SImode without modification.
32146         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
32147         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
32148         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
32149         passing and returning SVE modes.
32150         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
32151         rather than GEN_INT.
32152         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
32153         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
32154         (aarch64_allocate_and_probe_stack_space): Likewise.
32155         (aarch64_layout_frame): Cope with polynomial offsets.
32156         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
32157         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
32158         polynomial offsets.
32159         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
32160         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
32161         poly_int64 rather than a HOST_WIDE_INT.
32162         (aarch64_get_separate_components, aarch64_process_components)
32163         (aarch64_expand_prologue, aarch64_expand_epilogue)
32164         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
32165         (aarch64_anchor_offset): New function, split out from...
32166         (aarch64_legitimize_address): ...here.
32167         (aarch64_builtin_vectorization_cost): Handle polynomial
32168         TYPE_VECTOR_SUBPARTS.
32169         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
32170         GET_MODE_NUNITS.
32171         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
32172         number of elements from the PARALLEL rather than the mode.
32173         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
32174         rather than GET_MODE_BITSIZE.
32175         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
32176         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
32177         (aarch64_expand_vec_perm_const_1): Handle polynomial
32178         d->perm.length () and d->perm elements.
32179         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
32180         Apply to_constant to d->perm elements.
32181         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
32182         polynomial CONST_VECTOR_NUNITS.
32183         (aarch64_move_pointer): Take amount as a poly_int64 rather
32184         than an int.
32185         (aarch64_progress_pointer): Avoid temporary variable.
32186         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
32187         the mode attribute instead of GET_MODE.
32189 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32190             Alan Hayward  <alan.hayward@arm.com>
32191             David Sherwood  <david.sherwood@arm.com>
32193         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
32194         x exists before using it.
32195         (aarch64_add_constant_internal): Rename to...
32196         (aarch64_add_offset_1): ...this.  Replace regnum with separate
32197         src and dest rtxes.  Handle the case in which they're different,
32198         including when the offset is zero.  Replace scratchreg with an rtx.
32199         Use 2 additions if there is no spare register into which we can
32200         move a 16-bit constant.
32201         (aarch64_add_constant): Delete.
32202         (aarch64_add_offset): Replace reg with separate src and dest
32203         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
32204         Use aarch64_add_offset_1.
32205         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
32206         an rtx rather than an int.  Take the delta as a poly_int64
32207         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
32208         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
32209         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
32210         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
32211         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
32212         and aarch64_add_sp.
32213         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
32214         aarch64_add_constant.
32216 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32218         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
32219         Use scalar_float_mode.
32221 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32223         * config/aarch64/aarch64-simd.md
32224         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
32225         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
32226         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
32227         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
32228         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
32229         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
32230         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
32231         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
32232         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
32233         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
32235 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32237         PR target/83514
32238         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
32239         targ_options->x_arm_arch_string is non NULL.
32241 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
32243         * config/aarch64/aarch64.h
32244         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
32246 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
32248         PR target/82096
32249         * expmed.c (emit_store_flag_force): Swap if const op0
32250         and change VOIDmode to mode of op0.
32252 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32254         PR rtl-optimization/83761
32255         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
32256         than bytes to mode_for_size.
32258 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32260         PR middle-end/83189
32261         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
32262         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
32263         profile.
32265 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32267         PR middle-end/83575
32268         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
32269         when in layout mode.
32270         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
32271         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
32272         partition fixup.
32274 2018-01-10  Michael Collison  <michael.collison@arm.com>
32276         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
32277         * config/aarch64/aarch64-option-extension.def: Add
32278         AARCH64_OPT_EXTENSION of 'fp16fml'.
32279         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32280         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
32281         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
32282         * config/aarch64/constraints.md (Ui7): New constraint.
32283         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
32284         (VFMLA_SEL_W): Ditto.
32285         (f16quad): Ditto.
32286         (f16mac1): Ditto.
32287         (VFMLA16_LOW): New int iterator.
32288         (VFMLA16_HIGH): Ditto.
32289         (UNSPEC_FMLAL): New unspec.
32290         (UNSPEC_FMLSL): Ditto.
32291         (UNSPEC_FMLAL2): Ditto.
32292         (UNSPEC_FMLSL2): Ditto.
32293         (f16mac): New code attribute.
32294         * config/aarch64/aarch64-simd-builtins.def
32295         (aarch64_fmlal_lowv2sf): Ditto.
32296         (aarch64_fmlsl_lowv2sf): Ditto.
32297         (aarch64_fmlalq_lowv4sf): Ditto.
32298         (aarch64_fmlslq_lowv4sf): Ditto.
32299         (aarch64_fmlal_highv2sf): Ditto.
32300         (aarch64_fmlsl_highv2sf): Ditto.
32301         (aarch64_fmlalq_highv4sf): Ditto.
32302         (aarch64_fmlslq_highv4sf): Ditto.
32303         (aarch64_fmlal_lane_lowv2sf): Ditto.
32304         (aarch64_fmlsl_lane_lowv2sf): Ditto.
32305         (aarch64_fmlal_laneq_lowv2sf): Ditto.
32306         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
32307         (aarch64_fmlalq_lane_lowv4sf): Ditto.
32308         (aarch64_fmlsl_lane_lowv4sf): Ditto.
32309         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
32310         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
32311         (aarch64_fmlal_lane_highv2sf): Ditto.
32312         (aarch64_fmlsl_lane_highv2sf): Ditto.
32313         (aarch64_fmlal_laneq_highv2sf): Ditto.
32314         (aarch64_fmlsl_laneq_highv2sf): Ditto.
32315         (aarch64_fmlalq_lane_highv4sf): Ditto.
32316         (aarch64_fmlsl_lane_highv4sf): Ditto.
32317         (aarch64_fmlalq_laneq_highv4sf): Ditto.
32318         (aarch64_fmlsl_laneq_highv4sf): Ditto.
32319         * config/aarch64/aarch64-simd.md:
32320         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
32321         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
32322         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
32323         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
32324         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
32325         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
32326         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
32327         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
32328         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
32329         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
32330         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
32331         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
32332         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
32333         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
32334         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
32335         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
32336         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
32337         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
32338         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
32339         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
32340         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
32341         (vfmlsl_low_u32): Ditto.
32342         (vfmlalq_low_u32): Ditto.
32343         (vfmlslq_low_u32): Ditto.
32344         (vfmlal_high_u32): Ditto.
32345         (vfmlsl_high_u32): Ditto.
32346         (vfmlalq_high_u32): Ditto.
32347         (vfmlslq_high_u32): Ditto.
32348         (vfmlal_lane_low_u32): Ditto.
32349         (vfmlsl_lane_low_u32): Ditto.
32350         (vfmlal_laneq_low_u32): Ditto.
32351         (vfmlsl_laneq_low_u32): Ditto.
32352         (vfmlalq_lane_low_u32): Ditto.
32353         (vfmlslq_lane_low_u32): Ditto.
32354         (vfmlalq_laneq_low_u32): Ditto.
32355         (vfmlslq_laneq_low_u32): Ditto.
32356         (vfmlal_lane_high_u32): Ditto.
32357         (vfmlsl_lane_high_u32): Ditto.
32358         (vfmlal_laneq_high_u32): Ditto.
32359         (vfmlsl_laneq_high_u32): Ditto.
32360         (vfmlalq_lane_high_u32): Ditto.
32361         (vfmlslq_lane_high_u32): Ditto.
32362         (vfmlalq_laneq_high_u32): Ditto.
32363         (vfmlslq_laneq_high_u32): Ditto.
32364         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
32365         (AARCH64_FL_FOR_ARCH8_4): New.
32366         (AARCH64_ISA_F16FML): New ISA flag.
32367         (TARGET_F16FML): New feature flag for fp16fml.
32368         (doc/invoke.texi): Document new fp16fml option.
32370 2018-01-10  Michael Collison  <michael.collison@arm.com>
32372         * config/aarch64/aarch64-builtins.c:
32373         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
32374         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32375         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
32376         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
32377         (AARCH64_ISA_SHA3): New ISA flag.
32378         (TARGET_SHA3): New feature flag for sha3.
32379         * config/aarch64/iterators.md (sha512_op): New int attribute.
32380         (CRYPTO_SHA512): New int iterator.
32381         (UNSPEC_SHA512H): New unspec.
32382         (UNSPEC_SHA512H2): Ditto.
32383         (UNSPEC_SHA512SU0): Ditto.
32384         (UNSPEC_SHA512SU1): Ditto.
32385         * config/aarch64/aarch64-simd-builtins.def
32386         (aarch64_crypto_sha512hqv2di): New builtin.
32387         (aarch64_crypto_sha512h2qv2di): Ditto.
32388         (aarch64_crypto_sha512su0qv2di): Ditto.
32389         (aarch64_crypto_sha512su1qv2di): Ditto.
32390         (aarch64_eor3qv8hi): Ditto.
32391         (aarch64_rax1qv2di): Ditto.
32392         (aarch64_xarqv2di): Ditto.
32393         (aarch64_bcaxqv8hi): Ditto.
32394         * config/aarch64/aarch64-simd.md:
32395         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
32396         (aarch64_crypto_sha512su0qv2di): Ditto.
32397         (aarch64_crypto_sha512su1qv2di): Ditto.
32398         (aarch64_eor3qv8hi): Ditto.
32399         (aarch64_rax1qv2di): Ditto.
32400         (aarch64_xarqv2di): Ditto.
32401         (aarch64_bcaxqv8hi): Ditto.
32402         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
32403         (vsha512h2q_u64): Ditto.
32404         (vsha512su0q_u64): Ditto.
32405         (vsha512su1q_u64): Ditto.
32406         (veor3q_u16): Ditto.
32407         (vrax1q_u64): Ditto.
32408         (vxarq_u64): Ditto.
32409         (vbcaxq_u16): Ditto.
32410         * config/arm/types.md (crypto_sha512): New type attribute.
32411         (crypto_sha3): Ditto.
32412         (doc/invoke.texi): Document new sha3 option.
32414 2018-01-10  Michael Collison  <michael.collison@arm.com>
32416         * config/aarch64/aarch64-builtins.c:
32417         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
32418         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32419         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
32420         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
32421         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
32422         (AARCH64_ISA_SM4): New ISA flag.
32423         (TARGET_SM4): New feature flag for sm4.
32424         * config/aarch64/aarch64-simd-builtins.def
32425         (aarch64_sm3ss1qv4si): Ditto.
32426         (aarch64_sm3tt1aq4si): Ditto.
32427         (aarch64_sm3tt1bq4si): Ditto.
32428         (aarch64_sm3tt2aq4si): Ditto.
32429         (aarch64_sm3tt2bq4si): Ditto.
32430         (aarch64_sm3partw1qv4si): Ditto.
32431         (aarch64_sm3partw2qv4si): Ditto.
32432         (aarch64_sm4eqv4si): Ditto.
32433         (aarch64_sm4ekeyqv4si): Ditto.
32434         * config/aarch64/aarch64-simd.md:
32435         (aarch64_sm3ss1qv4si): Ditto.
32436         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
32437         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
32438         (aarch64_sm4eqv4si): Ditto.
32439         (aarch64_sm4ekeyqv4si): Ditto.
32440         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
32441         (sm3part_op): Ditto.
32442         (CRYPTO_SM3TT): Ditto.
32443         (CRYPTO_SM3PART): Ditto.
32444         (UNSPEC_SM3SS1): New unspec.
32445         (UNSPEC_SM3TT1A): Ditto.
32446         (UNSPEC_SM3TT1B): Ditto.
32447         (UNSPEC_SM3TT2A): Ditto.
32448         (UNSPEC_SM3TT2B): Ditto.
32449         (UNSPEC_SM3PARTW1): Ditto.
32450         (UNSPEC_SM3PARTW2): Ditto.
32451         (UNSPEC_SM4E): Ditto.
32452         (UNSPEC_SM4EKEY): Ditto.
32453         * config/aarch64/constraints.md (Ui2): New constraint.
32454         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
32455         * config/arm/types.md (crypto_sm3): New type attribute.
32456         (crypto_sm4): Ditto.
32457         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
32458         (vsm3tt1aq_u32): Ditto.
32459         (vsm3tt1bq_u32): Ditto.
32460         (vsm3tt2aq_u32): Ditto.
32461         (vsm3tt2bq_u32): Ditto.
32462         (vsm3partw1q_u32): Ditto.
32463         (vsm3partw2q_u32): Ditto.
32464         (vsm4eq_u32): Ditto.
32465         (vsm4ekeyq_u32): Ditto.
32466         (doc/invoke.texi): Document new sm4 option.
32468 2018-01-10  Michael Collison  <michael.collison@arm.com>
32470         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
32471         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
32472         (AARCH64_FL_FOR_ARCH8_4): New.
32473         (AARCH64_FL_V8_4): New flag.
32474         (doc/invoke.texi): Document new armv8.4-a option.
32476 2018-01-10  Michael Collison  <michael.collison@arm.com>
32478         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32479         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
32480         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
32481         * config/aarch64/aarch64-option-extension.def: Add
32482         AARCH64_OPT_EXTENSION of 'sha2'.
32483         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
32484         (crypto): Disable sha2 and aes if crypto disabled.
32485         (crypto): Enable aes and sha2 if enabled.
32486         (simd): Disable sha2 and aes if simd disabled.
32487         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
32488         New flags.
32489         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
32490         (TARGET_SHA2): New feature flag for sha2.
32491         (TARGET_AES): New feature flag for aes.
32492         * config/aarch64/aarch64-simd.md:
32493         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
32494         conditional on TARGET_AES.
32495         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
32496         (aarch64_crypto_sha1hsi): Make pattern conditional
32497         on TARGET_SHA2.
32498         (aarch64_crypto_sha1hv4si): Ditto.
32499         (aarch64_be_crypto_sha1hv4si): Ditto.
32500         (aarch64_crypto_sha1su1v4si): Ditto.
32501         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
32502         (aarch64_crypto_sha1su0v4si): Ditto.
32503         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
32504         (aarch64_crypto_sha256su0v4si): Ditto.
32505         (aarch64_crypto_sha256su1v4si): Ditto.
32506         (doc/invoke.texi): Document new aes and sha2 options.
32508 2018-01-10  Martin Sebor  <msebor@redhat.com>
32510         PR tree-optimization/83781
32511         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
32512         as string arrays.
32514 2018-01-11  Martin Sebor  <msebor@gmail.com>
32515             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32517         PR tree-optimization/83501
32518         PR tree-optimization/81703
32520         * tree-ssa-strlen.c (get_string_cst): Rename...
32521         (get_string_len): ...to this.  Handle global constants.
32522         (handle_char_store): Adjust.
32524 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
32525             Jim Wilson  <jimw@sifive.com>
32527         * config/riscv/riscv-protos.h (riscv_output_return): New.
32528         * config/riscv/riscv.c (struct machine_function): New naked_p field.
32529         (riscv_attribute_table, riscv_output_return),
32530         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
32531         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
32532         (riscv_compute_frame_info): Only compute frame->mask if not a naked
32533         function.
32534         (riscv_expand_prologue): Add early return for naked function.
32535         (riscv_expand_epilogue): Likewise.
32536         (riscv_function_ok_for_sibcall): Return false for naked function.
32537         (riscv_set_current_function): New.
32538         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
32539         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
32540         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
32541         * doc/extend.texi (RISC-V Function Attributes): New.
32543 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
32545         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
32546         check for 128-bit long double before checking TCmode.
32547         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
32548         128-bit long doubles before checking TFmode or TCmode.
32549         (FLOAT128_IBM_P): Likewise.
32551 2018-01-10  Martin Sebor  <msebor@redhat.com>
32553         PR tree-optimization/83671
32554         * builtins.c (c_strlen): Unconditionally return zero for the empty
32555         string.
32556         Use -Warray-bounds for warnings.
32557         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
32558         for non-constant array indices with COMPONENT_REF, arrays of
32559         arrays, and pointers to arrays.
32560         (gimple_fold_builtin_strlen): Determine and set length range for
32561         non-constant character arrays.
32563 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
32565         PR middle-end/81897
32566         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
32567         empty blocks.
32569 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
32571         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
32573 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
32575         PR target/83399
32576         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
32577         VECTOR_MEM_ALTIVEC_OR_VSX_P.
32578         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
32579         indexed_or_indirect_operand predicate.
32580         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
32581         (*vsx_le_perm_load_v8hi): Likewise.
32582         (*vsx_le_perm_load_v16qi): Likewise.
32583         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
32584         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
32585         (*vsx_le_perm_store_v8hi): Likewise.
32586         (*vsx_le_perm_store_v16qi): Likewise.
32587         (eight unnamed splitters): Likewise.
32589 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
32591         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
32592         * config/rs6000/emmintrin.h: Likewise.
32593         * config/rs6000/mmintrin.h: Likewise.
32594         * config/rs6000/xmmintrin.h: Likewise.
32596 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
32598         PR c++/43486
32599         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
32600         "public_flag".
32601         * tree.c (tree_nop_conversion): Return true for location wrapper
32602         nodes.
32603         (maybe_wrap_with_location): New function.
32604         (selftest::check_strip_nops): New function.
32605         (selftest::test_location_wrappers): New function.
32606         (selftest::tree_c_tests): Call it.
32607         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
32608         (maybe_wrap_with_location): New decl.
32609         (EXPR_LOCATION_WRAPPER_P): New macro.
32610         (location_wrapper_p): New inline function.
32611         (tree_strip_any_location_wrapper): New inline function.
32613 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
32615         PR target/83735
32616         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
32617         stack_realign_offset for the largest alignment of stack slot
32618         actually used.
32619         (ix86_find_max_used_stack_alignment): New function.
32620         (ix86_finalize_stack_frame_flags): Use it.  Set
32621         max_used_stack_alignment if we don't realign stack.
32622         * config/i386/i386.h (machine_function): Add
32623         max_used_stack_alignment.
32625 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
32627         * config/arm/arm.opt (-mbranch-cost): New option.
32628         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
32629         account.
32631 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
32633         PR target/83629
32634         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
32635         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
32637 2018-01-10  Richard Biener  <rguenther@suse.de>
32639         PR debug/83765
32640         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
32641         early out so it also covers the case where we have a non-NULL
32642         origin.
32644 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
32646         PR tree-optimization/83753
32647         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
32648         for non-strided grouped accesses if the number of elements is 1.
32650 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32652         PR target/81616
32653         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
32654         * i386.h (TARGET_USE_GATHER): Define.
32655         * x86-tune.def (X86_TUNE_USE_GATHER): New.
32657 2018-01-10  Martin Liska  <mliska@suse.cz>
32659         PR bootstrap/82831
32660         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
32661         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
32662         partitioning.
32663         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
32664         CLEANUP_NO_PARTITIONING is not set.
32666 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
32668         * doc/rtl.texi: Remove documentation of (const ...) wrappers
32669         for vectors, as a partial revert of r254296.
32670         * rtl.h (const_vec_p): Delete.
32671         (const_vec_duplicate_p): Don't test for vector CONSTs.
32672         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
32673         * expmed.c (make_tree): Likewise.
32675         Revert:
32676         * common.md (E, F): Use CONSTANT_P instead of checking for
32677         CONST_VECTOR.
32678         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
32679         checking for CONST_VECTOR.
32681 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
32683         PR middle-end/83575
32684         * predict.c (force_edge_cold): Handle in more sane way edges
32685         with no prediction.
32687 2018-01-09  Carl Love  <cel@us.ibm.com>
32689         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
32690         V4SI, V4SF types.
32691         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
32692         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
32693         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
32694         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
32695         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
32696         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
32697         * config/rs6000/rs6000-protos.h: Add extern defition for
32698         rs6000_generate_float2_double_code.
32699         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
32700         function.
32701         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
32702         (float2_v2df): Add define_expand.
32704 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
32706         PR target/83628
32707         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
32708         op_mode in the force_to_mode call.
32710 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
32712         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
32713         instead of checking each element individually.
32714         (aarch64_evpc_uzp): Likewise.
32715         (aarch64_evpc_zip): Likewise.
32716         (aarch64_evpc_ext): Likewise.
32717         (aarch64_evpc_rev): Likewise.
32718         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
32719         instead of checking each element individually.  Return true without
32720         generating rtl if
32721         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
32722         whether all selected elements come from the same input, instead of
32723         checking each element individually.  Remove calls to gen_rtx_REG,
32724         start_sequence and end_sequence and instead assert that no rtl is
32725         generated.
32727 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
32729         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
32730         order of HIGH and CONST checks.
32732 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
32734         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
32735         if the destination isn't an SSA_NAME.
32737 2018-01-09  Richard Biener  <rguenther@suse.de>
32739         PR tree-optimization/83668
32740         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
32741         move prologue...
32742         (canonicalize_loop_form): ... here, renamed from ...
32743         (canonicalize_loop_closed_ssa_form): ... this and amended to
32744         swap successor edges for loop exit blocks to make us use
32745         the RPO order we need for initial schedule generation.
32747 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
32749         PR tree-optimization/64811
32750         * match.pd: When optimizing comparisons with Inf, avoid
32751         introducing or losing exceptions from comparisons with NaN.
32753 2018-01-09  Martin Liska  <mliska@suse.cz>
32755         PR sanitizer/82517
32756         * asan.c (shadow_mem_size): Add gcc_assert.
32758 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
32760         Don't save registers in main().
32762         PR target/83738
32763         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
32764         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
32765         * config/avr/avr.c (avr_set_current_function): Don't error if
32766         naked, OS_task or OS_main are specified at the same time.
32767         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
32768         OS_main.
32769         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
32770         attribute.
32771         * common/config/avr/avr-common.c (avr_option_optimization_table):
32772         Switch on -mmain-is-OS_task for optimizing compilations.
32774 2018-01-09  Richard Biener  <rguenther@suse.de>
32776         PR tree-optimization/83572
32777         * graphite.c: Include cfganal.h.
32778         (graphite_transform_loops): Connect infinite loops to exit
32779         and remove fake edges at the end.
32781 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
32783         * ipa-inline.c (edge_badness): Revert accidental checkin.
32785 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
32787         PR ipa/80763
32788         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
32789         symbols; not inline clones.
32791 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
32793         PR target/83507
32794         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
32795         hard registers.  Formatting fixes.
32797         PR preprocessor/83722
32798         * gcc.c (try_generate_repro): Pass
32799         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
32800         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
32801         do_report_bug.
32803 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
32804             Kito Cheng  <kito.cheng@gmail.com>
32806         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
32807         (riscv_leaf_function_p): Delete.
32808         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
32810 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32812         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
32813         function.
32814         (do_ifelse): New function.
32815         (do_isel): New function.
32816         (do_sub3): New function.
32817         (do_add3): New function.
32818         (do_load_mask_compare): New function.
32819         (do_overlap_load_compare): New function.
32820         (expand_compare_loop): New function.
32821         (expand_block_compare): Call expand_compare_loop() when appropriate.
32822         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
32823         option description.
32824         (-mblock-compare-inline-loop-limit): New option.
32826 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32828         PR target/83677
32829         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
32830         Reverse order of second and third operands in first alternative.
32831         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
32832         of first and second elements in UNSPEC_VPERMR vector.
32833         (altivec_expand_vec_perm_le): Likewise.
32835 2018-01-08  Jeff Law  <law@redhat.com>
32837         PR rtl-optimizatin/81308
32838         * tree-switch-conversion.c (cfg_altered): New file scoped static.
32839         (process_switch): If group_case_labels makes a change, then set
32840         cfg_altered.
32841         (pass_convert_switch::execute): If a switch is converted, then
32842         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
32844         PR rtl-optimization/81308
32845         * recog.c (split_all_insns): Conditionally cleanup the CFG after
32846         splitting insns.
32848 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
32850         PR target/83663 - Revert r255946
32851         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
32852         generation for cases where splatting a value is not useful.
32853         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
32854         across a vec_duplicate and a paradoxical subreg forming a vector
32855         mode to a vec_concat.
32857 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32859         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
32860         -march=armv8.3-a variants.
32861         * config/arm/t-multilib: Likewise.
32862         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
32864 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32866         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
32867         to generate rtl.
32868         (cceq_ior_compare_complement): Give it a name so I can use it, and
32869         change boolean_or_operator predicate to boolean_operator so it can
32870         be used to generate a crand.
32871         (eqne): New code iterator.
32872         (bd/bd_neg): New code_attrs.
32873         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
32874         a single define_insn.
32875         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
32876         decrement (bdnzt/bdnzf/bdzt/bdzf).
32877         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
32878         with the new names of the branch decrement patterns, and added the
32879         names of the branch decrement conditional patterns.
32881 2018-01-08  Richard Biener  <rguenther@suse.de>
32883         PR tree-optimization/83563
32884         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
32885         cache.
32887 2018-01-08  Richard Biener  <rguenther@suse.de>
32889         PR middle-end/83713
32890         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
32892 2018-01-08  Richard Biener  <rguenther@suse.de>
32894         PR tree-optimization/83685
32895         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
32896         references to abnormals.
32898 2018-01-08  Richard Biener  <rguenther@suse.de>
32900         PR lto/83719
32901         * dwarf2out.c (output_indirect_strings): Handle empty
32902         skeleton_debug_str_hash.
32903         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
32905 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
32907         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
32908         (emit_store_direct): Likewise.
32909         (arc_trampoline_adjust_address): Likewise.
32910         (arc_asm_trampoline_template): New function.
32911         (arc_initialize_trampoline): Use asm_trampoline_template.
32912         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
32913         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
32914         * config/arc/arc.md (flush_icache): Delete pattern.
32916 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
32918         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
32919         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
32920         munaligned-access.
32922 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
32924         PR target/83681
32925         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
32926         by not USED_FOR_TARGET.
32927         (make_pass_resolve_sw_modes): Likewise.
32929 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
32931         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
32932         USED_FOR_TARGET.
32934 2018-01-08  Richard Biener  <rguenther@suse.de>
32936         PR middle-end/83580
32937         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
32939 2018-01-08  Richard Biener  <rguenther@suse.de>
32941         PR middle-end/83517
32942         * match.pd ((t * 2) / 2) -> t): Add missing :c.
32944 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
32946         PR middle-end/81897
32947         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
32948         basic blocks with a small number of successors.
32949         (convert_control_dep_chain_into_preds): Improve handling of
32950         forwarder blocks.
32951         (dump_predicates): Split apart into...
32952         (dump_pred_chain): ...here...
32953         (dump_pred_info): ...and here.
32954         (can_one_predicate_be_invalidated_p): Add debugging printfs.
32955         (can_chain_union_be_invalidated_p): Improve check for invalidation
32956         of paths.
32957         (uninit_uses_cannot_happen): Avoid unnecessary if
32958         convert_control_dep_chain_into_preds yielded nothing.
32960 2018-01-06  Martin Sebor  <msebor@redhat.com>
32962         PR tree-optimization/83640
32963         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
32964         subtracting negative offset from size.
32965         (builtin_access::overlap): Adjust offset bounds of the access to fall
32966         within the size of the object if possible.
32968 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
32970         PR rtl-optimization/83699
32971         * expmed.c (extract_bit_field_1): Restrict the vector usage of
32972         extract_bit_field_as_subreg to cases in which the extracted
32973         value is also a vector.
32975         * lra-constraints.c (process_alt_operands): Test for the equivalence
32976         substitutions when detecting a possible reload cycle.
32978 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
32980         PR debug/83480
32981         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
32982         by default if flag_selective_schedling{,2}.  Formatting fixes.
32984         PR rtl-optimization/83682
32985         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
32986         if it has non-VECTOR_MODE element mode.
32987         (vec_duplicate_p): Likewise.
32989         PR middle-end/83694
32990         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
32991         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
32993 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
32995         PR target/83604
32996         * config/i386/i386-builtin.def
32997         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
32998         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
32999         Require also OPTION_MASK_ISA_AVX512F in addition to
33000         OPTION_MASK_ISA_GFNI.
33001         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
33002         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
33003         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
33004         to OPTION_MASK_ISA_GFNI.
33005         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
33006         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
33007         OPTION_MASK_ISA_AVX512BW.
33008         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
33009         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
33010         addition to OPTION_MASK_ISA_GFNI.
33011         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
33012         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
33013         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
33014         to OPTION_MASK_ISA_GFNI.
33015         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
33016         a requirement for all ISAs rather than any of them with a few
33017         exceptions.
33018         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
33019         processing.
33020         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
33021         bitmasks to be enabled with 3 exceptions, instead of requiring any
33022         enabled ISA with lots of exceptions.
33023         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
33024         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
33025         Change avx512bw in isa attribute to avx512f.
33026         * config/i386/sgxintrin.h: Add license boilerplate.
33027         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
33028         to __AVX512F__ and __AVX512VL to __AVX512VL__.
33029         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
33030         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
33031         defined.
33032         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
33033         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
33034         temporarily sse2 rather than sse if not enabled already.
33036         PR target/83604
33037         * config/i386/sse.md (VI248_VLBW): Rename to ...
33038         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
33039         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
33040         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
33041         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
33042         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
33043         mode iterator instead of VI248_VLBW.
33045 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
33047         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
33048         (record_modified): Skip clobbers; add debug output.
33049         (param_change_prob): Use sreal frequencies.
33051 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
33053         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
33054         punt for user-aligned variables.
33056 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
33058         * tree-chrec.c (chrec_contains_symbols): Return true for
33059         POLY_INT_CST.
33061 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
33063         PR target/82439
33064         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
33065         of (x|y) == x for BICS pattern.
33067 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
33069         PR tree-optimization/83605
33070         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
33071         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
33072         can throw.
33074 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
33076         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
33077         * config/epiphany/rtems.h: New file.
33079 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33080             Uros Bizjak  <ubizjak@gmail.com>
33082         PR target/83554
33083         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
33084         QIreg_operand instead of register_operand predicate.
33085         * config/i386/i386.c (ix86_rop_should_change_byte_p,
33086         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
33087         comments instead of -fmitigate[-_]rop.
33089 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33091         PR bootstrap/81926
33092         * cgraphunit.c (symbol_table::compile): Switch to text_section
33093         before calling assembly_start debug hook.
33094         * run-rtl-passes.c (run_rtl_passes): Likewise.
33095         Include output.h.
33097 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33099         * tree-vrp.c (extract_range_from_binary_expr_1): Check
33100         range_int_cst_p rather than !symbolic_range_p before calling
33101         extract_range_from_multiplicative_op_1.
33103 2018-01-04  Jeff Law  <law@redhat.com>
33105         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
33106         redundant test in assertion.
33108 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33110         * doc/rtl.texi: Document machine_mode wrapper classes.
33112 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33114         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
33115         using tree_to_uhwi.
33117 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33119         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
33120         the VEC_PERM_EXPR fold to fail.
33122 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33124         PR debug/83585
33125         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
33126         to switched_sections.
33128 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33130         PR target/83680
33131         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
33132         test for d.testing.
33134 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
33136         PR target/83387
33137         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
33138         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
33140 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33142         PR debug/83666
33143         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
33144         is BLKmode and bitpos not zero or mode change is needed.
33146 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33148         PR target/83675
33149         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
33150         TARGET_VIS2.
33152 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
33154         PR target/83628
33155         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
33156         instead of MULT rtx.  Update all corresponding splitters.
33157         (*saddl_se): Ditto.
33158         (*ssub<modesuffix>): Ditto.
33159         (*ssubl_se): Ditto.
33160         (*cmp_sadd_di): Update split patterns.
33161         (*cmp_sadd_si): Ditto.
33162         (*cmp_sadd_sidi): Ditto.
33163         (*cmp_ssub_di): Ditto.
33164         (*cmp_ssub_si): Ditto.
33165         (*cmp_ssub_sidi): Ditto.
33166         * config/alpha/predicates.md (const23_operand): New predicate.
33167         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
33168         Look for ASHIFT, not MULT inner operand.
33169         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
33171 2018-01-04  Martin Liska  <mliska@suse.cz>
33173         PR gcov-profile/83669
33174         * gcov.c (output_intermediate_file): Add version to intermediate
33175         gcov file.
33176         * doc/gcov.texi: Document new field 'version' in intermediate
33177         file format. Fix location of '-k' option of gcov command.
33179 2018-01-04  Martin Liska  <mliska@suse.cz>
33181         PR ipa/82352
33182         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
33184 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33186         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
33188 2018-01-03  Martin Sebor  <msebor@redhat.com>
33190         PR tree-optimization/83655
33191         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
33192         checking calls with invalid arguments.
33194 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33196         * tree-vect-stmts.c (vect_get_store_rhs): New function.
33197         (vectorizable_mask_load_store): Delete.
33198         (vectorizable_call): Return false for masked loads and stores.
33199         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
33200         instead of gimple_assign_rhs1.
33201         (vectorizable_load): Handle IFN_MASK_LOAD.
33202         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
33204 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33206         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
33207         split out from..,
33208         (vectorizable_mask_load_store): ...here.
33209         (vectorizable_load): ...and here.
33211 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33213         * tree-vect-stmts.c (vect_build_all_ones_mask)
33214         (vect_build_zero_merge_argument): New functions, split out from...
33215         (vectorizable_load): ...here.
33217 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33219         * tree-vect-stmts.c (vect_check_store_rhs): New function,
33220         split out from...
33221         (vectorizable_mask_load_store): ...here.
33222         (vectorizable_store): ...and here.
33224 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33226         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
33227         split out from...
33228         (vectorizable_mask_load_store): ...here.
33230 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33232         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
33233         (vect_model_store_cost): Take a vec_load_store_type instead of a
33234         vect_def_type.
33235         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
33236         (vect_model_store_cost): Take a vec_load_store_type instead of a
33237         vect_def_type.
33238         (vectorizable_mask_load_store): Update accordingly.
33239         (vectorizable_store): Likewise.
33240         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
33242 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33244         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
33245         IFN_MASK_LOAD calls here rather than...
33246         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
33248 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33249             Alan Hayward  <alan.hayward@arm.com>
33250             David Sherwood  <david.sherwood@arm.com>
33252         * expmed.c (extract_bit_field_1): For vector extracts,
33253         fall back to extract_bit_field_as_subreg if vec_extract
33254         isn't available.
33256 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33257             Alan Hayward  <alan.hayward@arm.com>
33258             David Sherwood  <david.sherwood@arm.com>
33260         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
33261         they are variable or constant sized.
33262         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
33263         slots for constant-sized data.
33265 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33266             Alan Hayward  <alan.hayward@arm.com>
33267             David Sherwood  <david.sherwood@arm.com>
33269         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
33270         handling COND_EXPRs with boolean comparisons, try to find a better
33271         basis for the mask type than the boolean itself.
33273 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33275         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
33276         is calculated and how it can be overridden.
33277         * genmodes.c (max_bitsize_mode_any_mode): New variable.
33278         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
33279         if defined.
33280         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
33281         if nonzero.
33283 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33284             Alan Hayward  <alan.hayward@arm.com>
33285             David Sherwood  <david.sherwood@arm.com>
33287         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
33288         Remove the mode argument.
33289         (aarch64_simd_valid_immediate): Remove the mode and inverse
33290         arguments.
33291         * config/aarch64/iterators.md (bitsize): New iterator.
33292         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
33293         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
33294         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
33295         aarch64_simd_valid_immediate.
33296         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
33297         (aarch64_reg_or_bic_imm): Likewise.
33298         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
33299         with an insn_type enum and msl with a modifier_type enum.
33300         Replace element_width with a scalar_mode.  Change the shift
33301         to unsigned int.  Add constructors for scalar_float_mode and
33302         scalar_int_mode elements.
33303         (aarch64_vect_float_const_representable_p): Delete.
33304         (aarch64_can_const_movi_rtx_p)
33305         (aarch64_simd_scalar_immediate_valid_for_move)
33306         (aarch64_simd_make_constant): Update call to
33307         aarch64_simd_valid_immediate.
33308         (aarch64_advsimd_valid_immediate_hs): New function.
33309         (aarch64_advsimd_valid_immediate): Likewise.
33310         (aarch64_simd_valid_immediate): Remove mode and inverse
33311         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
33312         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
33313         and aarch64_float_const_representable_p on the result.
33314         (aarch64_output_simd_mov_immediate): Remove mode argument.
33315         Update call to aarch64_simd_valid_immediate and use of
33316         simd_immediate_info.
33317         (aarch64_output_scalar_simd_mov_immediate): Update call
33318         accordingly.
33320 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33321             Alan Hayward  <alan.hayward@arm.com>
33322             David Sherwood  <david.sherwood@arm.com>
33324         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
33325         (mode_nunits): Likewise CONST_MODE_NUNITS.
33326         * machmode.def (ADJUST_NUNITS): Document.
33327         * genmodes.c (mode_data::need_nunits_adj): New field.
33328         (blank_mode): Update accordingly.
33329         (adj_nunits): New variable.
33330         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
33331         parameter.
33332         (emit_mode_size_inline): Set need_bytesize_adj for all modes
33333         listed in adj_nunits.
33334         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
33335         listed in adj_nunits.  Don't emit case statements for such modes.
33336         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
33337         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
33338         nothing if adj_nunits is nonnull.
33339         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
33340         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
33341         (emit_mode_fbit): Update use of print_maybe_const_decl.
33342         (emit_move_size): Likewise.  Treat the array as non-const
33343         if adj_nunits.
33344         (emit_mode_adjustments): Handle adj_nunits.
33346 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33348         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
33349         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
33350         (VECTOR_MODES): Use it.
33351         (make_vector_modes): Take the prefix as an argument.
33353 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33354             Alan Hayward  <alan.hayward@arm.com>
33355             David Sherwood  <david.sherwood@arm.com>
33357         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
33358         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
33359         for MODE_VECTOR_BOOL.
33360         * machmode.def (VECTOR_BOOL_MODE): Document.
33361         * genmodes.c (VECTOR_BOOL_MODE): New macro.
33362         (make_vector_bool_mode): New function.
33363         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
33364         MODE_VECTOR_BOOL.
33365         * lto-streamer-in.c (lto_input_mode_table): Likewise.
33366         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
33367         Likewise.
33368         * stor-layout.c (int_mode_for_mode): Likewise.
33369         * tree.c (build_vector_type_for_mode): Likewise.
33370         * varasm.c (output_constant_pool_2): Likewise.
33371         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
33372         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
33373         for MODE_VECTOR_BOOL.
33374         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
33375         of mode class checks.
33376         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
33377         instead of a list of mode class checks.
33378         (expand_vector_scalar_condition): Likewise.
33379         (type_for_widest_vector_mode): Handle BImode as an inner mode.
33381 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33382             Alan Hayward  <alan.hayward@arm.com>
33383             David Sherwood  <david.sherwood@arm.com>
33385         * machmode.h (mode_size): Change from unsigned short to
33386         poly_uint16_pod.
33387         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
33388         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
33389         or if measurement_type is not polynomial.
33390         (fixed_size_mode::includes_p): Check for constant-sized modes.
33391         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
33392         return a poly_uint16 rather than an unsigned short.
33393         (emit_mode_size): Change the type of mode_size from unsigned short
33394         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
33395         (emit_mode_adjustments): Cope with polynomial vector sizes.
33396         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
33397         for GET_MODE_SIZE.
33398         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
33399         for GET_MODE_SIZE.
33400         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
33401         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
33402         * caller-save.c (setup_save_areas): Likewise.
33403         (replace_reg_with_saved_mem): Likewise.
33404         * calls.c (emit_library_call_value_1): Likewise.
33405         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
33406         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
33407         (gen_lowpart_for_combine): Likewise.
33408         * convert.c (convert_to_integer_1): Likewise.
33409         * cse.c (equiv_constant, cse_insn): Likewise.
33410         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
33411         (cselib_subst_to_values): Likewise.
33412         * dce.c (word_dce_process_block): Likewise.
33413         * df-problems.c (df_word_lr_mark_ref): Likewise.
33414         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
33415         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
33416         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
33417         (rtl_for_decl_location): Likewise.
33418         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
33419         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
33420         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
33421         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
33422         (expand_expr_real_1): Likewise.
33423         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
33424         (pad_below): Likewise.
33425         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
33426         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
33427         * ira.c (get_subreg_tracking_sizes): Likewise.
33428         * ira-build.c (ira_create_allocno_objects): Likewise.
33429         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
33430         (ira_sort_regnos_for_alter_reg): Likewise.
33431         * ira-costs.c (record_operand_costs): Likewise.
33432         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
33433         (resolve_simple_move): Likewise.
33434         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
33435         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
33436         (lra_constraints): Likewise.
33437         (CONST_POOL_OK_P): Reject variable-sized modes.
33438         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
33439         (add_pseudo_to_slot, lra_spill): Likewise.
33440         * omp-low.c (omp_clause_aligned_alignment): Likewise.
33441         * optabs-query.c (get_best_extraction_insn): Likewise.
33442         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
33443         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
33444         (expand_mult_highpart, valid_multiword_target_p): Likewise.
33445         * recog.c (offsettable_address_addr_space_p): Likewise.
33446         * regcprop.c (maybe_mode_change): Likewise.
33447         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
33448         * regrename.c (build_def_use): Likewise.
33449         * regstat.c (dump_reg_info): Likewise.
33450         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
33451         (find_reloads, find_reloads_subreg_address): Likewise.
33452         * reload1.c (eliminate_regs_1): Likewise.
33453         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
33454         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
33455         (simplify_binary_operation_1, simplify_subreg): Likewise.
33456         * targhooks.c (default_function_arg_padding): Likewise.
33457         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
33458         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
33459         (verify_gimple_assign_ternary): Likewise.
33460         * tree-inline.c (estimate_move_cost): Likewise.
33461         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33462         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
33463         (get_address_cost_ainc): Likewise.
33464         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
33465         (vect_supportable_dr_alignment): Likewise.
33466         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
33467         (vectorizable_reduction): Likewise.
33468         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
33469         (vectorizable_operation, vectorizable_load): Likewise.
33470         * tree.c (build_same_sized_truth_vector_type): Likewise.
33471         * valtrack.c (cleanup_auto_inc_dec): Likewise.
33472         * var-tracking.c (emit_note_insn_var_location): Likewise.
33473         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
33474         (ADDR_VEC_ALIGN): Likewise.
33476 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33477             Alan Hayward  <alan.hayward@arm.com>
33478             David Sherwood  <david.sherwood@arm.com>
33480         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
33481         unsigned short.
33482         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
33483         or if measurement_type is polynomial.
33484         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
33485         * combine.c (make_extraction): Likewise.
33486         * dse.c (find_shift_sequence): Likewise.
33487         * dwarf2out.c (mem_loc_descriptor): Likewise.
33488         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
33489         (extract_bit_field, extract_low_bits): Likewise.
33490         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
33491         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
33492         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
33493         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
33494         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
33495         * reload.c (find_reloads): Likewise.
33496         * reload1.c (alter_reg): Likewise.
33497         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
33498         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
33499         * tree-if-conv.c (predicate_mem_writes): Likewise.
33500         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
33501         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
33502         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
33503         * valtrack.c (dead_debug_insert_temp): Likewise.
33504         * varasm.c (mergeable_constant_section): Likewise.
33505         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
33507 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33508             Alan Hayward  <alan.hayward@arm.com>
33509             David Sherwood  <david.sherwood@arm.com>
33511         * expr.c (expand_assignment): Cope with polynomial mode sizes
33512         when assigning to a CONCAT.
33514 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33515             Alan Hayward  <alan.hayward@arm.com>
33516             David Sherwood  <david.sherwood@arm.com>
33518         * machmode.h (mode_precision): Change from unsigned short to
33519         poly_uint16_pod.
33520         (mode_to_precision): Return a poly_uint16 rather than an unsigned
33521         short.
33522         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
33523         or if measurement_type is not polynomial.
33524         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
33525         in which the mode is already known to be a scalar_int_mode.
33526         * genmodes.c (emit_mode_precision): Change the type of mode_precision
33527         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
33528         initializer.
33529         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
33530         for GET_MODE_PRECISION.
33531         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
33532         for GET_MODE_PRECISION.
33533         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
33534         as polynomial.
33535         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
33536         (expand_field_assignment, make_extraction): Likewise.
33537         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
33538         (get_last_value): Likewise.
33539         * convert.c (convert_to_integer_1): Likewise.
33540         * cse.c (cse_insn): Likewise.
33541         * expr.c (expand_expr_real_1): Likewise.
33542         * lra-constraints.c (simplify_operand_subreg): Likewise.
33543         * optabs-query.c (can_atomic_load_p): Likewise.
33544         * optabs.c (expand_atomic_load): Likewise.
33545         (expand_atomic_store): Likewise.
33546         * ree.c (combine_reaching_defs): Likewise.
33547         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
33548         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
33549         * tree.h (type_has_mode_precision_p): Likewise.
33550         * ubsan.c (instrument_si_overflow): Likewise.
33552 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33553             Alan Hayward  <alan.hayward@arm.com>
33554             David Sherwood  <david.sherwood@arm.com>
33556         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
33557         polynomial numbers of units.
33558         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
33559         (valid_vector_subparts_p): New function.
33560         (build_vector_type): Remove temporary shim and take the number
33561         of units as a poly_uint64 rather than an int.
33562         (build_opaque_vector_type): Take the number of units as a
33563         poly_uint64 rather than an int.
33564         * tree.c (build_vector_from_ctor): Handle polynomial
33565         TYPE_VECTOR_SUBPARTS.
33566         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
33567         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
33568         (build_vector_from_val): If the number of units is variable,
33569         use build_vec_duplicate_cst for constant operands and
33570         VEC_DUPLICATE_EXPR otherwise.
33571         (make_vector_type): Remove temporary is_constant ().
33572         (build_vector_type, build_opaque_vector_type): Take the number of
33573         units as a poly_uint64 rather than an int.
33574         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
33575         VECTOR_CST_NELTS.
33576         * cfgexpand.c (expand_debug_expr): Likewise.
33577         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
33578         (store_constructor, expand_expr_real_1): Likewise.
33579         (const_scalar_mask_from_tree): Likewise.
33580         * fold-const-call.c (fold_const_reduction): Likewise.
33581         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
33582         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
33583         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
33584         (fold_relational_const): Likewise.
33585         (native_interpret_vector): Likewise.  Change the size from an
33586         int to an unsigned int.
33587         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
33588         TYPE_VECTOR_SUBPARTS.
33589         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
33590         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
33591         duplicating a non-constant operand into a variable-length vector.
33592         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
33593         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
33594         * ipa-icf.c (sem_variable::equals): Likewise.
33595         * match.pd: Likewise.
33596         * omp-simd-clone.c (simd_clone_subparts): Likewise.
33597         * print-tree.c (print_node): Likewise.
33598         * stor-layout.c (layout_type): Likewise.
33599         * targhooks.c (default_builtin_vectorization_cost): Likewise.
33600         * tree-cfg.c (verify_gimple_comparison): Likewise.
33601         (verify_gimple_assign_binary): Likewise.
33602         (verify_gimple_assign_ternary): Likewise.
33603         (verify_gimple_assign_single): Likewise.
33604         * tree-pretty-print.c (dump_generic_node): Likewise.
33605         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33606         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
33607         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
33608         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
33609         (vect_shift_permute_load_chain): Likewise.
33610         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
33611         (expand_vector_condition, optimize_vector_constructor): Likewise.
33612         (lower_vec_perm, get_compute_type): Likewise.
33613         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
33614         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
33615         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
33616         (vect_recog_mask_conversion_pattern): Likewise.
33617         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
33618         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
33619         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33620         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
33621         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
33622         (vectorizable_shift, vectorizable_operation, vectorizable_store)
33623         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
33624         (supportable_widening_operation): Likewise.
33625         (supportable_narrowing_operation): Likewise.
33626         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
33627         Likewise.
33628         * varasm.c (output_constant): Likewise.
33630 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33631             Alan Hayward  <alan.hayward@arm.com>
33632             David Sherwood  <david.sherwood@arm.com>
33634         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
33635         so that both the length == 3 and length != 3 cases set up their
33636         own permute vectors.  Add comments explaining why we know the
33637         number of elements is constant.
33638         (vect_permute_load_chain): Likewise.
33640 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33641             Alan Hayward  <alan.hayward@arm.com>
33642             David Sherwood  <david.sherwood@arm.com>
33644         * machmode.h (mode_nunits): Change from unsigned char to
33645         poly_uint16_pod.
33646         (ONLY_FIXED_SIZE_MODES): New macro.
33647         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
33648         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
33649         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
33650         New typedefs.
33651         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
33652         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
33653         or if measurement_type is not polynomial.
33654         * genmodes.c (ZERO_COEFFS): New macro.
33655         (emit_mode_nunits_inline): Make mode_nunits_inline return a
33656         poly_uint16.
33657         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
33658         Use ZERO_COEFFS when emitting initializers.
33659         * data-streamer.h (bp_pack_poly_value): New function.
33660         (bp_unpack_poly_value): Likewise.
33661         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
33662         for GET_MODE_NUNITS.
33663         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
33664         for GET_MODE_NUNITS.
33665         * tree.c (make_vector_type): Remove temporary shim and make
33666         the real function take the number of units as a poly_uint64
33667         rather than an int.
33668         (build_vector_type_for_mode): Handle polynomial nunits.
33669         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
33670         * emit-rtl.c (const_vec_series_p_1): Likewise.
33671         (gen_rtx_CONST_VECTOR): Likewise.
33672         * fold-const.c (test_vec_duplicate_folding): Likewise.
33673         * genrecog.c (validate_pattern): Likewise.
33674         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
33675         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
33676         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
33677         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
33678         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
33679         * rtlanal.c (subreg_get_info): Likewise.
33680         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33681         (vect_grouped_load_supported): Likewise.
33682         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
33683         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
33684         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
33685         (simplify_const_unary_operation, simplify_binary_operation_1)
33686         (simplify_const_binary_operation, simplify_ternary_operation)
33687         (test_vector_ops_duplicate, test_vector_ops): Likewise.
33688         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
33689         instead of CONST_VECTOR_NUNITS.
33690         * varasm.c (output_constant_pool_2): Likewise.
33691         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
33692         explicit-encoded elements in the XVEC for variable-length vectors.
33694 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33696         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
33698 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33699             Alan Hayward  <alan.hayward@arm.com>
33700             David Sherwood  <david.sherwood@arm.com>
33702         * coretypes.h (fixed_size_mode): Declare.
33703         (fixed_size_mode_pod): New typedef.
33704         * builtins.h (target_builtins::x_apply_args_mode)
33705         (target_builtins::x_apply_result_mode): Change type to
33706         fixed_size_mode_pod.
33707         * builtins.c (apply_args_size, apply_result_size, result_vector)
33708         (expand_builtin_apply_args_1, expand_builtin_apply)
33709         (expand_builtin_return): Update accordingly.
33711 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33713         * cse.c (hash_rtx_cb): Hash only the encoded elements.
33714         * cselib.c (cselib_hash_rtx): Likewise.
33715         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
33716         CONST_VECTOR encoding.
33718 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33719             Jeff Law  <law@redhat.com>
33721         PR target/83641
33722         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
33723         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
33724         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
33725         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
33727         PR target/83641
33728         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
33729         explicitly probe *sp in a noreturn function if there were any callee
33730         register saves or frame pointer is needed.
33732 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33734         PR debug/83621
33735         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
33736         BLKmode for ternary, binary or unary expressions.
33738         PR debug/83645
33739         * var-tracking.c (delete_vta_debug_insn): New inline function.
33740         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
33741         insns from get_insns () to NULL instead of each bb separately.
33742         Use delete_vta_debug_insn.  No longer static.
33743         (vt_debug_insns_local, variable_tracking_main_1): Adjust
33744         delete_vta_debug_insns callers.
33745         * rtl.h (delete_vta_debug_insns): Declare.
33746         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
33747         instead of variable_tracking_main.
33749 2018-01-03  Martin Sebor  <msebor@redhat.com>
33751         PR tree-optimization/83603
33752         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
33753         arguments past the endof the argument list in functions declared
33754         without a prototype.
33755         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
33756         Avoid checking when arguments are null.
33758 2018-01-03  Martin Sebor  <msebor@redhat.com>
33760         PR c/83559
33761         * doc/extend.texi (attribute const): Fix a typo.
33762         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
33763         issuing -Wsuggest-attribute for void functions.
33765 2018-01-03  Martin Sebor  <msebor@redhat.com>
33767         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
33768         offset_int::from instead of wide_int::to_shwi.
33769         (maybe_diag_overlap): Remove assertion.
33770         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
33771         * gimple-ssa-sprintf.c (format_directive): Same.
33772         (parse_directive): Same.
33773         (sprintf_dom_walker::compute_format_length): Same.
33774         (try_substitute_return_value): Same.
33776 2018-01-03  Jeff Law  <law@redhat.com>
33778         PR middle-end/83654
33779         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
33780         non-constant residual for zero at runtime and avoid probing in
33781         that case.  Reorganize code for trailing problem to mirror handling
33782         of the residual.
33784 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33786         PR tree-optimization/83501
33787         * tree-ssa-strlen.c (get_string_cst): New.
33788         (handle_char_store): Call get_string_cst.
33790 2018-01-03  Martin Liska  <mliska@suse.cz>
33792         PR tree-optimization/83593
33793         * tree-ssa-strlen.c: Include tree-cfg.h.
33794         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
33795         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
33796         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
33797         to false.
33798         (strlen_dom_walker::before_dom_children): Call
33799         gimple_purge_dead_eh_edges. Dump tranformation with details
33800         dump flags.
33801         (strlen_dom_walker::before_dom_children): Update call by adding
33802         new argument cleanup_eh.
33803         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
33805 2018-01-03  Martin Liska  <mliska@suse.cz>
33807         PR ipa/83549
33808         * cif-code.def (VARIADIC_THUNK): New enum value.
33809         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
33810         thunks.
33812 2018-01-03  Jan Beulich  <jbeulich@suse.com>
33814         * sse.md (mov<mode>_internal): Tighten condition for when to use
33815         vmovdqu<ssescalarsize> for TI and OI modes.
33817 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33819         Update copyright years.
33821 2018-01-03  Martin Liska  <mliska@suse.cz>
33823         PR ipa/83594
33824         * ipa-visibility.c (function_and_variable_visibility): Skip
33825         functions with noipa attribure.
33827 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33829         * gcc.c (process_command): Update copyright notice dates.
33830         * gcov-dump.c (print_version): Ditto.
33831         * gcov.c (print_version): Ditto.
33832         * gcov-tool.c (print_version): Ditto.
33833         * gengtype.c (create_file): Ditto.
33834         * doc/cpp.texi: Bump @copying's copyright year.
33835         * doc/cppinternals.texi: Ditto.
33836         * doc/gcc.texi: Ditto.
33837         * doc/gccint.texi: Ditto.
33838         * doc/gcov.texi: Ditto.
33839         * doc/install.texi: Ditto.
33840         * doc/invoke.texi: Ditto.
33842 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33844         * vector-builder.h (vector_builder::m_full_nelts): Change from
33845         unsigned int to poly_uint64.
33846         (vector_builder::full_nelts): Update prototype accordingly.
33847         (vector_builder::new_vector): Likewise.
33848         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
33849         (vector_builder::operator ==): Likewise.
33850         (vector_builder::finalize): Likewise.
33851         * int-vector-builder.h (int_vector_builder::int_vector_builder):
33852         Take the number of elements as a poly_uint64 rather than an
33853         unsigned int.
33854         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
33855         from unsigned int to poly_uint64.
33856         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
33857         (vec_perm_indices::new_vector): Likewise.
33858         (vec_perm_indices::length): Likewise.
33859         (vec_perm_indices::nelts_per_input): Likewise.
33860         (vec_perm_indices::input_nelts): Likewise.
33861         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
33862         number of elements per input as a poly_uint64 rather than an
33863         unsigned int.  Use the original encoding for variable-length
33864         vectors, rather than clamping each individual element.
33865         For the second and subsequent elements in each pattern,
33866         clamp the step and base before clamping their sum.
33867         (vec_perm_indices::series_p): Handle polynomial element counts.
33868         (vec_perm_indices::all_in_range_p): Likewise.
33869         (vec_perm_indices_to_tree): Likewise.
33870         (vec_perm_indices_to_rtx): Likewise.
33871         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
33872         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
33873         (tree_vector_builder::new_binary_operation): Handle polynomial
33874         element counts.  Return false if we need to know the number
33875         of elements at compile time.
33876         * fold-const.c (fold_vec_perm): Punt if the number of elements
33877         isn't known at compile time.
33879 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33881         * vec-perm-indices.h (vec_perm_builder): Change element type
33882         from HOST_WIDE_INT to poly_int64.
33883         (vec_perm_indices::element_type): Update accordingly.
33884         (vec_perm_indices::clamp): Handle polynomial element_types.
33885         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
33886         (vec_perm_indices::all_in_range_p): Likewise.
33887         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
33888         than shwi trees.
33889         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
33890         polynomial vec_perm_indices element types.
33891         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
33892         * fold-const.c (fold_vec_perm): Likewise.
33893         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
33894         * tree-vect-generic.c (lower_vec_perm): Likewise.
33895         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
33896         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
33897         element type to HOST_WIDE_INT.
33899 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33900             Alan Hayward  <alan.hayward@arm.com>
33901             David Sherwood  <david.sherwood@arm.com>
33903         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
33904         rather than an int.  Use plus_constant.
33905         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
33906         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
33908 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33909             Alan Hayward  <alan.hayward@arm.com>
33910             David Sherwood  <david.sherwood@arm.com>
33912         * calls.c (emit_call_1, expand_call): Change struct_value_size from
33913         a HOST_WIDE_INT to a poly_int64.
33915 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33916             Alan Hayward  <alan.hayward@arm.com>
33917             David Sherwood  <david.sherwood@arm.com>
33919         * calls.c (load_register_parameters): Cope with polynomial
33920         mode sizes.  Require a constant size for BLKmode parameters
33921         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
33922         forces a parameter to be padded at the lsb end in order to
33923         fill a complete number of words, require the parameter size
33924         to be ordered wrt UNITS_PER_WORD.
33926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33927             Alan Hayward  <alan.hayward@arm.com>
33928             David Sherwood  <david.sherwood@arm.com>
33930         * reload1.c (spill_stack_slot_width): Change element type
33931         from unsigned int to poly_uint64_pod.
33932         (alter_reg): Treat mode sizes as polynomial.
33934 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33935             Alan Hayward  <alan.hayward@arm.com>
33936             David Sherwood  <david.sherwood@arm.com>
33938         * reload.c (complex_word_subreg_p): New function.
33939         (reload_inner_reg_of_subreg, push_reload): Use it.
33941 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33942             Alan Hayward  <alan.hayward@arm.com>
33943             David Sherwood  <david.sherwood@arm.com>
33945         * lra-constraints.c (process_alt_operands): Reject matched
33946         operands whose sizes aren't ordered.
33947         (match_reload): Refer to this check here.
33949 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33950             Alan Hayward  <alan.hayward@arm.com>
33951             David Sherwood  <david.sherwood@arm.com>
33953         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
33954         that the mode size is in the set {1, 2, 4, 8, 16}.
33956 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33957             Alan Hayward  <alan.hayward@arm.com>
33958             David Sherwood  <david.sherwood@arm.com>
33960         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
33961         Use plus_constant instead of gen_rtx_PLUS.
33963 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33964             Alan Hayward  <alan.hayward@arm.com>
33965             David Sherwood  <david.sherwood@arm.com>
33967         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
33968         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
33969         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
33970         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
33971         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
33972         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
33973         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
33974         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
33975         * config/i386/i386.c (ix86_push_rounding): ...this new function.
33976         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
33977         a poly_int64.
33978         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
33979         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
33980         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
33981         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
33982         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
33983         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
33984         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
33985         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
33986         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
33987         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
33988         function.
33989         * expr.c (emit_move_resolve_push): Treat the input and result
33990         of PUSH_ROUNDING as a poly_int64.
33991         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
33992         (emit_push_insn): Likewise.
33993         * lra-eliminations.c (mark_not_eliminable): Likewise.
33994         * recog.c (push_operand): Likewise.
33995         * reload1.c (elimination_effects): Likewise.
33996         * rtlanal.c (nonzero_bits1): Likewise.
33997         * calls.c (store_one_arg): Likewise.  Require the padding to be
33998         known at compile time.
34000 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34001             Alan Hayward  <alan.hayward@arm.com>
34002             David Sherwood  <david.sherwood@arm.com>
34004         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
34005         Use plus_constant instead of gen_rtx_PLUS.
34007 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34008             Alan Hayward  <alan.hayward@arm.com>
34009             David Sherwood  <david.sherwood@arm.com>
34011         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
34012         rather than an int.
34014 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34015             Alan Hayward  <alan.hayward@arm.com>
34016             David Sherwood  <david.sherwood@arm.com>
34018         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
34019         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
34020         via stack temporaries.  Treat the mode size as polynomial too.
34022 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34023             Alan Hayward  <alan.hayward@arm.com>
34024             David Sherwood  <david.sherwood@arm.com>
34026         * expr.c (expand_expr_real_2): When handling conversions involving
34027         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
34028         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
34029         as a poly_uint64 too.
34031 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34032             Alan Hayward  <alan.hayward@arm.com>
34033             David Sherwood  <david.sherwood@arm.com>
34035         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
34037 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34038             Alan Hayward  <alan.hayward@arm.com>
34039             David Sherwood  <david.sherwood@arm.com>
34041         * combine.c (can_change_dest_mode): Handle polynomial
34042         REGMODE_NATURAL_SIZE.
34043         * expmed.c (store_bit_field_1): Likewise.
34044         * expr.c (store_constructor): Likewise.
34045         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
34046         and polynomial REGMODE_NATURAL_SIZE.
34047         (gen_lowpart_common): Likewise.
34048         * reginfo.c (record_subregs_of_mode): Likewise.
34049         * rtlanal.c (read_modify_subreg_p): Likewise.
34051 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34052             Alan Hayward  <alan.hayward@arm.com>
34053             David Sherwood  <david.sherwood@arm.com>
34055         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
34056         numbers of elements.
34058 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34059             Alan Hayward  <alan.hayward@arm.com>
34060             David Sherwood  <david.sherwood@arm.com>
34062         * match.pd: Cope with polynomial numbers of vector elements.
34064 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34065             Alan Hayward  <alan.hayward@arm.com>
34066             David Sherwood  <david.sherwood@arm.com>
34068         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
34069         in a POINTER_PLUS_EXPR.
34071 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34072             Alan Hayward  <alan.hayward@arm.com>
34073             David Sherwood  <david.sherwood@arm.com>
34075         * omp-simd-clone.c (simd_clone_subparts): New function.
34076         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
34077         (ipa_simd_modify_function_body): Likewise.
34079 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34080             Alan Hayward  <alan.hayward@arm.com>
34081             David Sherwood  <david.sherwood@arm.com>
34083         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
34084         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
34085         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
34086         (expand_vector_condition, vector_element): Likewise.
34087         (subparts_gt): New function.
34088         (get_compute_type): Use subparts_gt.
34089         (count_type_subparts): Delete.
34090         (expand_vector_operations_1): Use subparts_gt instead of
34091         count_type_subparts.
34093 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34094             Alan Hayward  <alan.hayward@arm.com>
34095             David Sherwood  <david.sherwood@arm.com>
34097         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
34098         (vect_compile_time_alias): ...this new function.  Do the calculation
34099         on poly_ints rather than trees.
34100         (vect_prune_runtime_alias_test_list): Update call accordingly.
34102 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34103             Alan Hayward  <alan.hayward@arm.com>
34104             David Sherwood  <david.sherwood@arm.com>
34106         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
34107         numbers of units.
34108         (vect_schedule_slp_instance): Likewise.
34110 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34111             Alan Hayward  <alan.hayward@arm.com>
34112             David Sherwood  <david.sherwood@arm.com>
34114         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
34115         constant and extern definitions for variable-length vectors.
34116         (vect_get_constant_vectors): Note that the number of units
34117         is known to be constant.
34119 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34120             Alan Hayward  <alan.hayward@arm.com>
34121             David Sherwood  <david.sherwood@arm.com>
34123         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
34124         of units as polynomial.  Choose between WIDE and NARROW based
34125         on multiple_p.
34127 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34128             Alan Hayward  <alan.hayward@arm.com>
34129             David Sherwood  <david.sherwood@arm.com>
34131         * tree-vect-stmts.c (simd_clone_subparts): New function.
34132         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
34134 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34135             Alan Hayward  <alan.hayward@arm.com>
34136             David Sherwood  <david.sherwood@arm.com>
34138         * tree-vect-stmts.c (vectorizable_call): Treat the number of
34139         vectors as polynomial.  Use build_index_vector for
34140         IFN_GOMP_SIMD_LANE.
34142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34143             Alan Hayward  <alan.hayward@arm.com>
34144             David Sherwood  <david.sherwood@arm.com>
34146         * tree-vect-stmts.c (get_load_store_type): Treat the number of
34147         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
34148         for variable-length vectors.
34149         (vectorizable_mask_load_store): Treat the number of units as
34150         polynomial, asserting that it is constant if the condition has
34151         already been enforced.
34152         (vectorizable_store, vectorizable_load): Likewise.
34154 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34155             Alan Hayward  <alan.hayward@arm.com>
34156             David Sherwood  <david.sherwood@arm.com>
34158         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
34159         of units as polynomial.  Punt if we can't tell at compile time
34160         which vector contains the final result.
34162 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34163             Alan Hayward  <alan.hayward@arm.com>
34164             David Sherwood  <david.sherwood@arm.com>
34166         * tree-vect-loop.c (vectorizable_induction): Treat the number
34167         of units as polynomial.  Punt on SLP inductions.  Use an integer
34168         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
34169         cast of such a series for variable-length floating-point
34170         reductions.
34172 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34173             Alan Hayward  <alan.hayward@arm.com>
34174             David Sherwood  <david.sherwood@arm.com>
34176         * tree.h (build_index_vector): Declare.
34177         * tree.c (build_index_vector): New function.
34178         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
34179         of units as polynomial, forcibly converting it to a constant if
34180         vectorizable_reduction has already enforced the condition.
34181         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
34182         to create a {1,2,3,...} vector.
34183         (vectorizable_reduction): Treat the number of units as polynomial.
34184         Choose vectype_in based on the largest scalar element size rather
34185         than the smallest number of units.  Enforce the restrictions
34186         relied on above.
34188 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34189             Alan Hayward  <alan.hayward@arm.com>
34190             David Sherwood  <david.sherwood@arm.com>
34192         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
34193         number of units as polynomial.
34195 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34196             Alan Hayward  <alan.hayward@arm.com>
34197             David Sherwood  <david.sherwood@arm.com>
34199         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
34200         * target.def (autovectorize_vector_sizes): Return the vector sizes
34201         by pointer, using vector_sizes rather than a bitmask.
34202         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
34203         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
34204         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
34205         Likewise.
34206         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
34207         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
34208         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
34209         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
34210         * omp-general.c (omp_max_vf): Likewise.
34211         * omp-low.c (omp_clause_aligned_alignment): Likewise.
34212         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
34213         * tree-vect-loop.c (vect_analyze_loop): Likewise.
34214         * tree-vect-slp.c (vect_slp_bb): Likewise.
34215         * doc/tm.texi: Regenerate.
34216         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
34217         to a poly_uint64.
34218         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
34219         the vector size as a poly_uint64 rather than an unsigned int.
34220         (current_vector_size): Change from an unsigned int to a poly_uint64.
34221         (get_vectype_for_scalar_type): Update accordingly.
34222         * tree.h (build_truth_vector_type): Take the size and number of
34223         units as a poly_uint64 rather than an unsigned int.
34224         (build_vector_type): Add a temporary overload that takes
34225         the number of units as a poly_uint64 rather than an unsigned int.
34226         * tree.c (make_vector_type): Likewise.
34227         (build_truth_vector_type): Take the number of units as a poly_uint64
34228         rather than an unsigned int.
34230 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34231             Alan Hayward  <alan.hayward@arm.com>
34232             David Sherwood  <david.sherwood@arm.com>
34234         * target.def (get_mask_mode): Take the number of units and length
34235         as poly_uint64s rather than unsigned ints.
34236         * targhooks.h (default_get_mask_mode): Update accordingly.
34237         * targhooks.c (default_get_mask_mode): Likewise.
34238         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
34239         * doc/tm.texi: Regenerate.
34241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34242             Alan Hayward  <alan.hayward@arm.com>
34243             David Sherwood  <david.sherwood@arm.com>
34245         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
34246         * omp-general.c (omp_max_vf): Likewise.
34247         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
34248         (expand_omp_simd): Handle polynomial safelen.
34249         * omp-low.c (omplow_simd_context): Add a default constructor.
34250         (omplow_simd_context::max_vf): Change from int to poly_uint64.
34251         (lower_rec_simd_input_clauses): Update accordingly.
34252         (lower_rec_input_clauses): Likewise.
34254 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34255             Alan Hayward  <alan.hayward@arm.com>
34256             David Sherwood  <david.sherwood@arm.com>
34258         * tree-vectorizer.h (vect_nunits_for_cost): New function.
34259         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
34260         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
34261         (vect_analyze_slp_cost): Likewise.
34262         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
34263         (vect_model_load_cost): Likewise.
34265 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34266             Alan Hayward  <alan.hayward@arm.com>
34267             David Sherwood  <david.sherwood@arm.com>
34269         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
34270         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
34271         from an unsigned int * to a poly_uint64_pod *.
34272         (calculate_unrolling_factor): New function.
34273         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
34275 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34276             Alan Hayward  <alan.hayward@arm.com>
34277             David Sherwood  <david.sherwood@arm.com>
34279         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
34280         from an unsigned int to a poly_uint64.
34281         (_loop_vec_info::slp_unrolling_factor): Likewise.
34282         (_loop_vec_info::vectorization_factor): Change from an int
34283         to a poly_uint64.
34284         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
34285         (vect_get_num_vectors): New function.
34286         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
34287         (vect_get_num_copies): Use vect_get_num_vectors.
34288         (vect_analyze_data_ref_dependences): Change max_vf from an int *
34289         to an unsigned int *.
34290         (vect_analyze_data_refs): Change min_vf from an int * to a
34291         poly_uint64 *.
34292         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
34293         than an unsigned HOST_WIDE_INT.
34294         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
34295         (vect_analyze_data_ref_dependence): Change max_vf from an int *
34296         to an unsigned int *.
34297         (vect_analyze_data_ref_dependences): Likewise.
34298         (vect_compute_data_ref_alignment): Handle polynomial vf.
34299         (vect_enhance_data_refs_alignment): Likewise.
34300         (vect_prune_runtime_alias_test_list): Likewise.
34301         (vect_shift_permute_load_chain): Likewise.
34302         (vect_supportable_dr_alignment): Likewise.
34303         (dependence_distance_ge_vf): Take the vectorization factor as a
34304         poly_uint64 rather than an unsigned HOST_WIDE_INT.
34305         (vect_analyze_data_refs): Change min_vf from an int * to a
34306         poly_uint64 *.
34307         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
34308         vfm1 as a poly_uint64 rather than an int.  Make the same change
34309         for the returned bound_scalar.
34310         (vect_gen_vector_loop_niters): Handle polynomial vf.
34311         (vect_do_peeling): Likewise.  Update call to
34312         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
34313         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
34314         be constant.
34315         * tree-vect-loop.c (vect_determine_vectorization_factor)
34316         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
34317         (vect_get_known_peeling_cost): Likewise.
34318         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
34319         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
34320         (vect_transform_loop): Likewise.  Use the lowest possible VF when
34321         updating the upper bounds of the loop.
34322         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
34323         rather than an int.
34324         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
34325         polynomial unroll factors.
34326         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
34327         (vect_make_slp_decision): Likewise.
34328         (vect_supported_load_permutation_p): Likewise, and polynomial
34329         vf too.
34330         (vect_analyze_slp_cost): Handle polynomial vf.
34331         (vect_slp_analyze_node_operations): Likewise.
34332         (vect_slp_analyze_bb_1): Likewise.
34333         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
34334         than an unsigned HOST_WIDE_INT.
34335         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
34336         (vectorizable_load): Handle polynomial vf.
34337         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
34338         a poly_uint64.
34339         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
34341 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34342             Alan Hayward  <alan.hayward@arm.com>
34343             David Sherwood  <david.sherwood@arm.com>
34345         * match.pd: Handle bit operations involving three constants
34346         and try to fold one pair.
34348 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34350         * tree-vect-loop-manip.c: Include gimple-fold.h.
34351         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
34352         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
34353         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
34354         Add a path that uses a step of VF instead of 1, but disable it
34355         for now.
34356         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
34357         and niters_no_overflow parameters.  Update calls to
34358         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
34359         Create a new SSA name if the latter choses to use a ste other
34360         than zero, and return it via niters_vector_mult_vf_var.
34361         * tree-vect-loop.c (vect_transform_loop): Update calls to
34362         vect_do_peeling, vect_gen_vector_loop_niters and
34363         slpeel_make_loop_iterate_ntimes.
34364         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
34365         (vect_gen_vector_loop_niters): Update declarations after above changes.
34367 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
34369         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
34370         128-bit round to integer instructions.
34371         (ceil<mode>2): Likewise.
34372         (btrunc<mode>2): Likewise.
34373         (round<mode>2): Likewise.
34375 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34377         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
34378         unaligned VSX load/store on P8/P9.
34379         (expand_block_clear): Allow the use of unaligned VSX
34380         load/store on P8/P9.
34382 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34384         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
34385         New function.
34386         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
34387         swap associated with both a load and a store.
34389 2018-01-02  Andrew Waterman  <andrew@sifive.com>
34391         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
34392         * config/riscv/riscv.md (clear_cache): Use it.
34394 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
34396         * web.c: Remove out-of-date comment.
34398 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34400         * expr.c (fixup_args_size_notes): Check that any existing
34401         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
34402         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
34403         (emit_single_push_insn): ...here.
34405 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34407         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
34408         (const_vector_encoded_nelts): New function.
34409         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
34410         (const_vector_int_elt, const_vector_elt): Declare.
34411         * emit-rtl.c (const_vector_int_elt_1): New function.
34412         (const_vector_elt): Likewise.
34413         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
34414         of CONST_VECTOR_ELT.
34416 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34418         * expr.c: Include rtx-vector-builder.h.
34419         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
34420         directly on the tree encoding.
34421         (const_vector_from_tree): Likewise.
34422         * optabs.c: Include rtx-vector-builder.h.
34423         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
34424         sequence of "u" values.
34425         * vec-perm-indices.c: Include rtx-vector-builder.h.
34426         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
34427         directly on the vec_perm_indices encoding.
34429 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34431         * doc/rtl.texi (const_vector): Describe new encoding scheme.
34432         * Makefile.in (OBJS): Add rtx-vector-builder.o.
34433         * rtx-vector-builder.h: New file.
34434         * rtx-vector-builder.c: Likewise.
34435         * rtl.h (rtx_def::u2): Add a const_vector field.
34436         (CONST_VECTOR_NPATTERNS): New macro.
34437         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
34438         (CONST_VECTOR_DUPLICATE_P): Likewise.
34439         (CONST_VECTOR_STEPPED_P): Likewise.
34440         (CONST_VECTOR_ENCODED_ELT): Likewise.
34441         (const_vec_duplicate_p): Check for a duplicated vector encoding.
34442         (unwrap_const_vec_duplicate): Likewise.
34443         (const_vec_series_p): Check for a non-duplicated vector encoding.
34444         Say that the function only returns true for integer vectors.
34445         * emit-rtl.c: Include rtx-vector-builder.h.
34446         (gen_const_vec_duplicate_1): Delete.
34447         (gen_const_vector): Call gen_const_vec_duplicate instead of
34448         gen_const_vec_duplicate_1.
34449         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
34450         (gen_const_vec_duplicate): Use rtx_vector_builder.
34451         (gen_const_vec_series): Likewise.
34452         (gen_rtx_CONST_VECTOR): Likewise.
34453         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
34454         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
34455         Build a new vector rather than modifying a CONST_VECTOR in-place.
34456         (handle_special_swappables): Update call accordingly.
34457         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
34458         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
34459         Build a new vector rather than modifying a CONST_VECTOR in-place.
34460         (handle_special_swappables): Update call accordingly.
34462 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34464         * simplify-rtx.c (simplify_const_binary_operation): Use
34465         CONST_VECTOR_ELT instead of XVECEXP.
34467 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34469         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
34470         the selector elements to be different from the data elements
34471         if the selector is a VECTOR_CST.
34472         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
34473         ssizetype for the selector.
34475 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34477         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
34478         before testing each element individually.
34479         * tree-vect-generic.c (lower_vec_perm): Likewise.
34481 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34483         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
34484         * selftest-run-tests.c (selftest::run_tests): Call it.
34485         * vector-builder.h (vector_builder::operator ==): New function.
34486         (vector_builder::operator !=): Likewise.
34487         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
34488         (vec_perm_indices::all_from_input_p): New function.
34489         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
34490         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
34491         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
34492         instead of reading the VECTOR_CST directly.  Detect whether both
34493         vector inputs are the same before constructing the vec_perm_indices,
34494         and update the number of inputs argument accordingly.  Use the
34495         utility functions added above.  Only construct sel2 if we need to.
34497 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34499         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
34500         the broadcast of the low byte.
34501         (expand_mult_highpart): Use an explicit encoding for the permutes.
34502         * optabs-query.c (can_mult_highpart_p): Likewise.
34503         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
34504         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34505         (vectorizable_bswap): Likewise.
34506         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
34507         explicit encoding for the power-of-2 permutes.
34508         (vect_permute_store_chain): Likewise.
34509         (vect_grouped_load_supported): Likewise.
34510         (vect_permute_load_chain): Likewise.
34512 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34514         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
34515         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
34516         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
34517         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
34518         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
34519         (vect_gen_perm_mask_any): Likewise.
34521 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34523         * int-vector-builder.h: New file.
34524         * vec-perm-indices.h: Include int-vector-builder.h.
34525         (vec_perm_indices): Redefine as an int_vector_builder.
34526         (auto_vec_perm_indices): Delete.
34527         (vec_perm_builder): Redefine as a stand-alone class.
34528         (vec_perm_indices::vec_perm_indices): New function.
34529         (vec_perm_indices::clamp): Likewise.
34530         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
34531         (vec_perm_indices::new_vector): New function.
34532         (vec_perm_indices::new_expanded_vector): Update for new
34533         vec_perm_indices class.
34534         (vec_perm_indices::rotate_inputs): New function.
34535         (vec_perm_indices::all_in_range_p): Operate directly on the
34536         encoded form, without computing elided elements.
34537         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
34538         encoding.  Update for new vec_perm_indices class.
34539         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
34540         the given vec_perm_builder.
34541         (expand_vec_perm_var): Update vec_perm_builder constructor.
34542         (expand_mult_highpart): Use vec_perm_builder instead of
34543         auto_vec_perm_indices.
34544         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
34545         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
34546         or double series encoding as appropriate.
34547         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
34548         vec_perm_indices instead of auto_vec_perm_indices.
34549         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34550         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34551         (vect_permute_store_chain): Likewise.
34552         (vect_grouped_load_supported): Likewise.
34553         (vect_permute_load_chain): Likewise.
34554         (vect_shift_permute_load_chain): Likewise.
34555         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
34556         (vect_transform_slp_perm_load): Likewise.
34557         (vect_schedule_slp_instance): Likewise.
34558         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34559         (vectorizable_mask_load_store): Likewise.
34560         (vectorizable_bswap): Likewise.
34561         (vectorizable_store): Likewise.
34562         (vectorizable_load): Likewise.
34563         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
34564         vec_perm_indices instead of auto_vec_perm_indices.  Use
34565         tree_to_vec_perm_builder to read the vector from a tree.
34566         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
34567         vec_perm_builder instead of a vec_perm_indices.
34568         (have_whole_vector_shift): Use vec_perm_builder and
34569         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
34570         truncation to calc_vec_perm_mask_for_shift.
34571         (vect_create_epilog_for_reduction): Likewise.
34572         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
34573         from auto_vec_perm_indices to vec_perm_indices.
34574         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
34575         instead of changing individual elements.
34576         (aarch64_vectorize_vec_perm_const): Use new_vector to install
34577         the vector in d.perm.
34578         * config/arm/arm.c (expand_vec_perm_d::perm): Change
34579         from auto_vec_perm_indices to vec_perm_indices.
34580         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
34581         instead of changing individual elements.
34582         (arm_vectorize_vec_perm_const): Use new_vector to install
34583         the vector in d.perm.
34584         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
34585         Update vec_perm_builder constructor.
34586         (rs6000_expand_interleave): Likewise.
34587         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
34588         (rs6000_expand_interleave): Likewise.
34590 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34592         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
34593         to qimode could truncate the indices.
34594         * optabs.c (expand_vec_perm_var): Likewise.
34596 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34598         * Makefile.in (OBJS): Add vec-perm-indices.o.
34599         * vec-perm-indices.h: New file.
34600         * vec-perm-indices.c: Likewise.
34601         * target.h (vec_perm_indices): Replace with a forward class
34602         declaration.
34603         (auto_vec_perm_indices): Move to vec-perm-indices.h.
34604         * optabs.h: Include vec-perm-indices.h.
34605         (expand_vec_perm): Delete.
34606         (selector_fits_mode_p, expand_vec_perm_var): Declare.
34607         (expand_vec_perm_const): Declare.
34608         * target.def (vec_perm_const_ok): Replace with...
34609         (vec_perm_const): ...this new hook.
34610         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
34611         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
34612         * doc/tm.texi: Regenerate.
34613         * optabs.def (vec_perm_const): Delete.
34614         * doc/md.texi (vec_perm_const): Likewise.
34615         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
34616         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
34617         expand_vec_perm for constant permutation vectors.  Assert that
34618         the mode of variable permutation vectors is the integer equivalent
34619         of the mode that is being permuted.
34620         * optabs-query.h (selector_fits_mode_p): Declare.
34621         * optabs-query.c: Include vec-perm-indices.h.
34622         (selector_fits_mode_p): New function.
34623         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
34624         is defined, instead of checking whether the vec_perm_const_optab
34625         exists.  Use targetm.vectorize.vec_perm_const instead of
34626         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
34627         fit in the vector mode before using a variable permute.
34628         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
34629         vec_perm_indices instead of an rtx.
34630         (expand_vec_perm): Replace with...
34631         (expand_vec_perm_const): ...this new function.  Take the selector
34632         as a vec_perm_indices rather than an rtx.  Also take the mode of
34633         the selector.  Update call to shift_amt_for_vec_perm_mask.
34634         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
34635         Use vec_perm_indices::new_expanded_vector to expand the original
34636         selector into bytes.  Check whether the indices fit in the vector
34637         mode before using a variable permute.
34638         (expand_vec_perm_var): Make global.
34639         (expand_mult_highpart): Use expand_vec_perm_const.
34640         * fold-const.c: Includes vec-perm-indices.h.
34641         * tree-ssa-forwprop.c: Likewise.
34642         * tree-vect-data-refs.c: Likewise.
34643         * tree-vect-generic.c: Likewise.
34644         * tree-vect-loop.c: Likewise.
34645         * tree-vect-slp.c: Likewise.
34646         * tree-vect-stmts.c: Likewise.
34647         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
34648         Delete.
34649         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
34650         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
34651         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
34652         (aarch64_vectorize_vec_perm_const): ...this new function.
34653         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
34654         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34655         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
34656         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
34657         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
34658         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34659         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
34660         into...
34661         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
34662         check for NEON modes.
34663         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
34664         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
34665         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
34666         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
34667         into...
34668         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
34669         the old VEC_PERM_CONST conditions.
34670         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
34671         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
34672         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
34673         (ia64_vectorize_vec_perm_const_ok): Merge into...
34674         (ia64_vectorize_vec_perm_const): ...this new function.
34675         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
34676         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
34677         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
34678         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
34679         * config/mips/mips.c (mips_expand_vec_perm_const)
34680         (mips_vectorize_vec_perm_const_ok): Merge into...
34681         (mips_vectorize_vec_perm_const): ...this new function.
34682         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
34683         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
34684         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
34685         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
34686         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
34687         (rs6000_expand_vec_perm_const): Delete.
34688         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
34689         Delete.
34690         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34691         (altivec_expand_vec_perm_const_le): Take each operand individually.
34692         Operate on constant selectors rather than rtxes.
34693         (altivec_expand_vec_perm_const): Likewise.  Update call to
34694         altivec_expand_vec_perm_const_le.
34695         (rs6000_expand_vec_perm_const): Delete.
34696         (rs6000_vectorize_vec_perm_const_ok): Delete.
34697         (rs6000_vectorize_vec_perm_const): New function.
34698         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
34699         an element count and rtx array.
34700         (rs6000_expand_extract_even): Update call accordingly.
34701         (rs6000_expand_interleave): Likewise.
34702         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
34703         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
34704         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
34705         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
34706         (rs6000_expand_vec_perm_const): Delete.
34707         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
34708         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34709         (altivec_expand_vec_perm_const_le): Take each operand individually.
34710         Operate on constant selectors rather than rtxes.
34711         (altivec_expand_vec_perm_const): Likewise.  Update call to
34712         altivec_expand_vec_perm_const_le.
34713         (rs6000_expand_vec_perm_const): Delete.
34714         (rs6000_vectorize_vec_perm_const_ok): Delete.
34715         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
34716         reference to the SPE evmerge intructions.
34717         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
34718         an element count and rtx array.
34719         (rs6000_expand_extract_even): Update call accordingly.
34720         (rs6000_expand_interleave): Likewise.
34721         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
34722         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
34723         new function.
34724         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34726 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34728         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
34729         vector mode and that that mode matches the mode of the data
34730         being permuted.
34731         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
34732         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
34733         directly using expand_vec_perm_1 when forcing selectors into
34734         registers.
34735         (expand_vec_perm_var): New function, split out from expand_vec_perm.
34737 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34739         * optabs-query.h (can_vec_perm_p): Delete.
34740         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
34741         * optabs-query.c (can_vec_perm_p): Split into...
34742         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
34743         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
34744         particular selector is valid.
34745         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34746         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34747         (vect_grouped_load_supported): Likewise.
34748         (vect_shift_permute_load_chain): Likewise.
34749         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
34750         (vect_transform_slp_perm_load): Likewise.
34751         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34752         (vectorizable_bswap): Likewise.
34753         (vect_gen_perm_mask_checked): Likewise.
34754         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
34755         implementations of variable permutation vectors into account
34756         when deciding which selector to use.
34757         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
34758         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
34759         with a false third argument.
34760         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
34761         to test whether the constant selector is valid and can_vec_perm_var_p
34762         to test whether a variable selector is valid.
34764 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34766         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
34767         * optabs-query.c (can_vec_perm_p): Likewise.
34768         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
34769         instead of vec_perm_indices.
34770         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
34771         (vect_gen_perm_mask_checked): Likewise,
34772         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
34773         (vect_gen_perm_mask_checked): Likewise,
34775 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34777         * optabs-query.h (qimode_for_vec_perm): Declare.
34778         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
34779         (qimode_for_vec_perm): ...this new function.
34780         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
34782 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34784         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
34785         does not have a conditional at the top.
34787 2018-01-02  Richard Biener  <rguenther@suse.de>
34789         * ipa-inline.c (big_speedup_p): Fix expression.
34791 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
34793         PR target/81616
34794         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
34795         for generic 4->6.
34797 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
34799         PR target/81616
34800         Generic tuning.
34801         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
34802         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
34803         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
34804         cond_taken_branch_cost 3->4.
34806 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
34808         PR tree-optimization/83581
34809         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
34810         TODO_cleanup_cfg if any changes have been made.
34812         PR middle-end/83608
34813         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
34814         convert_modes if target mode has the right side, but different mode
34815         class.
34817         PR middle-end/83609
34818         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
34819         last argument when extracting from CONCAT.  If either from_real or
34820         from_imag is NULL, use expansion through memory.  If result is not
34821         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
34822         the parts directly to inner mode, if even that fails, use expansion
34823         through memory.
34825         PR middle-end/83623
34826         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
34827         check for bswap in mode rather than HImode and use that in expand_unop
34828         too.
34830 Copyright (C) 2018 Free Software Foundation, Inc.
34832 Copying and distribution of this file, with or without modification,
34833 are permitted in any medium without royalty provided the copyright
34834 notice and this notice are preserved.