20180-12-08 Steven G. Kargl <kargl@gcc.gnu.org>
[official-gcc.git] / gcc / ChangeLog
blobf97a51ef7ffcd7528bfe036914f0500353500440
1 2018-12-08  Jakub Jelinek  <jakub@redhat.com>
3         PR fortran/88304
4         * tree-nested.c (convert_local_reference_stmt): Handle clobbers where
5         lhs is not a decl normally, don't call use_pointer_in_frame on that
6         lhs.
8 2018-12-08  Eric Botcazou  <ebotcazou@adacore.com>
10         PR rtl-optimization/88390
11         * dwarf2cfi.c (struct dw_cfi_row): Add window_save field.
12         (cfi_row_equal_p): Compare it.
13         (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter.
14         If FAKE is false, set window_save of the current row.
15         (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above
16         function with FAKE parameter set to true.
17         <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false.
18         (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary.
20 2018-12-07  Peter Bergner  <bergner@linux.ibm.com>
22         PR target/87496
23         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
24         -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
25         Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
26         * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
27         require -mlong-double-128.
29 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31         * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add
32         sve_width entry.
33         (aarch64_parse_sve_width_string): Define.
35 2018-12-07  Jeff Law  <law@redhat.com>
37         PR middle-end/87813
38         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add
39         m_update_global_ranges member.  Add corresponding argument to ctor.
40         * gimple-ssa-evrp-analyze.c
41         (evrp_range_analyzer::evrp_range_analyzer): Add new argument and
42         initialize m_update_global_ranges.
43         (evrp_range_analyzer::set_ssa_range_info): Assert that we are
44         updating global ranges.
45         (evrp_range_analyzer::record_ranges_from_incoming_edge): Only
46         update global ranges if explicitly requested.
47         (evrp_range_analyzer::record_ranges_from_phis): Similarly.
48         (evrp_range_analyzer::record_ranges_from_stmt): Similarly.
49         * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to
50         evrp_range_analyzer ctor.
51         * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly.
52         * tree-ssa-dom.c (dom_opt_dom_walker): Similarly.
54 2018-12-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
56         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum):
57         Add SVE_NOT_IMPLEMENTED value.
58         * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width
59         field.
60         * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings,
61         cortexa53_tunings, cortexa57_tunings, cortexa72_tunings,
62         cortexa73_tunings, exynosm1_tunings, thunderx_tunings,
63         thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings,
64         saphira_tunings, thunderx2t99_tunings, emag_tunings):
65         Specify sve_width.
66         (aarch64_estimated_poly_value): Define.
67         (TARGET_ESTIMATED_POLY_VALUE): Define.
69 2018-12-07  Paul A. Clarke  <pc@us.ibm.com>
71         PR target/88408
72         * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector".
74 2018-12-07  Vladimir Makarov  <vmakarov@redhat.com>
76         PR rtl-optimization/88349
77         * ira-costs.c (record_operand_costs): Check bigger reg class on
78         NO_REGS.
80 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
82         * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3):
83         Split the patterns after reload if we don't need the predicate
84         operand.
85         (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns.
87 2018-12-07  Richard Sandiford  <richard.sandiford@arm.com>
89         * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code
90         iterator.
91         (sve_fp_op): Handle minus and mult.
92         * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3)
93         (*mul<mode>3): Split the patterns after reload if we don't
94         need the predicate operand.
95         (*post_ra_<sve_fp_op><mode>3): New pattern.
97 2018-12-07  Bin Cheng  <bin.cheng@linux.alibaba.com>
99         * profile-count.h (profile_count::oeprator>=): Fix typo by inverting
100         return condition when *this is precise zero.
102 2018-12-07  Jakub Jelinek  <jakub@redhat.com>
104         PR target/85593
105         * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
106         for functions with naked attribute.
108         PR c/88367
109         * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR
110         with -fno-delete-null-pointer-checks, set_nonnull only if the pointer
111         is non-NULL and offset is known to have most significant bit clear.
112         * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR
113         of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with
114         most significant bit clear.  If offset does have most significant bit
115         set and -fno-delete-null-pointer-checks, don't return true even if
116         the base pointer is non-NULL.
118 2018-12-06  Alexandre Oliva  <aoliva@redhat.com>
120         * cselib.c (cselib_record_sets): Skip strict low part sets
121         with NULL src_elt.
123 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
125         PR target/88316
126         * config/rs6000/smmintrin.h: New file.
127         * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*.
129 2018-12-06  Paul A. Clarke  <pc@us.ibm.com>
131         PR target/88316
132         * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
133         (_mm_unpacklo_pi8): Likewise.
134         (_mm_mulhi_pi16): Likewise.
135         (_mm_packs_pi16): Fix for big-endian. Use preferred API.
136         (_mm_packs_pi32): Likewise.
137         (_mm_packs_pu16): Likewise.
138         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
139         (_mm_cvtss_si64): Likewise.
140         (_mm_cvtpi32x2_ps): Likewise.
141         (_mm_shuffle_ps): Likewise.
142         (_mm_movemask_pi8): Likewise.
143         (_mm_mulhi_pu16): Likewise.
144         (_mm_sad_pu8): Likewise.
145         (_mm_sad_pu8): Likewise.
146         (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
147         (_mm_cvtpu8_ps): Likewise.
148         (_mm_movemask_ps): Better #else case for big-endian (no functional
149         change).
150         (_mm_shuffle_pi16): Likewise.
151         * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
152         Better #else case for big-endian (no functional change).
153         (_mm_movemask_epi8): Likewise.
154         (_mm_shufflehi_epi16): Likewise.
155         (_mm_shufflelo_epi16): Likewise.
156         (_mm_shuffle_epi32): Likewise.
157         (_mm_mul_epu32): Fix for big-endian.
158         (_mm_bsrli_si128): Likewise.
159         (_mm_cvtps_pd): Better #else case for big endian.
160         (_mm_mulhi_epi16): Likewise.
161         (_mm_mul_epu32): Likewise.
162         (_mm_slli_si128): Likewise.
163         (_mm_sll_epi16): Likewise.
164         (_mm_sll_epi32): Likewise.
165         (_mm_sra_epi16): Likewise.
166         (_mm_sra_epi32): Likewise.
167         (_mm_srl_epi16): Likewise.
168         (_mm_srl_epi32): Likewise.
169         (_mm_mulhi_epu16): Likewise.
170         (_mm_sad_epu8): Likewise.
171         * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
172         (_mm_sub_ps): Likewise.
173         * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
174         * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN
175         macros consistently (no functional changes).
176         (_mm_alignr_pi8): Likewise.
178 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
180         PR c++/87380
181         * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
182         default.
183         * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
185 2018-12-06  Vladimir Makarov  <vmakarov@redhat.com>
187         PR target/88282
188         * ira.c (ira_init_register_move_cost): Use info from
189         hard_regno_mode_ok instead of contains_reg_of_mode.
190         * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger
191         hard register class for some fixed hard registers.
193 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
195         * doc/extend.texi (Using Assembly Language with C): Document asm inline.
196         (Size of an asm): Fix typo.  Document asm inline.
197         * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
198         * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
199         (gimple_asm_set_volatile): Fix typo.
200         (gimple_asm_inline_p): New.
201         (gimple_asm_set_inline): New.
202         * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
203         tree to gimple.
204         * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
205         gimple_asm_inline_p flag, too.
206         * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
207         in an ASM_EXPR.
208         * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
209         a minimum size for an asm.
210         * tree.h (ASM_INLINE_P): New.
212 2018-12-06  Segher Boessenkool  <segher@kernel.crashing.org>
214         PR inline-asm/55681
215         * doc/extend.texi (Basic Asm): Update grammar.
216         (Extended Asm): Update grammar.
218 2018-12-06  Iain Sandoe  <iain@sandoe.co.uk>
220         PR target/78444
221         * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
222         * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
223         stack alignment in non-leaf functions.
225 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
227         PR target/87598
228         * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
229         call output_operand_lossage on VOIDmode CONST_INTs.  After
230         output_operand_lossage do return false.
232 2018-12-06  Richard Biener  <rguenther@suse.de>
234         * df-problems.c (df_rd_local_compute): Use bitmap_release.
235         (df_live_free): Likewise.
236         (df_md_local_compute): Likewise.
237         (df_md_free): Release df_md_scratch bitmap.
238         * loop-invariant.c (calculate_loop_reg_pressure): Use
239         bitmap_release.
240         * sched-deps.c (true_dependency_cache, output_dependency_cache,
241         anti_dependency_cache, control_dependency_cache,
242         spec_dependency_cache): Use bitmap instead of bitmap_head *.
243         * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY
244         dont_calc_deps as bitmap allocated from obstack not GC.
245         (schedule_ebbs_finish): Use bitmap_release.
246         * sched-rgn.c (schedule_insns): Initialize non-GTY
247         not_in_df as bitmap allocated from obstack not GC.
248         Use bitmap_release.
249         * sel-sched.c (_forced_ebb_heads): Remove premature optimization.
250         (sel_region_init): Allocate forced_ebb_heads.
251         (sel_region_finish): Free forced_ebb_heads.
253 2018-12-06  Richard Biener  <rguenther@suse.de>
255         * bitmap.c (bitmap_head::crashme): Define.
256         * bitmap.h (bitmap_head): Add constexpr default constructor
257         poisoning the obstack member.
258         (bitmap_head::crashme): Declare.
259         (bitmap_release): New function clearing a bitmap and poisoning
260         the obstack member.
261         * gengtype.c (main): Make it recognize CONSTEXPR.
262         * lra-constraints.c (lra_inheritance): Use bitmap_release
263         instead of bitmap_clear.
264         * ira.c (ira): Work around class-memaccess warning.
265         * regrename.c (create_new_chain): Likewise.
267 2018-12-06  Jakub Jelinek  <jakub@redhat.com>
269         PR tree-optimization/85726
270         * generic-match-head.c (optimize_successive_divisions_p): New function.
271         * gimple-match-head.c (optimize_successive_divisions_p): Likewise.
272         * match.pd: Don't combine successive divisions if they aren't exact
273         and optimize_successive_divisions_p is false.
275         PR fortran/88304
276         * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers
277         for non-local automatic decls.
279 2018-12-05  David Edelsohn  <dje.gcc@gmail.com>
281         * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor
282         default. 
284 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
286         * dwarf2out.c (field_byte_offset): Remove
287         #ifdef PCC_BITFIELD_TYPE_MATTERS check.
289 2018-12-05  Richard Biener  <rguenther@suse.de>
290             Martin Sebor  <msebor@redhat.com>
292         PR c/87028
293         * calls.c (get_attr_nonstring_decl): Avoid setting *REF to
294         SSA_NAME_VAR.
295         * gcc/gimple-low.c (lower_stmt): Fold builtin calls here.
296         * gimplify (maybe_fold_stmt): Avoid folding builtin calls.
298 2018-12-05  Iain Sandoe  <iain@sandoe.co.uk>
300         * configure.ac (gcc_cv_otool): Set.
301         * configure: Regenerate.
303 2018-12-05  Uros Bizjak  <ubizjak@gmail.com>
305         * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove.
306         (GROUP_BITFIELDS_BY_ALIGN): Ditto.
308 2018-12-05  Richard Sandiford  <richard.sandiford@arm.com>
310         PR tree-optimization/88064
311         * tree-vect-patterns.c (vect_convert_input): Convert the result of
312         an existing cast if it has the right width but the wrong sign.
313         Do not test the signedness of the required result when
314         considering whether to split an existing cast; instead split to
315         a type with the same signedness as the source of the cast, then
316         convert it to the opposite signedness where necessary.
317         (vect_recog_over_widening_pattern): Handle sign changes between
318         the final PLUS_EXPR and the RSHIFT_EXPR.
319         (vect_recog_average_pattern): Use an unsigned operation when
320         truncating an addition, subtraction or multiplication.  Cast the
321         result back to the "real" signedness before promoting.
323 2018-12-05  Richard Biener  <rguenther@suse.de>
325         PR tree-optimization/86637
326         * tree-vectorizer.c (pass_slp_vectorize::execute): Reset
327         vect_location at the end.
329 2018-12-05  Jakub Jelinek  <jakub@redhat.com>
331         PR tree-optimization/87360
332         * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
333         data dependencies, don't return false, just continue.  Formatting
334         fixes.
335         (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p,
336         fuse_loops): Formatting fixes.
338         PR sanitizer/88333
339         * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with
340         aligning frame offset to ASAN_RED_ZONE_SIZE bytes.
342 2018-12-05  Xianmiao Qu  <xianmiao_qu@c-sky.com>
344         * config.gcc (csky-*-linux-gnu*): Force .init_array support.
346 2018-12-05  Segher Boessenkool  <segher@kernel.crashing.org>
348         * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special
349         aggregate handling on actual AIX, not on somewhat similar systems.
350         (rs6000_arg_partial_bytes): Ditto.
352 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
354         PR rtl-optimization/88317
355         * lra-constraints.c (split_reg): Don't set up check_only_regs if
356         we are outside the inheritance pass.
358 2018-12-04  Jan Hubicka  <hubicka@ucw.cz>
360         * ipa-prop.c (jump_function_useful_p): New.
361         (ipa_read_node_info): Do not allocated useless jump functions.
363 2018-12-04  David Malcolm  <dmalcolm@redhat.com>
365         PR c/82967
366         * spellcheck.c (selftest::test_suggestions): Add another assertion.
368 2018-12-04  Martin Sebor  <msebor@redhat.com>
370         * doc/extend.texi (__builtin_has_attribute): Fix typo.
372 2018-12-04  Jakub Jelinek  <jakub@redhat.com>
374         PR target/88188
375         * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion
376         failures.
377         (print_operand): Use REG_P instead of GET_CODE == REG.
378         <case 'D'>: Also check CR_REGNO_P (REGNO (x)).
379         <case 't'>: Likewise.  Remove GET_MODE check.
380         * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't
381         return -1.
383 2018-12-04  Martin Sebor  <msebor@redhat.com>
385         * doc/extend.texi (attribute aligned): Expand.
387 2018-12-04  David Edelsohn  <dje.gcc@gmail.com>
389         PR target/61976
390         * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
391         in FPRs on AIX.
392         (rs6000_arg_partial_bytes): Same.
394 2018-12-04  Vladimir Makarov  <vmakarov@redhat.com>
396         PR target/88282
397         * ira-costs.c (exec): Try bigger class to use smaller register
398         move cost.
400 2018-12-04  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
402         PR ipa/88297
403         * ipa-cp.c (create_specialized_node): Track clone counters by
404         node assembler names.
405         (ipcp_driver): Change type of clone_num_suffixes key to const char*.
407 2018-12-04  Richard Biener  <rguenther@suse.de>
409         * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations
410         for epilogue vectorization.
412 2018-12-04  Richard Biener  <rguenther@suse.de>
414         PR tree-optimization/88301
415         * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion
416         issues in last commit.
418 2018-12-04  Richard Biener  <rguenther@suse.de>
420         PR tree-optimization/88315
421         * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify
422         and fix initialization vector for SAD and DOT_PROD SLP reductions.
424 2018-12-03  Sandra Loosemore  <sandra@codesourcery.com>
426         PR c/59039
427         * doc/extend.texi (Nonlocal gotos): New section.
429 2018-12-03  Uros Bizjak  <ubizjak@gmail.com>
431         * config/i386/sse.md (vec_concatv2df): Change (v,xm,C)
432         alternative to (v,vm,c).
434 2018-12-03  Segher Boessenkool  <segher@kernel.crashing.org>
436         * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*.
437         (powerpc*-*-linux*): Don't handle it here.
439 2018-12-03  Jeff Law  <law@redhat.com>
441         * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep
442         warnings quiet.
444         * config/h8300/h8300.md (low byte masking peepholes): Only accept
445         nonimmediate_operand for the to-be-masked operand.
447 2018-12-03  Richard Biener  <rguenther@suse.de>
449         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
450         allow expected function-pointer cast re-instantiation.
452 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
454         * common/config/s390/s390-common.c (s390_option_init_struct):
455         Use section anchors by default.
456         * config/s390/s390.c (s390_check_symref_alignment): Handle
457         anchors.
458         (TARGET_MAX_ANCHOR_OFFSET): Use short displacement.
459         * output.h (assemble_align): Pass `align' as unsigned int, so
460         that the value 0x80000000, which corresponds to `aligned(1 <<
461         28)', would pass the `align > BITS_PER_UNIT' check.
462         * varasm.c (assemble_align): Likewise.
464 2018-12-03  Julian Brown  <julian@codesourcery.com>
466         * tree-pretty-print.c (dump_omp_clause): Make default case
467         gcc_unreachable.
469 2018-12-03  Richard Biener  <rguenther@suse.de>
471         PR tree-optimization/88301
472         * tree-vrp.c (register_edge_assert_for_2): Handle conversions
473         that do not change the value by registering the same assert
474         for the operand.
476 2018-12-03  Richard Biener  <rguenther@suse.de>
478         * tree-vect-loop.c (vect_transform_loop): Properly compute
479         upper bound for the epilogue when doing epilogue vectorization.
481 2018-12-03  Ilya Leoshkevich  <iii@linux.ibm.com>
483         PR target/80080
484         * cfgcleanup.c (class pass_postreload_jump): New pass.
485         (pass_postreload_jump::execute): Likewise.
486         (make_pass_postreload_jump): Likewise.
487         * passes.def: Add pass_postreload_jump before
488         pass_postreload_cse.
489         * tree-pass.h (make_pass_postreload_jump): New pass.
491 2018-12-02  Uros Bizjak  <ubizjak@gmail.com>
493         * config/i386/i386.c (inline_memory_move_cost): Correctly
494         compute number of TFmode GPR moves.
496 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
498         PR target/88278
499         * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns.
501 2018-12-02  Jeff Law  <law@redhat.com>
503         * config/h8300/h8300.md (call, call_value): Drop mode from
504         operand containing number of bytes pushed.
505         (call_insn, call_value_insn): Use mode iterator for number
506         of bytes pushed operand.  Make pattern name vary appropriately.
508 2018-12-02  Jan Hubicka  <jh@suse.cz>
510         * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges):
511         trust precise profiles.
513 2018-12-02  Jakub Jelinek  <jakub@redhat.com>
515         * tree-nested.c (convert_nonlocal_omp_clauses,
516         convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION,
517         OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses.
518         (convert_nonlocal_reference_stmt, convert_local_reference_stmt):
519         Convert clauses for GIMPLE_OMP_TASKGROUP.
521         * omp-low.c (check_omp_nesting_restrictions): Allow cancel or
522         cancellation point with taskgroup clause inside of taskloop.  Consider
523         a taskloop construct without nogroup clause as implicit taskgroup for
524         diagnostics if cancel/cancellation point with taskgroup clause is
525         closely nested inside of taskgroup region.
527         PR sanitizer/88291
528         * asan.c (asan_clear_shadow): Move assert that len is multiple of 4
529         to the start of the function.
530         (asan_emit_stack_protection): When emitting clearing sequence for
531         epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity,
532         add last_size_aligned which is last_size padded to multiples of
533         ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte
534         boundaries.
536 2018-12-01  Jakub Jelinek  <jakub@redhat.com>
538         PR sanitizer/88289
539         * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up
540         an off-by-one for BYTES_BIG_ENDIAN.
542         PR target/54589
543         * combine.c (find_split_point): For invalid memory address
544         nonobj + obj + const, if reg + obj + const is valid addressing
545         mode, split at nonobj.  Use if rather than else if for the
546         fallback.  Comment fixes.
548 2018-11-30  Indu Bhagat  <indu.bhagat@oracle.com>
550         * coverage.c (get_coverage_counts): Use from_function_decl for precise
551         function location.
552         * profile-count.c (profile_count::dump): Add handling for precise
553         profile quality.
554         * profile.c (compute_branch_probabilities): Rely on exec_counts instead
555         of profile_info to set profile_status of function.
556         (branch_prob): Do not set profile_status of function based on
557         profile_info. Done above based on exec_counts.
559 2018-11-30  Jakub Jelinek  <jakub@redhat.com>
561         PR tree-optimization/88274
562         * tree-ssa-reassoc.c (optimize_range_tests_xor,
563         optimize_range_tests_diff): If type has smaller precision than
564         corresponding mode or if it has non-standard min/max, compute
565         everything in a standard type for the precision.
567         PR testsuite/85368
568         * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
569         * tree-ssa-ifcombine.c (ifcombine_ifandif): If
570         --param logical-op-non-short-circuit is present, override
571         LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
572         * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
574 2018-11-30  Jeff Law  <law@redhat.com>
576         * optabs.c (expand_binop): Use "machine_mode" rather than
577         "enum machine mode" in most recent change.
579 2018-11-30  Wilco Dijkstra  <wdijkstr@arm.com>
581         PR middle-end/64242
582         * builtins.c (expand_builtin_longjmp): Use a temporary when restoring
583         the frame pointer.
584         (expand_builtin_nonlocal_goto): Likewise.
586 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
588         * diagnostic-core.h (emit_diagnostic): New decl.
589         * diagnostic.c (emit_diagnostic): New overload, taking a
590         rich_location *.
592 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
594         * pretty-print.c (class selftest::test_pretty_printer): New
595         subclass of pretty_printer.
596         (selftest::test_prefixes_and_wrapping): New test.
597         (selftest::pretty_print_c_tests): Call it.
599 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
601         Minimize clone counter memory usage in create_virtual_clone.
602         * cgraph.h (clone_function_name): Add a variant that takes a
603         tree decl.
604         * cgraph.h (cgraph_node::create_virtual_clone): Add a new
605         argument: num_suffix.
606         * cgraphclones.c (cgraph_node::create_virtual_clone): Pass
607         num_suffix to clone_function_name.
608         (clone_function_name): Add a variant that takes a tree decl.
609         * ipa-cp.c (create_specialized_node): Keep track of clone
610         counters in clone_num_suffixes hash map.
611         (ipcp_driver): Free the counter hash map.
612         * ipa-hsa.c (process_hsa_functions): Creates at most one hsa
613         clone per function.
615 2018-11-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
616         Make function assembly more independent.
618         This is achieved by having clone_function_name assign unique clone
619         numbers for each function independently.
621         * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids;
622         hash map.
623         (clone_function_name_numbered): Use clone_fn_ids.
625 2018-11-30  Vladimir Makarov  <vmakarov@redhat.com>
627         PR rtl-optimization/88179
628         * lra-constraints.c (address_eliminator::address_eliminator):
629         Don't eleminate regs in illegitimate address.
631 2018-11-30  David Malcolm  <dmalcolm@redhat.com>
633         PR preprocessor/88257
634         * input.c (get_substring_ranges_for_loc): Fix indentation.
635         Bulletproof against getting a different files back from
636         linemap_resolve_location and expand_location_to_spelling_point.
638 2018-11-30  Alexander Monakov  <amonakov@ispras.ru>
640         PR gcov-profile/88279
641         * tree-profile.c (init_ic_make_global_vars): Call
642         decl_default_tls_model for variable and not it's type.
644 2018-11-30  Sam Tebbs  <sam.tebbs@arm.com>
646         * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint.
648 2018-11-30  Martin Liska  <mliska@suse.cz>
650         PR sanitizer/81715
651         * asan.c (asan_shadow_cst): Remove, partially transform
652         into flush_redzone_payload.
653         (RZ_BUFFER_SIZE): New.
654         (struct asan_redzone_buffer): New.
655         (asan_redzone_buffer::emit_redzone_byte): Likewise.
656         (asan_redzone_buffer::flush_redzone_payload): Likewise.
657         (asan_redzone_buffer::flush_if_full): Likewise.
658         (asan_emit_stack_protection): Use asan_redzone_buffer class
659         that is responsible for proper aligned stores and flushing
660         of shadow memory payload.
661         * asan.h (ASAN_MIN_RED_ZONE_SIZE): New.
662         (asan_var_and_redzone_size): Likewise.
663         * cfgexpand.c (expand_stack_vars): Use smaller alignment
664         (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory
665         for automatic variables more compact.
667 2018-11-30  Alan Modra  <amodra@gmail.com>
669         * config/rs6000/predicates.md (easy_fp_constant): Avoid long
670         dependent insn sequences.
671         * config/rs6000/rs6000.c (num_insns_constant): Support long
672         double constants.
673         * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust
674         length attribute.
676 2018-11-30  Alan Modra  <amodra@gmail.com>
678         * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from
679         num_insns_constant_wide.  Make static.  Revise comment.
680         (num_insns_constant_multi): New function.
681         (num_insns_constant): Formatting.  Correct CONST_WIDE_INT
682         calculation.  Simplify and extract code common to both
683         CONST_INT and CONST_DOUBLE.  Add gcc_unreachable for unhandled
684         const_double modes.
685         * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete.
687 2018-11-30  Richard Biener  <rguenther@suse.de>
689         PR tree-optimization/88274
690         * tree-vrp.c (ranges_from_anti_range): Fix handling of
691         TYPE_MIN/MAX_VALUE.
693 2018-11-29  Andi Kleen  <ak@linux.intel.com>
695         * config/i386/i386.c (current_fentry_section): Add.
696         (x86_function_profiler): Handle fentry section.
697         (ix86_attribute_table): Add fentry section.
698         * config/i386/i386.opt: Add -mfentry-section.
699         * doc/extend.texi: Document fentry_section attribute.
700         * doc/invoke.texi: Document -mfentry-section.
702 2018-11-29  Andi Kleen  <ak@linux.intel.com>
704         * config/i386/i386.c (x86_print_call_or_nop): Handle nop name.
705         (current_fentry_name): Add.
706         (x86_function_profiler): Handle fentry_name attribute.
707         (ix86_handle_fentry_name): Add.
708         (ix86_attribute_table): Add fentry_name.
709         * config/i386/i386.opt: Add -mfentry-name
710         * doc/extend.texi: Document fentry_name.
711         * doc/invoke.texi: Document minstrument-return.
713 2018-11-29  Andi Kleen  <ak@linux.intel.com>
715         * config/i386/i386-opts.h (enum instrument_return): Add.
716         * config/i386/i386.c (output_return_instrumentation): Add.
717         (ix86_output_function_return): Call output_return_instrumentation.
718         (ix86_output_call_insn): Call output_return_instrumentation.
719         * config/i386/i386.opt: Add -minstrument-return=.
720         * doc/invoke.texi (-minstrument-return): Document.
722 2018-11-29  Eric Botcazou  <ebotcazou@adacore.com>
724         PR target/87807
725         * config/sparc/sparc-modes.def: Minor tweak.
726         * config/sparc/sparc.c: Minor reordering.
727         (sparc_pass_by_reference): Move around.
728         (traverse_record_type): Change offset from HOST_WIDE_INT to int.
729         (classify_registers): Likewise for bitpos.
730         (function_arg_slotno): Remove dead test and tweak comments.
731         <MODE_RANDOM>: Remove useless assertion and test whether the
732         parameter is named in order to pass it in FP registers.  Return
733         the regno for floating-point vector types.
734         (compute_int_layout): Change bitpos from HOST_WIDE_INT to int.
735         (compute_fp_layout): Likewise.
736         (count_registers): Likewise.
737         (assign_int_registers): Likewise.
738         (assign_fp_registers): Likewise.
739         (assign_registers): Likewise.
740         (function_arg_record_value): Change size from HOST_WIDE_INT to int
741         and use CEIL_NWORDS to compute the number of registers.
742         (function_arg_union_value): Minor tweaks.
743         (function_arg_vector_value): Add slotno and named parameters, use
744         CEIL_NWORDS to compute the number of registers.
745         (sparc_function_arg_1): Rework handling of vector types.  Change
746         size from HOST_WIDE_INT to int.
747         (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal
748         with vector types for the 64-bt ABI.
749         (sparc_function_arg_advance): Likewise.
750         (sparc_return_in_memory): Add reference to -fpcc-struct-return.
751         (sparc_struct_value_rtx): Return NULL_RTX instead of 0.
752         (sparc_function_value_1): Rework handling of vector types.  Change
753         size from HOST_WIDE_INT to int.
755 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
757         PR target/88152
758         * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt,
759         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt,
760         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift,
761         *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift,
762         *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New
763         define_insn_and_split patterns.
765         PR target/54700
766         * config/i386/sse.md
767         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
768         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
769         *<sse4_1_avx2>_pblendvb_lt): Change define_insn into
770         define_insn_and_split.
772 2018-11-29  Peter Bergner  <bergner@linux.ibm.com>
774         PR target/87496
775         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
776         -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
778 2018-11-29  Uros Bizjak  <ubizjak@gmail.com>
780         * config/i386/i386.c (inline_memory_move_cost):
781         Check "in" for 2 in MMX_CLASS_P case.
782         * config/i386/mmx.md (*mov<mode>_internal): Correct
783         TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC
784         alternatives in preferred_for_speed attribute calculation.
786 2018-11-29  Martin Sebor  <msebor@redhat.com>
788         PR c/88172
789         PR testsuite/88208
790         * doc/extend.texi (attribute constructor): Clarify.
792 2018-11-29  Martin Liska  <mliska@suse.cz>
794         PR middle-end/88246
795         * builtins.c (expand_movstr): Fix thinko introduced
796         when switching to the new enum.
798 2018-11-29  qing zhao  <qing.zhao@oracle.com>
800         * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
801         * common.opt: Add -flive-patching flag.
802         * doc/invoke.texi: Document -flive-patching.
803         * flag-types.h (enum live_patching_level): New enum.
804         * ipa-inline.c (can_inline_edge_p): Disable external functions from
805         inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
806         * opts.c (control_options_for_live_patching): New function.
807         (finish_options): Make flag_live_patching incompatible with flag_lto.
808         Control IPA optimizations based on different levels of
809         flag_live_patching.
811 2018-11-29  Giuliano Belinassi  <giuliano.belinassi@usp.br>
813         * match.pd (sinh (atanh (x))): New simplification rules.
814         (cosh (atanh (x))): Likewise.
816 2018-11-29  Jakub Jelinek  <jakub@redhat.com>
818         PR target/54700
819         * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
820         (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
821         (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
822         *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
823         *<sse4_1_avx2>_pblendvb_lt): New define_insns.
825         PR target/88152
826         * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
827         * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
828         * match.pd (define_predicates): Add uniform_integer_cst_p.
829         (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
830         so that it works also for vector comparisons with uniform constants
831         with INTEGER_CST element.
833         PR target/88234
834         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
835         vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
836         in unsigned_type_for instead of vector integral type where overflow
837         doesn't wrap.
839 2018-11-29  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
841         There can be at most one .resolver clone per function
842         * config/rs6000/rs6000.c (make_resolver_func): Generate
843         resolver symbol with clone_function_name instead of
844         clone_function_name_numbered.
846 2018-11-29  Richard Biener  <rguenther@suse.de>
848         PR tree-optimization/88243
849         * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
850         type of all pattern-sequence stmts to vect_internal_def.
852 2018-11-29  Andre Vieira  <andre.simoesdiasvieira@arm.com>
854         PR target/88224
855         * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
856         (cortex-r7, cortex-r8): Update default and add new configuration.
857         * doc/invoke.texi (armv7-r): Add two new vfp options.
858         (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
859         support this option.
861 2018-11-29  Alan Modra  <amodra@gmail.com>
863         * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
864         double split for targets other than Darwin.
866 2018-11-29  Alan Modra  <amodra@gmail.com>
868         * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
870 2018-11-29  Alan Modra  <amodra@gmail.com>
872         * config/rs6000/predicates.md (easy_fp_constant): Remove code
873         dealing with integers in const_double.  Assert on unexpected
874         modes.  Delete superfluous ABI_V4 test.
875         * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
876         Don't call easy_fp_constant for Pmode.
877         (rs6000_hash_constant): Remove code dealing with integers in
878         const_double.
879         (rs6000_legitimate_constant_p): Likewise.
880         (output_toc): Formatting, use CONST_DOUBLE_P.
882 2018-11-29  Alan Modra  <amodra@gmail.com>
884         * config/rs6000/constraints.md (G, H): Comment on purpose of
885         constraint.  Correct mode comments and "G" mode test.
886         * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
887         alternative handling "F".  Add length attr.  Formatting.
888         (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
889         alternative, and move "F" to separate alternative.  Correct
890         insn lengths.
892 2018-11-29  Alan Modra  <amodra@gmail.com>
894         * config/rs6000/predicates.md (easy_vector_constant): Don't call
895         easy_fp_constant.
897 2018-11-29  Alan Modra  <amodra@gmail.com>
899         * config.in (HAVE_AS_PLTSEQ): Add.
900         * config/rs6000/predicates.md (indirect_call_operand): New.
901         * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
902         (rs6000_sibcall_sysv): Declare.
903         * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
904         CALL_LONG for -fno-plt.
905         (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
906         (rs6000_indirect_call_template_1): Emit .reloc directives for
907         UNSPEC_PLTSEQ calls.
908         (rs6000_pltseq_template): New function.
909         (rs6000_longcall_ref): Add arg parameter.  Use PLT16 insns if
910         relocs supported by assembler.  Move SYMBOL_REF test to callers.
911         (rs6000_call_aix): Adjust rs6000_longcall_ref call.  Package
912         insns in UNSPEC_PLTSEQ, preserving original func_desc.
913         (rs6000_call_sysv): Likewise.
914         (rs6000_sibcall_sysv): New function.
915         * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
916         * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
917         UNSPEC_PLT16_LO): New.
918         (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
919         (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
920         cookie in constraints.  Test explicitly for flags in length attr.
921         Handle unspec operand 1.
922         (call_value_indirect_nonlocal_sysv): Likewise.
923         (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
924         (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
925         (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
926         (sibcall_indirect_nonlocal_sysv): New pattern.
927         (sibcall_value_indirect_nonlocal_sysv): Likewise.
928         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
929         call alternatives.
930         * configure.ac: Check for gas plt sequence marker support.
931         * configure: Regenerate.
933 2018-11-29  Alan Modra  <amodra@gmail.com>
935         * config/rs6000/predicates.md (unspec_tls): New.
936         * config/rs6000/rs6000-protos.h (rs6000_call_template),
937         (rs6000_sibcall_template): Update prototype.
938         (rs6000_longcall_ref): Delete.
939         (rs6000_call_sysv): Declare.
940         * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
941         (global_tlsarg): New variable.
942         (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
943         handling.
944         (print_operand): Extract UNSPEC_TLSGD address operand.
945         (rs6000_call_template, rs6000_sibcall_template): Remove arg
946         parameter, extract from second call operand instead.
947         (rs6000_longcall_ref): Make static, localize vars.
948         (rs6000_call_aix): Rename parameter to reflect new usage.  Take
949         tlsarg from global_tlsarg.  Don't create unused rtl or nop insns.
950         (rs6000_sibcall_aix): Rename parameter to reflect new usage.  Take
951         tlsarg from global_tlsarg.
952         (rs6000_call_sysv): New function.
953         * config/rs6000/rs6000.md: Adjust rs6000_call_template and
954         rs6000_sibcall_template throughout.
955         (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
956         (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
957         (tls_gdld_nomark): New insn.
958         (tls_gd): Swap operand order.  Simplify mode selection.
959         (tls_gd_high, tls_gd_low): Swap operand order.
960         (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
961         Simplify mode selection.
962         (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
963         (call, call_value): Don't assert for second call operand.
964         Use rs6000_call_sysv.
966 2018-11-29  Alan Modra  <amodra@gmail.com>
968         * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
969         (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
970         (call_value_nonlocal_darwin64): Remove constraints from second call
971         arg, the rounded_stack_size_rtx arg.
972         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
973         (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
974         (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
975         (call_value_local64, call_indirect_nonlocal_sysv),
976         (call_nonlocal_sysv, call_nonlocal_sysv_secure),
977         (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
978         (call_value_nonlocal_sysv_secure, call_local_aix),
979         (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
980         (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
981         (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
982         (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
983         (sibcall_value_aix): Likewise.
985 2018-11-29  Alan Modra  <amodra@gmail.com>
987         * config/rs6000/rs6000.md (TLSmode): Delete mode iterator.  Replace
988         with P throughout except for call mems which should use SI.
989         (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
990         attributes.  Replace with bits, mode and ptrload respectively.
992 2018-11-29  Alan Modra  <amodra@gmail.com>
994         * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
995         (rs6000_indirect_sibcall_template): Declare.
996         * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
997         (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
998         New functions.
999         * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
1000         (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
1001         (call_indirect_aix, call_value_indirect_aix): Use
1002         rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
1003         call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
1004         handle both speculation and non-speculation cases.
1005         (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
1006         (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
1008 2018-11-29  Alan Modra  <amodra@gmail.com>
1010         * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
1011         (rs6000_sibcall_template): Declare.
1012         (macho_call_template): Rename from output_call.
1013         * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
1014         (rs6000_call_template, rs6000_sibcall_template): Likewise.
1015         (macho_call_template): Rename from output_call.
1016         * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
1017         (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
1018         (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
1019         (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
1020         (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
1021         (call_value_nonlocal_aix): Use rs6000_call_template and update
1022         occurrences of output_call to macho_call_template.
1023         (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
1024         (sibcall_value_aix): Use rs6000_sibcall_template.
1026 2018-11-28  Aaron Sawdey  <acsawdey@linux.ibm.com>
1028         * config/rs6000/rs6000-string.c (expand_block_clear): Change how
1029         we determine if unaligned vsx is ok.
1031 2018-11-28  Jan Hubicka  <jh@suse.cz>
1033         * predict.c (determine_unlikely_bbs): Forward declare; also determine
1034         edges that are always known to be taken because it is only likely
1035         edge out of given BB.
1036         (tree_estimate_probability): Call before profile guessing to get bit
1037         of extra precision.
1039 2018-11-28  Jan Hubicka  <jh@suse.cz>
1041         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
1042         profile_probability::always better.
1044 2018-11-28  Jan Hubicka  <jh@suse.cz>
1046         * profile-count.h (profile_count::split): Give better result when
1047         splitting profile_probability::always.
1049 2018-11-28  Vladimir Makarov  <vmakarov@redhat.com>
1051         PR target/88207
1052         * ira-costs.c (scan_one_insn): Process subregs when updating costs
1053         for pseudos and allocnos from insn.
1055 2018-11-28  David Edelsohn  <dje.gcc@gmail.com>
1057         * config/rs6000/aix72.h: Update to match aix71.h changes.
1059 2018-11-28  Jeff Law  <law@redhat.com>
1061         * config/h8300/constraints.md: Add "C" constraint for call insns.
1062         * config/h8300/h8300.md (call, call_value): Turn into a define_expand
1063         and define_insn pair.  Move invalid call targets into a register in
1064         the expander and fix constraints in the matching pattern.
1065         * config/h8300/predicates.md (call_expander_operand): Renamed from
1066         call_insn_operand.  Reject things we shouldn't be trying to handle.
1067         (call_insn_operand): New predicate for use by the call/call_value
1068         insns.
1069         (small_call_insn_operand): Update appropriately.
1071 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
1073         * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
1074         calls to strtok with strtok_r.
1076 2018-11-28  Richard Biener  <rguenther@suse.de>
1078         PR tree-optimization/88223
1079         * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
1080         over a stored-same value may-alias store make sure to consider
1081         partial overlaps which are valid when TBAA reasonings do not
1082         apply and byte-granular overlaps are possible at all.
1084 2018-11-28  Richard Biener  <rguenther@suse.de>
1086         PR tree-optimization/88217
1087         * vr-values.c (vr_values::extract_range_from_phi_node): Make
1088         sure to handle results > +INF and < -INF correctly when
1089         trying to drop down to +INF - 1 or -INF + 1.
1091 2018-11-28  Alan Modra  <amodra@gmail.com>
1093         * xcoffout.c (do_block): Signed/unsigned warning fix.
1095 2018-11-28  Richard Biener  <rguenther@suse.de>
1097         PR tree-optimization/79351
1098         * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
1099         empty CONSTRUCTORs ensure the store is at a constant position.
1101 2018-11-28  Richard Biener  <rguenther@suse.de>
1103         PR tree-optimization/88229
1104         * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
1105         is an INTEGER_CST before accessing it so.
1107 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
1109         * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
1111 2018-11-28  Jakub Jelinek  <jakub@redhat.com>
1113         PR target/88189
1114         * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
1115         SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1.  Formatting fixes.
1116         * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
1118 2018-11-27  Martin Liska  <mliska@suse.cz>
1120         * gcov.c (generate_results): Append current_working_directory
1121         only when exists.
1123 2018-11-27  Mike Gulick  <mgulick@mathworks.com>
1125         PR preprocessor/83173
1126         * input.c (dump_location_info): Dump reason and included_from
1127         fields from line_map_ordinary struct.  Fix indentation when
1128         location > 5 digits.
1129         * diagnostic-show-locus.c (num_digits, num_digits): Move to
1130         diagnostic.c to allow it to be utilized by input.c.
1131         * diagnostic.c (num_digits, selftest::test_num_digits): Moved
1132         here.
1133         (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
1134         * diagnostic.h (num_digits): Add extern definition.
1136 2018-11-27  Fredrik Noring  <noring@nocrew.org>
1138         * config/mips/mips.c (mips_reorg_process_insns)
1139         (mips_option_override): Handle `-mfix-r5900'.
1140         * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
1141         `mno-fix-r5900'.
1142         * config/mips/mips.opt (mfix-r5900): New option.
1143         * doc/invoke.texi: Document the `r5900' processor name, and
1144         `-mfix-r5900' and `-mno-fix-r5900' options.
1146 2018-11-27  Jakub Jelinek  <jakub@redhat.com>
1148         PR target/88188
1149         * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
1150         output_operand_lossage instead of gcc_assert.
1151         <case 't'>: Likewise.
1152         <case 'z'>: Likewise.
1153         <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
1155 2018-11-27  Jeff Law  <law@redhat.com>
1157         * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
1158         in call to move_by_pieces.
1160         * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
1161         to move_by_pieces.
1163         * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
1164         call to move_by_pieces.
1166         * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
1167         in call to move_by_pieces.
1169         * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
1170         RETURN_BEGIN in call to move_by_pieces.
1171         (microblaze_expand_block_move): Likewise.
1173 2018-11-27  Tamar Christina  <tamar.christina@arm.com>
1175         * doc/md.texi (xorsign): Document it.
1177 2018-11-27  Martin Liska  <mliska@suse.cz>
1179         * config/gnu-user.h (TARGET_F951_OPTIONS): New.
1180         * gcc.c (find_fortran_preinclude_file): New function
1181         to handle Fortran pre-include.
1183 2018-11-27  Martin Liska  <mliska@suse.cz>
1185         * asan.c (asan_emit_stack_protection): Use new enum values
1186         instead of int constants.
1187         * builtins.c (expand_builtin_memory_copy_args): Replace int
1188         type with memop_ret enum type.
1189         (expand_builtin_mempcpy_args): Likewise.
1190         (expand_builtin_memcpy): Use new enum values
1191         instead of int constants. Likewise.
1192         (expand_builtin_mempcpy): Likewise.
1193         (expand_movstr): Likewise.
1194         (expand_builtin_strcpy_args): Likewise.
1195         (expand_builtin_stpcpy_1): Likewise.
1196         (expand_builtin_strncpy): Likewise.
1197         (expand_builtin_memset_args): Likewise.
1198         * expr.c (move_by_pieces_d::finish_endp): Rename to ...
1199         (move_by_pieces_d::finish_retmode): ... this.
1200         (move_by_pieces): Change last argument type to memop_ret.
1201         (store_by_pieces): Use new enum values
1202         instead of int constants.
1203         (emit_block_move_hints): Likewise.
1204         (emit_push_insn): Likewise.
1205         (store_expr): Likewise.
1206         * expr.h (store_by_pieces): Change int to newly added enum
1207         type.
1208         * rtl.h (enum memop_ret): Define.
1209         (move_by_pieces): Use the enum type.
1211 2018-11-27  Alan Modra  <amodra@gmail.com>
1213         * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
1214         cpu here.
1215         (ASM_CPU_SPEC): Do so here.  Rewrite using if .. else if .. specs
1216         form.  Error on missing -mcpu case.
1217         * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
1218         (asm_names <!_AIX>): Add missing cpus.  Update NULL case.  Apply
1219         PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
1220         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
1221         else if .. specs form.  Error on missing -mcpu case.  Don't output
1222         duplicate -maltivec.  Apply PR63177 fix for -mcpu=powerpc64le.
1224 2018-11-26  David Malcolm  <dmalcolm@redhat.com>
1226         * dump-context.h (dump_context::dump_loc): Convert 1st param from
1227         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
1228         const dump_location_t & to const dump_user_location_t &.
1229         (dump_context::dump_loc_immediate): Convert 2nd param from
1230         const dump_location_t & to const dump_user_location_t &.
1231         (dump_context::dump_gimple_stmt): Convert 1st param from
1232         dump_flags_t to const dump_metadata_t &.
1233         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
1234         2nd param from const dump_location_t & to
1235         const dump_user_location_t &.
1236         (dump_context::dump_gimple_expr): Convert 1st param from
1237         dump_flags_t to const dump_metadata_t &.
1238         (dump_context::dump_gimple_expr_loc): Likewise; convert
1239         2nd param from const dump_location_t & to
1240         const dump_user_location_t &.
1241         (dump_context::dump_generic_expr): Convert 1st param from
1242         dump_flags_t to const dump_metadata_t &.
1243         (dump_context::dump_generic_expr_loc): Likewise; convert
1244         2nd param from const dump_location_t & to
1245         const dump_user_location_t &.
1246         (dump_context::dump_printf_va): Convert 1st param from
1247         dump_flags_t to const dump_metadata_t &.
1248         (dump_context::dump_printf_loc_va): Likewise; convert
1249         2nd param from const dump_location_t & to
1250         const dump_user_location_t &.
1251         (dump_context::dump_dec): Convert 1st param from
1252         dump_flags_t to const dump_metadata_t &.
1253         (dump_context::dump_symtab_node): Likewise.
1254         (dump_context::begin_scope): Split out 2nd param into
1255         user and impl locations.
1256         (dump_context::ensure_pending_optinfo): Add metadata param.
1257         (dump_context::begin_next_optinfo): Replace dump_location_t param
1258         with metadata and user location.
1259         * dumpfile.c (dump_context::dump_loc): Convert 1st param from
1260         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
1261         const dump_location_t & to const dump_user_location_t &.
1262         (dump_context::dump_loc_immediate): Convert 2nd param from
1263         const dump_location_t & to const dump_user_location_t &.
1264         (dump_context::dump_gimple_stmt): Convert 1st param from
1265         dump_flags_t to const dump_metadata_t &.
1266         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
1267         2nd param from const dump_location_t & to
1268         const dump_user_location_t &.
1269         (dump_context::dump_gimple_expr): Convert 1st param from
1270         dump_flags_t to const dump_metadata_t &.
1271         (dump_context::dump_gimple_expr_loc): Likewise; convert
1272         2nd param from const dump_location_t & to
1273         const dump_user_location_t &.
1274         (dump_context::dump_generic_expr): Convert 1st param from
1275         dump_flags_t to const dump_metadata_t &.
1276         (dump_context::dump_generic_expr_loc): Likewise; convert
1277         2nd param from const dump_location_t & to
1278         const dump_user_location_t &.
1279         (dump_context::dump_printf_va): Convert 1st param from
1280         dump_flags_t to const dump_metadata_t &.
1281         (dump_context::dump_printf_loc_va): Likewise; convert
1282         2nd param from const dump_location_t & to
1283         const dump_user_location_t &.
1284         (dump_context::dump_dec): Convert 1st param from
1285         dump_flags_t to const dump_metadata_t &.
1286         (dump_context::dump_symtab_node): Likewise.
1287         (dump_context::begin_scope): Split out 2nd param into
1288         user and impl locations.
1289         (dump_context::ensure_pending_optinfo): Add metadata param.
1290         (dump_context::begin_next_optinfo): Replace dump_location_t param
1291         with metadata and user location.
1292         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
1293         const dump_metadata_t &.
1294         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
1295         const dump_location_t & to const dump_user_location_t &.
1296         (dump_gimple_expr): Convert 1st param from dump_flags_t to
1297         const dump_metadata_t &.
1298         (dump_gimple_expr_loc): Likewise; convert 2nd param from
1299         const dump_location_t & to const dump_user_location_t &.
1300         (dump_generic_expr): Convert 1st param from dump_flags_t to
1301         const dump_metadata_t &.
1302         (dump_generic_expr_loc): Likewise; convert 2nd param from
1303         const dump_location_t & to const dump_user_location_t &.
1304         (dump_printf): Convert 1st param from dump_flags_t to
1305         const dump_metadata_t &.
1306         (dump_printf_loc): Likewise; convert 2nd param from
1307         const dump_location_t & to const dump_user_location_t &.
1308         (dump_dec): Convert 1st param from dump_flags_t to
1309         const dump_metadata_t &.
1310         (dump_symtab_node): Likewise.
1311         (dump_begin_scope): Split out 2nd param into user and impl
1312         locations.
1313         (selftest::assert_impl_location_eq): New function.
1314         (ASSERT_IMPL_LOCATION_EQ): New macro.
1315         (selftest::test_impl_location): Update to use
1316         ASSERT_IMPL_LOCATION_EQ.
1317         (selftest::test_capture_of_dump_calls): Convert "loc" to
1318         dump_user_location_t.  Add ASSERT_IMPL_LOCATION_EQ throughout,
1319         verifying line numbers of dump emissions.
1320         * dumpfile.h (class dump_metadata_t): New class.
1321         (dump_printf): Convert 1st param from dump_flags_t to
1322         const dump_metadata_t &.
1323         (dump_printf_loc): Likewise; convert 2nd param from
1324         const dump_location_t & to const dump_user_location_t &.
1325         (dump_generic_expr_loc): Likewise.
1326         (dump_generic_expr): Convert 1st param from dump_flags_t to
1327         const dump_metadata_t &.
1328         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
1329         const dump_location_t & to const dump_user_location_t &.
1330         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
1331         const dump_metadata_t &.
1332         (dump_gimple_expr_loc): Likewise; convert 2nd param from
1333         const dump_location_t & to const dump_user_location_t &.
1334         (dump_gimple_expr): Convert 1st param from dump_flags_t to
1335         const dump_metadata_t &.
1336         (dump_symtab_node): Likewise.
1337         (dump_dec): Likewise.
1338         (dump_begin_scope): Split out 2nd param into user and impl
1339         locations.
1340         (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
1341         location and impl_location, and capture the impl_location.
1342         (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
1343         * loop-unroll.c (report_unroll): Update for changes to
1344         dump_printf_loc and dump_printf.
1345         * opt-problem.cc (opt_problem::opt_problem): Update for change to
1346         dump_loc.
1347         * optinfo-emit-json.cc
1348         (selftest::test_building_json_from_dump_calls): Convert "loc" from
1349         dump_location_t to dump_user_location_t.
1350         * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
1351         dump_loc_immediate.
1352         * profile.c (compute_branch_probabilities): Update for change to
1353         dump_printf_loc.
1354         * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
1355         * tree-vect-slp.c (vect_print_slp_tree): Update for change to
1356         dump_printf_loc.
1358 2018-11-27  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1360         * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
1361         and check_effective_target_ptr_eq_long.
1363 2018-11-27  Alan Modra  <amodra@gmail.com>
1365         * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
1366         (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
1367         (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
1368         (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
1369         (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
1370         * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
1371         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
1372         (ASM_APP_ON, ASM_APP_OFF): Don't define.
1373         * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
1374         * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
1375         * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1376         * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
1377         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
1378         * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1379         * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
1380         (GNU_USER_TARGET_CRTN): Redefine.
1381         (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
1382         (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
1383         (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
1384         (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
1385         (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
1386         (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
1387         (LINK_EH_SPEC): Don't define.
1389 2018-11-26  Martin Sebor  <msebor@redhat.com>
1391         PR 87756
1392         * expr.c (string_constant): Handle top-level decls of all character
1393         types and subobjects of narrow character type.
1395 2018-11-27  Alan Modra  <amodra@gmail.com>
1397         * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
1398         * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
1399         GNU_USER_TARGET_OS_CPP_BUILTINS.
1400         (RS6000_ABI_NAME): Define.
1401         * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
1402         GNU_USER_TARGET_OS_CPP_BUILTINS.
1403         (MUSL_DYNAMIC_LINKER32): Undef before defining.
1404         (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
1405         (CHOOSE_DYNAMIC_LINKER): Don't define.
1406         (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
1407         * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
1408         (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
1409         not already defined.
1410         (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
1411         TARGET_OS_CPP_BUILTINS.
1413 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1415         * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
1417 2018-11-26  Jakub Jelinek  <jakub@redhat.com>
1419         PR target/88195
1420         * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
1421         and !TARGET_64BIT, return NULL_TREE.
1423         PR c++/86900
1424         * dwarf2out.c (secname_for_decl): For functions with
1425         DECL_SECTION_NAME if in_cold_section_p, try to return
1426         current_function_section's name if it is a named section.
1428 2018-11-26  Maya Rashish  <coypu@sdf.org>
1430         PR target/58397
1431         * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
1432         * config/x-netbsd: New file.
1433         * config/host-netbsd: New file.
1435 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1437         * doc/sourcebuild.texi: Document check_effective_target_size20plus.
1438         Clarify documentation for check_effective_target_size32plus.
1440 2018-11-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
1442         * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
1443         and modify expand_strncmp_vec_sequence.
1444         (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
1445         (generate_6432_conversion): New function.
1446         (expand_block_compare): Add support for vsx.
1447         (expand_block_compare_gpr): New function.
1448         * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
1449         default limit to 63 because of more compact vsx code.
1451 2018-11-26  Uros Bizjak  <ubizjak@gmail.com>
1453         PR target/88178
1454         * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
1455         for registers for which no debug information can be generated.
1456         Use INVALID_REGNUM for non-existent registers.
1457         (dbx64_register_map): Ditto.
1458         (svr4_dbx_register_map): Ditto.
1460 2018-11-26  Sandra Loosemore  <sandra@codesourcery.com>
1462         PR c/57182
1463         * doc/invoke.texi (Optimize Options): Document that
1464         -freorder-functions also uses "hot" and "cold" attributes.
1466 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1468         * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
1470 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1472         * config/arm/types.md (mul64): Rename to...
1473         (widen_mul64): ... This.
1474         * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
1475         * config/arm/cortex-a15.md: Likewise.
1476         * config/arm/cortex-a5.md: Likewise.
1477         * config/arm/cortex-a53.md: Likewise.
1478         * config/arm/cortex-a57.md: Likewise.
1479         * config/arm/cortex-a7.md: Likewise.
1480         * config/arm/cortex-m4.md: Likewise.
1481         * config/arm/exynos-m1.md: Likewise.
1482         * config/arm/marvell-pj4.md: Likewise.
1483         * config/arm/xgene1.md: Likewise.
1485 2018-11-26  Richard Biener  <rguenther@suse.de>
1487         PR tree-optimization/88182
1488         * tree-vect-loop.c (vectorizable_reduction): Pick up single
1489         correct reduc_def_info.
1490         * tree-vect-slp.c (vect_analyze_slp_instance): Set
1491         STMT_VINFO_REDUC_DEF of the first stmt.
1493 2018-11-26  Andreas Krebbel  <krebbel@linux.ibm.com>
1495         * doc/invoke.texi: Document z14/arch12 -march option.
1497 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
1499         PR c/57166
1500         * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
1501         (Warning Options): Likewise.
1503 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
1505         PR web/79738
1506         * doc/extend.texi (Common Function Attributes): Clarify that
1507         functions with "const" attribute can read const global variables.
1509 2018-11-26  Alan Modra  <amodra@gmail.com>
1511         * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
1513 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
1515         PR other/54265
1516         * doc/extend.texi (Common Variable Attributes): Use preferred
1517         placement of type attributes in examples, plus whitespace fixes.
1518         (Type Attributes): Clarify why placement of attributes
1519         immediately after struct/union/enum keyword is preferred.
1520         (Common Type Attributes): Use preferred placement of type
1521         attributes in examples, plus more whitespace fixes.
1523 2018-11-25  Paul Koning  <ni1d@arrl.net>
1525         * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
1527 2018-11-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1529         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
1530         attributes): Document d_runtime.
1532 2018-11-25  Vladimir Makarov  <vmakarov@redhat.com>
1534         PR bootstrap/88157
1535         * ira-costs.c (record_operand_costs): Initiate register move cost
1536         for mode.
1538 2018-11-23  Jeff Law  <law@redhat.com>
1540         PR rtl-optimization/87468
1541         * tree-ssa-threadupdate.c (create_block_for_threading): Clear
1542         EDGE_IGNORE on all outgoing edges of the duplicate block.
1544 2018-11-23  Vladimir Makarov  <vmakarov@redhat.com>
1546         PR bootstrap/88157
1547         * ira-costs.c (record_operand_costs): Use bigger hard reg class if
1548         its mode does not fit to the original class.
1550 2018-11-23  Martin Sebor  <msebor@redhat.com>
1552         PR tree-optimization/87756
1553         * expr.c (string_constant): Synthesize a string literal from
1554         the address of a constant character.
1555         * tree.c (build_string_literal): Add an argument.
1556         * tree.h (build_string_literal): Same.
1558 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
1560         * config/aarch64/aarch64-cores.def: Define emag.
1561         * config/aarch64/aarch64-tune.md: Regenerated with emag.
1562         * config/aarch64/aarch64.c (emag_tunings): New struct.
1563         * doc/invoke.texi: Document mtune value.
1565 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
1566             Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
1568         * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
1569         for GCC 9.
1570         * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
1571         prefetch tunings.
1572         * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
1573         costs.
1574         * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
1575         for XGene1.
1577 2018-11-23  Richard Biener  <rguenther@suse.de>
1579         PR tree-optimization/88149
1580         * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
1581         the case where there are two different def types for the
1582         same operand at different operand position in the same stmt.
1584 2018-23-11  Mihail Ionescu  <mihail.ionescu@arm.com>
1586         * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
1587         gen_atomic_compare_swap_1.
1588         (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
1589         (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
1590         (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
1591         (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
1592         (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
1593         checks.
1594         * config/arm/arm.md (neon_vext<mode>),
1595         (neon_vrev64<mode>, neon_vrev32<mode>),
1596         (neon_vrev16<mode>, neon_vtrn<mode>_internal),
1597         (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
1598         before the pattern name.
1599         * config/arm/sync.md:
1600         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
1601         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
1603 2018-11-23  Jakub Jelinek  <jakub@redhat.com>
1605         PR tree-optimization/86614
1606         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
1607         if TREE_NO_WARNING is set on ref.ref.
1609 2018-11-22  Sandra Loosemore  <sandra@codesourcery.com>
1610             Alan Coopersmith  <alan.coopersmith@oracle.com>
1612         PR c/53608
1613         * doc/extend.texi (Designated Inits): Clarify handling of multiple
1614         initializers for unions.
1616 2018-11-22  Jan Hubicka  <jh@suse.cz>
1618         PR lto/88142
1619         * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
1620         parameters; do not warn here.
1621         (odr_subtypes_equivalent_p): Likewise.
1622         (warn_odr): Fix typo.
1623         (warn_types_mismatch): Do not output confused warnings on integer types.
1624         (odr_types_equivalent_p): Update.
1626 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
1628         * config/i386/i386.c (ix86_option_override_internal,
1629         ix86_can_inline_p, classify_argument, construct_container,
1630         ix86_expand_prologue, ix86_expand_split_stack_prologue,
1631         ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
1632         expand_setmem_epilogue_via_loop, promote_duplicated_reg,
1633         ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
1634         Formatting fixes.
1636         * config/i386/i386.c (ix86_option_override_internal): For
1637         stack_protector_guard related options, use opts_set->x_ instead
1638         of global_options_set. and prefix options with opts->x_ .  Move
1639         defaults for offset and reg into else block.
1641 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
1643         * config/i386/i386.c (ix86_check_avx_upper_register):
1644         Return true for all SSE registers with mode bitsize > 128.
1646 2018-11-22  Vladimir Makarov  <vmakarov@redhat.com>
1648         PR rtl-optimization/87718
1649         * ira-costs.c: Remove trailing white-spaces.
1650         (record_operand_costs): Add a special treatment for moves
1651         involving a hard register.
1653 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
1655         * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
1656         (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
1658 2018-11-22  Martin Liska  <mliska@suse.cz>
1660         * common/config/i386/i386-common.c (processor_names): Add
1661         static assert and add missing "znver2".
1662         (ix86_get_valid_option_values): Add checking assert for null
1663         values and add "native" value if feasible.
1664         * config/i386/i386.h: Do not declare size of processor_names.
1665         * common/config/i386/i386-common.c:
1666         * config/i386/i386.c: Add static assert for size
1667         of processor_cost_table.
1669 2018-11-22  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
1671         * target-insns.def (stack_protect_combined_set): Define new standard
1672         pattern name.
1673         (stack_protect_combined_test): Likewise.
1674         * cfgexpand.c (stack_protect_prologue): Try new
1675         stack_protect_combined_set pattern first.
1676         * function.c (stack_protect_epilogue): Try new
1677         stack_protect_combined_test pattern first.
1678         * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
1679         parameters to control which register to use as PIC register and force
1680         reloading PIC register respectively.  Insert in the stream of insns if
1681         possible.
1682         (legitimize_pic_address): Expose above new parameters in prototype and
1683         adapt recursive calls accordingly.  Use pic_reg if non null instead of
1684         cached one.
1685         (arm_load_pic_register): Add pic_reg parameter and use it if non null.
1686         (arm_legitimize_address): Adapt to new legitimize_pic_address
1687         prototype.
1688         (thumb_legitimize_address): Likewise.
1689         (arm_emit_call_insn): Adapt to require_pic_register prototype change.
1690         (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
1691         (thumb1_expand_prologue): Likewise.
1692         * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
1693         change.
1694         (arm_load_pic_register): Likewise.
1695         * config/arm/predicated.md (guard_addr_operand): New predicate.
1696         (guard_operand): New predicate.
1697         * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
1698         prototype change.
1699         (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
1700         prototype change.
1701         (stack_protect_combined_set): New expander..
1702         (stack_protect_combined_set_insn): New insn_and_split pattern.
1703         (stack_protect_set_insn): New insn pattern.
1704         (stack_protect_combined_test): New expander.
1705         (stack_protect_combined_test_insn): New insn_and_split pattern.
1706         (arm_stack_protect_test_insn): New insn pattern.
1707         * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
1708         * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
1709         (UNSPEC_SP_TEST): Likewise.
1710         * doc/md.texi (stack_protect_combined_set): Document new standard
1711         pattern name.
1712         (stack_protect_set): Clarify that the operand for guard's address is
1713         legal.
1714         (stack_protect_combined_test): Document new standard pattern name.
1715         (stack_protect_test): Clarify that the operand for guard's address is
1716         legal.
1718 2018-11-22  Richard Biener  <rguenther@suse.de>
1720         PR tree-optimization/88148
1721         * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
1722         valueization callback parameter and handle it.
1723         * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
1724         * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
1726 2018-11-22  Richard Biener  <rguenther@suse.de>
1728         PR lto/87229
1729         PR lto/88112
1730         * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
1731         which can appear in size expressions.
1732         * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
1733         CALL_EXPR_BY_DESCRIPTOR.
1734         (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
1735         * tree-streamer-out.c (pack_ts_base_value_fields): Stream
1736         CALL_EXPR_BY_DESCRIPTOR.
1737         (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
1739         Revert
1740         PR lto/87229
1741         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
1742         sizepos values.
1744 2018-11-22  Richard Biener  <rguenther@suse.de>
1746         PR tree-optimization/88069
1747         * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
1748         apply to default defs.
1750 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
1752         PR target/85644
1753         PR target/86832
1754         * config/i386/i386.c (ix86_option_override_internal): Default
1755         ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
1756         is defined.
1757         * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
1758         stack_protect_test, stack_protect_test_<mode>): Use empty condition
1759         instead of TARGET_SSP_TLS_GUARD.
1761 2018-11-22  Martin Liska  <mliska@suse.cz>
1763         PR sanitizer/88017
1764         * config/i386/i386.c (ix86_option_override_internal): Error
1765         about usage -mabi=ms and -fsanitize=thread.
1767 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
1769         Revert the revert:
1770         2013-10-26  Vladimir Makarov  <vmakarov@redhat.com>
1772         Revert:
1773         2013-10-25  Vladimir Makarov  <vmakarov@redhat.com>
1775         * lra-spills.c (lra_final_code_change): Remove useless move insns.
1777 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
1779         PR middle-end/88129
1780         * function.c (expand_function_end): Do not emit extra blockage insn.
1782 2018-11-21  Lokesh Janghel  <lokeshjanghel91@gmail.com>
1784         PR target/85667
1785         * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
1786         of FIRST_SSE_REG for 4 or 8 byte modes.
1788 2018-11-21  Jan Hubicka  <jh@suse.cz>
1790         PR lto/87957
1791         * tree.c (fld_decl_context): Break out from ...
1792         (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
1793         DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
1794         (fld_incomplete_type_of): Build copy of TYP_DECL.
1795         * ipa-devirt.c (free_enum_values): Rename to ...
1796         (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
1797         and TREE_TYPEs of TYPE_DECLs.
1798         (get_odr_type): Initialize odr_vtable_hash if needed.
1800 2018-11-21  Alexandre Oliva <oliva@adacore.com>
1802         * final.c (compute_discriminator): Declare.  Renamed from...
1803         (maybe_set_discriminator): ... this.  Set and return a local.
1804         (override_discriminator): New.
1805         (final_scan_insn_1): Set it.
1806         (notice_source_line): Adjust.  Always set discriminator.
1808 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
1810         PR target/87839
1811         * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
1812         rIJ constraint for aarch64_plus_operand rather than rn.
1814 2018-11-21  Renlin Li  <renlin.li@arm.com>
1816         PR middle-end/84877
1817         * explow.h (get_dynamic_stack_size): Declare it as external.
1818         * explow.c (record_new_stack_level): Remove function static attribute.
1819         * function.c (assign_stack_local_1): Dynamically align the stack slot
1820         addr for parameter copy on the stack.
1822 2018-11-21  Richard Biener  <rguenther@suse.de>
1824         PR bootstrap/88133
1825         * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
1826         * Makefile.in (bitmap.o-warn): Remove again.
1828 2018-11-20  Jeff Law  <law@redhat.com>
1830         PR tree-optimization/88069
1831         * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
1832         degenerate virtual PHIs.
1834 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
1836         PR target/87317
1837         * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
1838         nonimmediate_operand with register_operand.
1839         (avx2_<code>v8qiv8si2<mask_name>): Likewise.
1840         (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
1841         (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
1842         (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
1843         (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
1844         (avx2_<code>v4qiv4di2<mask_name>): Likewise.
1845         (avx2_<code>v4hiv4di2<mask_name>): Likewise.
1846         (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
1847         (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
1848         (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
1849         (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
1850         (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
1851         (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
1852         (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
1853         (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
1854         (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
1855         (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
1856         (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
1857         (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
1858         (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
1859         (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
1860         (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
1861         (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
1862         (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
1863         (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
1864         (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
1865         (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
1867 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
1869         * read-rtl.c (apply_subst_iterator): Handle define_split and
1870         define_insn_and_split.
1872 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
1874         PR rtl-optimization/87817
1875         * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
1876         *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
1877         in the pattern to avoid triggering UB when operands[2] is zero.
1878         (tbm_bextri_<mode>): New expander.  Renamed the old define_insn to ...
1879         (*tbm_bextri_<mode>): ... this.
1881 2018-11-21  Tom de Vries  <tdevries@suse.de>
1883         PR driver/79855
1884         * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
1885         period.
1887 2018-11-21  Andreas Krebbel  <krebbel@linux.ibm.com>
1889         * configure.ac: Add check for Binutils to determine whether vector
1890         load/store alignments hints are being supported.
1891         * config.in: Regenerate.
1892         * configure: Regenerate.
1893         * config/s390/s390.c (print_operand): Support new output
1894         modifier A.
1895         * config/s390/s390.md ("movti"): Append alignment hint output
1896         using the new output modifier 'A'.
1897         * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
1898         ("*vec_ti_to_v1ti"): Likewise.
1900 2018-11-20  Martin Sebor  <msebor@redhat.com>
1902         * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
1904 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1906         PR lto/84044
1907         * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
1908         compare ENUM values.
1910 2018-11-20  Jakub Jelinek  <jakub@redhat.com>
1912         PR tree-optimization/87895
1913         * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
1914         or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
1915         (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
1916         redirects edges to EXIT to edges to incr_bb, iterate while EXIT
1917         has any preds and always use EDGE_PRED (, 0).
1919 2018-11-20  Uros Bizjak  <ubizjak@gmail.com>
1921         PR target/88070
1922         * mode-switching.c (create_pre_exit): After reload, always split the
1923         fallthrough edge to the exit block.
1925 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1927         * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
1928         * tree.c (build_array_type_1): Forward declare.
1929         (fld_type_variant_equal_p): Add INNER_TYPE parameter.
1930         (fld_type_variant): Likewise.
1931         (fld_simplified_types): New hash.
1932         (fld_process_array_type): New function.
1933         (fld_incomplete_type_of): Handle array and enumeration types.
1934         (fld_simplified_type): Handle simplification of arrays.
1935         (free_lang_data): Allocate and free simplified types hash.
1937 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1939         PR lto/87957
1940         * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
1941         when sybtype already violates ODR.
1942         (get_odr_type): Do not ICE when insert is false and type duplicate
1943         is not registered yet.
1944         (register_odr_type): Be sure to register subtypes first.
1946 2018-11-20  Andreas Krebbel  <krebbel@linux.ibm.com>
1948         * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
1949         DImode parts of the target operand.
1951 2018-11-20  Nathan Sidwell  <nathan@acm.org>
1953         PR 87926
1954         * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
1956 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1958         PR lto/87997
1959         * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
1960         type checking to...
1961         (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
1963 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
1965         PR ipa/87706
1966         * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
1967         * ipa.c (possible_inline_candidate_p): Break out from ..
1968         (process_references): ... here ; drop before_inlining_p;
1969         cleanup handling of alises.
1970         (walk_polymorphic_call_targets): Likewise.
1971         (symbol_table::remove_unreachable_nodes): Likewise.
1972         * passes.c (pass_data_ipa_remove_symbols): New structure.
1973         (pass_ipa_remove_symbols): New pass.
1974         (make_pass_ipa_remove_symbols): New function.
1975         * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
1976         * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
1978 2018-11-20  Richard Biener  <rguenther@suse.de>
1980         * tree-vect-stmts.c (vectorizable_condition): Do not get
1981         at else_clause vect def for EXTRACT_LAST_REDUCTION.  Remove
1982         pointless vect_is_simple_use calls.
1984 2018-11-20  Richard Biener  <rguenther@suse.de>
1986         PR tree-optimization/88074
1987         * tree-ssa-dom.c (pass_dominator::execute): Do not walk
1988         backedges.
1990 2018-11-20  Richard Biener  <rguenther@suse.de>
1992         PR tree-optimization/88069
1993         * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
1994         virtual PHI arguments.
1996 2018-11-20  Ilya Leoshkevich  <iii@linux.ibm.com>
1998         PR target/88083
1999         * config/s390/s390.md: Skip LT(G) peephole when literal pool is
2000         involved.
2001         * rtl.h (contains_constant_pool_address_p): New function.
2002         * rtlanal.c (contains_constant_pool_address_p): Likewise.
2004 2018-11-20  Richard Biener  <rguenther@suse.de>
2006         PR middle-end/83215
2007         * alias.c (component_uses_parent_alias_set_from): Remove
2008         alias-set zero and TYPE_TYPELESS_STORAGE case both already
2009         handled in other ways.
2011 2018-11-20  Richard Biener  <rguenther@suse.de>
2013         PR tree-optimization/88087
2014         * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
2015         call fntype.
2016         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
2017         call fntype.
2019 2018-11-20  Richard Biener  <rguenther@suse.de>
2021         PR middle-end/88089
2022         * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
2024 2018-11-19  Jan Hubicka  <hubicka@ucw.cz>
2026         PR lto/87957
2027         * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
2029 2018-11-19  Sandra Loosemore  <sandra@codesourcery.com>
2031         PR driver/50250
2032         * doc/invoke.texi (Link Options): Mention shared libraries
2033         in documentation for the -l option.  Simplify discussion and
2034         point to the system linker documentation for details.
2036 2018-11-19  Peter Bergner  <bergner@linux.ibm.com>
2038         PR rtl-optimization/88033
2039         * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
2040         to itself.  Use HARD_REGISTER_NUM_P.
2042 2018-11-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2044         * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
2045         of the loop if vf is non-constant.
2047 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
2049         PR tree-optimization/87025
2050         * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
2051         immediately after creating the scope optinfo.
2052         (selftest::test_pr87025): New function.
2053         (selftest::dumpfile_c_tests): Call it.
2054         * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
2055         that we're not popping the top-level records array.
2056         * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
2057         not changing the kind of a "scope" optinfo.
2059 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
2061         PR tree-optimization/87025
2062         * dump-context.h: Include "optinfo.h".
2063         (class optrecord_json_writer): New forward decl.
2064         (dump_context::forcibly_enable_optinfo_p): Delete.
2065         (dump_context::optinfo_enabled_p): New member function.
2066         (dump_context::optimization_records_enabled_p): New member
2067         function.
2068         (dump_context::set_json_writer): New member function.
2069         (dump_context::emit_optinfo): New member function.
2070         (dump_context::m_forcibly_enable_optinfo): Delete.
2071         (dump_context::m_json_writer): New member data.
2072         * dumpfile.c (dump_context::set_json_writer): New member function.
2073         (dump_context::finish_any_json_writer): New member function.
2074         (dump_context::end_scope): Replace call to
2075         optimization_records_maybe_pop_dump_scope with call to
2076         m_json_writer->pop_scope.
2077         (dump_context::optinfo_enabled_p): New member function.
2078         (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
2079         to dump_context::emit_optinfo.
2080         (dump_context::emit_optinfo): New member function.
2081         (temp_dump_context::temp_dump_context): Replace
2082         m_forcibly_enable_optinfo with call to set_json_writer.
2083         (temp_dump_context::~temp_dump_context): Clean up any json writer.
2084         * optinfo-emit-json.cc (class optrecord_json_writer): Move to
2085         optinfo-emit-json.h
2086         (the_json_writer): Delete.
2087         (optimization_records_start): Delete.
2088         (optimization_records_finish): Delete.
2089         (optimization_records_enabled_p): Delete, in favor of
2090         dump_context::optimization_records_enabled_p.
2091         (optimization_records_maybe_record_optinfo): Delete.
2092         (optimization_records_maybe_pop_dump_scope): Delete.
2093         * optinfo-emit-json.h: Include "json.h".  Delete forward
2094         decl of opt_pass.
2095         (optimization_records_start): Delete.
2096         (optimization_records_finish): Delete.
2097         (optimization_records_enabled_p): Delete.
2098         (optimization_records_maybe_record_optinfo): Delete.
2099         (optimization_records_maybe_pop_dump_scope): Delete.
2100         (class optrecord_json_writer): Move here from
2101         optinfo-emit-json.cc.
2102         * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
2103         to optinfo::emit with call to dump_context::emit_optinfo.
2104         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
2105         (optinfo_enabled_p): Delete, in favor of
2106         dump_context::optinfo_enabled_p.
2107         (optinfo_wants_inlining_info_p): Update for conversion o
2108         optimization_records_enabled_p to a member function of
2109         dump_context.
2110         * optinfo.h (optinfo_enabled_p): Delete, in favor of
2111         dump_context::optinfo_enabled_p.
2112         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
2113         * toplev.c: Include "dump-context.h".
2114         (compile_file): Replace call to optimization_records_finish with
2115         dump_context::finish_any_json_writer.
2116         (do_compile): Replace call to optimization_records_start with
2117         conditionally creating a optrecord_json_writer for the
2118         dump_context.
2120 2018-11-19  Jakub Jelinek  <jakub@redhat.com>
2122         PR tree-optimization/88071
2123         * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
2124         of false as last argument to gsi_remove.
2125         * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
2126         false as last argument to gsi_replace.
2128         PR debug/87039
2129         * omp-expand.c: Don't include debug.h.
2130         (adjust_context_and_scope): Add REGION argument.  Find DECL_CONTEXT
2131         from innermost outer parallel, task, teams or target that has a
2132         child_fn set, or, if there is no such outer region, use
2133         current_function_decl.  Do the DECL_CONTEXT adjustment regardless of
2134         whether a suitable BLOCK is found or not.
2135         (expand_parallel_call, expand_teams_call): Don't call
2136         adjust_context_and_scope here.
2137         (grid_expand_target_grid_body): Revert 2017-01-25 changes.
2138         (expand_omp_taskreg, expand_omp_target): Likewise.  Call
2139         adjust_context_and_scope.
2140         * dwarf2out.c (dwarf2out_early_global_decl): For
2141         decl_function_context recurse instead of calling dwarf2out_decl.
2143 2018-11-20  Eric Botcazou  <ebotcazou@adacore.com>
2145         PR rtl-optimization/85925
2146         * rtl.h (word_register_operation_p): New predicate.
2147         * combine.c (record_dead_and_set_regs_1): Only apply specific handling
2148         for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
2149         * rtlanal.c (nonzero_bits1): Likewise.  Adjust couple of comments.
2150         (num_sign_bit_copies1): Likewise.
2152 2018-11-19  Richard Biener  <rguenther@suse.de>
2154         PR lto/87229
2155         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
2156         sizepos values.
2158 2018-11-19  Eric Botcazou  <ebotcazou@adacore.com>
2160         * compare-elim.c (struct comparison): Add not_in_a field.
2161         (is_not): New static function.
2162         (strip_not): Likewise.
2163         (conforming_compare): Handle a NOT in the first operand.
2164         (can_eliminate_compare): Likewise.
2165         (find_comparison_dom_walker::before_dom_children): Likewise.
2166         (try_eliminate_compare): Likewise.
2167         * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
2168         (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
2170 2018-11-19  Jonathan Wakely  <jwakely@redhat.com>
2172         * doc/extend.texi (Common Type Attributes): Fix typos.
2174 2018-11-19  Martin Liska  <mliska@suse.cz>
2176         PR gcov-profile/88045
2177         * coverage.c (coverage_begin_function): Add assert.
2179 2018-11-18  Sandra Loosemore  <sandra@codesourcery.com>
2181         PR other/40498
2182         * doc/extend.texi (Common Function Attributes): Document that
2183         no_instrument_function applies to -p and -pg, too.
2184         * doc/invoke.texi (Instrumentation Options): Add cross-references
2185         to docs for -p, -pg, and -finstrument-functions.
2187 2018-11-18  Alan Modra  <amodra@gmail.com>
2189         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
2190         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
2191         (GNU_USER_TARGET_CRTN): Define.
2192         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
2193         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
2195 2018-11-18  Alan Modra  <amodra@gmail.com>
2197         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
2198         !HAVE_LD_PIE variant.
2199         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
2201 2018-11-17  Nick Clifton  <nickc@redhat.com>
2202             Sandra Loosemore  <sandra@codesourcery.com>
2204         PR driver/31357
2205         * doc/invoke.texi (Overall Options): Document that --help and
2206         --help=<value> options cannot be combined.
2208 2018-11-17  Jim Wilson  <jimw@sifive.com>
2210         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
2211         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
2212         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
2213         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
2214         to step2 if frame_pointer_needed and step1 is 0.
2216 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
2218         PR c++/4225
2219         * doc/cpp.texi (System Headers): Add note about implicit
2220         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
2222 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
2224         PR c++/4025
2225         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
2226         -fno-implicit-templates.
2228 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
2230         PR tree-optimization/87546
2231         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
2232         min_precision variable, initially set it to orig_precision, only does
2233         something if op_type's precision is <= min_precision and update
2234         min_precision whenever calling set_op.
2236 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
2238         PR ipa/87957
2239         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
2241 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
2243         * doc/invoke.texi (Option Summary): Fix whitespace and line
2244         breaks in @gccoptlist environments.
2245         (Warning Options): Likewise.
2246         (Optimize Options): Likewise.
2247         (PowerPC SPE Options): Likewise.
2248         (RS/6000 and PowerPC Options): Likewise.
2250 2018-11-16  Jeff Law  <law@redhat.com>
2252         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
2253         gcc_assert.
2255 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
2257         PR middle-end/23197
2258         * doc/invoke.texi (Optimize Options): Update options enabled by
2259         fprofile-generate, -fprofile-use, and -fauto-profile.
2261 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2263         PR target/87927
2264         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
2265         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
2266         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
2267         the byte_op hook.
2268         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
2269         * targhooks.c (default_print_patchable_function_entry): Assert
2270         asm_int_op does not return a NULL string.
2271         * varasm.c (integer_asm_op): Return the op for a partial int type
2272         when the requested size does not correspond to an integer type.
2273         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
2274         * doc/tm.texi: Regenerate.
2276 2018-11-12  Jason Merrill  <jason@redhat.com>
2278         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
2280 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
2282         * config/rs6000/constraints.md (wF constraint): Remove power9
2283         fusion documentation.  Just document wF's use for power8 fusion.
2284         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
2285         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
2286         fusion support to require the upper 12 bits to be all 0's or all
2287         1's.
2288         (fusion_gpr_mem_load): Add comment.
2289         (fusion_addis_mem_combo_load): Remove power9 fusion support.
2290         (fusion_addis_mem_combo_store): Delete.
2291         (fusion_offsettable_mem_operand): Delete.
2292         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
2293         power8 fusion here.
2294         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
2295         (POWERPC_MASKS): Delete power9 fusion.
2296         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
2297         (fusion_p9_p): Delete.
2298         (expand_fusion_p9_load): Delete.
2299         (expand_fusion_p9_store): Delete.
2300         (emit_fusion_p9_load): Delete.
2301         (emit_fusion_p9_store): Delete.
2302         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
2303         fusion support.
2304         (rs6000_option_override_internal): Set power8 fusion based on
2305         whether we are tuning for power8.  Delete power9 fusion support.
2306         (rs6000_opt_masks): Delete -mpower9-fusion switch.
2307         (emit_fusion_load): Rename emit_fusion_load_store to
2308         emit_fusion_load, and drop fusion store support.  Update callers.
2309         (emit_fusion_load_store): Likewise.
2310         (emit_fusion_gpr_load): Likewise.
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.md (UNSPEC_FUSION_P9): Delete.
2317         (GPR_FUSION): Delete.
2318         (FPR_FUSION): Delete.
2319         (power9 fusion peephole2s): Delete.
2320         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
2321         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
2322         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
2323         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
2324         (fusion_p9_<mode>_constant): Delete.
2325         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
2326         power9 fusion switch.
2327         * doc/md.texi (PowerPC constraints): Update wF constraint
2328         documentation for power8 fusion only.
2330 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
2332         PR rtl-optimization/87475
2333         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
2334         CROSSING_JUMP_P insns.
2335         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
2337 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
2339         PR target/88051
2340         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
2341         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
2342         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
2344 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
2346         PR middle-end/88032
2347         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
2348         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
2350 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
2352         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
2353         location_t parameter.
2354         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2355         (switch_decision_tree::do_jump_if_equal): Likewise.
2356         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
2357         of switch statement to emit_case_nodes.
2358         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
2359         set it on the newly built GIMPLE comparison statement.
2360         (switch_decision_tree::do_jump_if_equal): Likewise.
2361         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
2362         in calls to do_jump_if_equal as well as recursive calls.
2364 2018-11-16  Andi Kleen  <ak@linux.intel.com>
2366         * config/i386/i386.md: Allow memory operands to ptwrite.
2368 2018-11-16  Michael Matz  <matz@suse.de>
2370         * system.h (PRsa): New macro.
2371         (SIZE_AMOUNT): Cast number to uint64_t.
2372         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
2373         (pool_usage::dump_footer): Likewise and also use PRIu64.
2374         * bitmap.h (bitmap_usage::dump): Likewise.
2375         * ggc-common.c (ggc_usage::dump): Likewise.
2376         * ggc-page.c (ggc_print_statistics): Likewise.
2377         * input.c (dump_line_table_statistics): Likewise.
2378         * mem-stats.h (mem_usage::dump): Likewise.
2379         (mem_usage::dump_footer): Likewise.
2380         * rtl.c (dump_rtx_statistics): Likewise.
2381         * tree-cfg.c (dump_cfg_stats): Likewise.
2382         * tree-dfa.c (dump_dfa_stats): Likewise.
2383         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2384         * tree-ssanames (ssanames_print_statistics): Likewise.
2385         * vec.c (vec_usage::dump): Likewise.
2386         (vec_usage::dump_footer): Likewise.
2388 2018-11-16  Richard Biener  <rguenther@suse.de>
2390         PR tree-optimization/88011
2391         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
2392         replacing set_value_range_to_undefined and
2393         set_value_range_to_varying with method calls.
2395 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
2397         * config/s390/s390.md
2398         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
2399         assembler.
2400         (*r<noxa>sbg_<mode>_sll): Likewise.
2401         (*r<noxa>sbg_<mode>_srl): Likewise.
2402         (*r<noxa>sbg_sidi_srl): New pattern.
2404 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
2406         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
2407         Default to TARGET_VXWORKS_RTP.
2408         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
2409         of TARGET_VXWORKS_RTP.
2410         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
2411         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
2412         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
2414 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
2416         PR c++/25759
2417         * doc/extend.texi (Common Type Attributes): Make it explicit
2418         that attribute "packed" can apply to C++ classes.
2420 2018-11-15  Martin Sebor  <msebor@redhat.com>
2422         PR c++/87541
2423         PR c++/87542
2424         * tree.c (type_argument_type): New function.
2425         * tree.h (type_argument_type): Declare it.
2426         * gcc/doc/extend.texi (alloc_align): Update and clarify.
2427         (alloc_size, nonnull, sentinel): Same.
2429 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
2430             Kwok Cheung Yeung  <kcy@codesourcery.com>
2432         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
2433         int_mode_for_size fails.
2434         (vectorizable_load): Likewise.
2436 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
2438         * doc/ux.texi (Group logically-related diagnostics): Move
2439         discussion of auto_diagnostic_group into this new subsection.
2440         Give an example of where this grouping is used.
2442 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
2444         PR other/19165
2445         * Makefile.in (OBJS): Move json.o to...
2446         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
2447         * common.opt (fdiagnostics-format=): New option.
2448         (diagnostics_output_format): New enum.
2449         * diagnostic-format-json.cc: New file.
2450         * diagnostic.c (default_diagnostic_final_cb): New function, taken
2451         from start of diagnostic_finish.
2452         (diagnostic_initialize): Initialize final_cb to
2453         default_diagnostic_final_cb.
2454         (diagnostic_finish): Move "being treated as errors" messages to
2455         default_diagnostic_final_cb.  Call any final_cb.
2456         (default_diagnostic_finalizer): Add diagnostic_t param.
2457         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
2458         diagnostic_finalizer callback.
2459         * diagnostic.h (enum diagnostics_output_format): New enum.
2460         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
2461         (struct diagnostic_context): Add "final_cb".
2462         (default_diagnostic_finalizer): Add diagnostic_t param.
2463         (diagnostic_output_format_init): New decl.
2464         * doc/invoke.texi (-fdiagnostics-format): New option.
2465         * dwarf2out.c (gen_producer_string): Ignore
2466         OPT_fdiagnostics_format_.
2467         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
2468         * lto-wrapper.c (append_diag_options): Ignore it.
2469         * opts.c (common_handle_option): Handle it.
2471 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
2473         PR tree-optimization/88015
2474         * graphite-isl-ast-to-gimple.c
2475         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
2476         for dump_enabled_p.
2477         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
2479 2018-11-15  Richard Biener  <rguenther@suse.de>
2481         PR middle-end/88029
2482         * gimple.c (gimple_call_flags): Union flags from decl, type
2483         and call fntype.
2484         * trans-mem.c (is_tm_pure_call): Simplify.
2486 2018-11-15  Richard Biener  <rguenther@suse.de>
2488         PR tree-optimization/88031
2489         * tree-vect-loop.c (vectorizable_reduction): Move check
2490         for multiple types earlier so we get the expected dump.
2491         Simplify calls to vectorizable_condition.
2492         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
2493         * tree-vect-stmts.c (vectorizable_condition): Instead of
2494         reduc_def and reduc_index take just a flag.  Simplify
2495         code-generation now that we can rely on the defs being set up.
2496         (vectorizable_comparison): Remove unused argument.
2498 2018-11-15  Richard Biener  <rguenther@suse.de>
2500         PR tree-optimization/88030
2501         * tree-complex.c (need_eh_cleanup): New global.
2502         (update_complex_assignment): Mark blocks that need EH update.
2503         (expand_complex_comparison): Likewise.
2504         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
2505         perform EH cleanup and schedule CFG cleanup if that did anything.
2507 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
2509         PR rtl-optimization/88018
2510         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
2511         on the presence of fallthru edge, rather than if it is a USE or not.
2513 2018-11-15  Richard Biener  <rguenther@suse.de>
2515         PR middle-end/87917
2516         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
2517         analyze_subscript_affine_affine properly.
2519 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2521         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
2522         (GLIBC_DYNAMIC_LINKER): Define.
2523         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
2525 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
2527         PR tree-optimization/84648
2528         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
2529         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
2530         by calling adjust_cond_for_loop_until_wrap.
2532 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
2534         PR other/56334
2535         * doc/extend.texi (Common Function Attributes): Clarify linker
2536         restrictions on "aligned" attribute.
2537         (Common Variable Attributes): Likewise.  Mention that linker
2538         restrictions don't apply to stack-allocated variables.
2540 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
2542         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
2543         as an lvalue.
2545 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
2547         PR lto/55102
2548         PR lto/56700
2549         * doc/invoke.texi (Optimize Options): Remove bad example about
2550         interaction between -flto and -O.  Replace it with a note that
2551         you need to compile with -O and not just link.  Copy-edit -flto
2552         discussion to reduce verbiage and improve flow.
2554 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
2556         PR middle-end/59658
2557         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
2558         also suppress many optimizations.  Alphabetize option lists for
2559         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
2560         correct documentation for those options to say that.
2561         * opts.c (default_options_table): Sort table by level and option
2562         name, to make it easier to correlate to the manual.
2564 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
2566         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
2567         printing of the names of segment registers.
2569 2018-11-14  Nathan Sidwell  <nathan@acm.org>
2571         PR debug/88006
2572         PR debug/87462
2573         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
2574         type list.
2576 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
2578         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
2580 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
2582         P1236R1 - Signed integers are two's complement
2583         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
2585         PR bootstrap/86739
2586         * hash-map.h (hash_map::iterator::reference_pair): New class.
2587         (hash_map::iterator::operator*): Return it rather than std::pair.
2589 2018-11-14  Jeff Law  <law@redhat.com>
2591         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
2592         iff the operand is a constant.
2594 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
2596         * gimple-ssa-evrp-analyze.c
2597         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
2598         ignore_equivs_equal_p to equal_p.
2599         * ipa-cp.c (meet_with_1): Use equal_p instead of
2600         ignore_equivs_equal_p.
2601         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
2602         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
2603         (value_range::operator==): Remove.
2604         (value_range::operator!=): Remove.
2605         (vrp_prop::visit_stmt): Use equal_p.
2606         * tree-vrp.h (value_range): Remove operator==, operator!=,
2607         ignore_equivs_equal_p.
2608         * vr-values.c (update_value_range): Use equal_p.
2610 2018-11-14  Michael Matz  <matz@suse.de>
2612         PR middle-end/86575
2613         * gimplify.c (collect_fallthrough_labels): Add new argument,
2614         return location via that, don't modify statements.
2615         (warn_implicit_fallthrough_r): Adjust call, don't use
2616         statement location directly.
2618 2018-11-14  Richard Biener  <rguenther@suse.de>
2620         PR middle-end/87985
2621         * tree-data-ref.c (split_constant_offset): Add wrapper
2622         allocating a cache hash-map.
2623         (split_constant_offset_1): Cache results of expanding
2624         expressions from SSA def stmts.
2626 2018-11-14  Richard Biener  <rguenther@suse.de>
2628         PR middle-end/88021
2629         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
2630         to lambda_int.
2631         (lambda_vector_mult_const): Likewise.
2632         (lambda_matrix_right_hermite): Use lambda_int temporaries.
2634 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
2635             Jackson Woodruff  <jackson.woodruff@arm.com>
2637         PR 71026/tree-optimization
2638         * match.pd: Simplify floating point comparisons.
2640 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
2642         PR rtl-optimization/87817
2643         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
2644         last argument has low 8 bits clear, fold to 0.
2646 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
2648         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
2649         MAX_OFILE_ALIGNMENT as needed.
2651 2018-11-14  Richard Biener  <rguenther@suse.de>
2653         PR tree-optimization/88019
2654         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
2655         COND_EXPR arguments to non-trapping overflow.
2657 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
2659         * config/arc/arc.h (reg_class): Reorder registers classes, remove
2660         unused register classes.
2661         (REG_CLASS_NAMES): Likewise.
2662         (REG_CLASS_CONTENTS): Likewise.
2663         (FIXED_REGISTERS): Make lp_count fixed.
2664         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
2665         (PROGRAM_COUNTER_REGNO): Remove.
2666         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
2667         register classes, use constants for register numbers, remove
2668         reg_class_contents references.
2669         (arc_process_double_reg_moves): Add asserts.
2670         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
2671         lra_in_progress predicate.
2672         (arc_init_reg_tables): Remove unused register classes.
2673         (arc_register_move_cost): Likewise.
2674         (arc_preferred_reload_class): Likewise.
2675         (hwloop_optimize): Update rtx patterns involving lp_count register.
2676         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
2677         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
2678         CHEAP_CORE_REGS.  Former one will be used for LRA.
2679         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
2680         one will be used for LRA.
2681         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
2682         one will be used for LRA.
2683         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
2684         Former one will be used for LRA.
2685         ("f"): Delete constraint.
2686         ("k"): Likewise.
2687         ("e"): Likewise.
2688         ("l"): Change it from register constraint to constraint.
2689         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
2690         (movhi_insn): Likewise.
2691         (movsi_insn): Update pattern.
2692         (arc_lp): Likewise.
2693         (dbnz): Likewise.
2694         (stack_tie): Remove 'b' constraint letter.
2695         (R4_REG): Define.
2696         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
2697         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
2698         (R57_REG, R59_REG, PCL_REG): Likewise.
2699         (ILINK1_REGNUM): Renamed to ILINK1_REG.
2700         (ILINK2_REGNUM): Renamed to ILINK2_REG.
2701         (Rgp): Remove.
2702         (SP_REGS): Likewise.
2703         (Rcw): Remove unused reg classes.
2704         * config/arc/predicates.md (dest_reg_operand): Just default on
2705         register_operand predicate.
2706         (mpy_dest_reg_operand): Likewise.
2707         (move_dest_operand): Use macros instead of constants.
2708         * config/arc/arc.opt (mlra): Switch to lra as default.
2710 2018-11-14  Richard Biener  <rguenther@suse.de>
2712         PR tree-optimization/87974
2713         * tree-vect-loop.c (vectorizable_reduction): When computing
2714         the vectorized reduction PHI vector type ignore constant
2715         and external defs.
2717 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
2719         PR tree-optimization/87977
2720         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
2721         stmt, build a new one and replace the old one with it.  Formatting fix.
2722         Call release_ssa_name (x) if !has_other_use and !delete_div.
2723         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
2724         verify lhs of stmt is still def.
2726 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
2728         PR rtl-optimization/87507
2729         * lower-subreg.c (operand_for_swap_move_operator): New function.
2730         (simple_move): Strip simple operators.
2731         (find_pseudo_copy): Likewise.
2732         (resolve_operand_for_swap_move_operator): New function.
2733         (resolve_simple_move): Strip simple operators and swap operands.
2735 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
2737         PR tree-optimization/87898
2738         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
2739         (ipa_simd_modify_function_body): Remove debug stmts where the first
2740         argument was changed into a non-decl.
2742 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
2744         PR rtl-optimization/87899
2745         * lra-lives.c (start_living): Update white space in comment.
2746         (enum point_type): New.
2747         (sparseset_contains_pseudos_p): New function.
2748         (update_pseudo_point): Likewise.
2749         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
2750         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
2751         handling.  Move early exit after adding conflicts.
2752         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
2753         if regno is already live.  Remove all handling of program points.
2754         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
2755         after adding conflicts.  Remove all handling of program points and
2756         ignore_reg_for_conflicts.
2757         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
2758         and do not guard call to mark_pseudo_live.
2759         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
2760         and do not guard call to mark_pseudo_dead.
2761         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
2762         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
2763         Use new function update_pseudo_point.  Handle register copies by
2764         removing the source register from the live set.  Handle INOUT operands.
2765         Update to the next program point using the unused_set, dead_set and
2766         start_dying sets.
2767         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
2769 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2771         * builtins.c: Replace "source_location" with "location_t".
2772         * diagnostic-show-locus.c: Likewise.
2773         * diagnostic.c: Likewise.
2774         * dumpfile.c: Likewise.
2775         * gcc-rich-location.h: Likewise.
2776         * genmatch.c: Likewise.
2777         * gimple.h: Likewise.
2778         * gimplify.c: Likewise.
2779         * input.c: Likewise.
2780         * input.h: Likewise.  Eliminate the typedef.
2781         * omp-expand.c: Likewise.
2782         * selftest.h: Likewise.
2783         * substring-locations.h (get_source_location_for_substring):
2784         Rename to...
2785         (get_location_within_string): ...this.
2786         * tree-cfg.c: Replace "source_location" with "location_t".
2787         * tree-cfgcleanup.c: Likewise.
2788         * tree-diagnostic.c: Likewise.
2789         * tree-into-ssa.c: Likewise.
2790         * tree-outof-ssa.c: Likewise.
2791         * tree-parloops.c: Likewise.
2792         * tree-phinodes.c: Likewise.
2793         * tree-phinodes.h: Likewise.
2794         * tree-ssa-loop-ivopts.c: Likewise.
2795         * tree-ssa-loop-manip.c: Likewise.
2796         * tree-ssa-phiopt.c: Likewise.
2797         * tree-ssa-phiprop.c: Likewise.
2798         * tree-ssa-threadupdate.c: Likewise.
2799         * tree-ssa.c: Likewise.
2800         * tree-ssa.h: Likewise.
2801         * tree-vect-loop-manip.c: Likewise.
2803 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2805         * doc/extend.texi: Fix typo in the weakref description.
2807 2018-11-13  Richard Biener  <rguenther@suse.de>
2809         PR tree-optimization/86991
2810         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
2811         group building until we have successfully detected the SLP
2812         reduction.
2813         (vect_is_simple_reduction): Remove fixup code here.
2815 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2817         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
2818         (dump_gimple_stmt): Use it.
2819         (dump_gimple_stmt_loc): Likewise.
2820         (dump_gimple_expr): Likewise.
2821         (dump_gimple_expr_loc): Likewise.
2822         (dump_generic_expr): Likewise.
2823         (dump_generic_expr_loc): Likewise.
2824         (dump_printf): Likewise.
2825         (dump_printf_loc): Likewise.
2826         (dump_dec): Likewise.
2827         (dump_dec): Likewise.
2828         (dump_hex): Likewise.
2829         (dump_symtab_node): Likewise.
2830         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
2831         Guard dump call with dump_enabled_p.
2832         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
2833         * graphite-optimize-isl.c (optimize_isl): Likewise.
2834         * graphite.c (graphite_transform_loops): Likewise.
2835         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
2836         * tree-parloops.c (parallelize_loops): Likewise.
2837         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
2838         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
2839         (vect_prune_runtime_alias_test_list): Likewise.
2840         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
2841         (vect_estimate_min_profitable_iters): Likewise.
2842         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
2843         (vect_build_slp_tree_2): Likewise.
2844         (vect_supported_load_permutation_p): Likewise.
2845         (vect_slp_analyze_operations): Likewise.
2846         (vect_slp_analyze_bb_1): Likewise.
2847         (vect_slp_bb): Likewise.
2848         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
2849         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
2850         (pass_slp_vectorize::execute): Likewise.
2851         (increase_alignment): Likewise.
2853 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2855         PR ipa/87955
2856         * ipa-inline.c (report_inline_failed_reason): Guard calls to
2857         cl_target_option_print_diff and cl_optimization_print_diff with
2858         if (dump_file).
2860 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
2862         * doc/invoke.texi (-fsave-optimization-record): Note that the
2863         output is compressed.
2864         * optinfo-emit-json.cc: Include <zlib.h>.
2865         (optrecord_json_writer::write): Compress the output.
2867 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
2869         * tree-vrp.c (value_range_base::dump): Dump type.
2870         Do not use INF nomenclature for 1-bit types.
2871         (dump_value_range): Group all variants to common dumping code.
2872         (debug): New overloaded functions for value_ranges.
2873         (value_range_base::dump): Remove no argument version.
2874         (value_range::dump): Same.
2876 2018-11-13  Richard Biener  <rguenther@suse.de>
2878         PR tree-optimization/87931
2879         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
2880         nested cycles we support to latch computations vectorizable_reduction
2881         handles.
2883 2018-11-13  Martin Liska  <mliska@suse.cz>
2885         PR tree-optimization/87885
2886         * cfghooks.c (account_profile_record): Rename
2887         to ...
2888         (profile_record_check_consistency): ... this.
2889         Calculate missing num_mismatched_freq_in.
2890         (profile_record_account_profile): New function
2891         that calculates time and size of a function.
2892         * cfghooks.h (struct profile_record): Remove
2893         all tuples.
2894         (struct cfg_hooks): Remove after_pass flag.
2895         (account_profile_record): Rename to ...
2896         (profile_record_check_consistency): ... this.
2897         (profile_record_account_profile): New.
2898         * cfgrtl.c (rtl_account_profile_record): Remove
2899         after_pass flag.
2900         * passes.c (check_profile_consistency): Do only
2901         checking.
2902         (account_profile): Calculate size and time of
2903         function only.
2904         (pass_manager::dump_profile_report): Reformat
2905         output.
2906         (execute_one_ipa_transform_pass): Call
2907         consistency check before clean upand call account_profile
2908         after a clean up is done.
2909         (execute_one_pass): Call check_profile_consistency and
2910         account_profile instead of using after_pass flag..
2911         * tree-cfg.c (gimple_account_profile_record): Likewise.
2913 2018-11-13  Martin Liska  <mliska@suse.cz>
2915         PR sanitizer/87930
2916         * config/i386/i386.c (ix86_option_override_internal): Error
2917         about usage -mabi=ms and -fsanitize={,kernel-}address.
2919 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
2921         * config/aarch64/aarch64.c
2922         (aarch64_vectorize_preferred_vector_alignment): Change return type to
2923         poly_uint64.
2924         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
2925         alignment being a poly int.
2926         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
2927         return type to poly_uint64.
2928         * target.def (default_preferred_vector_alignment): Likewise.
2929         * targhooks.c (default_preferred_vector_alignment): Likewise.
2930         * targhooks.h (default_preferred_vector_alignment): Likewise.
2931         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
2932         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
2933         poly int.
2934         (vect_update_misalignment_for_peel): Likewise.
2935         (vect_enhance_data_refs_alignment): Likewise.
2936         (vect_find_same_alignment_drs): Likewise.
2937         (vect_duplicate_ssa_name_ptr_info): Likewise.
2938         (vect_setup_realignment): Likewise.
2939         (vect_can_force_dr_alignment_p): Change alignment parameter type to
2940         poly_uint64.
2941         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
2942         mask with a compile time variable vector alignment.
2943         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
2944         int.
2945         (vect_do_peeling): Exit early if vector alignment is not constant.
2946         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
2947         a poly int.
2948         (vectorizable_store): Likewise.
2949         (vectorizable_load): Likweise.
2950         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
2951         poly_uint64.
2952         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
2953         poly int.
2954         (vect_can_force_dr_alignment_p): Change alignment parameter type to
2955         poly_uint64.
2957 2018-11-13  Richard Biener  <rguenther@suse.de>
2959         PR tree-optimization/87962
2960         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
2961         detect outer reduction for disqualifying in-loop uses.
2963 2018-11-13  Richard Biener  <rguenther@suse.de>
2965         PR tree-optimization/87967
2966         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
2967         for constants for the scalar loop.
2969 2018-11-13  Alan Modra  <amodra@gmail.com>
2971         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
2972         integer constants when -mcmodel=medium.
2974 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
2976         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
2977         * config/arc/elf.h (PROFILE_HOOK): Define.
2978         * config/arc/linux.h (PROFILE_HOOK): Likewise.
2980 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
2982         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
2983         jump table data in the text section.
2985 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
2987         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
2988         to fit the eh_return pattern.
2989         * config/arc/arc.md (eh_return): Define.
2990         (VUNSPEC_ARC_EH_RETURN): Likewise.
2991         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
2992         new implementation.
2993         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
2996 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
2998         * common/config/arc/arc-common.c (arc_option_optimization_table):
2999         Millicode optimization is default on for size optimizations.
3000         * config/arc/arc-protos.h (arc_check_multi): New function.
3001         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
3002         (ENTER_LEAVE_START_REG): Define.
3003         (ENTER_LEAVE_END_REG): Likewise.
3004         (arc_override_options): Disable millicode when long calls option
3005         is on.
3006         (arc_frame_info): Change it from int to bool.
3007         (arc_compute_frame_size): Clean up.
3008         (arc_save_restore): Remove.
3009         (frame_save_reg): New function.
3010         (frame_restore_reg): Likewise.
3011         (arc_enter_leave_p): Likewise.
3012         (arc_save_callee_saves): Likewise.
3013         (arc_restore_callee_saves): Likewise.
3014         (arc_save_callee_enter): Likewise.
3015         (arc_restore_callee_leave): Likewise.
3016         (arc_save_callee_milli): Likewise.
3017         (arc_restore_callee_milli): Likewise.
3018         (arc_expand_prologue): Reimplement to emit enter/leave
3019         instructions.
3020         (arc_expand_epilogue): Likewise.
3021         (arc_check_multi): New function.
3022         * config/arc/arc.md (push_multi_fp): New pattern.
3023         (push_multi_fp_blink): Likewise.
3024         (pop_multi_fp): Likewise.
3025         (pop_multi_fp_blink): Likewise.
3026         (pop_multi_fp_ret): Likewise.
3027         (pop_multi_fp_blink_ret): Likewise.
3028         * config/arc/arc.opt (mmillicode): Update option.
3029         (mcode-density-frame): New option.
3030         * config/arc/predicates.md (push_multi_operand): New predicate.
3031         (pop_multi_operand): Likewise.
3032         * doc/invoke.texi (ARC): Update ARC options information.
3035 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
3037         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
3038         * config/arc/arc.c (operands_ok_ldd_std): New function.
3039         (mem_ok_for_ldd_std): Likewise.
3040         (gen_operands_ldd_std): Likewise.
3041         * config/arc/arc.md: Add peephole2 rules for std/ldd.
3043 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
3045         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
3046         test on stack_usage_file.
3047         (lang_dependent_init): Do not open the .su file if generating LTO.
3049 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
3051         PR rtl-optimization/87918
3052         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
3053         simplify_gen_relational rather than simplify_gen_binary.
3055 2018-11-13  Richard Biener  <rguenther@suse.de>
3057         * tree-ssanames.h (set_range_info): Use value_range_base.
3058         (get_range_info): Likewise.
3059         * tree-ssanames.c (set_range_info): Likewise.
3060         (get_range_info): Likewise.
3061         * tree-vrp.c (value_range_base::union_helper): Split
3062         out common parts of value_range[_base]::union_.
3063         (value_range_base::union_): Update.
3064         (value_range::union_): Likewise.
3065         (determine_value_range_1): Use value_range_base.
3066         (determine_value_range): Likewise.
3067         * tree-vrp.h (value_range_base::union_helper): Move ...
3068         (value_range::union_helper): ... from here.
3070 2018-11-13  Alan Modra  <amodra@gmail.com>
3072         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
3073         offset for PRE_DEC.
3074         (rs6000_secondary_reload_gpr): Don't call find_replacement.
3076 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
3078         PR middle-end/59634
3079         * doc/invoke.texi (Optimize Options): Clarify that the
3080         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
3081         apply to data cache size.
3083 2018-11-13  Alan Modra  <amodra@gmail.com>
3085         * config/rs6000/predicates.md (logical_const_operand),
3086         (logical_operand): Correct comment.
3087         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
3089 2018-11-13  Alan Modra  <amodra@gmail.com>
3091         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
3093 2018-11-13  Alan Modra  <amodra@gmail.com>
3095         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
3096         description.
3097         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
3098         covered by alternative.
3099         (movcc_internal1): Ignore h for register preference.
3100         (mov<mode>_hardfloat64): Likewise.
3101         (mov<mode>_softfloat): Ignore c, l, h for register preference.
3103 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
3105         PR preprocessor/47823
3106         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
3107         documentation to...
3108         (Pragmas): ...here.
3109         * doc/extend.texi (Pragmas): Note additional pragmas documented
3110         in the CPP manual.
3112 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
3114         PR middle-end/86677
3115         PR middle-end/87528
3116         * tree-scalar-evolution.c (expression_expensive_p): Make
3117         BUILTIN POPCOUNT as expensive when backend does not define it.
3119 2018-11-12  Fredrik Noring  <noring@nocrew.org>
3121         * config.gcc: Update with-llsc defaults for MIPS r5900.
3123 2018-11-12  Martin Liska  <mliska@suse.cz>
3125         PR gcov-profile/87442
3126         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
3127         options.
3128         * doc/invoke.texi: Document them.
3129         * tree-profile.c (parse_profile_filter): New.
3130         (parse_profile_file_filtering): Likewise.
3131         (release_profile_file_filtering): Likewise.
3132         (include_source_file_for_profile): Likewise.
3133         (tree_profiling): Filter source files based on the
3134         newly added options.
3136 2018-11-12  Martin Liska  <mliska@suse.cz>
3138         PR target/87903
3139         * doc/extend.texi: Add missing values for __builtin_cpu_is and
3140         __builtin_cpu_supports for x86 target.
3142 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
3144         PR middle-end/21110
3145         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
3146         have to be Pmode.
3147         (Arithmetic): Likewise for "lo_sum".
3149 2018-11-12  Renlin Li  <renlin.li@arm.com>
3151         PR target/87815
3152         * dse.c (get_stored_val): Add check for compile-time constantness
3153         of gap.
3155 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
3157         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
3158         (ARMv8_5a): New fgroup.
3159         (armv8.5-a): New arch.
3160         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
3161         sb and predres.
3162         * config/arm/arm-tables.opt: Regenerate.
3163         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
3164         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
3165         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
3166         Add matching rules for -march=armv8.5-a and extensions.
3167         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
3168         Add sb and predres to all armv8-a except armv8.5-a.
3170 2018-11-12  Richard Biener  <rguenther@suse.de>
3172         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
3173         overload for single value.
3174         (value_range[_base]::set_nonnull): New.
3175         (value_range[_base]::set_null): Likewise.
3176         (value_range): Document bitmap copying behavior, mark
3177         copy constructor and assignment operator deleted.
3178         (value_range::move): New.
3179         (value_range::set_and_canonicalize): Default bitmap to zero.
3180         (set_value_range_to_nonnull): Remove.
3181         (set_value_range_to_null): Likewise.
3182         (set_value_range): Likewise.
3183         (set_value_range_to_value): Likewise.
3184         (extract_range_from_unary_expr): Work on value_range_base.
3185         (extract_range_from_binary_expr_1): Likewise.  Rename to...
3186         (extract_range_from_binary_expr): ... this.
3187         * tree-vrp.c (value_range::update): Clear equiv bitmap
3188         if required.
3189         (value_range::move): New, move equiv bitmap.
3190         (value_range_base::set_undefined): Avoid assignment.
3191         (value_range::set_undefined): Likewise.
3192         (value_range_base::set_varying): Likewise.
3193         (value_range::set_varying): Likewise.
3194         (set_value_range): Remove.
3195         (value_range_base::set): New overload for value.
3196         (value_range::set): Likewise.
3197         (set_value_range_to_nonnull): Remove.
3198         (value_range_base::set_nonnull): New.
3199         (value_range::set_nonnull): Likewise.
3200         (set_value_range_to_null): Remove.
3201         (value_range_base::set_null): New.
3202         (value_range::set_null): Likewise.
3203         (range_is_null): Work on value_range_base.
3204         (range_is_nonnull): Likewise.
3205         (ranges_from_anti_range): Likewise.
3206         (extract_range_into_wide_ints): Likewise.
3207         (extract_range_from_multiplicative_op): Likewise.
3208         (extract_range_from_binary_expr): Likewise.  Update for API changes.
3209         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
3210         handling.
3211         (value_range::intersect_helper): Avoid copy and assignment.
3212         (value_range::union_helper): Likewise.
3213         (determine_value_range_1): Adjust.
3214         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
3215         Avoid assignment by using move.
3216         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
3217         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3218         Likewise.
3219         * tree-ssanames.c (get_range_info): Likewise.
3220         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
3221         * vr-values.c (vr_values::get_value_range): Adjust.
3222         (vr_values::update_value_range): Likewise.
3223         (symbolic_range_based_on_p): Work on value_range_base.
3224         (vr_values::extract_range_from_binary_expr): Use value_range_base.
3225         (vr_values::extract_range_from_unary_expr): Likewise.
3226         (vr_values::extract_range_from_cond_expr): Avoid assignment.
3227         (vr_values::extract_range_from_comparison): Adjust.
3228         (vr_values::check_for_binary_op_overflow): Use value_range_base.
3229         (vr_values::extract_range_basic): Adjust.
3230         (vr_values::adjust_range_with_scev): Likewise.
3231         (vr_values::vrp_visit_assignment_or_call): Likewise.
3232         (vr_values::get_vr_for_comparison): Change API to avoid
3233         assignment and copy construction.
3234         (vr_values::compare_name_with_value): Adjust accordingly.
3235         (vr_values::compare_names): Likewise.
3236         (vr_values::extract_range_from_phi_node): Avoid assignment and
3237         bogus in-place modify of equiv bitmap.
3238         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
3239         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
3240         for extract_range_from_unary_expr API change.
3241         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
3243 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
3245         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
3247 2018-11-12  Richard Biener  <rguenther@suse.de>
3249         * tree-vrp.h (value_range_base::symbolic_p,
3250         value_range_base::constant_p, value_range_base::zero_p,
3251         value_range_base::singleton_p): Move from value_range.
3252         (value_range::dump): Add.
3253         * gimple-ssa-evrp-analyze.c
3254         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
3255         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
3256         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3257         Use set_varying.
3258         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
3259         (value_range::constant_p): Likewise.
3260         (value_range::singleton_p): Likewise.
3261         (value_range_base::dump): Add.
3262         (set_value_range_to_undefined): Remove.
3263         (set_value_range_to_varying): Likewise.
3264         (range_int_cst_p): Take value_range_base argument.
3265         (range_int_cst_singleton_p): Likewise.
3266         (value_range_constant_singleton): Likewise.
3267         (vrp_set_zero_nonzero_bits): Likewise.
3268         (extract_range_from_multiplicative_op): Use set_varying.
3269         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
3270         (extract_range_from_unary_expr): Likewise.
3271         (dump_value_range_base): Change to overload of dump_value_range.
3272         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
3273         (vrp_prop::visit_stmt): Likewise.
3274         (value_range::intersect_helper): Likewise.
3275         (value_range::union_helper): Likewise.
3276         (determine_value_range_1): Likewise.
3278 2018-11-12  Richard Biener  <rguenther@suse.de>
3280         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
3281         (set_value_range_to_null): Likewise.
3282         * vr-values.c (vr_values::extract_range_from_comparison):
3283         Clear equiv for constant singleton ranges.
3285 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
3287         * config/i386/sse.md: Combine VFIXUPIMM* patterns
3288         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3289         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3290         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
3291         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3292         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3293         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
3295 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3297         PR c/69502
3298         * doc/extend.texi (Common Type Attributes): For the align type
3299         attribute, copy language about decreasing alignment from the
3300         corresponding variable attribute.
3302 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3304         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
3305         -frounding-math.
3307 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3309         PR c++/43105
3310         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
3311         -frtti and -fno-rtti code.
3313 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
3315         PR c/26366
3316         * doc/extend.texi (Other Builtins): Document probability associated
3317         with __builtin_expect.
3319 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
3321         PR target/87928
3322         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
3323         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
3324         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
3325         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
3327 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3329         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
3331 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3333         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
3335 2018-11-11  Richard Biener  <rguenther@suse.de>
3337         * tree-vrp.h (class value_range_base): New base class for
3338         value_range containing all but the m_equiv member.
3339         (dump_value_range_base): Add.
3340         (range_includes_zero_p): Work on value_range_base.
3341         * tree-vrp.c (value_range_base::set): Split out base handling
3342         from...
3343         (value_range::set): this.
3344         (value_range::set_equiv): New.
3345         (value_range_base::value_range_base): New constructors.
3346         (value_range_base::check): Split out base handling from...
3347         (value_range::check): this.
3348         (value_range::equal_p): Refactor in terms of
3349         ignore_equivs_equal_p which is now member of the base.
3350         (value_range_base::set_undefined): New.
3351         (value_range_base::set_varying): Likewise.
3352         (value_range_base::dump):Split out base handling from...
3353         (value_range::dump): this.
3354         (value_range_base::set_and_canonicalize): Split out base handling
3355         from...
3356         (value_range::set_and_canonicalize): this.
3357         (value_range_base::union_): New.
3358         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
3359         for m_vr.
3360         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
3361         instead of value_range everywhere.
3362         (ipcp_vr_lattice::print): Use dump_value_range_base.
3363         (ipcp_vr_lattice::meet_with): Adjust.
3364         (ipcp_vr_lattice::meet_with_1): Likewise.
3365         (ipa_vr_operation_and_type_effects): Likewise.
3366         (propagate_vr_across_jump_function): Likewise.
3367         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
3368         (ipa_get_value_range): Likewise.
3369         (ipa_set_jfunc_vr): Likewise.
3370         (ipa_compute_jump_functions_for_edge): Likewise.
3372 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
3374         PR middle-end/65703
3375         * doc/invoke.texi (Optimize Options): Add @opindex entries
3376         for the positive forms of -fno-xxx and -mno-xxx options
3377         that were lacking them.
3379 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
3381         * combine.c (make_more_copies): Only make an intermediate copy if the
3382         dest of a move is a pseudo.
3384 2018-11-09  Maya Rashish  <coypu@sdf.org>
3386         PR target/87221
3387         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
3388         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
3390 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
3392         PR driver/41179
3393         PR middle-end/65703
3394         * doc/invoke.texi (Optimize Options): Clarify default behavior
3395         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
3397 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3399         PR tree-optimization/87940
3400         * expr.c (string_constant): Don't strip NOPS in subexpressions.
3401         Fold PLUS_EXPR correctly.
3403 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
3405         PR target/87762
3406         * config/s390/s390.c (s390_safe_relative_long_p): New function.
3407         (annotate_constant_pool_refs): Skip insns which support
3408         relative addressing.
3409         (annotate_constant_pool_refs_1): New helper function.
3410         (find_constant_pool_ref): Skip insns which support relative
3411         addression.
3412         (find_constant_pool_ref_1): New helper function.
3413         (replace_constant_pool_ref): Skip insns which support
3414         relative addressing.
3415         (replace_constant_pool_ref_1): New helper function.
3416         (s390_mainpool_start): Adapt to the new signature.
3417         (s390_mainpool_finish): Likewise.
3418         (s390_chunkify_start): Likewise.
3419         (s390_chunkify_finish): Likewise.
3420         (pass_s390_early_mach::execute): Likewise.
3421         (s390_prologue_plus_offset): Likewise.
3422         (s390_emit_prologue): Likewise.
3423         (s390_emit_epilogue): Likewise.
3425 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
3427         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
3428         but unsupported lastprivate with conditional modifier.
3430 2018-11-09  Jeff Law  <law@redhat.com>
3432         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
3433         unused argument better.  Add gcc_unreachable to silence warning.
3435 2018-11-09  Martin Sebor  <msebor@redhat.com>
3437         PR middle-end/81824
3438         * attribs.c (has_attribute): New helper function.
3439         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
3440         * attribs.h (decls_mismatched_attributes): Declare.
3441         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
3442         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
3443         * common.opt (-Wattribute-alias): Take an argument.
3444         (-Wno-attribute-alias): New option.
3445         * doc/extend.texi (Common Function Attributes): Document copy.
3446         (Common Variable Attributes): Same.
3447         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
3448         (-Wattribute-alias): Document new option argument.
3450 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
3452         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
3453         matching criteria.  Remove unused array initializer.
3455 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
3456             Jinsong Ji  <jji@us.ibm.com>
3458         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
3459         constraints by introducing a new temporary.
3460         (_mm_cvtss_si64): Likewise.
3462 2018-11-09  Martin Liska  <mliska@suse.cz>
3464         * common.opt: Add -fipa-stack-alignment flag.
3465         * doc/invoke.texi: Document it.
3466         * final.c (rest_of_clean_state): Guard stack
3467         shrinking with flag.
3469 2018-11-09  Martin Liska  <mliska@suse.cz>
3471         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
3472         to ...
3473         (ipa_discover_variable_flags): ... this.
3474         * common.opt: Come up with new flag -fipa-reference-addressable.
3475         * doc/invoke.texi: Document it.
3476         * ipa-reference.c (propagate): Call the renamed fn.
3477         * ipa-visibility.c (whole_program_function_and_variable_visibility):
3478         Likewise.
3479         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
3480         ...
3481         (ipa_discover_variable_flags): ... this.  Discover
3482         non-addressable variables only with the newly added flag.
3483         * opts.c: Enable the newly added flag with -O1 and higher
3484         optimization level.
3486 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
3488         * json.cc (selftest::test_writing_literals): Fix comment.
3490 2018-11-09  Martin Liska  <mliska@suse.cz>
3492         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
3493         string to a stack buffer.
3494         (aarch64_parse_cpu): Likewise.
3495         (aarch64_parse_tune): Likewise.
3497 2018-11-09  Richard Biener  <rguenther@suse.de>
3499         PR tree-optimization/87953
3500         * tree-vect-loop.c (vectorizable_reduction): For analysis
3501         always pass ops[0] to vectorizable_condition.
3503 2018-11-09  Stafford Horne  <shorne@gmail.com>
3504             Richard Henderson  <rth@twiddle.net>
3505             Joel Sherrill  <joel@rtems.org>
3507         * common/config/or1k/or1k-common.c: New file.
3508         * config/or1k/*: New.
3509         * config.gcc (or1k*-*-*): New.
3510         * configure.ac (or1k*-*-*): New test for openrisc tls.
3511         * configure: Regenerated.
3512         * doc/install.texi: Document OpenRISC triplets.
3513         * doc/invoke.texi: Document OpenRISC arguments.
3514         * doc/md.texi: Document OpenRISC.
3516 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
3518         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
3519         (arm7tdmi-s): Delete CPU.
3520         (arm710t): Add aliases for arm720t and arm740t.
3521         (arm720t, arm740t): Delete CPUs.
3522         (arm920t): Add aliases for arm920, arm922t and arm940t.
3523         (arm920, arm922t, arm940t): Delete CPUs.
3524         (arm10tdmi): Add alias for arm1020t.
3525         (arm1020t): Delete CPU.
3526         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
3527         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
3528         (arm10e): Add aliases for arm1020e and arm1022e.
3529         (arm1020e, arm1022e): Delete CPU.
3530         * config/arm/arm.md (generic_sched): Remove entries that are now
3531         handled by aliases.
3532         (generic_vfp): Likewise.
3533         * config/arm/arm1020e.md: Simplify tuning selection based on alias
3534         changes.
3535         * config/arm/arm-tune.md: Regenerated.
3536         * config/arm/arm-tables.opt: Regenerated.
3538 2018-11-09  Richard Biener  <rguenther@suse.de>
3540         PR tree-optimization/87621
3541         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
3542         op with only phi inputs.
3543         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
3544         (ch_base::copy_headers): Run CSE on copied loop headers.
3545         (pass_ch_vect::process_loop_p): Simplify.
3547 2018-11-09  Alexandre Oliva <oliva@adacore.com>
3549         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
3550         for non-w64 x86_64 biarch.
3552 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
3554         PR rtl-optimization/86438
3555         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
3556         of in_b for the compare if in_b is SET_DEST.
3558         PR target/87793
3559         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
3560         non-toplevel UNSPEC.
3562 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3564         * tree-vrp.c (value_range::check): Do not access internals
3565         directly.
3566         (value_range::singleton_p): Same.
3567         (value_range::type): Same.
3568         (vrp_finalize): Use value_range API.
3570 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3572         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
3574 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3576         * vr-values.c (vr_values::get_value_range): Use value_range API
3577         instead of piecing together ranges.
3578         (vr_values::update_value_range): Same.
3580 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
3582         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
3583         of performing ad-hoc calculations.
3584         * tree-ssanames.c (set_range_info): New overloaded function
3585         accepting value_range &.
3586         (get_range_info): Same.
3587         * tree-ssanames.h (set_range_info_raw): Remove.
3588         (set_range_info): New prototype.
3589         (get_range_info): Same.
3590         * tree-vrp.h (value_range::null_p): Rename to zero_p.
3591         * tree-vrp.c (value_range::null_p): Same.
3593 2018-11-09  Jan Hubicka  <jh@suse.cz>
3595         * tree.c (fld_type_variant_equal_p): Test user align flag.
3596         (flt_type_variant): Copy user align flag.
3597         (fld_incomplete_type_of): Clear it.
3599 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3601         * config/arm/neon.md (div<mode>3): New pattern.
3603 2018-11-08  Andi Kleen  <ak@linux.intel.com>
3605         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
3606         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
3607         (ix86_handle_option): Handle OPT_mptwrite.
3608         * config/i386/cpuid.h (bit_PTWRITE): Add.
3609         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
3610         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
3611         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
3612         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
3613         (ix86_option_override_internal): Handle PTA_PTWRITE.
3614         (ix86_valid_target_attribute_inner_p): Define ptwrite.
3615         (def_builtin2): Force UINT64 to be 64bit only.
3616         * config/i386/i386.h (TARGET_PTWRITE): Add.
3617         (TARGET_PTWRITE_P): Add.
3618         (PTA_PTWRITE): Add.
3619         * config/i386/i386.md: Define ptwrite.
3620         * config/i386/i386.opt: Add -mptwrite.
3621         * config/i386/immintrin.h (_ptwrite64): Add.
3622         (_ptwrite32): Add
3623         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
3624         * doc/invoke.texi: Document -mptwrite.
3626 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
3628         PR rtl-optimization/87600
3629         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
3630         * lra-constraints.c (process_alt_operands): Skip illegal hard
3631         register usage.  Prefer reloading non hard register operands.
3633 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
3635         PR other/36572
3636         * doc/invoke.texi (Optimize Options): Clarify default behavior
3637         for -fno-sched-interblock and -fno-sched-spec.
3639 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
3641         * collect2.c (linker_select):  Add USE_LLD_LD.
3642         (ld_suffixes): Add ld.lld.
3643         (main): Handle -fuse-ld=lld.
3644         * common.opt (-fuse-ld=lld): New option.
3645         * doc/invoke.texi (-fuse-ld=lld): Document.
3646         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
3648 2018-11-08  Paul Koning  <ni1d@arrl.net>
3650         * config/pdp11/constraints.md: Add "Z" series constraints for use
3651         with pre-dec and post-inc addressing.
3652         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
3653         (pdp11_expand_operands): Add int argument (word count).
3654         (pdp11_sp_frame_offset): Delete.
3655         (pdp11_cmp_length): New function.
3656         (pushpop_regeq): New function.
3657         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
3658         Add hook.
3659         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
3660         frame layout.
3661         (pdp11_initial_elimination_offset): Ditto.
3662         (pdp11_expand_operands): Add word count argument.  Bugfixes.
3663         (output_move_multiple): Change how pointer adjustment is done.
3664         (pdp11_gen_int_label): Correct format.
3665         (output_ascii): Ditto.
3666         (pdp11_asm_output_var): Add code for DEC assembler case.
3667         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
3668         value.
3669         (legitimate_const_double_p): Ditto.
3670         (pdp11_register_move_cost): Adjust for new register classes.
3671         (pdp11_regno_reg_class): Ditto.
3672         (expand_block_move): Delete.
3673         (pushpop_regeq): New function.
3674         (pdp11_legitimate_address_p): Bugfix in check for constant
3675         offset.
3676         (pdp11_sp_frame_offset): Delete.
3677         (pdp11_reg_save_size): New helper function for new frame layout.
3678         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
3679         (pdp11_expand_shift): Bugfix in check for constant shift count.
3680         (pdp11_shift_length): Ditto.
3681         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
3682         (pdp11_cmp_length): New function.
3683         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
3684         some compile options.
3685         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
3686         (CALL_USED_REGISTERS): Ditto.
3687         (ELIMINABLE_REGS): Ditto.
3688         (REGISTER_NAMES): Ditto.
3689         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
3690         constraints.
3691         (REG_CLASS_NAMES): Ditto.
3692         (REG_CLASS_CONTENTS): Ditto.  Also remove
3693         HARD_FRAME_POINTER_REGNUM.
3694         (CPU_REG_CLASS): New macro.
3695         (CLASS_MAX_NREGS): Adjust for new register classes.
3696         (FUNCTION_PROFILER): Make no-op.
3697         (may_call_alloca): Remove unused declaration.
3698         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
3699         (ASM_OUTPUT_SKIP): Fix format.
3700         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
3701         (HARD_FRAME_POINTER_REGNUM): Remove.
3702         (return): Delete.
3703         (*rts): Rename.  Remove epilogue related checks.
3704         (cmpsi, cmpdi): New insn.
3705         (cbranch<mode>4): Change to apply to SI and DI modes as well.
3706         (mov<mode>): Change constraints to enforce that push/pop
3707         destination cannot use the same register as source.
3708         (*mov<mode><cc_cc>): Ditto.
3709         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
3710         at assembly output rather than as RTL expander.
3711         (zero_extendqihi2): Bugfix in check for same registers.
3712         (adddi3_nocc): Bugfix in check for constant operand.
3713         (addsi3_nocc): Ditto.
3714         (subdi3_nocc): Ditto.
3715         (subsi3_nocc): Ditto.
3716         (negdi2_nocc): Copy input to pdp11_expand_operands.
3717         (negsi2_nocc): Ditto.
3718         (bswap2_nocc): Ditto.
3719         * config/pdp11/pdp11.opt (mlra): Fix documentation.
3720         * config/pdp11/t-pdp11: Use -Os.
3722 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
3724         * config/arm/parsecpu.awk (/alias/): New parsing rule.
3725         (/begin cpu/): Check that the cpu name hasn't been previously defined.
3726         (gen_comm_data): Print out CPU alias tables.
3727         (check_cpu): Match aliases when checking the CPU name.
3728         * config/arm/arm-protos.h (cpu_alias): New structure.
3729         (cpu_option): Add entry for aliases.
3730         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
3731         strongarm1100 and strongarm1110.
3732         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
3733         (config/arm/arm-generic.md): Remove redundant references to
3734         strongarm110, strongarm1100 and strongarm1110.
3735         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3736         Scan aliases for additional hints.
3737         (arm_parse_cpu_option_name): Also match a cpu name against the list
3738         of aliases.
3739         * config/arm/arm-tables.opt: Regenerated.
3740         * config/arm/arm-tune.md: Regenerated.
3742 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
3744         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
3745         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
3746         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
3747         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
3748         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
3749         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
3750         * gengtype.c (open_base_files): Add omp-general.h.
3751         * gimple.c (gimple_build_omp_critical):
3752         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
3753         gimple_omp_taskgroup_set_clauses.
3754         (gimple_build_omp_atomic_load): Add mo argument, call
3755         gimple_omp_atomic_set_memory_order.
3756         (gimple_build_omp_atomic_store): Likewise.
3757         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
3758         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
3759         instead of GSS_OMP.
3760         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
3761         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
3762         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
3763         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
3764         different value for GF_OMP_ATOMIC_NEED_VALUE.
3765         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
3766         comments.
3767         (struct gimple_statement_omp_single_layout): And remove here.
3768         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
3769         than gimple_statement_omp_single_layout.
3770         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
3771         GIMPLE_OMP_TEAMS.
3772         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
3773         (gimple_omp_subcode): Formatting fix.
3774         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
3775         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
3776         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
3777         gimple_omp_teams_host, gimple_omp_teams_set_host,
3778         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
3779         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
3780         gimple_omp_taskgroup_set_clauses): New inline functions.
3781         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
3782         (gimple_build_omp_atomic_store): Likewise.
3783         (gimple_omp_atomic_seq_cst_p): Remove.
3784         (gimple_omp_atomic_memory_order): New function.
3785         (gimple_omp_atomic_set_seq_cst): Remove.
3786         (gimple_omp_atomic_set_memory_order): New function.
3787         (gimple_build_omp_taskgroup): Add clauses argument.
3788         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
3789         (dump_gimple_omp_task): Print taskwait with depend clauses.
3790         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
3791         dump_omp_atomic_memory_order.
3792         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
3793         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
3794         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
3795         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
3796         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
3797         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
3798         ORT_UNTIED_TASKLOOP enumerators.
3799         (enum gimplify_defaultmap_kind): New.
3800         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
3801         target_map_pointers_as_0len_arrays members, add defaultmap.
3802         (new_omp_context): Initialize defaultmap member.
3803         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3804         (maybe_fold_stmt): Don't fold even in host teams regions.
3805         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
3806         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
3807         ctx->omp_firstprivatize_variable.
3808         (omp_add_variable): Don't add private/firstprivate for VLAs in
3809         ORT_TASKGROUP.
3810         (omp_default_clause): Print "taskloop" rather than "task" if
3811         ORT_*TASKLOOP.
3812         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3813         Handle new defaultmap clause kinds.
3814         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
3815         iterator to be lastprivate or private.  Fix up diagnostics if linear
3816         is used on collapse>1 simd iterator.
3817         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3818         (gimplify_omp_depend): New function.
3819         (gimplify_scan_omp_clauses): Add shared clause on parallel for
3820         combined parallel master taskloop{, simd} if taskloop has
3821         firstprivate, lastprivate or reduction clause.  Handle
3822         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
3823         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
3824         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
3825         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
3826         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
3827         lastprivate.
3828         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
3829         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
3830         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
3831         OMP_CLAUSE_{TASK,IN}_REDUCTION.
3832         (gimplify_omp_task): Handle taskwait with depend clauses.
3833         (gimplify_omp_for): Add shared clause on parallel for combined
3834         parallel master taskloop{, simd} if taskloop has firstprivate,
3835         lastprivate or reduction clause.  Use ORT_TASKLOOP or
3836         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
3837         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
3838         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
3839         __for_end and __for_range temporaries on OMP_PARALLEL for
3840         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
3841         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
3842         sandwiched in between two taskloops.
3843         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
3844         instead of ctx->omp_firstprivatize_variable.
3845         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
3846         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
3847         host teams, use gimplify_and_return_first etc. for body like
3848         for target or target data constructs, and at the end call
3849         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
3850         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
3851         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
3852         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
3853         gimple_omp_atomic_set_seq_cst calls.
3854         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
3855         case, handle taskgroup clauses.
3856         * lto-streamer-out.c (hash_tree): Handle
3857         OMP_CLAUSE_{TASK,IN}_REDUCTION.
3858         * Makefile.in (GTFILES): Add omp-general.h.
3859         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
3860         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
3861         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
3862         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
3863         BUILT_IN_GOMP_LOOP_DOACROSS_START,
3864         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
3865         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
3866         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
3867         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
3868         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
3869         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
3870         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
3871         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
3872         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
3873         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
3874         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
3875         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
3876         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
3877         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
3878         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
3879         * omp-expand.c (workshare_safe_to_combine_p): Return false for
3880         non-worksharing loops.
3881         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
3882         (determine_parallel_type): Don't combine parallel with worksharing
3883         which has _reductemp_ clause.
3884         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
3885         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
3886         if there is nonmonotonic modifier or if there is no modifier and no
3887         ordered clause.  For dynamic and guided schedule without monotonic
3888         and nonmonotonic modifier, default to nonmonotonic.
3889         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
3890         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
3891         task reductions.
3892         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
3893         there are any reduction clauses.
3894         (expand_taskwait_call): New function.
3895         (expand_teams_call): New function.
3896         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
3897         expand_teams_call for it.  Formatting fix.  Handle taskwait with
3898         depend clauses.
3899         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
3900         of worksharing loops with task reductions.
3901         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
3902         expansion of worksharing loops with task reductions.
3903         (expand_omp_sections): Handle expansion of sections with task
3904         reductions.
3905         (expand_omp_synch): For host teams call expand_omp_taskreg.
3906         (omp_memory_order_to_memmodel): New function.
3907         (expand_omp_atomic_load, expand_omp_atomic_store,
3908         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
3909         instead of gimple_omp_atomic_seq_cst_p.
3910         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
3911         depend clauses as a standalone directive.
3912         * omp-general.c (enum omp_requires): New variable.
3913         (omp_extract_for_data): Initialize have_reductemp member.  Allow
3914         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
3915         GT_EXPR loops depending on incr sign.  Formatting fixes.
3916         * omp-general.h (struct omp_for_data): Add have_reductemp member.
3917         (enum omp_requires): New enum.
3918         (omp_requires_mask): Declare.
3919         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
3920         Fix comment typos.
3921         * omp-low.c (struct omp_context): Add task_reductions and
3922         task_reduction_map fields.
3923         (is_host_teams_ctx): New function.
3924         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
3925         (use_pointer_for_field): Use is_global_var instead of
3926         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
3927         in outer contexts.
3928         (build_outer_var_ref): Ignore taskgroup outer contexts.
3929         (delete_omp_context): Release task_reductions and task_reduction_map.
3930         (scan_sharing_clauses): Don't add any fields for reduction clause on
3931         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
3932         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
3933         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
3934         Handle OMP_CLAUSE_NONTEMPORAL.
3935         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
3936         needed.
3937         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
3938         clauses with task modifier.
3939         (scan_omp_task): Handle taskwait with depend clauses.
3940         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
3941         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
3942         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
3943         (scan_omp_for): Fix comment formatting.
3944         (scan_omp_teams): Handle host teams constructs.
3945         (check_omp_nesting_restrictions): Allow teams with no outer
3946         OpenMP context.  Adjust diagnostics for teams strictly nested into
3947         some explicit OpenMP construct other than target.  Allow OpenMP atomics
3948         inside of simd regions.
3949         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
3950         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
3951         taskreg_nesting_level while scanning host teams construct.
3952         (task_reduction_read): New function.
3953         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
3954         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
3955         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
3956         second argument create_tmp_var if it is NULL.  Don't ignore shared
3957         clauses in is_host_teams_ctx contexts.  Handle
3958         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
3959         clauses.
3960         (lower_reduction_clauses): Ignore reduction clauses with task
3961         modifier.  Remove second argument create_tmp_var if it is NULL.
3962         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
3963         (lower_send_clauses): Ignore reduction clauses with task modifier.
3964         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
3965         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
3966         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
3967         rely that it is the last stmt in body so far.  Ignore outer taskgroup
3968         contexts.
3969         (omp_task_reductions_find_first, omp_task_reduction_iterate,
3970         lower_omp_task_reductions): New functions.
3971         (lower_omp_sections): Handle reduction clauses with taskgroup
3972         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
3973         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
3974         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
3975         modifiers.
3976         (lower_omp_taskgroup): Handle taskgroup reductions.
3977         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
3978         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
3979         (lower_depend_clauses): If there are any
3980         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
3981         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
3982         seen, assume lowering is done already and return early.  Set kind
3983         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
3984         (lower_omp_taskreg): Handle reduction clauses with task modifier on
3985         parallel construct.  Handle reduction clause on taskloop construct.
3986         Handle taskwait with depend clauses.
3987         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
3988         for host teams constructs.
3989         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
3990         nontemporal and _reductemp_ clause entries.
3991         (omp_clause_code_name): Likewise.
3992         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
3993         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
3994         * tree-core.h (enum omp_clause_code): Add
3995         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
3996         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
3997         (struct tree_base): Add omp_atomic_memory_order field into union.
3998         Remove OMP_ATOMIC_SEQ_CST comment.
3999         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
4000         and OMP_CLAUSE_DEPEND_DEPOBJ.
4001         (struct tree_omp_clause): Add subcode.defaultmap_kind.
4002         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
4003         OpenMP constructs with body and clauses operands.
4004         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
4005         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
4006         (OMP_TASKGROUP_CLAUSES): Define.
4007         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
4008         OMP_CLAUSE__LOOPTEMP_.
4009         (OMP_ATOMIC_SEQ_CST): Remove.
4010         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
4011         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
4012         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
4013         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
4014         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
4015         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
4016         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
4017         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
4018         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
4019         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
4020         Define.
4021         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
4022         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4023         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
4024         (convert_local_omp_clauses): Likewise.  Remove useless test.
4025         * tree-parloops.c (create_call_for_reduction_1): Pass
4026         OMP_MEMORY_ORDER_RELAXED as new argument to
4027         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
4028         * tree-pretty-print.c (dump_omp_iterators): New function.
4029         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
4030         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
4031         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
4032         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
4033         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
4034         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
4035         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
4036         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
4037         (dump_omp_atomic_memory_order): New function.
4038         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
4039         taskwait with depend clauses.
4040         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
4041         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
4042         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
4043         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
4044         write_ts_omp_clause_tree_pointers): Likewise.
4046 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4048         PR ipa/86395
4049         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
4050         "-missed", "-note", and "-all" sub-options.
4051         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
4052         API.
4053         (can_early_inline_edge_p): Likewise.
4054         (want_early_inline_function_p): Likewise.
4055         (want_inline_self_recursive_call_p): Likewise.
4056         (recursive_inlining): Likewise.
4057         (inline_small_functions): Likewise.
4058         (flatten_function): Likewise.
4059         (ipa_inline): Likewise.
4060         (inline_always_inline_functions): Likewise.
4061         (early_inline_small_functions): Likewise.
4062         (early_inliner): Likewise.
4063         * tree-inline.c (expand_call_inline): Likewise.
4065 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4067         * pretty-print.c (pp_format): Handle %f.
4068         (selftest::test_pp_format): Add test of %f.
4069         * pretty-print.h (pp_double): New macro.
4071 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4073         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
4074         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
4075         (dump_pretty_printer::decode_format): Implement "%C" for
4076         cgraph_node *.
4077         (selftest::test_capture_of_dump_calls): Rename "where" to
4078         "stmt_loc".  Convert test_decl to a function decl and set its
4079         location.  Add a symbol_table_test RAII instance and a
4080         cgraph_node, using it to test "%C" and dump_symtab_node.
4082 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
4084         PR middle-end/87916
4085         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
4087 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
4089         * cgraph.c: Include "selftest.h".
4090         (saved_symtab): New variable.
4091         (selftest::symbol_table_test::symbol_table_test): New ctor.
4092         (selftest::symbol_table_test::~symbol_table_test): New dtor.
4093         (selftest::test_symbol_table_test): New test.
4094         (selftest::cgraph_c_tests): New.
4095         * cgraph.h (saved_symtab): New decl.
4096         (selftest::symbol_table_test): New class.
4097         * selftest-run-tests.c (selftest::run_tests): Call
4098         selftest::cgraph_c_tests.
4099         * selftest.h (selftest::cgraph_c_tests): New decl.
4101 2018-11-08  Richard Biener  <rguenther@suse.de>
4103         * tree-data-ref.h (lambda_int): New typedef.
4104         (lambda_vector_gcd): Adjust.
4105         (lambda_vector_new): Likewise.
4106         (lambda_matrix_new): Likewise.
4107         * tree-data-ref.c  (print_lambda_vector): Adjust.
4109 2018-11-08  Richard Biener  <rguenther@suse.de>
4111         PR tree-optimization/87929
4112         * tree-complex.c (expand_complex_comparison): Clean EH.
4114 2018-11-08  Martin Liska  <mliska@suse.cz>
4116         * doc/extend.texi: Reword.
4117         * predict.c (expr_expected_value_1): Likewise.
4119 2018-11-08  Richard Biener  <rguenther@suse.de>
4121         PR tree-optimization/87913
4122         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
4123         of extreme values to ordered comparisons.
4125 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4127         PR middle-end/42726
4128         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
4130 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4132         * doc/invoke.texi: Remove leading dash from @opindex entries
4133         throughout the file.
4135 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
4137         PR driver/80828
4138         * doc/invoke.texi (Option Summary): Add -e and --entry.
4139         (Link Options): Likewise.
4141 2018-11-07  Nathan Sidwell  <nathan@acm.org>
4143         PR 87926
4144         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
4145         --disable-checking bootstrap.
4147 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
4149         * configure: Regenerated.
4151 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4153         PR c/87691
4154         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
4155         to the mode of the widest field iff the widest field has mode class
4156         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
4157         reference.
4159 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
4161         * common.opt: Add -Wattribute-warning.
4162         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
4163         * expr.c (expand_expr_real_1): Add new attribute to warning_at
4164         call to allow user configure behavior of "warning" attribute.
4166 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
4168         * target.def: Put @: after every vs., e.g., and i.e. where it is
4169         followed by whitespace.
4170         * doc/extend.texi: Ditto.
4171         * doc/fragments.texi: Ditto.
4172         * doc/gimple.texi: Ditto.
4173         * doc/implement-c.texi: Ditto.
4174         * doc/install.texi: Ditto.
4175         * doc/invoke.texi: Ditto.
4176         * doc/md.texi: Ditto.
4177         * doc/plugins.texi: Ditto.
4178         * doc/rtl.texi: Ditto.
4179         * doc/sourcebuild.texi: Ditto.
4180         * doc/tm.texi.in: Ditto.
4181         * doc/ux.texi: Ditto.
4182         * doc/tm.texi: Regenerate.
4184 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4186         * config/arm/arm-cpus.in (ares): New entry.
4187         * config/arm/arm-tables.opt: Regenerate.
4188         * config/arm/arm-tune.md: Likewise.
4189         * doc/invoke.texi (ARM Options): Document ares.
4191 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4193         * config/aarch64/aarch64-cores.def (ares): Define.
4194         * config/aarch64/aarch64-tune.md: Regenerate.
4195         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
4197 2018-11-07  Jan Hubicka  <jh@suse.cz>
4199         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
4200         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
4201         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
4202         TYPE_DECL.
4204 2018-11-07  Richard Biener  <rguenther@suse.de>
4206         PR tree-optimization/87914
4207         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
4208         of nested cycles.
4209         (vectorizable_reduction): Handle shifts and rotates by dispatching
4210         to vectorizable_shift.
4211         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
4212         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
4213         def cases.
4214         (vectorizable_shift): Export and handle being called as
4215         vect_nested_cycle.
4216         (vect_analyze_stmt): Call vectorizable_shift after
4217         vectorizable_reduction.
4218         * tree-vectorizer.h (vectorizable_shift): Declare.
4220 2018-11-07  Jan Hubicka  <jh@suse.cz>
4222         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
4223         than const decls in TREE_VALUE of enum.
4224         (dump_type_inheritance_graph): Improve duplicate dumping.
4225         (free_enum_values): New.
4226         (build_type_inheritance_graph): Use it.
4227         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
4228         which are not main variants or not ODR types.
4229         (verify_type_variant): Expect variants to have no TYPE_VALUES.
4231 2018-11-07  Richard Biener  <rguenther@suse.de>
4233         * ipa-inline.c (want_inline_small_function_p): Compute
4234         big_speedup_p lazily and last.
4236 2018-11-07  Jan Hubicka  <jh@suse.cz>
4238         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
4239         building incomplete variant of complete type.
4240         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
4241         variant of complete type.
4243 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4245         * config/mips/mips.c: Fix typo in documentation of
4246         mips_loongson_ext2_prefetch_cookie.
4247         (mips_option_override): fix brain twister logical.
4248         * config/mips/mips.h: Fix typo in documentation of
4249         ISA_HAS_CTZ_CTO and define pattern.
4250         * config/mips/mips.md (prefetch): Hoist EXT2 above
4251         the 2EF/EXT block.
4252         (prefetch_indexed): Hoist EXT2 above the EXT block.
4254 2018-11-07  Jan Hubicka  <jh@suse.cz>
4256         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
4257         return and parameter types of function and method types.
4258         (free_lang_data_in_cgraph): Update.
4260 2018-11-07  Martin Liska  <mliska@suse.cz>
4262         PR rtl-optimization/87868
4263         * postreload-gcse.c (eliminate_partially_redundant_load): Set
4264         threshold to max_count if we would overflow.
4265         * profile-count.h: Make max_count a public constant.
4267 2018-11-07  Martin Liska  <mliska@suse.cz>
4269         * mem-stats.h: Fix GNU coding style.
4271 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4273         * config/mips/gs264e.md: New.
4274         * config/mips/mips-cpus.def: Define gs264e.
4275         * config/mips/mips-tables.opt: Regenerate.
4276         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
4277         gs264e.
4278         (mips_issue_rate): Add support for gs264e.
4279         (mips_multipass_dfa_lookahead): Likewise.
4280         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
4281         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
4282         (MIPS_ASE_MSA_SPEC): New.
4283         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
4284         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
4285         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
4286         * config/mips/mips.md: Include gs264e.md.
4287         (processor): Add gs264e.
4288         * config/mips/mips.opt (MSA): Use Mask instead of Var.
4289         * doc/invoke.texi: Add gs264e to supported architectures.
4291 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4293         * config/mips/gs464e.md: New.
4294         * config/mips/mips-cpus.def: Define gs464e.
4295         * config/mips/mips-tables.opt: Regenerate.
4296         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
4297         gs464e.
4298         (mips_issue_rate): Add support for gs464e.
4299         (mips_multipass_dfa_lookahead): Likewise.
4300         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
4301         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
4302         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
4303         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
4304         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
4305         * config/mips/mips.md: Include gs464e.md.
4306         (processor): Add gs464e.
4307         * doc/invoke.texi: Add gs464e to supported architectures.
4309 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4311         * config/mips/loongson3a.md: Rename to ...
4312         * config/mips/gs464.md: ... here.
4313         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
4314         as an alias of gs464 processor.
4315         * config/mips/mips-tables.opt: Regenerate.
4316         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
4317         instead of PROCESSOR_LOONGSON_3A.
4318         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
4319         TUNE_LOONGSON_3A.
4320         (mips_option_override): Enable MMI and EXT for gs464.
4321         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
4322         Rename TUNE_LOONGSON_3A to TUNE_GS464.
4323         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
4324         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
4325         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
4326         TARGET_LOONGSON_3A.
4327         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
4328         (processor): Add gs464;
4329         * doc/invoke.texi: Add gs464 to supported architectures.
4331 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4333         * config/mips/mips-protos.h
4334         (mips_loongson_ext2_prefetch_cookie): New prototype.
4335         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
4336         (mips_option_override): Enable TARGET_LOONGSON_EXT when
4337         TARGET_LOONGSON_EXT2 is true.
4338         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
4339         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
4340         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
4341         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
4342         TARGET_LOONGSON_EXT2.
4343         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
4344         (define_insn "ctz<mode>2"): New insn pattern.
4345         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
4346         (define_insn "prefetch_indexed_<mode>"): Include
4347         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
4348         * config/mips/mips.opt (-mloongson-ext2): Add option.
4349         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
4351 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4353         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
4354         __mips_loongson_ext.
4355         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
4356         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
4357         -mloongson-ext.
4358         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
4359         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
4360         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
4361         instead of TARGET_LOONGSON_3A.
4362         * config/mips/mips.opt (-mloongson-ext): Add option.
4363         * gcc/doc/invoke.texi (-mloongson-ext): Document.
4365 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
4367         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
4368         * config/mips/loongson.md: Move to ...
4369         * config/mips/loongson-mmi.md: here; Adjustment.
4370         * config/mips/loongson.h: Move to ...
4371         State as deprecated. Include loongson-mmiintrin.h for back
4372         compatibility and warning.
4373         * config/mips/loongson-mmiintrin.h: ... here.
4374         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
4375         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
4376         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
4377         (mips_option_override): Make sure MMI use hard float;
4378         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
4379         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
4380         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
4381         TARGET_LOONGSON_VECTORS.
4382         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
4383         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
4384         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
4385         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
4386         -mloongson-mmi.
4387         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
4388         TARGET_LOONGSON_VECTORS.
4389         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
4390         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
4391         (Loongson MMI patterns): Include loongson-mmi.md instead of
4392         loongson.md.
4393         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
4394         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
4396 2018-11-07  Richard Biener  <rguenther@suse.de>
4398         PR lto/87906
4399         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
4400         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
4402 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
4404         PR rtl-optimization/87874
4405         * lra.c (lra_substitute_pseudo): Do not create a subreg for
4406         const wide ints.
4408 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
4410         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
4411         if not in indexed or indirect form.
4412         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
4413         (bswap<mode>2_store): Ditto.
4415 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
4417         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
4418         the UNSPEC.
4420 2018-11-06  Richard Biener  <rguenther@suse.de>
4422         PR tree-optimization/86850
4423         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
4424         instead of src.m_vec.
4426 2018-11-06  Jan Hubicka  <jh@suse.cz>
4428         * tree.c (fld_simplified_type_name): Break out form ...
4429         (free_lang_data_in_type): ... here.
4430         (fld_type_variant_equal_p): Use it.
4432 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4434         * config/default-d.c: Include memmodel.h.
4436         * config/sol2-d.c: New file.
4437         * config/t-sol2 (sol2-d.o): New rule.
4438         * config.gcc <*-*-solaris2*>: Set d_target_objs,
4439         target_has_targetdm.
4441 2018-11-06  Jan Hubicka  <jh@suse.cz>
4443         * tree.c (fld_type_variant): Also copy alignment; be sure that
4444         new variant is equal.
4446 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4448         PR target/87762
4449         * config/s390/s390.md: Add relative_long attribute.
4451 2018-11-06  Jan Hubicka  <jh@suse.cz>
4453         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
4454         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
4455         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
4456         stream TYPE_NEEDS_CONSTRUCTING.
4457         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
4458         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
4460 2018-11-06  Richard Biener  <rguenther@suse.de>
4462         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
4463         dump-scope ...
4464         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
4466 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4468         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
4469         REG_CLASS_CONTENTS[GEN_REGS].
4470         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
4472 2018-11-06  Jan Hubicka  <jh@suse.cz>
4474          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
4476 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4478         PR sanitizer/80953
4479         * config/sol2.h (ASAN_CC1_SPEC): Define.
4480         (LD_WHOLE_ARCHIVE_OPTION): Define.
4481         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
4482         (ASAN_REJECT_SPEC): Provide default.
4483         (LIBASAN_EARLY_SPEC): Define.
4484         (LIBTSAN_EARLY_SPEC): Define.
4485         (LIBLSAN_EARLY_SPEC): Define.
4486         * config/i386/sol2.h (CC1_SPEC): Redefine.
4487         (ASAN_REJECT_SPEC): Define.
4489         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
4490         (TARGET_ASAN_SHADOW_OFFSET): Define.
4491         (sparc_asan_shadow_offset): New function.
4492         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
4493         (ASAN_REJECT_SPEC): Define.
4495 2018-11-06  Jan Hubicka  <jh@suse.cz>
4497         * tree.c (fld_type_variant): Copy canonical type.
4498         (fld_incomplete_type_of): Check that canonical types looks sane;
4499         copy canonical type.
4500         (verify_type): Accept when incomplete type has complete canonical type.
4502 2018-11-06  Jan Hubicka  <jh@suse.cz>
4504         * tree.c (free_lang_data): Reset overwite_assembler_name,
4505         print_xnode, print_decl, print_type and print_identifier of
4506         langhooks.
4508 2018-11-06  Richard Biener  <rguenther@suse.de>
4510         PR tree-optimization/87889
4511         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
4512         Do nothing if old and new arg are the same
4514 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
4516         PR target/87723
4517         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
4518         attributes for operands 3 and 4.
4520 2018-11-06  Richard Biener  <rguenther@suse.de>
4522         PR middle-end/18041
4523         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
4524         matching bitfield insertion.
4526 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
4528         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
4529         (reg_next_debug_use): New.
4530         (attempt_change): Propagate adjusted expression into affected
4531         debug insns.
4532         (merge_in_block): Track uses in debug insns.
4533         (pass_inc_dec::execute): Allocate and release
4534         reg_next_debug_use.
4536 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
4538         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
4539         (_mm512_fixupimm_round_pd): Update parameters and builtin.
4540         (_mm512_maskz_fixupimm_round_pd): Ditto.
4541         (_mm512_fixupimm_round_ps): Ditto.
4542         (_mm512_maskz_fixupimm_round_ps): Ditto.
4543         (_mm_fixupimm_round_sd): Ditto.
4544         (_mm_maskz_fixupimm_round_sd): Ditto.
4545         (_mm_fixupimm_round_ss): Ditto.
4546         (_mm_maskz_fixupimm_round_ss): Ditto.
4547         (_mm512_fixupimm_pd): Ditto.
4548         (_mm512_maskz_fixupimm_pd): Ditto.
4549         (_mm512_fixupimm_ps): Ditto.
4550         (_mm512_maskz_fixupimm_ps): Ditto.
4551         (_mm_fixupimm_sd): Ditto.
4552         (_mm_maskz_fixupimm_sd): Ditto.
4553         (_mm_fixupimm_ss): Ditto.
4554         (_mm_maskz_fixupimm_ss): Ditto.
4555         (_mm512_mask_fixupimm_round_pd): Update builtin.
4556         (_mm512_mask_fixupimm_round_ps): Ditto.
4557         (_mm_mask_fixupimm_round_sd): Ditto.
4558         (_mm_mask_fixupimm_round_ss): Ditto.
4559         (_mm512_mask_fixupimm_pd): Ditto.
4560         (_mm512_mask_fixupimm_ps): Ditto.
4561         (_mm_mask_fixupimm_sd): Ditto.
4562         (_mm_mask_fixupimm_ss): Ditto.
4563         * config/i386/avx512vlintrin.h:
4564         (_mm256_fixupimm_pd): Update parameters and builtin.
4565         (_mm256_maskz_fixupimm_pd): Ditto.
4566         (_mm256_fixupimm_ps): Ditto.
4567         (_mm256_maskz_fixupimm_ps): Ditto.
4568         (_mm_fixupimm_pd): Ditto.
4569         (_mm_maskz_fixupimm_pd): Ditto.
4570         (_mm_fixupimm_ps): Ditto.
4571         (_mm_maskz_fixupimm_ps): Ditto.
4572         (_mm256_mask_fixupimm_pd): Update builtin.
4573         (_mm256_mask_fixupimm_ps): Ditto.
4574         (_mm_mask_fixupimm_pd): Ditto.
4575         (_mm_mask_fixupimm_ps): Ditto.
4576         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
4577         * config/i386/i386-builtin.def: Update builtin definitions.
4578         * config/i386/i386.c: Handle new builtin types and remove useless ones.
4579         * config/i386/sse.md: Update VFIXUPIMM* patterns.
4580         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4581         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4582         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
4583         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4584         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4585         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
4586         * config/i386/subst.md:
4587         (round_saeonly_sd_mask_operand4): Add new subst_attr.
4588         (round_saeonly_sd_mask_op4): Ditto.
4589         (round_saeonly_expand_operand5): Ditto.
4590         (round_saeonly_expand): Update.
4592 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
4594         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
4596 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
4598         PR rtl-optimization/87871
4599         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
4601 2018-11-05  Paul Koning  <ni1d@arrl.net>
4603         * doc/sourcebuild.texi (target attributes): Document new "inf"
4604         effective target keyword.
4606 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
4608         * config/s390/s390.c (s390_register_move_cost): Increase costs for
4609         moves involving the CC reg.
4611 2018-11-05  Richard Biener  <rguenther@suse.de>
4613         PR tree-optimization/87873
4614         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
4615         argument.
4616         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
4617         * tree-vect-loop.c (vect_transform_loop): When splitting the
4618         loop exit also create forwarder PHIs for constants.
4619         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
4620         Handle constant to_arg, add extra checking we match up the correct
4621         PHIs.
4623 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
4625         * config/s390/s390.md: QImode and HImode for load on condition.
4627 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
4629         * config/s390/predicates.md: Fix typo.
4630         * config/s390/s390.md: Allow immediates for load on condition.
4632 2018-11-05  Martin Liska  <mliska@suse.cz>
4634         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
4635         * bitmap.h (struct bitmap_usage): Likewise.
4636         * ggc-common.c (SCALE): Remove.
4637         (LABEL): Likewise.
4638         (struct ggc_usage): Use SIZE_AMOUNT. And update
4639         compare method.
4640         * ggc-page.c (SCALE): Remove.
4641         (STAT_LABEL): Remove.
4642         (ggc_print_statistics): Use SIZE_AMOUNT.
4643         * gimple.h (SCALE): Remove.
4644         (LABEL): Likewise.
4645         * input.c (ONE_K): Remove.
4646         (ONE_M): Likewise.
4647         (SCALE): Likewise.
4648         (STAT_LABEL): Likewise.
4649         (FORMAT_AMOUNT): Likewise.
4650         (dump_line_table_statistics): Use SIZE_AMOUNT.
4651         * mem-stats.h (struct mem_usage): Likewise.
4652         * rtl.c (dump_rtx_statistics): Likewise.
4653         (rtx_alloc_counts): Change type to size_t.
4654         (rtx_alloc_sizes): Likewise.
4655         (rtx_count_cmp): New.
4656         (dump_rtx_statistics): Sort first based on counts.
4657         * tree.c (tree_nodes_cmp): New.
4658         (tree_codes_cmp): New.
4659         (dump_tree_statistics): Sort first based on counts.
4660         * system.h (ONE_K): New.
4661         (ONE_M): Likewise.
4662         (SIZE_SCALE): Likewise.
4663         (SIZE_LABEL): Likewise.
4664         (SIZE_AMOUNT): Likewise.
4665         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
4666         * tree-dfa.c (dump_dfa_stats): Likewise.
4667         * tree-phinodes.c (phinodes_print_statistics): Likewise.
4668         * tree-ssanames.c (ssanames_print_statistics): Likewise.
4669         * tree.c (dump_tree_statistics): Likewise.
4670         * vec.c (struct vec_usage): Likewise.
4671         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
4672         trigger a -Werror=format-overflow with
4673         --enable-gather-detailed-stats.
4675 2018-11-05  Martin Liska  <mliska@suse.cz>
4677         * mem-stats.h (mem_alloc_description::release_instance_overhead):
4678         Return T *.
4679         * vec.c (struct vec_usage): Register m_element_size.
4680         (vec_prefix::register_overhead): New arguments: elements and
4681         element_size.
4682         (vec_prefix::release_overhead): Subtract elements.
4683         * vec.h (struct vec_prefix): Change signature.
4684         (va_heap::reserve): Pass proper arguments.
4685         (va_heap::release): Likewise.
4687 2018-11-05  Martin Liska  <mliska@suse.cz>
4689         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
4690         style.
4691         * vec.c: Likewise.
4693 2018-11-05  Richard Biener  <rguenther@suse.de>
4695         * tree-scalar-evolution.h (final_value_replacement_loop): Update
4696         prototype.
4697         * tree-scalar-evolution.c (final_value_replacement_loop): Return
4698         whether anything was done.
4699         (scev_const_prop): Remove constant propagation part, fold
4700         remains into ...
4701         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
4702         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
4703         conditionally.
4705 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
4707         PR tree-optimization/87859
4708         * gimple-ssa-store-merging.c (struct merged_store_group): Add
4709         only_constants and first_nonmergeable_order members.
4710         (merged_store_group::merged_store_group): Initialize them.
4711         (merged_store_group::do_merge): Clear only_constants member if
4712         adding something other than INTEGER_CST store.
4713         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
4714         stores with order >= first_nonmergeable_order.  Use
4715         merged_store->only_constants instead of always recomputing it.
4716         Set merged_store->first_nonmergeable_order if we've skipped any
4717         stores.  Attempt to merge overlapping INTEGER_CST stores that
4718         we would otherwise skip.
4720         PR sanitizer/87837
4721         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
4723 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
4725         PR target/87853
4726         * config/i386/emmintrin.h (__v16qs): New to cope with option
4727         -funsigned-char.
4728         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
4729         (_mm_cmplt_epi8): Likewise.
4730         (_mm_cmpgt_epi8): Likewise.
4732 2018-11-05  Richard Biener  <rguenther@suse.de>
4734         PR rtl-optimization/87852
4735         * fwprop.c (use_killed_between): Only consider single-defs of the
4736         use whose definition statement dominates the use.
4738 2018-11-05  Martin Liska  <mliska@suse.cz>
4740         PR web/87829
4741         * doc/invoke.texi: Remove options that are
4742         not disabled with -Os.
4744 2018-11-05  Martin Liska  <mliska@suse.cz>
4746         PR c/87811
4747         * doc/extend.texi: Update constrain about the last argument
4748         of __builtin_expect_with_probability.
4750 2018-11-05  Martin Liska  <mliska@suse.cz>
4752         PR c/87811
4753         * predict.c (expr_expected_value_1): Verify
4754         that last argument is a real constants and emit
4755         error.
4757 2018-11-05  Martin Liska  <mliska@suse.cz>
4759         PR gcov-profile/77698
4760         * ipa-profile.c (ipa_profile): Adjust hotness threshold
4761         only in LTO mode.
4763 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4765         PR tree-optimization/86572
4766         * builtins.c (c_strlen): Handle negative offsets in a safe way.
4768 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4770         PR tree-optimization/87672
4771         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
4772         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
4774 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
4776         PR middle-end/58372
4777         * cfgexpand.c (pass_expand::execute): Move the call to
4778         finish_eh_generation in front of the call to expand_stack_alignment.
4780 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
4782         * common/config/i386/i386-common.c (processor_alias_table): Add
4783         znver2 entry.
4784         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
4785         (case ${target}): Add znver2.
4786         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
4787         -march=native recognize znver2 processors.
4788         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
4789         * config/i386/i386.c (m_znver2): New definition.
4790         (m_ZNVER): New definition.
4791         (m_AMD_MULTIPLE): Includes m_znver2.
4792         (processor_cost_table): Add znver2 entry.
4793         (processor_target_table): Add znver2 entry.
4794         (get_builtin_code_for_version): Set priority for
4795         PROCESSOR_ZNVER2.
4796         (processor_model): Add M_AMDFAM17H_ZNVER2.
4797         (arch_names_table): Ditto.
4798         (ix86_reassociation_width): Include znver2.
4799         * config/i386/i386.h (TARGET_znver2): New definition.
4800         (struct ix86_size_cost): Add TARGET_ZNVER2.
4801         (enum processor_type): Add PROCESSOR_ZNVER2.
4802         * config/i386/i386.md (define_attr "cpu"): Add znver2.
4803         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
4804         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
4805         (ix86_adjust_cost): Add znver2.
4806         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
4807         * gcc/doc/extend.texi: Add details about znver2.
4808         * gcc/doc/invoke.texi: Add details about znver2.
4810 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
4812         PR target/87079
4814         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
4815         pattern.
4817 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4819         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
4820         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
4822 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
4824         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
4825         shrink-wrapping when -mtrack-speculation.
4827 2018-11-02  Richard Biener  <rguenther@suse.de>
4829         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
4830         (pop_cost_one_pair): Do not free pair.
4831         (pop_best_coalesce): Likewise.
4832         (create_coalesce_list): Initialize obstack.
4833         (delete_coalesce_list): Free obstack.
4834         (find_coalesce_pair): Obstack-allocate coalesce pairs.
4835         (add_cost_one_coalesce): Likewise.
4836         (struct live_track): Remove bitmap pointer indirections.
4837         (new_live_track): Adjust.
4838         (delete_live_track): Likewise.
4839         (live_track_remove_partition): Likewise.
4840         (live_track_add_partition): Likewise.
4841         (live_track_live_p): Likewise.
4842         (live_track_process_def): Likewise.
4843         (live_track_clear_base_vars): Likewise.
4845 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4847         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
4848         * configure: Regenerate.
4849         * config.in: Regenerate.
4850         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
4851         HAVE_AS_REGISTER_PSEUDO_OP guard.
4852         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
4853         !HAVE_AS_REGISTER_PSEUDO_OP support.
4854         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
4855         guard.
4857 2018-11-02  Richard Biener  <rguenther@suse.de>
4859         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
4860         with is_type_die.
4862 2018-11-02  Richard Biener  <rguenther@suse.de>
4864         PR tree-optimization/87776
4865         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
4866         executable when iterating but running into rpo-vn-max-loop-depth
4867         and not eliding the iteration.
4869 2018-11-30  Jan Hubicka  <jh@suse.cz>
4871         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
4872         TYPE_DECL.
4874 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
4876         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
4877         prototype.
4878         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
4879         Combine with rs6000_address_for_fpconvert.
4880         (rs6000_address_for_fpconvert): Combine with
4881         rs6000_force_indexed_or_indirect_mem.
4882         (rs6000_expand_vector_init): Change function call from
4883         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
4884         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
4885         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
4886         (floatsi<mode>2_lfiwax_mem): Ditto.
4887         (floatunssi<mode>2_lfiwzx): Ditto.
4888         (floatunssi<mode>2_lfiwzx_mem): Ditto.
4889         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
4890         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
4891         (fix_trunc<mode>si2_stfiwx): Ditto.
4892         (fixuns_trunc<mode>si2_stfiwx): Ditto.
4893         (float_<mode>si2_hw): Ditto.
4894         (floatuns_<mode>si2_hw): Ditto.
4895         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
4896         (vsx_splat_<mode>): Ditto.
4898 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
4900         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
4901         at end of assembler input text.
4902         * configure: Regenerate.
4904 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
4906         PR tree-optimization/87826
4907         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
4908         negative or larger or equal to type's precision.
4910 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
4912         * opts.c (default_options_table): Do not enable
4913         OPT_fdelayed_branch at -Og.
4914         * doc/invoke.texi (-fdelayed-branch): Document it.
4916 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
4918         * optabs-libfuncs.c (build_libfunc_function_visibility):
4919         New, split out from...
4920         (build_libfunc_function): ... here.
4921         (init_one_libfunc_visibility): New, split out from ...
4922         (init_one_libfunc): ... here.
4924         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
4925         scratch register need not be early-clobber.  Document the reason
4926         why we cannot use ST<OP>.
4928 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
4930         PR bootstrap/82856
4931         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
4932         line for second argument of AC_DEFINE_UNQUOTED.
4933         * doc/install.texi (Tools/packages necessary for modifying GCC):
4934         Update to autoconf 2.69 and automake 1.15.1.
4935         * aclocal.m4, config.in, configure: Regenerate.
4937 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
4939         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
4940         initialization.
4942 2018-10-31  Martin Liska  <mliska@suse.cz>
4944         PR driver/83193
4945         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
4946         Add new argument invalid_extension.
4947         (aarch64_get_all_extension_candidates): New function.
4948         (aarch64_rewrite_selected_cpu): Add NULL to function call.
4949         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
4950         new argument.
4951         (aarch64_get_all_extension_candidates): New function.
4952         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
4953         argument invalid_extension.
4954         (aarch64_parse_cpu): Likewise.
4955         (aarch64_print_hint_for_extensions): New function.
4956         (aarch64_validate_mcpu): Provide hint about invalid extension.
4957         (aarch64_validate_march): Likewise.
4958         (aarch64_handle_attr_arch): Pass new argument.
4959         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
4960         (aarch64_handle_attr_isa_flags): Likewise.
4962 2018-10-31  Richard Biener  <rguenther@suse.de>
4964         PR middle-end/70359
4965         PR middle-end/86270
4966         * tree-outof-ssa.c (insert_backedge_copies): Restrict
4967         copy generation to useful cases.  Place the copy before
4968         the definition of the backedge value when possible.
4970 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
4972         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
4973         * config/arc/arc.c (arc_active_insn): New function.
4974         (check_store_cacheline_hazard): Likewise.
4975         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
4976         (arc_override_options): Disable delay slot scheduler for older
4977         A7.
4978         (arc_store_addr_hazard_p): New implementation, old one renamed to
4979         ...
4980         (arc_store_addr_hazard_internal_p): Renamed.
4981         (arc_reorg): Don't combine into brcc instructions which are part
4982         of hardware hazard solution.
4983         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
4984         (tune_arc700): Likewise.
4985         * config/arc/arc.opt (arc7xx): New tune value.
4986         * config/arc/arc700.md: Improve A7 scheduler.
4988 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
4990         * config/arc/arc.c (arc_override_options): Remove
4991         TARGET_COMPACT_CASESI.
4992         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
4993         (CASE_VECTOR_MODE): Likewise.
4994         (CASE_VECTOR_PC_RELATIVE): Likewise.
4995         (CASE_VECTOR_SHORTEN_MODE): Likewise.
4996         (CASE_VECTOR_SHORTEN_MODE1): Delete.
4997         (ADDR_VEC_ALIGN): Update.
4998         (ASM_OUTPUT_CASE_LABEL): Undefine.
4999         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
5000         (TARGET_BI_BIH): Define.
5001         (DEFAULT_BRANCH_INDEX): Likewise.
5002         * config/arc/arc.md (casesi): Rework to accept BI/BIH
5003         instructions, remove compact_casesi use case.
5004         (casesi_compact_jump): Remove.
5005         (casesi_dispatch): New pattern.
5006         * config/arc/arc.opt: Add mbranch-index option. Deprecate
5007         compact_casesi option.
5008         * doc/invoke.texi: Document mbranch-index option.
5010 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5012         * config/arc/arc.c (arc_get_tp): Remove function.
5013         (arc_emit_call_tls_get_addr): Likewise.
5014         (arc_call_tls_get_addr): New function.
5015         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
5016         * config/arc/arc.md (tls_load_tp_soft): Remove.
5017         (tls_gd_get_addr): Likewise.
5019 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
5021         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
5022         (mulsi3_600_lib): Remove pattern.
5023         (umulsi3_highpart_600_lib_le): Likewise.
5024         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
5025         (umulsidi3): Remove call to umulsidi3_600_lib.
5026         (umulsidi3_600_lib): Remove pattern.
5027         (peephole2): Remove peephole using the above deprecated patterns.
5029 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5031         PR target/87374
5032         * config/arm/arm.c (arm_option_check_internal): Disable the combined
5033         use of -mslow-flash-data and -mword-relocations.
5034         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
5035         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
5036         flag_pic.
5037         * doc/invoke.texi (-mword-relocations): Mention conflict with
5038         -mslow-flash-data.
5039         (-mslow-flash-data): Reciprocally.
5041 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
5043         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
5044         16-byte modes held in GP registers to use an even regno.
5046         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
5047         (aarch64_atomic_ldop_supported_p): Remove.
5048         (aarch64_gen_atomic_ldop): Remove.
5049         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
5050         Fully expand LSE operations here.
5051         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
5052         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
5053         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
5054         and use ATOMIC_LDOP instead; use register_operand for the input;
5055         drop the split and emit insns directly.
5056         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
5057         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
5058         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
5060         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
5061         (aarch64_gen_atomic_ldop): Don't call it.
5062         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
5063         Use aarch64_reg_or_zero.
5064         (aarch64_atomic_exchange<ALLI>): Likewise.
5065         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
5066         operand 0; use aarch64_reg_or_zero for input; merge ...
5067         (@aarch64_atomic_swp<ALLI>): ... this and remove.
5069         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
5070         (aarch64_split_compare_and_swap): Use it.
5071         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
5072         test oldval against the proper predicate.
5073         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
5074         Use nonmemory_operand for expected.
5075         (cas_short_expected_pred): New.
5076         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
5077         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
5078         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
5079         (aarch64_plushi_operand): New.
5081         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
5082         Force oldval into the rval register for TARGET_LSE; emit the compare
5083         during initial expansion so that it may be deleted if unused.
5084         (aarch64_gen_atomic_cas): Remove.
5085         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
5086         Change =&r to +r for operand 0; use match_dup for operand 2;
5087         remove is_weak and mod_f operands as unused.  Drop the split
5088         and merge with...
5089         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
5090         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
5091         (@aarch64_atomic_cas<GPI>): Similarly.
5093 2018-10-31  Richard Biener  <rguenther@suse.de>
5095         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
5096         using ABSU_EXPR.
5098 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5100         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
5101         * config/aarch64/aarch64.md: Include saphira.md
5102         * config/aarch64/saphira.md: New file for pipeline description.
5104 2018-10-30  Martin Sebor  <msebor@redhat.com>
5106         PR middle-end/87041
5107         * gimple-ssa-sprintf.c (format_directive): Use %G to include
5108         inlining context.
5109         (sprintf_dom_walker::compute_format_length):
5110         Avoid setting POSUNDER4K here.
5111         (get_destination_size): Handle null argument values.
5112         (get_user_idx_format): New function.
5113         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
5114         functions, including user-defined with attribute format printf.
5115         Use %G to include inlining context.
5116         Set POSUNDER4K here.
5118 2018-10-30  Jan Hubicka  <jh@suse.cz>
5120         * params.def (lto-partitions): Bump from 32 to 128.
5122 2018-10-30  Jan Hubicka  <jh@suse.cz>
5124         * tree.c
5125         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
5126         head in file.
5127         (free_lang_data_in_type): Forward declare.
5128         (fld_type_variant_equal_p): New function.
5129         (fld_type_variant): New function
5130         (fld_incomplete_types): New hash.
5131         (fld_incomplete_type_of): New function
5132         (fld_simplfied-type): New function.
5133         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
5134         (free_lang_data): Allocate and free fld_incomplete_type; update call
5135         of free_lang_data_in_decl.
5137 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
5139         * gcov.c (output_lines): Remove duplicate line.
5141 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5143         * config/rs6000/rs6000.md (bswapdi2): Force address into register
5144         if not in indexed or indirect form.
5145         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
5146         (bswapdi2_store): Ditto.
5147         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
5148         helper function.
5149         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
5150         Prototype for helper function.
5152 2018-10-30  Martin Sebor  <msebor@redhat.com>
5154         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
5155         (target, pragma GCC optimize, pragma GCC target): Ditto.
5157 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
5159         * doc/extend.texi: Fix prototype and description of
5160         __builtin_expect_with_probability.
5162 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
5164         * cgraph.h (clone_function_name_1): Replaced by new
5165           clone_function_name_numbered that takes name as string; for
5166           privatize_symbol_name_1 use only.
5167           (clone_function_name): Renamed to
5168           clone_function_name_numbered to be explicit about numbering.
5169           (clone_function_name): New two-argument function that does
5170           not number its output.
5171           (clone_function_name): New three-argument function that
5172           takes a number to append to its output.
5173         * cgraphclones.c (duplicate_thunk_for_node):
5174           (clone_function_name_1): Renamed.
5175           (clone_function_name_numbered): Two new functions.
5176           (clone_function_name): Improved documentation.
5177           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
5178         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
5179         * final.c (final_scan_insn_1): Use the new clone_function_name
5180           without numbering.
5181         * multiple_target.c (create_dispatcher_calls): Ditto.
5182           (create_target_clone): Ditto.
5183         * omp-expand.c (grid_expand_target_grid_body): Ditto.
5184         * omp-low.c (create_omp_child_function_name): Ditto.
5185         * omp-simd-clone.c (simd_clone_create): Ditto.
5186         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
5187           new clone_function_name without numbering.
5189 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
5191         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
5192         Assert that the allocation size is not zero.
5194 2018-10-30  Richard Biener  <rguenther@suse.de>
5196         PR tree-optimization/87800
5197         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
5198         non-induction or reduction PHIs.
5200 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
5202         * config/aarch64/falkor-tag-collision-avoidance.c
5203         (execute_tag_collision_avoidance): Call df_note_add_problem.
5205 2018-10-30  Martin Liska  <mliska@suse.cz>
5207         * doc/extend.texi: Fix typo in documentation
5208         of __builtin_expect_with_probability.
5210 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
5212         PR c++/87721
5213         * input.c (get_substring_ranges_for_loc): Detect if
5214         linemap_resolve_location gives us a NULL map, and reject
5215         this case.
5217 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
5219         * config.gcc (xstormy16-*-elf): Set tm_d_file.
5221 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
5222             Martin Sebor  <msebor@redhat.com>
5223             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
5225         * doc/ux.texi (Quoting): New subsection, adapted from material at
5226         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
5227         MartinSebor and ManuelLopezIbanez.
5228         (Fix-it hints): Note that fix-it hints shouldn't be marked for
5229         translation.
5231 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
5233         PR middle-end/87469
5234         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
5235         max value.
5237 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
5239         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
5241 2018-10-29  Paul Koning  <ni1d@arrl.net>
5243         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
5245 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
5247         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
5248         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
5249         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
5250         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
5251         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
5252         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
5253         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
5254         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
5255         Change 'vector' to '__vector'.
5256         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
5257         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
5258         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
5259         _mm_avg_pu16): Likewise.  And, whitespace corrections.
5261 2018-10-29  Richard Biener  <rguenther@suse.de>
5263         PR tree-optimization/87785
5264         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
5265         internal defs.
5267 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5269         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
5271 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5273         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
5274         ports configuration macro, defaults to "gnu".
5275         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
5276         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
5278 2018-10-29  Olivier Hainque  <hainque@adacore.com>
5280         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
5282 2018-10-29  Richard Biener  <rguenther@suse.de>
5284         PR tree-optimization/87790
5285         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
5286         (vect_make_slp_decision): Adjust.
5287         (vect_slp_analyze_bb_1): Likewise.
5288         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
5289         edges.
5291 2018-10-29  Richard Biener  <rguenther@suse.de>
5293         PR tree-optimization/87785
5294         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
5295         and processing.
5296         (vect_build_slp_tree): Likewise.
5297         (vect_gather_slp_loads): New function.
5298         (vect_analyze_slp_instance): Gather loads separately from the
5299         SLP tree build.
5301 2018-10-29  Martin Liska  <mliska@suse.cz>
5303         * Makefile.in: Make dependency to json.o.
5304         * doc/gcov.texi: Document new JSON format, remove
5305         old intermediate format documentation.
5306         * gcov.c (struct function_info): Come up with m_name and
5307         m_demangled_name.
5308         (function_info::function_info): Initialize it.
5309         (function_info::~function_info): Release it.
5310         (main): Rename flag_intermediate_format to flag_json_format.
5311         (print_usage): Describe --json-format.
5312         (process_args): Set flag_json_format.
5313         (output_intermediate_line): Remove.
5314         (output_intermediate_json_line): Likewise.
5315         (get_gcov_intermediate_filename): Return new extension
5316         ".gcov.json.gz".
5317         (output_intermediate_file): Implement JSON emission.
5318         (output_json_intermediate_file): Implement JSON emission.
5319         (generate_results): Use ::get_name for function name.
5320         Handle JSON output file.
5321         (read_graph_file): Use ::get_name instead of cplus_demangle.
5322         (read_count_file): Likewise.
5323         (solve_flow_graph): Likewise.
5324         (add_line_counts): Likewise.
5325         (accumulate_line_counts): Use new flag_json_format.
5326         (output_function_details): Use ::get_name instead of cplus_demangle.
5327         (output_lines): Likewise.
5328         * json.cc (test_writing_literals): Add new tests.
5329         * json.h (class literal): Add new boolean constructor.
5331 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
5333         PR rtl-optimization/87701
5334         PR rtl-optimization/87780
5335         * combine.c (make_more_copies): Rewrite.
5337 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
5339         * doc/generic.texi (ABSU_EXPR): Document.
5340         * match.pd (absu(x)*absu(x) -> x*x): Handle.
5341         (absu(absu(X)) -> absu(X)): Likewise.
5342         (absu(-X) -> absu(X)): Likewise.
5343         (absu(X)  where X is nonnegative -> X): Likewise.
5345 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
5347         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
5348         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
5349         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
5350         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
5351         (s-tm-texi): Also check timestamp on d-target.def.
5352         (generated_files): Add TM_D_H and d-target-hooks-def.h.
5353         (build/genhooks.o): Also depend on D_TARGET_DEF.
5354         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
5355         variables.
5356         * config/aarch64/aarch64-d.c: New file.
5357         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
5358         Define.
5359         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
5360         prototype.
5361         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
5362         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
5363         * config/arm/arm-d.c: New file.
5364         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
5365         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
5366         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
5367         * config/arm/t-arm (arm-d.o): New rule.
5368         * config/default-d.c: New file.
5369         * config/glibc-d.c: New file.
5370         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5371         * config/i386/i386-d.c: New file.
5372         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
5373         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
5374         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
5375         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
5376         * config/i386/t-i386 (i386-d.o): New rule.
5377         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5378         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5379         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
5380         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5381         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
5382         * config/mips/mips-d.c: New file.
5383         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
5384         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
5385         * config/mips/t-mips (mips-d.o): New rule.
5386         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5387         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5388         * config/powerpcspe/powerpcspe-d.c: New file.
5389         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
5390         New prototype.
5391         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5392         Support GNU D by using 0 as the language type.
5393         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
5394         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
5395         * config/riscv/riscv-d.c: New file.
5396         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
5397         prototype.
5398         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
5399         * config/riscv/t-riscv (riscv-d.o): New rule.
5400         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5401         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
5402         * config/rs6000/rs6000-d.c: New file.
5403         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
5404         prototype.
5405         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5406         Support GNU D by using 0 as the language type.
5407         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
5408         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
5409         * config/s390/s390-d.c: New file.
5410         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
5411         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
5412         * config/s390/t-s390 (s390-d.o): New rule.
5413         * config/sparc/sparc-d.c: New file.
5414         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
5415         prototype.
5416         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
5417         * config/sparc/t-sparc (sparc-d.o): New rule.
5418         * config/t-glibc (glibc-d.o): New rule.
5419         * configure: Regenerated.
5420         * configure.ac (tm_d_file): New variable.
5421         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
5422         * doc/contrib.texi (Contributors): Add self for the D frontend.
5423         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
5424         * doc/install.texi (Configuration): Mention libphobos as an option for
5425         --enable-shared.  Mention d as an option for --enable-languages.
5426         (Testing): Mention check-d as a target.
5427         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
5428         name suffixes.  Mention d as a -x option.
5429         * doc/sourcebuild.texi (Top Level): Mention libphobos.
5430         * doc/standards.texi (Standards): Add section on D language.
5431         * doc/tm.texi: Regenerated.
5432         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
5433         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
5434         * dwarf2out.c (is_dlang): New function.
5435         (gen_compile_unit_die): Use DW_LANG_D for D.
5436         (declare_in_namespace): Return module die for D, instead of adding
5437         extra declarations into the namespace.
5438         (gen_namespace_die): Generate DW_TAG_module for D.
5439         (gen_decl_die): Handle CONST_DECLSs for D.
5440         (dwarf2out_decl): Likewise.
5441         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
5442         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
5443         kinds of aggregates.
5444         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
5445         * genhooks.c: Include d/d-target.def.
5447 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
5449         PR target/85669
5450         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
5451         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
5452         (STACK_DYNAMIC_OFFSET): Likewise.
5454 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
5456         PR target/80024
5457         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
5458         error message.
5460 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
5462         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
5463         return type and other typos.
5465 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
5467         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
5468         a shorter sequence with fewer branches.
5469         (emit_final_str_compare_gpr): Ditto.
5471 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
5473         * config/rs6000/tmmintrin.h: New file.
5474         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
5476 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
5478         * config/rs6000/mmintrin.h: Enable 32bit compilation.
5479         * config/rs6000/xmmintrin.h: Likewise.
5481 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
5483         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
5485 2018-10-26  Richard Biener  <rguenther@suse.de>
5487         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
5488         and wrapper.
5489         (vect_mark_slp_stmts_relevant): Likewise.
5490         (vect_detect_hybrid_slp_stmts): Likewise.
5491         (vect_bb_slp_scalar_cost): Likewise.
5492         (vect_remove_slp_scalar_calls): Likewise.
5494 2018-10-26  Jan Hubicka  <jh@suse.cz>
5496         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
5497         (warn_types_mismatch): Fix walk of DECL_NAME.
5498         (odr_types_equivalent_p): Fix overactive assert.
5500 2018-10-26  Richard Biener  <rguenther@suse.de>
5502         PR tree-optimization/87105
5503         * tree-vectorizer.h (_slp_tree::refcnt): New member.
5504         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
5505         refcnt.
5506         (vect_create_new_slp_node): Initialize refcnt to one.
5507         (bst_traits): Move.
5508         (scalar_stmts_set_t, bst_fail): Remove.
5509         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
5510         (vect_build_slp_tree): Add bst_map argument and lookup
5511         already created SLP nodes.
5512         (vect_print_slp_tree): Handle a SLP graph, print SLP node
5513         addresses.
5514         (vect_slp_rearrange_stmts): Handle a SLP graph.
5515         (vect_analyze_slp_instance): Adjust and free SLP nodes from
5516         the CSE map.  Fix indenting.
5517         (vect_schedule_slp_instance): Add short-cut.
5519 2018-10-26  Martin Liska  <mliska@suse.cz>
5521         PR testsuite/86158
5522         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
5523         addr_expr and not with pointers.
5525 2018-10-26  Jan Hubicka  <jh@suse.cz>
5527         * tree.c (free_lang_data_in_type): Only check main variants.
5528         * ipa-devirt.c (warn_odr): Make static.
5529         (types_same_for_odr): Drop strict variant.
5530         (types_odr_comparable): Likewise.
5531         (odr_or_derived_type_p): Look for main variants.
5532         (odr_name_hasher::equal): Cleanup comment.
5533         (odr_subtypes_equivalent): Add warn and warned arguments; check main
5534         variants.
5535         (type_variants_equivalent_p): break out from ...
5536         (odr_types_equivalent): ... here; go for main variants where needed.
5537         (warn_odr): ... here; turn static.
5538         (warn_types_mismatch): Compare mangled names of main variants.
5539         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
5540         (type_with_linkage_p): Sanity check that we look at main variant.
5541         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
5542         * tree.h (types_same_for_odr): Drop strict argument.
5544 2018-10-26  Richard Biener  <rguenther@suse.de>
5546         PR tree-optimization/87746
5547         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
5548         Simplify and fix WRT strided store groups with size not
5549         equal to step in element count.
5550         (vect_analyze_group_access_1): Dump the whole group.
5552 2018-10-25  Carl Love  <cel@us.ibm.com>
5554         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
5555         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
5556         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
5557         precicion entry for each overloaded builtin.
5558         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
5559         VSCEDPUO): Rename overloaded name.
5560         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
5561         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
5562         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
5563         define_expand for xscmpexqp instruction.
5564         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
5566 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
5567             Jinsong Ji <jji@us.ibm.com>
5569         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
5570         function with vec_sl.
5571         (_mm_slli_epi32): Likewise.
5572         (_mm_slli_epi64): Likewise.
5573         (_mm_srai_epi16): Replace deprecated function with vec_sra.
5574         (_mm_srai_epi32): Likewise.
5575         (_mm_srli_epi16): Replace deprecated function with vec_sr.
5576         (_mm_srli_epi32): Likewise.
5577         (_mm_srli_epi64): Likewise.
5578         (_mm_sll_epi16): Replace deprecated function with vec_sl.
5579         (_mm_sll_epi32): Likewise.
5580         (_mm_sll_epi64): Likewise.
5581         (_mm_sra_epi16): Replace deprecated function with vec_sra.
5582         (_mm_sra_epi32): Likewise.
5583         (_mm_srl_epi16): Replace deprecated function with vec_sr.
5584         (_mm_srl_epi32): Likewise.
5585         (_mm_srl_epi64): Likewise.
5587 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
5588             Jinsong Ji <jji@us.ibm.com>
5590         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
5591         comparison operators with vec_cmp* for compatibility due to
5592         unfortunate history; clean up formatting and use types more
5593         appropriately.
5594         (_mm_sll_epi32): Likewise.
5595         (_mm_sll_epi64): Likewise.
5596         (_mm_srl_epi16): Likewise.
5597         (_mm_srl_epi32): Likewise.
5598         (_mm_srl_epi64): Likewise.
5600 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
5601             Jinsong Ji <jji@us.ibm.com>
5603         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
5604         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
5605         __vector __bool int.  Use vec_cmpgt in preference to deprecated
5606         function vec_vcmpgtfp.
5607         (_mm_max_ps): Likewise.
5609 2018-10-25  Jeff Law  <law@redhat.com>
5611         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
5612         if returning false.
5614 2018-10-25  Martin Sebor  <msebor@redhat.com>
5616         * doc/extend.texi (aligned): Expand attribute description.
5617         (Alignment): Rename section.  Discuss function arguments.
5619 2018-10-25  Jan Hubicka  <jh@suse.cz>
5621         * ipa-devirt.c (main_odr_variant): Remove.
5622         (hash_odr_name, types_same_for_odr, types_odr_comparable,
5623         odr_name_hasher::equal, odr_subtypes_equivalent_p):
5624         Drop use of main_odr_variant.
5625         (add_type_duplicate): Silence confused warnings on integer types.
5626         (get_odr_type): Always look for main variant.
5627         (register_odr_type): Simplify.
5629 2018-10-25  Richard Biener  <rguenther@suse.de>
5631         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
5632         Initialize ng to silence error with release checking bootstrap.
5634 2018-10-25  Richard Biener  <rguenther@suse.de>
5636         * tree-if-conv.c: Include tree-ssa-sccvn.h.
5637         (tree_if_conversion): Run CSE on the if-converted loop body.
5639 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5641         * config/s390/constraints.md (ZL): New constraint.
5642         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
5643         operands.
5644         * config/s390/s390.md (movdi_larl): Remove.
5645         (movdi_64): Add the LARL alternative.
5647 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
5649         PR bootstrap/87747
5650         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
5651         (RTX_CODE_HWINT_P): New macro.
5652         (rtx_code_size): Use RTX_CODE_HWINT_P ().
5654 2018-10-25  Jan Hubicka  <jh@suse.cz>
5656         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
5657         is anonymous.
5659 2018-10-25  Richard Biener  <rguenther@suse.de>
5661         PR tree-optimization/87665
5662         PR tree-optimization/87745
5663         * tree-vectorizer.h (get_earlier_stmt): Remove.
5664         (get_later_stmt): Pick up UID from the original non-pattern stmt.
5666 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
5668         * options.texi (Deprecated): Move list to Var section.
5670 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
5671             Jinsong Ji <jji@us.ibm.com>
5673         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
5674         __vector long to __vector long long.
5675         (_mm_cvtpd_ps): Likewise.
5676         (_mm_cvttpd_epi32): Likewise.
5677         (_mm_cvtpi32_pd): Likewise.
5678         (_mm_unpackhi_epi64): Likewise.
5679         (_mm_unpacklo_epi64): Likewise.
5681 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
5683         PR rtl-optimization/87720
5684         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
5686 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
5688         * gimple-ssa-isolate-paths.c
5689         (find_implicit_erroneous_behavior): Do not change code if the
5690         pass is running for warnings only.
5691         (find_explicit_erroneous_behavior): Likewise.
5693 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
5695         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
5696         Define as rs6000_mangle_decl_assembler_name.
5697         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
5698         long double to IEEE long double, switch the names of the long
5699         double built-in functions to be <func>f128 instead of <func>l.
5701 2018-10-24  Martin Sebor  <msebor@redhat.com>
5703         * doc/extend.texi (nonnull): List no-argument form.  Reference
5704         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
5706 2018-10-24  Richard Biener  <rguenther@suse.de>
5708         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
5710 2018-10-24  Martin Liska  <mliska@suse.cz>
5712         PR tree-optimization/84436
5713         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
5714         Remove.
5715         (switch_conversion::contains_linear_function_p): New.
5716         (switch_conversion::build_one_array): Support linear
5717         transformation on input.
5718         * tree-switch-conversion.h (struct switch_conversion): Add
5719         contains_linear_function_p declaration.
5721 2018-10-24  Richard Biener  <rguenther@suse.de>
5723         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
5724         if its argument is CONSTANT_CLASS_P.
5726 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
5728         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
5729         it is wrong for forward declarations.
5731 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5733         * config/s390/s390.c (s390_check_qrst_address): Add the missing
5734         SYMBOL_REF_P () check.
5736 2018-10-24  Richard Biener  <rguenther@suse.de>
5738         PR tree-optimization/87105
5739         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
5740         dump classification.
5741         (vect_analyze_data_ref_accesses): Handle duplicate loads and
5742         stores by splitting the affected group after the fact.
5743         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
5744         fail the SLP build because of size constraints.
5746 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5748         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
5749         * configure: Regenerate.
5750         * config.in: Regenerate.
5751         * varasm.c (mergeable_string_section): Use readonly_data_section
5752         if linker doesn't support SHF_MERGE with alignment > 8.
5753         (mergeable_constant_section): Likewise.
5755 2018-10-24  Richard Biener  <rguenther@suse.de>
5757         PR tree-optimization/84013
5758         * tree-ssa-structalias.c (struct msdi_data): New struct for
5759         marshalling data to walk_stmt_load_store_ops.
5760         (maybe_set_dependence_info): Refactor as callback for
5761         walk_stmt_load_store_ops.
5762         (compute_dependence_clique): Set restrict info on all stmt kinds.
5764 2018-10-24  Martin Liska  <mliska@suse.cz>
5766         * cgraph.c (cgraph_node::dump):
5767         Remove reduntant dumps and make tp_first_run dump more compact.
5769 2018-10-24  Richard Biener  <rguenther@suse.de>
5771         PR tree-optimization/87665
5772         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
5773         to reflect reality.
5775 2018-10-12  Jeff Law  <law@redhat.com>
5777         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
5778         for H8/S.
5780 2018-10-23  Richard Biener  <rguenther@suse.de>
5782         * tree-vrp.c (add_assert_info): Guard dump_printf with
5783         dump_enabled_p.
5784         * gimple-ssa-evrp-analyze.c
5785         (evrp_range_analyzer::record_ranges_from_incoming_edge):
5786         Use value_range::ignore_equivs_equal_p.
5788 2018-10-23  Richard Biener  <rguenther@suse.de>
5790         PR tree-optimization/87105
5791         PR tree-optimization/87608
5792         * passes.def (pass_all_early_optimizations): Add early phi-opt
5793         after dce.
5794         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
5795         addition to debug stmts.
5796         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
5797         and abs replacement early.
5798         * tree-cfg.c (gimple_empty_block_p): Likewise.
5800 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
5802         PR target/86383
5803         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
5804         specified to configure.
5805         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
5807 2018-10-23  Richard Biener  <rguenther@suse.de>
5809         PR tree-optimization/87700
5810         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
5812 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
5814         PR target/87674
5815         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
5816         second argument from __mmask16 to __mmask8.
5817         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
5818         _mm_mask_packs_epi32): Likewise.
5819         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
5820         Likewise.
5821         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
5823 2018-10-23  Richard Biener  <rguenther@suse.de>
5825         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
5827 2018-10-23  Richard Biener  <rguenther@suse.de>
5829         PR tree-optimization/86144
5830         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
5831         over simd attribute.
5833 2018-10-23  Richard Biener  <rguenther@suse.de>
5835         PR tree-optimization/87693
5836         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
5837         the case we do not find the taken edge.
5839 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
5840             Jinsong Ji  <jji@us.ibm.com>
5842         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
5843         (_mm_store_pd): Use unaligned vector type for pointer cast.
5844         (_mm_maskmoveu_si128): Likewise.
5845         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
5846         (_mm_store_ps): Use unaligned vector type for pointer cast.
5848 2018-10-22  Paul Koning  <ni1d@arrl.net>
5850         * symtab.c (symtab_node::increase_alignment): Correct max
5851         alignment check.
5853 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
5855         PR tree-optimization/87633
5856         * match.pd: Do not generate unordered integer comparisons.
5858 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
5860         PR rtl-optimization/87600
5861         * combine.c: Add include of expr.h.
5862         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
5863         register to a pseudo.
5864         (make_more_copies): New function, add a copy to a new pseudo after
5865         the moves from hard registers into pseudos.
5866         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
5867         later.  Call make_more_copies.
5869 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
5871         * lra-constraints.c (process_alt_operands): New local array,
5872         matching_early_clobber.  Check matching_early_clobber before
5873         decrementing reject, and set matching_early_clobber after.
5875 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
5877         PR target/87598
5878         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
5879         call output_addr_const and hope for the best.
5881 2018-10-22  Richard Biener  <rguenther@suse.de>
5883         * gimple-ssa-evrp-analyze.c
5884         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
5885         smarter about what ranges to use.
5886         * tree-vrp.c (add_assert_info): Dump here.
5887         (register_edge_assert_for_2): Instead of here at multiple but
5888         not all places.
5890         * gcc.dg/tree-ssa/evrp12.c: New testcase.
5891         * gcc.dg/predict-6.c: Adjust.
5892         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
5893         * gcc.dg/tree-ssa/vrp02.c: Likewise.
5894         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
5896 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
5897         Richard Biener  <rguenther@suse.de>
5899         * bitmap.h: Update data structure documentation, including a
5900         description of bitmap views as either linked-lists or splay trees.
5901         (struct bitmap_element_def): Update comments for splay tree bitmaps.
5902         (struct bitmap_head_def): Likewise.
5903         (bitmap_list_view, bitmap_tree_view): New prototypes.
5904         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
5905         tree_form fields.
5906         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
5907         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
5908         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
5909         released bitmap element here.
5910         (bitmap_element_free): Remove.
5911         (bitmap_elt_clear_from): Work on splay tree bitmaps.
5912         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
5913         this function similar ones such that linked-list bitmap implementation
5914         functions are grouped.
5915         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
5916         and moved for grouping.
5917         (bitmap_list_insert_element_after): Renamed from
5918         bitmap_elt_insert_after, and moved for grouping.
5919         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
5920         (bitmap_tree_link_left, bitmap_tree_link_right,
5921         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
5922         bitmap_tree_link_element, bitmap_tree_unlink_element,
5923         bitmap_tree_find_element): New functions for splay-tree bitmap
5924         implementation.
5925         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
5926         Renamed and moved, see above entries.
5927         (bitmap_tree_listify_from): New function to convert part of a splay
5928         tree bitmap to a linked-list bitmap.
5929         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
5930         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
5931         (bitmap_find_bit): Remove.
5932         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
5933         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
5934         Handle splay tree bitmaps.
5935         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
5936         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
5937         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
5938         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
5939         bitmap_intersect_compl_p, bitmap_ior_and_compl,
5940         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
5941         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
5942         corresponding changes to use linked-list specific bitmap_element
5943         manipulation functions as applicable for efficiency.
5944         (bitmap_tree_to_vec): New function.
5945         (debug_bitmap_elt_file): New function split out from ...
5946         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
5947         (bitmap_print): Likewise.
5949         PR tree-optimization/63155
5950         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
5951         SSA edge worklists.
5952         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
5953         in tree-view.
5955 2018-10-22  Martin Liska  <mliska@suse.cz>
5957         PR tree-optimization/87686
5958         Revert
5959         2018-08-29  Martin Liska  <mliska@suse.cz>
5961         * tree-switch-conversion.c (switch_conversion::expand):
5962         Strenghten assumption about gswitch statements.
5964 2018-10-22  Martin Liska  <mliska@suse.cz>
5966         * ipa-icf.c (sem_item::compare_attributes): Remove.
5967         (sem_item::compare_referenced_symbol_properties): Use
5968         attribute_list_equal instead.
5969         (sem_function::equals_wpa): Likewise.
5970         * ipa-icf.h: Remove compare_attributes.
5972 2018-10-22  Richard Biener  <rguenther@suse.de>
5974         PR middle-end/87682
5975         * mem-stats.h (mem_usage::operator==): Fix pasto.
5977 2018-10-22  Richard Biener  <rguenther@suse.de>
5979         PR tree-optimization/87640
5980         * tree-vrp.c (set_value_range_with_overflow): Decompose
5981         incomplete result.
5982         (extract_range_from_binary_expr_1): Adjust.
5984 2018-10-22  Martin Jambor  <mjambor@suse.cz>
5986         * tree-eh.h (stmt_could_throw_p): Add function parameter.
5987         (stmt_can_throw_external): Likewise.
5988         (stmt_can_throw_internal): Likewise.
5989         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
5990         (lower_eh_constructs_2): Likewise.
5991         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
5992         (stmt_can_throw_external): Likewise.
5993         (stmt_can_throw_internal): Likewise.
5994         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
5995         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
5996         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
5997         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
5998         (pass_lower_eh_dispatch::execute): Pass cfun to
5999         stmt_can_throw_external.
6000         (cleanup_empty_eh): Likewise.
6001         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
6002         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
6003         stmt_can_throw_external instead of pushing it to cfun.
6004         (symbol_table::create_edge): Likewise.
6005         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
6006         stmt_can_throw_internal.
6007         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
6008         to stmt_could_throw_p.
6009         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
6010         stmt_can_throw_internal.
6011         (pass_store_merging::execute): Likewise.
6012         * gimple-ssa-strength-reduction.c
6013         (find_candidates_dom_walker::before_dom_children): Pass cfun to
6014         stmt_could_throw_p.
6015         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
6016         stmt_can_throw_internal.
6017         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
6018         to stmt_can_throw_external.
6019         (check_stmt): Pass cfun to stmt_could_throw_p.
6020         (check_stmt): Pass cfun to stmt_can_throw_external.
6021         (pass_nothrow::execute): Likewise.
6022         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
6023         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
6024         stmt_can_throw_internal.
6025         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
6026         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
6027         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
6028         * tree-complex.c (expand_complex_libcall): Pass cfun to
6029         stmt_could_throw_p and to stmt_can_throw_internal.
6030         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
6031         * tree-inline.c (copy_edges_for_bb): Likewise.
6032         (maybe_move_debug_stmts_to_successors): Likewise.
6033         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
6034         stmt_could_throw_p.
6035         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
6036         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
6037         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
6038         stmt_can_throw_internal.
6039         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
6040         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
6041         stmt_could_throw_p.
6042         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
6043         stmt_can_throw_internal.
6044         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
6045         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
6046         stmt_could_throw_p.
6047         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
6048         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
6049         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6050         (convert_mult_to_fma_1): Likewise.
6051         (convert_to_divmod): Likewise.
6052         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
6053         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
6054         * tree-ssa-propagate.c
6055         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
6056         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
6057         (maybe_optimize_range_tests): Likewise.
6058         (linearize_expr_tree): Likewise.
6059         (reassociate_bb): Likewise.
6060         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6061         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
6062         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
6063         (handle_char_store): Likewise.
6064         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
6065         stmt_can_throw_internal.
6066         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
6067         stmt_could_throw_p.
6068         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6069         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
6070         (vectorizable_simd_clone_call): Likewise.
6071         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
6072         (gimple_stringop_fixed_value): Likewise.
6074 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
6076         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
6077         literal pool references.
6078         (s390_check_qrst_address): Adapt to the new behavior of
6079         s390_loadrelative_operand_p ().
6081 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6083         PR target/72782
6084         * config/i386/sse.md (*andnot<mode>3_bcst): New.
6086 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6088         PR target/72782
6089         * config/i386/sse.md (*<code><mode>3_bcst): New.
6091 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
6093         PR target/72782
6094         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
6095         V4DI, V16SI and V8DI.
6096         (*sub<mode>3<mask_name>_bcst): New.
6097         (*add<mode>3<mask_name>_bcst): Likewise.
6099 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
6100             Jinsong Ji  <jji@us.ibm.com>
6102         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
6103         __m64 with __vector unsigned long long for compatibility.
6104         (_mm_movemask_epi8): Likewise.
6105         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
6106         (_mm_cvttps_pi32): Likewise.
6107         (_mm_cvtpi32_ps): Likewise.
6108         (_mm_cvtps_pi16): Likewise.
6109         (_mm_loadh_pi): Likewise.
6110         (_mm_storeh_pi): Likewise.
6111         (_mm_movehl_ps): Likewise.
6112         (_mm_movelh_ps): Likewise.
6113         (_mm_loadl_pi): Likewise.
6114         (_mm_storel_pi): Likewise.
6115         (_mm_movemask_ps): Likewise.
6116         (_mm_shuffle_pi16): Likewise.
6118 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6120         PR target/72782
6121         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
6122         __builtin_ia32_vfnmsubpd512_mask.
6123         (_mm512_mask_fnmsub_round_pd): Likewise.
6124         (_mm512_fnmsub_pd): Likewise.
6125         (_mm512_mask_fnmsub_pd): Likewise.
6126         (_mm512_maskz_fnmsub_round_pd): Use
6127         __builtin_ia32_vfnmsubpd512_maskz.
6128         (_mm512_maskz_fnmsub_pd): Likewise.
6129         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
6130         (_mm512_mask_fnmsub_round_ps): Likewise.
6131         (_mm512_fnmsub_ps): Likewise.
6132         (_mm512_mask_fnmsub_ps): Likewise.
6133         (_mm512_maskz_fnmsub_round_ps): Use
6134         __builtin_ia32_vfnmsubps512_maskz.
6135         (_mm512_maskz_fnmsub_ps): Likewise.
6136         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
6137         __builtin_ia32_vfnmsubpd256_mask.
6138         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
6139         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
6140         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
6141         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
6142         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
6143         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
6144         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
6145         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
6146         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
6147         __builtin_ia32_vfnmsubpd.
6148         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
6149         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
6150         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
6151         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
6152         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
6153         * config/i386/i386-builtin.def: Add
6154         __builtin_ia32_vfnmsubpd256_mask,
6155         __builtin_ia32_vfnmsubpd256_maskz,
6156         __builtin_ia32_vfnmsubpd128_mask,
6157         __builtin_ia32_vfnmsubpd128_maskz,
6158         __builtin_ia32_vfnmsubps256_mask,
6159         __builtin_ia32_vfnmsubps256_maskz,
6160         __builtin_ia32_vfnmsubps128_mask,
6161         __builtin_ia32_vfnmsubps128_maskz,
6162         __builtin_ia32_vfnmsubpd512_mask,
6163         __builtin_ia32_vfnmsubpd512_maskz,
6164         __builtin_ia32_vfnmsubps512_mask,
6165         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
6166         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
6167         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
6168         __builtin_ia32_vfnmsubpd256.
6169         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
6170         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
6171         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
6172         Likewise.
6173         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
6174         Likewise.
6175         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
6176         Likewise.
6177         (fmai_vmfnmsub_<mode><round_name>): Likewise.
6179 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6181         PR target/72782
6182         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
6183         __builtin_ia32_vfnmaddpd512_mask.
6184         (_mm512_mask_fnmadd_round_pd): Likewise.
6185         (_mm512_fnmadd_pd): Likewise.
6186         (_mm512_mask_fnmadd_pd): Likewise.
6187         (_mm512_maskz_fnmadd_round_pd): Use
6188         __builtin_ia32_vfnmaddpd512_maskz.
6189         (_mm512_maskz_fnmadd_pd): Likewise.
6190         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
6191         (_mm512_mask_fnmadd_round_ps): Likewise.
6192         (_mm512_fnmadd_ps): Likewise.
6193         (_mm512_mask_fnmadd_ps): Likewise.
6194         (_mm512_maskz_fnmadd_round_ps): Use
6195         __builtin_ia32_vfnmaddps512_maskz.
6196         (_mm512_maskz_fnmadd_ps): Likewise.
6197         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
6198         __builtin_ia32_vfnmaddpd256_mask.
6199         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
6200         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
6201         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
6202         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
6203         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
6204         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
6205         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
6206         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
6207         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
6208         __builtin_ia32_vfnmaddpd.
6209         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
6210         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
6211         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
6212         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
6213         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
6214         * config/i386/i386-builtin.def: Add
6215         __builtin_ia32_vfnmaddpd256_mask,
6216         __builtin_ia32_vfnmaddpd256_maskz,
6217         __builtin_ia32_vfnmaddpd128_mask,
6218         __builtin_ia32_vfnmaddpd128_maskz,
6219         __builtin_ia32_vfnmaddps256_mask,
6220         __builtin_ia32_vfnmaddps256_maskz,
6221         __builtin_ia32_vfnmaddps128_mask,
6222         __builtin_ia32_vfnmaddps128_maskz,
6223         __builtin_ia32_vfnmaddpd512_mask,
6224         __builtin_ia32_vfnmaddpd512_maskz,
6225         __builtin_ia32_vfnmaddps512_mask,
6226         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
6227         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
6228         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
6229         __builtin_ia32_vfnmaddpd256.
6230         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
6231         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
6232         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
6233         Likewise.
6234         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
6235         Likewise.
6236         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
6237         Likewise.
6238         (fmai_vmfnmadd_<mode><round_name>): Likewise.
6240 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6242         PR target/72782
6243         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
6244         __builtin_ia32_vfmsubpd512_mask.
6245         (_mm512_mask_fmsub_round_pd): Likewise.
6246         (_mm512_fmsub_pd): Likewise.
6247         (_mm512_mask_fmsub_pd): Likewise.
6248         (_mm512_maskz_fmsub_round_pd): Use
6249         __builtin_ia32_vfmsubpd512_maskz.
6250         (_mm512_maskz_fmsub_pd): Likewise.
6251         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
6252         (_mm512_mask_fmsub_round_ps): Likewise.
6253         (_mm512_fmsub_ps): Likewise.
6254         (_mm512_mask_fmsub_ps): Likewise.
6255         (_mm512_maskz_fmsub_round_ps): Use
6256         __builtin_ia32_vfmsubps512_maskz.
6257         (_mm512_maskz_fmsub_ps): Likewise.
6258         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
6259         __builtin_ia32_vfmsubpd256_mask.
6260         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
6261         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
6262         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
6263         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
6264         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
6265         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
6266         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
6267         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
6268         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
6269         __builtin_ia32_vfmsubpd.
6270         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
6271         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
6272         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
6273         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
6274         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
6275         * config/i386/i386-builtin.def: Add
6276         __builtin_ia32_vfmsubpd256_mask,
6277         __builtin_ia32_vfmsubpd256_maskz,
6278         __builtin_ia32_vfmsubpd128_mask,
6279         __builtin_ia32_vfmsubpd128_maskz,
6280         __builtin_ia32_vfmsubps256_mask,
6281         __builtin_ia32_vfmsubps256_maskz,
6282         __builtin_ia32_vfmsubps128_mask,
6283         __builtin_ia32_vfmsubps128_maskz,
6284         __builtin_ia32_vfmsubpd512_mask,
6285         __builtin_ia32_vfmsubpd512_maskz,
6286         __builtin_ia32_vfmsubps512_mask,
6287         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
6288         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
6289         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
6290         __builtin_ia32_vfmsubpd256.
6291         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
6292         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
6293         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
6294         Likewise.
6295         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
6296         Likewise.
6297         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
6298         Likewise.
6299         (fmai_vmfmsub_<mode><round_name>): Likewise.
6301 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6303         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
6304         Remove plus.  Renamed to ...
6305         (*sub<mode>3<mask_name>_bcst): This.
6306         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
6307         (*add<mode>3<mask_name>_bcst): This.
6309 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6311         PR target/72782
6312         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
6314 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
6316         PR target/87662
6317         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
6318         (_mm_or_epi32): Likewise.
6319         (_mm256_xor_epi32): Likewise.
6320         (_mm_xor_epi32): Likewise.
6321         (_mm256_or_epi64): Likewise.
6322         (_mm_or_epi64): Likewise.
6323         (_mm256_xor_epi64): Likewise.
6324         (_mm_xor_epi64): Likewise.
6326 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
6328         PR target/72782
6329         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
6331 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
6333         PR middle-end/87647
6334         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
6336 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
6338         * doc/ux.texi: Move @section directly after @node.
6340 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
6342         PR middle-end/85488
6343         PR middle-end/87649
6344         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
6345         depend closely nested inside of loop with ordered clause with
6346         a parameter.
6348 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
6350         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
6351         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
6352         * doc/ux.texi: New file.
6354 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
6356         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
6357         be the first CR field allocated.
6359 2018-10-19  Richard Biener  <rguenther@suse.de>
6361         PR target/87657
6362         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
6363         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
6365 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
6367         PR target/72782
6368         * config/i386/sse.md
6369         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
6370         (*add<mode>3<mask_name>_bcst_2): Likewise.
6372 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
6374         * config/i386/sse.md
6375         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
6376         Replace nonimmediate_operand with register_operand.
6377         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
6378         Likewise.
6379         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
6380         Likewise.
6382 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
6384         PR rtl-optimization/87596
6385         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
6386         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
6387         for instructions in FROM..TO range.
6389 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
6391         * cfgexpand.c (expand_one_var): Use specific wording in error message
6392         for non-local frame variables.
6393         * stor-layout.c (layout_decl): Do not issue a warning for them.
6395 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
6397         * haifa-sched.c (priority): Add force_recompute parameter.
6398         (apply_replacement): Call priority () with force_recompute = true.
6399         (restore_pattern): Likewise.
6401 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6403         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
6404         HOST_BITS_PER_WIDE_INT.
6405         (test_vector_ops_duplicate): Likewise.
6407 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6409         PR target/72782
6410         * config/i386/sse.md (VF_AVX512): New.
6411         (avx512bcst): Likewise.
6412         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
6413         Likewise.
6414         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
6415         Likewise.
6416         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
6417         Likewise.
6419 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
6421         * doc/invoke.texi (-dumpversion): Improve grammar.
6422         (-dumpfullversion): Make more consistent with -dumpversion.
6424 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
6426         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
6427         Set JUMP_LABEL to the jump insn.
6428         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
6429         Predict emitted jump and add label to jump insn.
6431 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
6433         PR tree-optimization/87562
6434         * input.c (get_substring_ranges_for_loc): Use
6435         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
6436         getting the linemap for the endpoint.  Verify that it's either
6437         in the same linemap as the start point's spelling location, or
6438         at least in the same file.
6440 2018-10-18  Richard Biener  <rguenther@suse.de>
6442         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
6443         feed width-specific load/store costs through ix86_vec_cost.
6444         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
6445         (k8_cost): Likewise.
6446         (bdver_cost): Likewise.
6447         (znver1_cost): Likewise.
6448         (btver1_cost): Likewise.
6449         (btver2_cost): Likewise.
6451 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6453         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
6454         to simplify subreg of vec_merge.
6456 2018-10-18  Richard Biener  <rguenther@suse.de>
6458         * config/i386/i386.c: Fix costing of vector FMA.
6460 2018-10-18  Richard Biener  <rguenther@suse.de>
6462         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
6463         and argument.
6464         (ix86_builtin_vectorization_cost): For vec_construct properly
6465         cost insertion into SSE regs.
6466         (...): Adjust calls to ix86_vec_cost.
6468 2018-10-18  Richard Biener  <rguenther@suse.de>
6470         PR middle-end/87087
6471         Revert
6472         2018-02-07  Richard Biener  <rguenther@suse.de>
6474         PR tree-optimization/84204
6475         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6476         this place.
6478 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
6480         PR target/87537
6481         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
6482         of vec_duplicate.
6483         (test_vector_ops_duplicate): Add test for a scalar subreg of a
6484         VEC_MERGE of a VEC_DUPLICATE.
6486 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
6488         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
6489         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
6490         * doc/standards.texi (C Language): Document C2X.
6491         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
6492         (rl78_option_override): Handle "GNU C2X" language name.
6494 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
6496         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
6497         Document C17 as published in 2018.
6499 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
6501         PR middle-end/87623
6502         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
6503         bail out if both sides do not have the same storage order.
6505 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
6507         * bitmap.c (bitmap_head::dump): New.
6508         * bitmap.h (bitmap_head): Add dump().
6509         * gimple-ssa-evrp-analyze.c
6510         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
6511         (evrp_range_analyzer::set_ssa_range_info): Same.
6512         (evrp_range_analyzer::record_ranges_from_phis): Same.
6513         (evrp_range_analyzer::record_ranges_from_stmt): Same.
6514         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
6515         * gimple-ssa-sprintf.c (get_int_range): Same.
6516         (format_integer): Same.
6517         (sprintf_dom_walker::handle_gimple_call): Same.
6518         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
6519         (ipcp_vr_lattice::top_p): Same.
6520         (ipcp_vr_lattice::bottom_p): Same.
6521         (ipcp_vr_lattice::set_to_bottom): Same.
6522         (ipa_vr_operation_and_type_effects): Same.
6523         (propagate_vr_across_jump_function): Same.
6524         (ipcp_store_vr_results): Same.
6525         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
6526         (ipa_print_node_jump_functions_for_edge): Same.
6527         (ipa_get_value_range): Same.
6528         (ipa_compute_jump_functions_for_edge): Same.
6529         (ipa_write_jump_function): Same.
6530         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
6531         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
6532         Same.
6533         * vr-values.c (set_value_range_to_nonnegative): Same.
6534         (set_value_range_to_truthvalue): Same.
6535         (vr_values::get_value_range): Same.
6536         (vr_values::set_defs_to_varying): Same.
6537         (vr_values::update_value_range): Same.
6538         (symbolic_range_based_on_p): Same.
6539         (vr_values::op_with_boolean_value_range_p): Same.
6540         (vr_values::extract_range_for_var_from_comparison_expr): Same.
6541         (vr_values::extract_range_from_ssa_name): Same.
6542         (vr_values::extract_range_from_binary_expr): Same.
6543         (vr_values::extract_range_from_unary_expr): Same.
6544         (vr_values::extract_range_from_cond_expr): Same.
6545         (vr_values::extract_range_from_comparison): Same.
6546         (vr_values::check_for_binary_op_overflow): Same.
6547         (vr_values::extract_range_basic): Same.
6548         (vr_values::extract_range_from_assignment): Same.
6549         (compare_ranges): Same.
6550         (compare_range_with_value): Same.
6551         (vr_values::adjust_range_with_scev): Same.
6552         (vrp_valueize): Same.
6553         (vrp_valueize_1): Same.
6554         (vr_values::get_vr_for_comparison): Same.
6555         (vr_values::compare_name_with_value): Same.
6556         (vr_values::compare_names): Same.
6557         (vr_values::vrp_evaluate_conditional): Same.
6558         (find_case_label_ranges): Same.
6559         (vr_values::vrp_visit_switch_stmt): Same.
6560         (vr_values::extract_range_from_phi_node): Same.
6561         (vr_values::simplify_div_or_mod_using_ranges): Same.
6562         (vr_values::simplify_bit_ops_using_ranges): Same.
6563         (test_for_singularity): Same.
6564         (range_fits_type_p): Same.
6565         (vr_values::simplify_cond_using_ranges_1): Same.
6566         (vr_values::simplify_switch_using_ranges): Same.
6567         (vr_values::simplify_float_conversion_using_ranges): Same.
6568         (vr_values::two_valued_val_range_p): Same.
6569         (vr_values::add_equivalence): Move to value_range::equiv_add.
6570         * vr-values.h (vr_values::add_equivalence): Remove.
6571         (VR_INITIALIZER): Remove.
6572         * tree-vrp.c (value_range::set): New.
6573         (value_range::equiv_add): New.
6574         (value_range::value_range): New.
6575         (value_range::deep_copy): New.
6576         (value_range::check): New.
6577         (value_range::equal_p): New.
6578         (value_range::ignore_equivs_equal_p): New.
6579         (value_range::operator==): New.
6580         (value_range::operator!=): New.
6581         (value_range::symbolic_p): New.
6582         (value_range::numeric_p): New.
6583         (value_range::set_undefined): New.
6584         (value_range::set_varying): New.
6585         (value_range::may_contain_p): New.
6586         (value_range::equiv_clear): New.
6587         (value_range::singleton_p): New.
6588         (value_range::intersect): New.
6589         (value_range::dump): New.
6590         (value_range::set_and_canonicalize): New.
6591         (set_value_range): Adjust for value_range API.
6592         (set_value_range_to_undefined): Same.
6593         (set_value_range_to_varying): Same.
6594         (set_and_canonicalize_value_range): Same.
6595         (set_value_range_to_nonnull): Same.
6596         (set_value_range_to_null): Same.
6597         (range_is_null): Same.
6598         (range_is_nonnull): Same.
6599         (range_int_cst_p): Same.
6600         (range_int_cst_singleton_p): Same.
6601         (symbolic_range_p): Same.
6602         (range_includes_zero_p): Same.
6603         (value_range_constant_singleton): Same.
6604         (vrp_set_zero_nonzero_bits): Same.
6605         (ranges_from_anti_range): Same.
6606         (extract_range_into_wide_ints): Same.
6607         (extract_range_from_multiplicative_op): Same.
6608         (set_value_range_with_overflow): Same.
6609         (extract_range_from_binary_expr_1): Same.
6610         (extract_range_from_unary_expr): Same.
6611         (dump_value_range): Same.
6612         (debug_value_range): Same.
6613         (vrp_prop::check_array_ref): Same.
6614         (vrp_prop::check_mem_ref): Same.
6615         (vrp_prop::vrp_initialize): Same.
6616         (vrp_prop::visit_stmt): Same.
6617         (intersect_ranges): Same.
6618         (vrp_prop::visit_phi): Same.
6619         (vrp_prop::vrp_finalize): Same.
6620         (determine_value_range_1): Same.
6621         (determine_value_range): Same.
6622         (vrp_intersect_ranges_1): Rename to...
6623         (vrp_intersect_1): this.
6624         (vrp_intersect_ranges): Rename to...
6625         (value_range::intersect_helper): ...this.
6626         (vrp_meet_1): Rename to...
6627         (value_range::union_helper): ...this.
6628         (vrp_meet): Rename to...
6629         (value_range::union_): ...this.
6630         (copy_value_range): Remove.
6631         * tree-vrp.h (struct value_range): Rewrite into a proper class.
6632         (value_range::vrtype): New.
6633         (value_range::type): New.
6634         (value_range::equiv): New.
6635         (value_range::min): New.
6636         (value_range::max): New.
6637         (value_range::varying_p): New.
6638         (value_range::undefined_p): New.
6639         (value_range::null_p): New.
6640         (value_range::equiv_add): New.
6641         (copy_value_range): Remove.
6643 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
6645         * Makefile.in (SELFTEST_TARGETS): New.
6646         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
6647         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
6648         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
6649         c/Make-lang.in.
6650         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
6651         (selftest-c++-gdb, selftest-c++-valgrind): Move to
6652         cp/Make-lang.in.
6653         * configure: Regenerate.
6654         * configure.ac (selftest_languages): New.
6656 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
6658         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
6659         overflow wraps argument.
6660         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
6661         wide_int_range_multiplicative_op.
6662         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
6663         overflow wraps argument.
6664         (wide_int_range_multiplicative_op): Same.
6665         (wide_int_range_lshift): Same.
6666         (wide_int_range_div): Same.
6667         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
6668         (wide_int_range_lshift): Same.
6669         (wide_int_range_div): Same.
6671 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
6673         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
6674         use sign as argument.
6675         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
6676         wide_int_range_shift_undefined_p.
6678 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6680         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
6681         Rename to...
6682         (@despeculate_copy<ALLI_TI:mode>): ... This.
6683         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
6684         switch statement.
6686 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6688         * config.gcc: Obsolete *-*-solaris2.10*.
6689         * doc/install.texi (Specific, *-*-solaris2*): Document it.
6691 2018-10-12  Jeff Law  <law@redhat.com>
6693         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
6694         reg + sym +- const_int addressing modes.
6696 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
6698         * common.opt (fdiagnostics-minimum-margin-width=): New option.
6699         * diagnostic-show-locus.c (layout::layout): Apply the minimum
6700         margin width.
6701         (layout::start_annotation_line): Only print up to 3 of the
6702         margin character, to avoid touching the left-hand side.
6703         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
6704         minimum margin width, as set by test_diagnostic_context's ctor.
6705         (selftest::test_fixit_insert_containing_newline): Likewise.
6706         (selftest::test_fixit_insert_containing_newline_2): Likewise.
6707         (selftest::test_line_numbers_multiline_range): Clear
6708         dc.min_margin_width.
6709         * diagnostic.c (diagnostic_initialize): Initialize
6710         min_margin_width.
6711         * diagnostic.h (struct diagnostic_context): Add field
6712         "min_margin_width".
6713         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
6714         * opts.c (common_handle_option): Handle
6715         OPT_fdiagnostics_minimum_margin_width_.
6716         * selftest-diagnostic.c
6717         (selftest::test_diagnostic_context::test_diagnostic_context):
6718         Initialize min_margin_width to 6.
6719         * toplev.c (general_init): Initialize global_dc->min_margin_width.
6721 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
6723         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
6724         Fix usage of "error_at_rich_loc" in the comment.
6726 2018-10-15  Renlin Li  <renlin.li@arm.com>
6728         PR target/87563
6729         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
6730         if-conversioned loop when it contains ifn with types not
6731         supported by backend.
6732         * internal-fn.c (expand_direct_optab_fn): Add an assert.
6733         (direct_internal_fn_supported_p): New helper function.
6734         * internal-fn.h (direct_internal_fn_supported_p): Declare.
6736 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
6738         PR target/87572
6739         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
6740         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
6741         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
6743 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
6745         PR tree-optimization/87022
6746         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
6747         bits in dist vector rather than the first one.
6749 2018-10-15  Richard Biener  <rguenther@suse.de>
6751         PR middle-end/87610
6752         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
6753         (visit_loadstore): When a used restrict tag escaped verify that
6754         the points-to solution of "other" pointers do not include
6755         escaped.
6756         (compute_dependence_clique): If a used restrict tag escaped
6757         communicated that down to visit_loadstore.
6759 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
6761         * config/s390/s390.c (s390_expand_vec_init): Force vector element
6762         into reg if it isn't a general operand.
6764 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
6766         PR target/87599
6767         * config/i386/sse.md (*vec_dupv2di): Add register source to
6768         movddup.
6770 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
6772         PR target/87572
6773         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
6774         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
6775         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
6776         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
6778 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
6780         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
6781         (notice_args_size): Set it in the current trace if no insn that can
6782         throw internally has been seen yet.
6783         (connect_traces): When connecting args_size between traces, allow the
6784         incoming values not to match if there is an insn setting it before the
6785         first insn that can throw internally; in that case, force the creation
6786         of a CFI note on this latter insn.
6788 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
6790         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
6791         the base class, because using the injected-class-name was not clearly
6792         specified until DR 176.
6794 2018-10-12  Paul Koning  <ni1d@arrl.net>
6796         * config/pdp11/pdp11.md (doloop_end): New expander.
6797         (doloop_end_insn): renamed from "doloop_end".
6798         (addqi3): New pattern.
6799         (subqi3): New pattern.
6800         * config/pdp11/predicates.md (incdec_operand): New predicate.
6802 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
6804         PR middle-end/81376
6805         * real.c (format_helper::can_represent_integral_type_p): New function
6806         * real.h (format_helper::can_represent_integral_type_p): Ditto.
6807         * match.pd: New pattern.
6809 2018-10-12  Alexandre Oliva <oliva@adacore.com>
6811         * configure.ac: Introduce --enable-large-address-aware
6812         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
6813         * doc/install.texi: Document it.
6814         * configure, config.in: Rebuilt.
6815         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
6816         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
6817         (LINK_SPEC): Insert it.
6818         * config/i386/mingw-w64.h: Likewise.
6820         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
6822 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
6824         PR rtl-optimization/87600
6825         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
6827 2018-10-12  Paul Koning  <ni1d@arrl.net>
6829         * doc/md.texi (doloop_end): Document that the pattern code may
6830         need to check operand mode.
6832 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
6834         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
6835         to zero-extend between int and floating-point registers.
6836         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
6837         ldp into floating-point registers.  Add type and arch attributes.
6838         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
6839         Use f_loads for type attribute.
6841 2018-10-11  Martin Sebor  <msebor@redhat.com>
6843         * doc/extend.texi (attribute packed): Correct typos.
6845 2018-10-11  Martin Sebor  <msebor@redhat.com>
6847         * doc/extend.texi (attribute flatten): Mention interaction with
6848         noinline.
6850 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
6852         PR target/87156
6853         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6854         Set new_decl virtual flag to zero.
6856 2018-10-11  Martin Sebor  <msebor@redhat.com>
6858         PR middle-end/87593
6859         * doc/extend.texi (attribute format_arg): Discuss using multiple
6860         attributes on a single function.
6862 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6864         PR tree-optimization/86829
6865         * match.pd (sin (atan (x))): New simplification rules.
6866         (cos (atan (x))): Likewise.
6867         * real.c (build_sinatan_real): New function.
6868         * real.h (build_sinatan_real): Prototype.
6870 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
6872         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
6873         function.
6874         (fold_mergeeo_helper): New helper function.
6875         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
6876         intrinsics.  Correct some whitespace indentation issues.
6878 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
6880         PR target/87511
6881         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
6882         Use HOST_WIDE_INT_1U for shift.
6884 2018-10-11  Doug Rupp  <rupp@adacore.com>
6885             Olivier Hainque  <hainque@adacore.com>
6887         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
6888         Pass --relax to the linker for RTPs.
6889         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
6891 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
6892             Jan Hubicka  <jh@suse.cz>
6893             Martin Jambor  <mjambor@suse.cz>
6895         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
6896         the same elements are repeated rather than printing all of them.
6897         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
6898         "repeated" elements.
6899         * read-rtl-function.c (test_loading_repeat): New function.
6900         (read_rtl_function_c_tests): Call test_loading_repeat.
6901         * rtl-tests.c (test_dumping_repeat): New function.
6902         (rtl_tests_c_tests): Call test_dumping_repeat.
6904 2018-10-11  Richard Biener  <rguenther@suse.de>
6906         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
6907         bdver?_cost): Unify to ...
6908         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
6909         * config/i386/i386.c (processor_cost_table): Adjust.
6911 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
6913         PR middle-end/87574
6914         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
6915         the thunk when expanding to GIMPLE.
6917 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6919         PR bootstrap/87551
6920         * varasm.c (mergeable_string_section): Don't try to move zero-length
6921         strings to the merge section.
6923 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
6925         PR target/87573
6926         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
6928 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
6930         PR target/87550
6931         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
6932         to special_args set.
6934 2018-10-10  Richard Biener  <rguenther@suse.de>
6936         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
6937         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
6938         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
6939         and recursing and pattern terminating the recursion on SSE
6940         vector width using ix86_expand_reduc.
6941         (reduc_sminmax_scal_<mode>): Split into part reducing to half
6942         width and recursing and SSE2 vector variant doing the final
6943         reduction with ix86_expand_reduc.
6944         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
6945         with terminating the recursion at AVX level, splitting that
6946         to SSE there.
6948 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
6950         * genmatch.c (error_cb): Rename to...
6951         (diagnostic_cb): ...this, converting int params to enums.
6952         (fatal_at): Update for renaming.
6953         (warning_at): Likewise.
6954         (main): Likewise.
6955         * input.c (selftest::ebcdic_execution_charset::apply):
6956         Update for renaming of...
6957         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
6958         to...
6959         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
6960         converting level and reason to enums.
6961         (class selftest::lexer_error_sink): Rename to...
6962         (class selftest::lexer_test_options): ...this, renaming field
6963         "m_errors" to "m_diagnostics".
6964         (selftest::lexer_test_options::apply): Update for renaming of...
6965         (selftest::lexer_test_options::on_error): ...this, renaming to...
6966         (selftest::lexer_test_options::on_diagnostic): ...this
6967         converting level and reason to enums.
6968         (selftest::test_lexer_string_locations_raw_string_unterminated):
6969         Update for renamings.
6970         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
6971         "reason".
6973 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
6975         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
6976         * config/rs6000/pmmintrin.h: New file.
6978 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
6980         PR tree-optimization/86659
6981         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
6983 2018-10-09  Richard Biener  <rguenther@suse.de>
6985         PR tree-optimization/63155
6986         * tree-ssa-structalias.c: Include tree-ssa.h.
6987         (get_constraint_for_ssa_var): For undefs return nothing_id.
6988         (find_func_aliases): Cleanup PHI handling.
6990 2018-10-09  Richard Biener  <rguenther@suse.de>
6992         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
6993         replacements.
6995 2018-10-09  Martin Liska  <mliska@suse.cz>
6997         * asan.c (asan_emit_stack_protection): If a stack variable
6998         is located in a same file as current function, then emit
6999         line info into variable definition string.
7001 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7003         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
7004         information.
7006 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7008         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
7009         on the thunk.
7011 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
7013         PR tree-optimization/86659
7014         * gimple-match.h (struct gimple_match_op): Add reverse field.
7015         (gimple_match_op::set_op): New overloaded method.
7016         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
7017         the REF_REVERSE_STORAGE_ORDER flag on the value.
7018         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
7019         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
7021 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
7023         PR middle-end/63155
7024         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
7025         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
7027 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
7029         PR target/87517
7030         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
7031         Defined with __builtin_ia32_vfmaddsubpd512_mask.
7033 2018-10-08  Richard Biener  <rguenther@suse.de>
7035         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
7036         cost the same as AVX128 ones.
7038 2018-10-08  Paul Koning  <ni1d@arrl.net>
7040         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
7041         (expand_block_move): New function.
7042         * config/pdp11/pdp11.c (output_block_move): Remove.
7043         (expand_block_move): New function.
7044         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
7045         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
7046         (*movmemhi1): Remove.
7048 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
7050         * config/s390/2827.md: Increase latencies for some FP instructions.
7052 2018-10-08  Richard Biener  <rguenther@suse.de>
7054         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7055         Open a dump scope.
7056         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
7057         * tree-vectorizer.h (dump_stmt_cost): Adjust.
7058         (add_stmt_cost): Dump return value of the hook.
7060 2018-10-08  Richard Biener  <rguenther@suse.de>
7062         PR tree-optimization/63155
7063         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
7064         (ssa_propagation_engine::ssa_propagate): Remove redundant
7065         bitmap bit clearing.
7067 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
7069         PR rtl-optimization/86939
7070         PR rtl-optimization/87479
7071         * ira.h (non_conflicting_reg_copy_p): New prototype.
7072         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
7073         (make_hard_regno_dead): Don't add conflicts for register
7074         ignore_reg_for_conflicts.
7075         (make_object_dead): Likewise.
7076         (non_conflicting_reg_copy_p): New function.
7077         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
7078         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
7079         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
7080         (make_hard_regno_dead): Don't add conflicts for register
7081         ignore_reg_for_conflicts.  Remove special conflict handling of
7082         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
7083         check_pic_pseudo_p and update callers.
7084         (mark_pseudo_dead): Don't add conflicts for register
7085         ignore_reg_for_conflicts.
7086         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
7088 2018-10-05  Andrew Waterman  <andrew@sifive.com>
7089             Jim Wilson  <jimw@sifive.com>
7091         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
7092         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
7093         new pattern using HONOR_SNANS that emits one extra instruction.
7095 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
7097         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
7098         patterns): Merge SI and DI patterns to a GPR pattern.
7099         (unnamed define_insn and define_split for record form of that): Merge
7100         to a single define_insn_and_split pattern.
7102 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
7104         PR c++/56856
7105         * input.c (expand_location_to_spelling_point): Add param "aspect"
7106         and use rather than hardcoding LOCATION_ASPECT_CARET.
7107         (get_substring_ranges_for_loc): Handle the case of a single token
7108         within a macro expansion.
7109         * input.h (expand_location_to_spelling_point): Add "aspect" param,
7110         defaulting to LOCATION_ASPECT_CARET.
7112 2018-10-05  Paul Koning  <ni1d@arrl.net>
7114         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
7115         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
7116         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
7117         (pdp11_guard_type): New function.
7119 2018-10-05  Paul Koning  <ni1d@arrl.net>
7121         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
7122         * config/pdp11/pdp11.opt (mfloat32): Remove.
7123         (mfloat64): Remove.
7124         * doc/invoke.texi (pdp11 -mfloat32): Remove:
7125         (pdp11 -mfloat64): Remove.
7127 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7129         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
7130         (*cmp<mode>_cc_i387): Ditto.
7131         (*cmpu<mode>_cc_i387): Ditto.
7132         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
7133         * config/i386/i386.c (ix86_expand_fp_compare): Remove
7134         "scratch" argument.
7135         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
7136         Emit x86_sahf_1 pattern.
7137         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
7138         (ix86_expand_carry_flag_compare): Ditto.
7140 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7142         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
7143         to reg_or_0_operand.  Add "C" constraint.
7144         (*cmpxf_cc_i387): Ditto.
7145         (*cmp<mode>_i387): Change operand 2 predicate
7146         to nonimm_or_0_operand.  Add "C" constraint.
7147         (*cmp<mode>_cc_i387): Ditto.
7148         (*cmp<mode>_0_i387): Remove insn pattern.
7149         (*cmp<mode>_0_cc_i387): Ditto.
7151 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
7153         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
7154         * config/i386/predicates.md (nonimm_or_0_operand): Rename
7155         from vector_move_operand.  Update all uses.
7157 2018-10-05  Martin Sebor  <msebor@redhat.com>
7159         PR tree-optimization/87490
7160         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
7161         consistently.
7163 2018-10-05  Richard Biener  <rguenther@suse.de>
7165         PR tree-optimization/63155
7166         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
7167         vertical space in dumpfiles.
7168         * tree-ssa-propagate.h
7169         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
7170         * tree-ssa-propagate.c (cfg_blocks_back): New global.
7171         (ssa_edge_worklist_back): Likewise.
7172         (curr_order): Likewise.
7173         (cfg_blocks_get): Remove abstraction.
7174         (cfg_blocks_add): Likewise.
7175         (cfg_blocks_empty_p): Likewise.
7176         (add_ssa_edge): Add to current or next worklist based on
7177         RPO index.
7178         (add_control_edge): Likewise.
7179         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
7180         into ...
7181         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
7182         iteration from CFG and SSA edge worklist so we process
7183         everything in RPO order, prioritizing forward progress
7184         over iteration.
7185         (ssa_prop_init): Allocate new worklists, do not dump
7186         immediate uses.
7187         (ssa_prop_fini): Free new worklists.
7189 2018-10-05  Richard Biener  <rguenther@suse.de>
7191         * tree-core.h (tree_block::abstract_flag): Remove.
7192         (tree_block::block_num): Make full 32bits.
7193         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
7194         * tree.h (BLOCK_ABSTRACT): Remove.
7195         * dwarf2out.c (gen_lexical_block_die): Remove dead code
7196         resulting from BLOCK_ABSTRACT being always false.
7197         (gen_inlined_subroutine_die): Likewise.
7198         (gen_block_die): Likewise.
7199         * tree.c (block_ultimate_origin): Likewise.
7200         * tree-pretty-print.c (dump_block_node): Remove code dealing
7201         with BLOCK_ABSTRACT.
7202         * tree-ssa-live.c (dump_scope_block): Likewise.
7203         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
7204         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
7206 2018-10-05   Richard Biener  <rguenther@suse.de>
7208         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
7209         is asked for initialize mode to the component mode of the
7210         vector type.
7212 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
7214         PR target/87522
7215         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
7216         assembler for -mavx.
7217         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
7219 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
7221         PR target/87509
7222         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
7223         RS6000_BTM_DFP.
7224         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
7225         to be DImode.  When using mffscrn, force the operand to a register.
7227 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
7229         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
7230         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
7231         X87MODEF mode iterator.
7232         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
7233         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
7234         X87MODEF mode iterator.
7236 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
7238         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
7239         -Wno-prio-ctor-dtor.
7241 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7243         * Makefile.in (OBJS): Add opt-problem.o.
7244         * dump-context.h: Include "selftest.h.
7245         (selftest::temp_dump_context): New forward decl.
7246         (class dump_context): Make friend of class
7247         selftest::temp_dump_context.
7248         (dump_context::dump_loc_immediate): New decl.
7249         (class dump_pretty_printer): Move here from dumpfile.c.
7250         (class temp_dump_context): Move to namespace selftest.
7251         (temp_dump_context::temp_dump_context): Add param
7252         "forcibly_enable_dumping".
7253         (selftest::verify_dumped_text):
7254         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
7255         (selftest::verify_item):
7256         (ASSERT_IS_TEXT): Move here from dumpfile.c.
7257         (ASSERT_IS_TREE): Likewise.
7258         (ASSERT_IS_GIMPLE): Likewise.
7259         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
7260         to...
7261         (dump_context::dump_loc_immediate): ...this new function.
7262         (class dump_pretty_printer): Move to dump-context.h.
7263         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
7264         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
7265         (temp_dump_context::temp_dump_context): Move to "selftest"
7266         namespace.  Add param "forcibly_enable_dumping", and use it to
7267         conditionalize the use of m_pp;
7268         (selftest::verify_dumped_text): Make non-static.
7269         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
7270         (selftest::verify_item): Make non-static.
7271         (ASSERT_IS_TEXT): Move to dump-context.h.
7272         (ASSERT_IS_TREE): Likewise.
7273         (ASSERT_IS_GIMPLE): Likewise.
7274         (selftest::test_capture_of_dump_calls): Pass "true" for new
7275         param of temp_dump_context.
7276         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
7277         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
7278         TDF_COMPARE_DEBUG.
7279         * opt-problem.cc: New file.
7280         * opt-problem.h: New file.
7281         * optinfo-emit-json.cc
7282         (selftest::test_building_json_from_dump_calls): Pass "true" for
7283         new param of temp_dump_context.
7284         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
7285         (optinfo::emit_for_opt_problem): New function.
7286         (optinfo::emit): Clarity which emit_item is used.
7287         * optinfo.h (optinfo::get_dump_location): New accessor.
7288         (optinfo::emit_for_opt_problem): New decl.
7289         (optinfo::emit): Make const.
7290         * selftest-run-tests.c (selftest::run_tests): Call
7291         selftest::opt_problem_cc_tests.
7292         * selftest.h (selftest::opt_problem_cc_tests): New decl.
7293         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
7294         bool to opt_result, converting fprintf messages to
7295         opt_result::failure_at calls.  Add "stmt" param for use by the
7296         failure_at calls.
7297         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
7298         (runtime_alias_check_p): Convert return type from bool to
7299         opt_result, converting dump_printf calls to
7300         opt_result::failure_at, using the statement DDR_A for their
7301         location.
7302         (find_data_references_in_stmt): Convert return type from bool to
7303         opt_result, converting "return false" to opt_result::failure_at
7304         with a new message.
7305         * tree-data-ref.h: Include "opt-problem.h".
7306         (dr_analyze_innermost): Convert return type from bool to opt_result,
7307         and add a const gimple * param.
7308         (find_data_references_in_stmt): Convert return type from bool to
7309         opt_result.
7310         (runtime_alias_check_p): Likewise.
7311         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
7312         dr_analyze_innermost.
7313         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
7314         Convert return type from bool to opt_result, adding a message for
7315         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
7316         (vect_analyze_data_ref_dependence): Convert return type from bool
7317         to opt_result.  Change sense of return type from "false"
7318         effectively meaning "no problems" to "false" meaning a problem,
7319         so that "return false" becomes "return opt_result::success".
7320         Convert "return true" calls to opt_result::failure_at, using
7321         the location of statement A rather than vect_location.
7322         (vect_analyze_data_ref_dependences): Convert return type from bool
7323         to opt_result.
7324         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
7325         calls to opt_result::failure_at, using the stmt location rather
7326         than vect_location.
7327         (vect_verify_datarefs_alignment): Convert return type from bool
7328         to opt_result.
7329         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
7330         into multiple more-tightly-scoped copies.
7331         (vect_analyze_data_refs_alignment): Convert return type from bool
7332         to opt_result.
7333         (vect_analyze_data_ref_accesses): Likewise, converting a
7334         "return false" to a "return opt_result::failure_at", adding a
7335         new message.
7336         (vect_prune_runtime_alias_test_list): Convert return type from
7337         bool to opt_result, converting dump_printf_loc to
7338         opt_result::failure_at.  Add a %G to show the pertinent statement,
7339         and use the stmt's location rather than vect_location.
7340         (vect_find_stmt_data_reference): Convert return type from
7341         bool to opt_result, converting dump_printf_loc to
7342         opt_result::failure_at, using stmt's location.
7343         (vect_analyze_data_refs):  Convert return type from bool to
7344         opt_result.  Convert "return false" to "return
7345         opt_result::failure_at", adding messages as needed.
7346         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
7347         type from bool to opt_result.
7348         (vect_determine_vf_for_stmt): Likewise.
7349         (vect_determine_vectorization_factor): Likewise, converting
7350         dump_printf_loc to opt_result::failure_at, using location of phi
7351         rather than vect_location.
7352         (vect_analyze_loop_form_1): Convert return type from bool to
7353         opt_result, converting dump_printf_loc calls, retaining the use of
7354         vect_location.
7355         (vect_analyze_loop_form): Convert return type from loop_vec_info
7356         to opt_loop_vec_info.
7357         (vect_analyze_loop_operations): Convert return type from bool to
7358         opt_result, converting dump_printf_loc calls, using the location
7359         of phi/stmt rather than vect_location where available.  Convert
7360         various "return false" to "return opt_result::failure_at" with
7361         "unsupported phi" messages.
7362         (vect_get_datarefs_in_loop): Convert return type from bool to
7363         opt_result.  Add a message for the
7364         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
7365         (vect_analyze_loop_2): Convert return type from bool to
7366         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
7367         each "goto again;", adding new messages where needed.
7368         Add "unsupported grouped {store|load}" messages.
7369         (vect_analyze_loop): Convert return type from loop_vec_info to
7370         opt_loop_vec_info.
7371         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
7372         bool to opt_result.
7373         * tree-vect-stmts.c (process_use): Likewise, converting
7374         dump_printf_loc call and using stmt location, rather than
7375         vect_location.
7376         (vect_mark_stmts_to_be_vectorized): Likeise.
7377         (vect_analyze_stmt): Likewise, adding a %G.
7378         (vect_get_vector_types_for_stmt): Convert return type from bool to
7379         opt_result, converting dump_printf_loc calls and using stmt
7380         location, rather than vect_location.
7381         (vect_get_mask_type_for_stmt): Convert return type from tree to
7382         opt_tree, converting dump_printf_loc calls and using stmt location.
7383         * tree-vectorizer.c: Include "opt-problem.h.
7384         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
7385         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
7386         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
7387         enabled, use it to report at the top level "couldn't vectorize
7388         loop" followed by the problem.
7389         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
7390         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
7391         to opt_result.
7392         (vect_analyze_stmt): Likewise.
7393         (vect_get_vector_types_for_stmt): Likewise.
7394         (tree vect_get_mask_type_for_stmt): Likewise.
7395         (vect_analyze_data_ref_dependences): Likewise.
7396         (vect_enhance_data_refs_alignment): Likewise.
7397         (vect_analyze_data_refs_alignment): Likewise.
7398         (vect_verify_datarefs_alignment): Likewise.
7399         (vect_analyze_data_ref_accesses): Likewise.
7400         (vect_prune_runtime_alias_test_list): Likewise.
7401         (vect_find_stmt_data_reference): Likewise.
7402         (vect_analyze_data_refs): Likewise.
7403         (vect_analyze_loop): Convert return type from loop_vec_info to
7404         opt_loop_vec_info.
7405         (vect_analyze_loop_form): Likewise.
7406         (vect_analyze_slp): Convert return type from bool to opt_result.
7408 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7410         * doc/invoke.texi (-fopt-info): Document new "internals"
7411         sub-option.
7412         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
7413         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
7414         MSG_ALL_KINDS.
7415         (optinfo_verbosity_options): Add "internals".
7416         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
7417         (dump_context::apply_dump_filter_p): New member function.
7418         (dump_context::dump_loc): Use apply_dump_filter_p rather than
7419         explicitly masking the dump_kind.
7420         (dump_context::begin_scope): Increment the scope depth first.  Use
7421         apply_dump_filter_p rather than explicitly masking the dump_kind.
7422         (dump_context::emit_item): Use apply_dump_filter_p rather than
7423         explicitly masking the dump_kind.
7424         (dump_dec): Likewise.
7425         (dump_hex): Likewise.
7426         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
7427         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
7428         (opt_info_switch_p): Update handling of default
7429         MSG_OPTIMIZED_LOCATIONS to cope with default of
7430         MSG_PRIORITY_USER_FACING.
7431         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
7432         masking the dump_kind.
7433         (selftest::test_capture_of_dump_calls): Update test_dump_context
7434         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
7435         than MSG_ALL.  Generalize scope test to be run at all four
7436         combinations of with/without MSG_PRIORITY_USER_FACING and
7437         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
7438         for each of the two values.
7439         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
7440         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
7441         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
7442         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
7443         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
7444         with MSG_PRIORITY_*.
7445         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
7446         dump messages as MSG_PRIORITY_USER_FACING.
7447         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
7448         about the interaction with MSG_PRIORITY_*.
7450 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7452         * varasm.c (output_constant): Add new parameter merge_strings.
7453         Make strings properly zero terminated in merge string sections.
7454         (mergeable_string_section): Don't fail if the last char is non-zero.
7455         (assemble_variable_contents): Handle merge string sections.
7456         (assemble_variable): Likewise.
7457         (assemble_constant_contents): Likewise.
7458         (output_constant_def_contents): Likewise.
7459         (output_constructor_array_range,
7460         output_constructor_regular_field): Adjust call to output_constant.
7461         (output_object_block): Adjust call to assemble_constant_contents
7462         and assemble_variable_contents.
7464 2018-10-04  Martin Liska  <mliska@suse.cz>
7466         PR c/87483
7467         * cgraphunit.c (process_function_and_variable_attributes):
7468         Warn about a function with alias attribute and a body.
7470 2018-10-04  Martin Liska  <mliska@suse.cz>
7472         PR ipa/82625
7473         * multiple_target.c (redirect_to_specific_clone): New function.
7474         (ipa_target_clone): Use it.
7475         * tree-inline.c: Fix comment.
7477 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
7479         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
7480         fields.
7481         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
7482         (gcc::dump_manager::register_pass): New member function, adapted
7483         from loop body in gcc::pass_manager::register_pass, adding a
7484         call to update_dfi_for_opt_info.
7485         (gcc::dump_manager::opt_info_enable_passes): Store the
7486         -fopt-info options into the new fields.  Move the loop
7487         bodies into...
7488         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
7489         function.
7490         * dumpfile.h (struct opt_pass): New forward decl.
7491         (gcc::dump_manager::register_pass): New decl.
7492         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
7493         (class gcc::dump_manager): Add fields "m_optgroup_flags",
7494         "m_optinfo_flags", and "m_optinfo_filename".
7495         * passes.c (gcc::pass_manager::register_pass): Move all of the
7496         dump-handling code to gcc::dump_manager::register_pass.
7498 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
7500         PR rtl-optimization/87466
7501         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
7502         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
7503         * doc/tm.texi: Regenerate.
7504         * ira-lives.c (process_bb_node_lives): Use the new target hook.
7505         * lra-lives.c (process_bb_lives): Likewise.
7506         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
7507         Define.
7509 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
7511         * params.c (add_params): Fix initialization.
7513 2018-10-04  Martin Liska  <mliska@suse.cz>
7515         PR gcov-profile/84107
7516         * tree-profile.c (init_ic_make_global_vars):
7517         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
7518         Come up with new ic_tuple* variables.  Emit
7519         __gcov_indirect_call{,_topn} variables.
7520         (gimple_gen_ic_profiler): Access the variable
7521         and emit gimple.
7522         (gimple_gen_ic_func_profiler): Access
7523         __gcov_indirect_call.callee field.
7524         (gimple_init_gcov_profiler): Use ptr_type_node.
7525         * value-prof.c (gimple_ic): Use ptr_type_node.
7527 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
7529         PR tree-optimization/85787
7530         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
7531         malloc_candidate_p into this function and add support for
7532         detecting multiple phis.
7533         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
7535 2018-10-04  Martin Liska  <mliska@suse.cz>
7537         PR ipa/87491
7538         * ipa-inline.c (inline_to_all_callers_1):
7539         Call ultimate_alias_target for node being inlined.
7541 2018-10-03  Jeff Law  <law@redhat.com>
7543         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
7544         target's wchar_t.
7545         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
7546         * tree.h (get_typenode_from_name): Prototype.
7548 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
7550         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
7551         Change operand 2 predicate to nonimmediate_operand.
7552         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
7554 2018-10-03  Martin Sebor  <msebor@redhat.com>
7555             Jeff Law  <law@redhat.com>
7557         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
7558         initialize it.
7559         (get_string_length): Detect unterminated arrays.
7560         (format_string): Same.
7561         (format_directive): Warn about unterminated arrays.
7562         (handle_gimple_call): Mark statements with no_warning as needed.
7564 2018-10-03  Jim Wilson  <jimw@sifive.com>
7566         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
7567         also define __riscv_abi_rve.  Delete trailing white space.
7569 2018-10-03  Paul Koning  <ni1d@arrl.net>
7571         Enable LRA register allocator for PDP11.
7572         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
7573         (R): Likewise.
7574         (D): Likewise.
7575         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
7576         * config/pdp11/pdp11.opt (-mlra): New option.
7577         * doc/invoke.texi (PDP-11 Options): Document -mlra.
7579 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
7581         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
7582         (*<absneg:code>extend<mode>xf2): Ditto.
7584 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
7586         PR tree-optimization/87415
7587         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
7588         precision fields.
7590 2018-10-02  Jeff Law  <law@redhat.com>
7592         * gimple-fold.c (get_range_strlen): Only set *nonstr when
7593         an unterminated string is discovered.  Bubble up range
7594         even for unterminated strings.
7595         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
7596         indicates the string was not terminated via NONSTR.
7598 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
7600         * tree-vrp.c (extract_range_from_unary_expr): Special case all
7601         pointer conversions.
7602         Do not do anything special for anti-ranges.
7604 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
7606         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
7607         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
7609 2018-10-03  Martin Liska  <mliska@suse.cz>
7611         PR gcov-profile/86109
7612         * coverage.c (coverage_begin_function): Do not
7613         mark lambdas as artificial.
7614         * tree-core.h (struct GTY): Remove tm_clone_flag
7615         and introduce new lambda_function.
7616         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
7618 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7620         PR target/87474
7621         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
7622         P8_VECTOR and VSX are enabled.
7624 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
7626         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
7627         0x3907 as CPU model number.
7629 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
7631         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
7632         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
7633         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
7634         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
7635         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
7636         * config/s390/s390.md: Likewise. Rename also the cpu attribute
7637         value from arch12 to z14.
7639 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
7641         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
7642         (isinfxf2): Ditto.
7643         (isinf<mode>2): Ditto.
7645 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
7647         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
7648         before emitting fxam.  Perform calculations in XFmode.
7650 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
7652         * match.pd (((X /[ex] A) +- B) * A): New transformation.
7654 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
7656         PR middle-end/87319
7657         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
7658         * tree.c (signed_or_unsigned_type_for): Handle complex.
7660 2018-10-02  Jeff Law  <law@redhat.com>
7662         * gimple-fold.c (get_range_strlen): Remove dead code.
7664 2018-10-02  Martin Sebor  <msebor@redhat.com>
7665             Jeff Law  <law@redhat.com>
7667         * builtins.c (unterminated_array): Add new arguments.
7668         If argument is not terminated, bubble up size and exact
7669         state to callers.
7670         (expand_builtin_strnlen): Detect, avoid expanding
7671         and diagnose unterminated arrays.
7672         (c_strlen): Fill in offset of start of unterminated strings.
7673         * builtins.h (unterminated_array): Update prototype.
7675 2018-10-02  Richard Biener  <rguenther@suse.de>
7677         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
7678         of haddv4df, first reduce to SSE width and exploit the fact
7679         that we only need element zero with the reduction result.
7680         (reduc_plus_scal_v2df): Likewise.
7682 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
7684         * dojump.h (do_jump): Delete.
7685         (do_jump_1): Likewise.
7686         (split_comparison): Move around.
7687         * dojump.c (do_jump): Make static.
7688         (do_jump_1): Likewise.
7689         (jumpifnot): Move around.
7690         (jumpifnot_1): Likewise.
7691         (jumpif): Likewise.
7692         (jumpif_1): Likewise.
7693         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
7695 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
7697         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
7698         insns in the delay slot and add_insn_after for the jump insn.
7700 2018-10-02  Richard Biener  <rguenther@suse.de>
7702         * tree-inline.c (expand_call_inline): Use the location of
7703         the callee declaration for the inline-entry marker.
7704         * final.c (notice_source_line): Remove special-casing of
7705         NOTE_INSN_INLINE_ENTRY.
7707 2018-10-01  Carl Love  <cel@us.ibm.com>
7709         PR 69431
7710         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
7711         (__builtin_mtfsb0): New.
7712         (__builtin_mtfsb1): New.
7713         ( __builtin_set_fpscr_rn): New.
7714         (__builtin_set_fpscr_drn): New.
7715         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
7716         (rs6000_expand_set_fpscr_rn_builtin): Add.
7717         (rs6000_expand_set_fpscr_drn_builtin): Add.
7718         (rs6000_expand_builtin): Add case statement entries for
7719         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
7720         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
7721         RS6000_BUILTIN_MFFSL.
7722         (rs6000_init_builtins): Add ftype initialization and def_builtin
7723         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
7724         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
7725         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
7726         rs6000_mffscdrn): Add define_insn.
7727         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
7728         * doc/extend.texi: Add documentation for the builtins.
7730 2018-10-01  Richard Biener  <rguenther@suse.de>
7732         PR tree-optimization/87465
7733         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
7734         causing branch miscounts.
7736 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7738         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
7739         aarch64_option_default_param):  New.
7740         (params.h): Include.
7741         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
7742         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
7743         stack-clash protection validation code.
7745 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7747         * params.c (validate_param): New.
7748         (add_params): Use it.
7749         (set_param_value): Refactor param validation into validate_param.
7750         (diagnostic.h): Include.
7751         * diagnostic.h (diagnostic_ready_p): New.
7753 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7755         * params.c (set_param_value):
7756         Add index of parameter being validated.
7757         * common/common-target.def (option_validate_param): New.
7758         * common/common-targhooks.h (default_option_validate_param): New.
7759         * common/common-targhooks.c (default_option_validate_param): New.
7760         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
7761         * doc/tm.texi: Regenerate.
7763 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7765         PR target/86486
7766         * config/aarch64/aarch64.c (aarch64_override_options_internal):
7767         Add validation for stack-clash parameters and set defaults.
7769 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7771         PR target/86486
7772         * configure.ac: Add stack-clash-protection-guard-size.
7773         * doc/install.texi: Document it.
7774         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
7775         * params.def: Update comment for guard-size.
7776         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
7777         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
7778         * configure: Regenerate.
7780 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7782         PR target/86486
7783         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
7784         STACK_DYNAMIC_OFFSET): New.
7785         * config/aarch64/aarch64.c (aarch64_layout_frame):
7786         Update outgoing args size.
7787         (aarch64_stack_clash_protection_alloca_probe_range,
7788         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
7790 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7792         PR target/86486
7793         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
7794         probe ranges.
7795         * target.def (stack_clash_protection_alloca_probe_range): New.
7796         (stack_clash_protection_final_dynamic_probe): Remove.
7797         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
7798         (default_stack_clash_protection_final_dynamic_probe): Remove.
7799         * targhooks.c: Likewise.
7800         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
7801         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
7802         * doc/tm.texi: Regenerate.
7804 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7806         PR target/86486
7807         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
7808         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
7809         aarch64_clamp_to_uimm12_shift): New.
7810         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
7811         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
7813 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
7815         PR target/86486
7816         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
7818 2018-10-01  Jeff Law  <law@redhat.com>
7819             Richard Sandiford <richard.sandiford@linaro.org>
7820             Tamar Christina  <tamar.christina@arm.com>
7822         PR target/86486
7823         * config/aarch64/aarch64.md
7824         (probe_stack_range): Add k (SP) constraint.
7825         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
7826         STACK_CLASH_MAX_UNROLL_PAGES): New.
7827         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
7828         stack probes for stack clash.
7829         (aarch64_allocate_and_probe_stack_space): New.
7830         (aarch64_expand_prologue): Use it.
7831         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
7832         (aarch64_sub_sp): Add emit_move_imm optional param.
7834 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
7836         PR tree-optimization/87261
7837         * match.pd: Remove trailing whitespace.
7838         Add (x & y) | ~(x | y) -> ~(x ^ y),
7839         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
7841 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
7843         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
7844         constraints.
7845         (ashlsi3_insn): Update instruction constraints.
7846         (ashrsi3_insn): Likewise.
7847         (rotrsi3): Likewise.
7848         (add_shift): Likewise.
7849         * config/arc/constraints.md (Csz): New 32 bit constraint. It
7850         avoids placing in the limm field small constants which, otherwise,
7851         could end into a small instruction.
7853 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
7855         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
7856         destination register is not odd-even.
7857         (umaddsidi4_split): Likewise.
7859 2018-10-01  Richard Biener  <rguenther@suse.de>
7861         * tree-inline.c (expand_call_inline): Store origin of fn
7862         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
7863         * tree.c (block_ultimate_origin): Simplify and do some
7864         checking.
7866 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
7868         * config/i386/mmx.md (EMMS): New int iterator.
7869         (emms): New int attribute.
7870         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
7871         EMMS int iterator.  Explicitly declare clobbers.
7872         (mmx_emms): Remove expander.
7873         (mmx_femms): Ditto.
7874         * config/i386/predicates.md (emms_operation): Remove predicate.
7875         (vzeroall_pattern): New predicate.
7876         (vzeroupper_pattern): Rename from vzeroupper_operation.
7877         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
7878         vzeroupper_pattern and vzeroall_pattern predicates.
7880 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
7882         PR rtl-optimization/86939
7883         * ira-lives.c (make_hard_regno_born): Rename from this...
7884         (make_hard_regno_live): ... to this.  Remove update to conflict
7885         information.  Update function comment.
7886         (make_hard_regno_dead): Add conflict information update.  Update
7887         function comment.
7888         (make_object_born): Rename from this...
7889         (make_object_live): ... to this.  Remove update to conflict information.
7890         Update function comment.
7891         (make_object_dead):  Add conflict information update.  Update function
7892         comment.
7893         (mark_pseudo_regno_live): Call make_object_live.
7894         (mark_pseudo_regno_subword_live): Likewise.
7895         (mark_hard_reg_dead): Update function comment.
7896         (mark_hard_reg_live): Call make_hard_regno_live.
7897         (process_bb_node_lives): Likewise.
7898         * lra-lives.c (make_hard_regno_born): Rename from this...
7899         (make_hard_regno_live): ... to this.  Remove update to conflict
7900         information.  Remove now uneeded check_pic_pseudo_p argument.
7901         Update function comment.
7902         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
7903         to conflict information.  Update function comment.
7904         (mark_pseudo_live): Remove update to conflict information.  Update
7905         function comment.
7906         (mark_pseudo_dead): Add conflict information update.
7907         (mark_regno_live): Call make_hard_regno_live.
7908         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
7909         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
7911 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
7913         PR target/87370
7914         * config/i386/i386.c (construct_container): Use TImode for
7915         BLKmode values in 2 integer registers.
7917 2018-09-29  Jeff Law  <law@redhat.com>
7919         * builtins.c (unterminated_array): Pass in c_strlen_data * to
7920         c_strlen rather than just a tree *.
7921         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
7922         Update recursive calls appropriately.  If caller did not provide a
7923         suitable data pointer, create a local one.  When a non-terminated
7924         string is discovered, bubble up information about the string via the
7925         c_strlen_data object.
7926         * builtins.h (c_strlen): Update prototype.
7927         (c_strlen_data): New structure.
7928         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
7929         For a type 2 call, if c_strlen indicates a non-terminated string
7930         use the length of the non-terminated string.
7931         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
7933 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
7935         PR target/87467
7936         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
7937         __m512d type for __A argument rather than __m512.
7939 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
7941         * match.pd (simple_comparison): Don't optimize if either operand is
7942         a function pointer when target needs function pointer canonicalization.
7944 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
7946         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
7947         power5 .. power9 to remove indirection.
7948         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
7949         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
7950         ASM_CPU_476_SPEC): Delete.
7951         (ASM_CPU_SPEC): Adjust.
7952         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
7953         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
7955 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
7957         * config.in: Delete HAVE_AS_DCI.
7958         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
7959         * config/rs6000/rs6000.h: Ditto.
7960         * configure.ac: Delete HAVE_AS_DCI.
7961         * configure: Regenerate.
7963 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
7965         * config.in (HAVE_AS_LWSYNC): Delete.
7966         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
7967         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
7968         do it as a .long .
7969         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
7970         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
7971         as a .long .
7972         * configure.ac: Delete HAVE_AS_LWSYNC.
7973         * configure: Regenerate.
7975 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
7976             Pierre-Marie de Rodat  <derodat@adacore.com>
7978         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
7979         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
7980         (cgraph_node::create_thunk): Add indirect_offset parameter.
7981         (thunk_adjust): Likewise.
7982         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
7983         and initialize the corresponding field with it.
7984         (cgraph_node::dump): Dump indirect_offset field.
7985         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
7986         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
7987         (thunk_adjust): Add indirect_offset parameter and deal with it.
7988         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
7989         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
7990         or if the thunk is external or local.  Fix formatting.  Do not chain
7991         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
7992         if any, in the GIMPLE representation.
7993         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
7994         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
7995         (input_node): Read indirect_offset field.
7996         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
7997         call to thunk_adjust.
7998         * tree-nested.c (struct nesting_info): Add thunk_p field.
7999         (create_nesting_tree): Set it.
8000         (convert_all_function_calls): Copy static chain from targets to thunks.
8001         (finalize_nesting_tree_1): Return early for thunks.
8002         (unnest_nesting_tree_1): Do not finalize thunks.
8003         (gimplify_all_functions): Do not gimplify thunks.
8005 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
8007         * opt-suggestions.c (option_proposer::build_option_suggestions):
8008         Release "option_values".
8010 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
8012         * coverage.c (get_coverage_counts): Convert problem-reporting dump
8013         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
8014         * dumpfile.c (kind_as_string): New function.
8015         (dump_loc): Rather than a hardcoded prefix of "note: ", use
8016         kind_as_string to vary the prefix based on dump_kind.
8017         (selftest::test_capture_of_dump_calls): Update for above.
8019 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
8021         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
8022         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
8024 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
8026         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
8027         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
8028         INVALID_REGNUM instead of FPSR_REG.
8029         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
8030         * config/i386/i386.md: Update comment of FP compares.
8031         (fldenv): Do not clobber FPSR_REG.
8033 2018-09-28  Richard Biener  <rguenther@suse.de>
8035         * tree.h (BLOCK_ORIGIN): New.
8036         * omp-expand.c (grid_expand_target_grid_body): Assign
8037         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
8038         * tree-inline.c (remap_block): Likewise.
8039         * auto-profile.c (get_function_decl_from_block): Simplify
8040         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
8041         * langhooks.c (lhd_print_error_function): Likewise.
8042         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
8043         Likewise.
8044         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8045         * tree.c (block_nonartificial_location): Likewise.
8046         (block_ultimate_origin): Likewise.
8047         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
8048         no longer needed LTO case.
8050 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
8051             Jan Hubicka  <jh@suse.cz>
8052             Martin Jambor  <mjambor@suse.cz>
8054         * simplify-rtx.c (simplify_merge_mask): New function.
8055         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
8056         same masks are used in op1 or op2.
8057         (test_vec_merge): New function.
8058         (test_vector_ops): Call test_vec_merge.
8060 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
8062         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
8063         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
8064         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
8065         (cypress_costs): Set it.
8066         (supersparc_costs): Likewise.
8067         (hypersparc_costs): Likewise.
8068         (leon_cost): Likewise.
8069         (leon3_costs): Likewise.
8070         (sparclet_costs): Likewise.
8071         (ultrasparc_costs): Likewise.
8072         (ultrasparc_costs): Likewise.
8073         (niagara_costs): Likewise.
8074         (niagara2_costs): Likewise.
8075         (niagara3_costs): Likewise.
8076         (niagara4_costs): Likewise.
8077         (niagara7_costs): Likewise.
8078         (m8_costs): Likewise.
8079         (TARGET_CAN_FOLLOW_JUMP): Define.
8080         (pass_work_around_errata::gate): Minor tweak.
8081         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
8082         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
8083         Automaitcally clear MASK_FSMULD mask for V7 processors.
8084         (sparc_can_follow_jump): New static function.
8085         (output_ubranch): Deal with CROSSING_JUMP_P.
8086         (sparc_use_sched_lookahead): Rewrite using switch statement.
8087         (sparc_issue_rate): Reorder.
8088         (sparc_branch_cost): New function.
8090 2018-09-27  Martin Sebor  <msebor@redhat.com>
8092         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
8093         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
8094         (int_fits_type_p): Same.
8096 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
8098         * config/i386/i386.md (FPCR_REG): Remove.
8099         (UNSPEC_FLDCW): Remove.
8100         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
8101         (x86_fldcw_1): Remove insn pattern.
8102         (fnstenv): Do not clobber FPCR_REG.
8103         (fldenv): Ditto.
8104         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
8105         (CALL_USED_REGISTERS): Ditto.
8106         (REG_ALLOC_ORDER): Ditto.
8107         (REG_CLASS_CONTENTS): Ditto.
8108         (HI_REGISTER_NAMES): Ditto.
8109         (ADDITIONAL_REGISTER_NAMES): Use defines instead
8110         of numerical constants.
8111         * config/i386/i386.c (regclass_map): Remove fpsr register.
8112         (dbx_register_map): Ditto.
8113         (dbx64_register_map): Ditto.
8114         (svr4_dbx_register_map): Ditto.
8115         (print_reg): Do not handle FPCR_REG.
8117 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
8119         PR target/87149
8120         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
8121         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
8122         Delete, always treat as true.
8123         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8124         Ditto.  Simplify remaining code.
8125         * config/powerpcspe/powerpcspe.h: Ditto.
8126         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
8127         Simplify remaining code.
8128         (rs6000_expand_builtin): Ditto.
8129         * config/rs6000/rs6000.h: Ditto.
8130         * configure.ac: Ditto.
8131         * configure: Regenerate.
8133 2018-09-27  Martin Liska  <mliska@suse.cz>
8135         * coverage.c (get_coverage_counts): Revert the formatting
8136         of missing profile opt info.
8138 2018-09-27  Richard Biener  <rguenther@suse.de>
8140         PR debug/37801
8141         PR debug/87440
8142         * dwarf2out.c (set_block_origin_self): Do not mark outermost
8143         block as we do not output that.
8144         (gen_inlined_subroutine_die): Elide the originally outermost
8145         block, matching what we do for concrete instances.
8146         (decls_for_scope): Add parameter specifying whether to recurse
8147         to subblocks.
8149 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
8150             Tom de Vries  <tom@codesourcery.com>
8152         PR 82089
8154         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
8155         STORE_FLAG_VALUE == 1.
8157 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
8159         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
8160         constant definitions.
8161         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
8162         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
8163         ("speculation_barrier"): New expander definition.
8165 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
8167         PR gcov-profile/86957
8168         * common.opt: New warning option -Wmissing-profile.
8169         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
8170         * doc/invoke.texi: Document -Wmissing-profile.
8172 2018-09-26  Jim Wilson  <jimw@sifive.com>
8174         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
8175         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
8177 2018-09-26  Martin Sebor  <msebor@redhat.com>
8179         * tree.c (zerop): Change return type to bool.
8180         (integer_zerop, integer_onep, integer_each_onep): Same.
8181         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
8182         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
8183         (real_onep, real_minus_onep, chain_index): Same.
8184         (print_type_hash_statistics, type_list_equal): Same.
8185         * tree.h (zerop): Same.
8186         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
8187         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
8188         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
8189         (real_onep, real_minus_onep, chain_index): Same.
8190         (print_type_hash_statistics, type_list_equal): Same.
8192 2018-09-26  Jim Wilson  <jimw@sifive.com>
8194         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
8196 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
8198         PR target/87414
8199         * config/i386/i386.c: Include debug.h and dwarf2out.h.
8200         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
8201         call.
8203 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
8205         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
8207 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
8209         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
8210         and FP_SECOND_SSE_REGS.
8211         (REG_CLASS_NAMES): Ditto.
8212         (REG_CLASS_CONTENTS): Ditto.
8213         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
8214         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
8215         (ix86_preferred_output_reload_class): Ditto.
8216         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
8217         clobber constraint to "=&f".
8218         (fix_truncdi_i387): Ditto.
8219         (lrintxfdi2): Ditto.
8220         (fistdi2_<rounding>): Ditto.
8221         (fpremxf4_i387): Change "=u" constraint to "=f".
8222         (fprem1xf4_i387): Ditto.
8223         (sincosxf3): Ditto.
8224         (fptanxf4_i387): Ditto.
8225         (fxtractxf3_i387): Ditto.
8226         (fscalexf4_i387): Ditto.
8227         (atan2xf3): Change "u" constraint to "f".
8228         (fyl2xxf3_i387): Ditto.
8229         (fyl2xp1xf3_i387): Ditto.
8231 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
8233         PR target/87439
8234         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
8235         for removed I387_MASK_PM entity.
8238 2018-09-26  Jeff Law  <law@redhat.com>
8239         Revert
8240         2018-09-26  Alexey Neyman  <stilor@att.net>
8242         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
8243         headers are no longer pulled in by <isl/val.h>.
8245 2018-09-26  Richard Biener  <rguenther@suse.de>
8247         PR debug/87443
8248         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
8249         or concrete instance DIE to the tree.  Create abstract origin
8250         attributes also for concrete instances.
8252 2018-09-26  Alexey Neyman  <stilor@att.net>
8254         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
8255         headers are no longer pulled in by <isl/val.h>.
8257 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
8259         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
8260         Use new helper functions.
8261         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
8262         Use new helper functions.
8263         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
8264         aarch_mm_needs_release): New declarations.
8265         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
8266         aarch_mm_needs_release): New.
8268 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
8270         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
8271         (arm32_output_mi_thunk): Deal with long calls.
8273 2018-09-26  Richard Biener  <rguenther@suse.de>
8275         PR debug/87428
8276         PR debug/87362
8277         * tree-inline.c (expand_call_inline): When the location
8278         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
8279         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
8280         the inserted BLOCK to make inlined_function_outer_scope_p
8281         recognize it.
8282         * dwarf2out.c (add_call_src_coords_attributes): Do not add
8283         coords for reserved locations.
8285 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8287         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
8288         (*call_indirect_nonlocal_sysv<mode>): Ditto.
8289         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
8290         (*sibcall_nonlocal_sysv<mode>): Ditto.
8291         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
8292         (<bd>_<mode>): Ditto.
8293         (<bd>tf_<mode>): Ditto.
8295 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8297         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
8298         control string as a list of templates instead of as C code.
8299         (*altivec_movti): Ditto.
8300         * config/rs6000/darwin.md (movdf_low_di): Ditto.
8302 2018-09-25  Jim Wilson  <jimw@sifive.com>
8304         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
8305         when target symbol is weak.
8307 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8309         PR c/87387
8310         * builtins.c (unterminated_array): Simplify.
8311         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
8312         where pointer arithmetic is safe.
8314 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
8316         PR target/86987
8317         * config/rs6000/altivec.md (altivec_vspltb): Use
8318         const_0_to_15_operand instead of u5bit_cint_operand.
8319         (*altivec_vspltb_internal): Ditto.
8320         (altivec_vspltb_direct): Ditto.
8321         (altivec_vsplth): Use const_0_to_7_operand instead of
8322         u5bit_cint_operand.
8323         (*altivec_vsplth_internal): Ditto.
8324         (altivec_vsplth_direct): Ditto.
8325         (altivec_vspltw): Use const_0_to_3_operand instead of
8326         u5bit_cint_operand.
8327         (*altivec_vspltw_internal): Ditto.
8328         (altivec_vspltw_direct): Ditto.
8329         (altivec_vspltsf): Ditto.
8330         (*altivec_vspltsf_internal): Ditto.
8331         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
8332         various splats with the proper size immediate.  Reorder the various
8333         cases by ascending size of immediate, and put all such together.
8335 2018-09-25  Richard Biener  <rguenther@suse.de>
8337         PR debug/83941
8338         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
8339         GC-ification.
8340         (maybe_create_die_with_external_ref): Do not create
8341         DW_TAG_imported_unit here.
8342         (add_abstract_origin_attribute): Handle external BLOCK refs.
8343         (dwarf2out_abstract_function): Simplify LTO case.
8344         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
8345         rather than using maybe_create_die_with_external_ref.
8347 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
8349         PR target/71278
8350         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
8351         (frndintxf2_mask_pm_i387): Ditto.
8352         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
8353         Enable for !flag_trapping_math.
8354         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
8355         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
8356         Change operand 1 predicate to nonimmediate_operand.
8357         (attr "i387_cw"): Remove mask_pm.
8358         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
8359         (enum ix86_entity): Remove I387_MASK_PM.
8360         * config/i386/i386.c (ix86_i387_mode_needed): Do not
8361         handle I387_MASK_PM.
8362         (ix86_mode_needed): Ditto.
8363         (ix86_mode_after): Ditto.
8364         (ix86_mode_entry): Ditto.
8365         (ix86_mode_exit): Ditto.
8366         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
8368 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
8370         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
8371         to_update_switch_stmts to vNULL instead of calling create on them
8372         immediately.
8374 2018-09-25  Richard Biener  <rguenther@suse.de>
8376         PR tree-optimization/87402
8377         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
8378         (visit_phi): Re-instantiate handling of supposed to be VARYING
8379         but non-VARYING backedge value.
8381 2018-09-25  Richard Biener  <rguenther@suse.de>
8383         PR debug/83941
8384         * dwarf2out.c (struct sym_off_pair): New.
8385         (external_die_map): New global.
8386         (lookup_decl_die): When in LTO create DIEs lazily from the
8387         external_die_map.
8388         (lookup_block_die): New function, create DIEs lazily in LTO.
8389         (equate_block_to_die): New function.
8390         (dwarf2out_die_ref_for_decl): During WPA get the association
8391         from the external DIE map.
8392         (dwarf2out_register_external_die): Record mapping into the
8393         external DIE map.
8394         (maybe_create_die_with_external_ref): New function split out from
8395         DIE generation part of old dwarf2out_register_external_die.
8396         (add_abstract_origin_attribute): Do not return the DIE.  When
8397         in LTO reference externals directly.
8398         (dwarf2out_abstract_function): When in LTO ignore calls for
8399         decls with external DIEs (already present abstract instances).
8400         (gen_call_site_die): Adjust.
8401         (add_high_low_attributes): Likewise.
8402         (gen_lexical_block_die): Likewise.
8403         (gen_inlined_subroutine_die): Likewie.
8404         (gen_block_die): Likewise.
8405         (dwarf2out_inline_entry): Likewise.
8406         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
8407         DIEs.
8409 2018-09-25  Martin Liska  <mliska@suse.cz>
8411         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
8412         integers and not by a float value.
8414 2018-09-25  Martin Liska  <mliska@suse.cz>
8416         PR fortran/87394
8417         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
8418         instead of NULL.
8419         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
8420         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
8421         * godump.c (go_format_type): Remove extra parenthesis.
8423 2018-09-25  Martin Liska  <mliska@suse.cz>
8425         * alias.c (set_dest_equal_p): Remove unused function.
8426         * config/i386/i386.c (def_builtin_pure2): Likewise.
8427         * diagnostic-show-locus.c (class layout): Remove
8428         unused field.
8429         (layout::layout): Likewise here.
8430         * dump-context.h (class temp_dump_context): Likewise.
8431         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
8432         (add_AT_loclistsptr): Likewise.
8433         (add_AT_offset): Likewise.
8434         (get_AT_hi_pc): Likewise.
8435         (is_comdat_die): Likewise.
8436         (type_is_enum): Likewise.
8437         (ceiling): Likewise.
8438         (add_AT_vms_delta): Likewise.
8439         (is_class_die): Likewise.
8440         * edit-context.c (class line_event): Remove unused field.
8441         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
8442         unused function.
8443         * ipa-cp.c (ipa_get_vr_lat): Likewise.
8444         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
8445         (ok_for_base_p_nonstrict): Likewise.
8446         * tree-chrec.c (is_not_constant_evolution): Likewise.
8447         (chrec_fold_poly_cst): Likewise.
8448         * tree-if-conv.c (has_pred_critical_p): Likewise.
8449         * tree-ssa-coalesce.c (print_exprs): Likewise.
8450         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
8451         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
8452         * tree-vrp.c (value_ranges_intersect_p): Likewise.
8453         (value_range_nonnegative_p): Likewise.
8455 2018-09-25  Martin Liska  <mliska@suse.cz>
8457         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
8458         Do not handle "GNU Pascal".
8459         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
8460         Likewise.
8461         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
8462         from documentation. Likewise.
8463         * dbxout.c (dbxout_range_type): Likewise.
8464         * doc/cpp.texi: Likewise.
8465         * doc/extend.texi: Likewise.
8466         * doc/frontends.texi: Likewise.
8467         * doc/invoke.texi: Remove Pascal entry.
8468         * tree.def (CLEANUP_POINT_EXPR): Likewise.
8469         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
8471 2018-09-25  Martin Liska  <mliska@suse.cz>
8473         PR middle-end/86078
8474         * doc/invoke.texi: Document all parameters and remove default
8475         of the parameters.
8477 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
8479         PR bootstrap/87417
8480         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
8481         contains HOST_WIDE_INTs when computing its size.
8483 2018-09-24  Jim Wilson  <jimw@sifive.com>
8485         PR target/87391
8486         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
8487         not TARGET_RVE.
8488         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
8490 2018-09-24  Andrew Pinski  <apinski@marvell.com>
8492         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
8493         access prev before checking it for NULLness in the
8494         AARCH64_FUSE_CMP_BRANCH case.
8496 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
8498         PR target/82699
8499         * config/i386/i386.c (rest_of_insert_endbranch): Set
8500         endbr_queued_at_entrance to true and don't insert ENDBR if
8501         x86_function_profiler will be called.
8502         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
8503         is true.
8504         * config/i386/i386.h (machine_function): Add
8505         endbr_queued_at_entrance.
8507 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
8509         * genattrtab.c (mk_attr_alt): Use alternative_mask.
8510         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
8511         types.
8512         (check_attr_test): Use alternative_mask.
8513         (get_attr_value): Likewise.
8514         (compute_alternative_mask): Use alternative_mask and XWINT.
8515         (make_alternative_compare): Use alternative_mask.
8516         (attr_alt_subset_p): Use XWINT.
8517         (attr_alt_subset_of_compl_p): Likewise.
8518         (attr_alt_intersection): Use alternative_mask and XWINT.
8519         (attr_alt_union): Likewise.
8520         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
8521         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
8522         (simplify_test_exp): Use alternative_mask and XWINT.
8523         (write_test_expr): Use alternative_mask and XWINT, adjust bit
8524         number calculation to support 64 bits.  Generate code that
8525         checks 64-bit masks.
8526         (main): Use alternative_mask.
8527         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
8529 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
8531         PR target/80080
8532         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
8533         RETURN+USE when returning via %r14.
8535 2018-09-24  Martin Liska  <mliska@suse.cz>
8537         * gcov.c (output_lines): Print colorization legend
8538         for both flag_use_colors and flag_use_hotness_colors.
8539         Reword the help.
8541 2018-09-24  Martin Liska  <mliska@suse.cz>
8543         * coverage.c (get_coverage_counts): Use warning_at
8544         with current_function_decl location. Use %qD in warning
8545         message.
8547 2018-09-24  Martin Liska  <mliska@suse.cz>
8549         * memory-block.h (memory_block_pool::release): Annotate with
8550         valgrind that the memory is not accessible.
8552 2018-09-24  Martin Liska  <mliska@suse.cz>
8554         PR sanitizer/85774
8555         * asan.c: Make asan_handled_variables extern.
8556         * asan.h: Likewise.
8557         * cfgexpand.c (expand_stack_vars): Make sure
8558         a representative is unpoison if another
8559         variable in the partition is handled by
8560         use-after-scope sanitization.
8562 2018-09-24  Richard Biener  <rguenther@suse.de>
8564         PR tree-optimization/63155
8565         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
8566         the worklist when the edge of the respective argument isn't
8567         executable.
8569 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
8571         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
8572         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
8573         (MASK_CLASS_P): Update for rename.
8574         (MAYBE_MASK_CLASS_P): Ditto.
8575         (REG_CLASS_NAMES): Update.
8576         (REG_CLASS_CONTENT): Update.
8577         * config/i386/i386.c (regclass_map): Update for MASK_REG
8578         and ALL_MASK_REGS rename.
8579         * config/i386/constraints.md (Yk): Update for rename.
8580         (k): Ditto.
8582 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
8584         * config/i386/i386.h (enum reg_class): Remove
8585         EVEX_SSE_REGS and MOD4_SSE_REGS.
8586         (REG_CLASS_NAMES): Update.
8587         (REG_CLASS_CONTENT): Update.
8588         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
8589         registers as ALL_SSE_REGS.
8590         (ix86_additional_allocno_class_p): Remove.
8591         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
8592         (ix86_register_priority): Lower priority of EVEX SSE registers.
8593         Use IN_RANGE macro where appropriate.
8594         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
8595         AVX-5124VNNIW checks.
8596         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
8597         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
8598         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
8599         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
8600         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
8601         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
8602         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
8603         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
8604         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
8605         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
8606         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
8607         * config/i386/constraints.md (Yh): Remove.
8609 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
8611         * config/i386/i386.c (regclass_map): Declare integer REX registers
8612         as GENERAL_REGS.
8614 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
8616         * doc/service.texi (Service): Switch the fsf.org link to https.
8618 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
8620         PR target/86798
8621         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8622         Define to speculation_safe_value_not_needed.
8624 2018-09-21  Florian Weimer  <fweimer@redhat.com>
8626         PR middle-end/81035
8627         * doc/extend.texi (Common Function Attributes): Mention that
8628         noreturn suppresses tail call optimization.
8630 2018-09-21  Jeff Law  <law@redhat.com>
8632         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
8633         vr_values::cleanup_edges_and_switches.
8634         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
8635         vr_values class.
8636         (identify_jump_threads): Remove EDGE_IGNORE handling.
8637         (execute_vrp): Move handling of to_remove_edges and
8638         to_update_switch_stmts into vr_values class member functions.
8639         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
8640         (to_update_switch_stmts): Likewise.
8641         * vr-values.c: Include cfghooks.h.
8642         (vr_values::vr_values): Initialize to_remove_edges and
8643         to_update_switch_stmts.
8644         (vr_values::~vr_values): Verify to_remove_edges and
8645         to_update_switch_stmts are empty.
8646         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
8647         (vr_values::cleanup_edges_and_switches): New member function.
8648         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
8649         function.  Add new data members.
8651 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
8653         PR tree-optimization/87309
8654         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
8655         calls with pflags and alt_flags.
8656         (selftest::test_capture_of_dump_calls): Add test of interaction of
8657         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
8659 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8661         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
8663 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8665         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
8667 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8669         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
8670         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
8671         (VXWORKS_LIBS_RTP): Minor reordering.
8673 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8675         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
8676         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
8678 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8680         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
8681         (PTRDIFF_TYPE): Likewise.
8683 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8685         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
8686         triplet, similar to support for VxWorks7.
8687         * config/vxworks-dummy.h: Provide a default definition
8688         of TARGET_VXWORKS64 to 0.
8690 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8692         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
8693         * config/vxworks-dummy.h: here.
8695 2018-09-21  Olivier Hainque  <hainque@adacore.com>
8697         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
8699 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
8700             Bo Zhou  <zbo.zhou@hisilicon.com>
8702         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
8703         * config/aarch64/aarch64-tune.md: Regenerated.
8704         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
8705         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
8706         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
8708 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
8709             Julian Brown  <julian@codesourcery.com>
8711         * builtins.c (get_builtin_sync_mem): Handle address spaces.
8713 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
8715         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
8716         if the call takes a static chain.
8718 2018-09-21  Martin Liska  <mliska@suse.cz>
8720         * auto-profile.c (autofdo_source_profile::read): Do not
8721         set sum_all.
8722         (read_profile): Do not add working sets.
8723         (read_autofdo_file): Remove sum_all.
8724         (afdo_callsite_hot_enough_for_early_inline): Remove const
8725         qualifier.
8726         * coverage.c (struct counts_entry): Remove gcov_summary.
8727         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
8728         do not support GCOV_TAG_PROGRAM_SUMMARY.
8729         (get_coverage_counts): Remove summary and expected
8730         arguments.
8731         * coverage.h (get_coverage_counts): Likewise.
8732         * doc/gcov-dump.texi: Remove -w option.
8733         * gcov-dump.c (dump_working_sets): Remove.
8734         (main): Do not support '-w' option.
8735         (print_usage): Likewise.
8736         (tag_summary): Likewise.
8737         * gcov-io.c (gcov_write_summary): Do not dump
8738         histogram.
8739         (gcov_read_summary): Likewise.
8740         (gcov_histo_index): Remove.
8741         (gcov_histogram_merge): Likewise.
8742         (compute_working_sets): Likewise.
8743         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
8744         it not obsolete.
8745         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
8746         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
8747         (GCOV_HISTOGRAM_SIZE): Remove.
8748         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
8749         (struct gcov_summary): Simplify rapidly just
8750         to runs and sum_max fields.
8751         (gcov_histo_index): Remove.
8752         (NUM_GCOV_WORKING_SETS): Likewise.
8753         (compute_working_sets): Likewise.
8754         * gcov-tool.c (print_overlap_usage_message): Remove
8755         trailing empty line.
8756         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
8757         (output_lines): Remove program related line.
8758         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
8759         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
8760         histogram.
8761         (input_profile_summary): Do not read it.
8762         (merge_profile_summaries): And do not merge it.
8763         (input_symtab): Do not call removed function.
8764         * modulo-sched.c (sms_schedule): Do not print sum_max.
8765         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
8766         removed when histogram method was invented.
8767         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
8768         mode.
8769         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
8770         GCOV coding style.
8771         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
8772         and dump selected value.
8773         * profile.c (add_working_set): Remove.
8774         (get_working_sets): Likewise.
8775         (find_working_set): Likewise.
8776         (get_exec_counts): Do not work with working sets.
8777         (read_profile_edge_counts): Do not inform as sum_max is removed.
8778         (compute_branch_probabilities): Likewise.
8779         (compute_value_histograms): Remove argument for call of
8780         get_coverage_counts.
8781         * profile.h: Do not make gcov_summary const.
8783 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
8785         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
8787 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
8789         PR tree-optimization/86990
8790         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
8791         Check that the entire merged store group is made of constants only for
8792         overlapping stores.
8794 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
8796         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
8797         (VTABLE_VERIFICATION_SPEC): Likewise.
8798         (SANITIZER_EARLY_SPEC): Likewise.
8799         (SANITIZER_SPEC): Likewise.
8800         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
8801         * doc/invoke.texi (Link Options): Document -r.
8803 2018-09-20  Richard Biener <rguenther@suse.de>
8805         PR middle-end/87054
8806         * gimplify.c (gimplify_expr): Retain alignment of
8807         addressable lvalue in dereference.
8809 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
8811         PR bootstrap/87013
8812         * configure.ac: Check for .loc is_stmt support.
8813         * configure, config.in: Rebuilt.
8814         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
8815         if not supported.
8817 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
8819         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
8820         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
8821         -misel=no.
8823 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
8825         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8826         VECTOR_OTHER.
8827         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
8828         case VECTOR_OTHER.
8830 2018-09-20  Marek Polacek  <polacek@redhat.com>
8832         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
8834 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
8836         PR tree-optimization/87288
8837         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
8838         into account when determining PEELING_FOR_NITERS.
8840 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
8842         PR tree-optimization/86877
8843         * tree-vect-loop.c (vect_analyze_loop_2): Call
8844         vect_verify_datarefs_alignment.
8846 2018-09-19  Marek Polacek  <polacek@redhat.com>
8848         * doc/invoke.texi: Document -Wclass-conversion.
8850 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
8852         * config/pa/pa.c (pa_adjust_priority): Delete.
8853         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
8855         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
8856         (atomic_storehi): Likewise.
8857         (atomic_storesi): Likewise.
8858         (atomic_loaddi): Restore compare and swap exchange loop code.
8860 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
8862         PR rtl-optimization/86902
8863         * combine.c (try_combine): When changing the CC mode used, don't change
8864         an unrelated mode in other_insn to that new CC mode.
8866 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
8868         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
8869         with %T in place of calls to dump_generic_expr.
8870         (prune_runtime_alias_test_list): Likewise.
8871         (create_runtime_alias_checks): Likewise.
8872         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
8873         (vect_analyze_data_ref_dependence): Likewise.
8874         (vect_slp_analyze_data_ref_dependence): Likewise.
8875         (vect_record_base_alignment): Likewise.  Use %G in place of call
8876         to dump_gimple_stmt.
8877         (vect_compute_data_ref_alignment): Likewise.
8878         (verify_data_ref_alignment): Likewise.
8879         (vect_find_same_alignment_drs): Likewise.
8880         (vect_analyze_group_access_1): Likewise.
8881         (vect_analyze_data_ref_accesses): Likewise.
8882         (dependence_distance_ge_vf): Likewise.
8883         (dump_lower_bound): Likewise.
8884         (vect_prune_runtime_alias_test_list): Likewise.
8885         (vect_find_stmt_data_reference): Likewise.
8886         (vect_analyze_data_refs): Likewise.
8887         (vect_create_addr_base_for_vector_ref): Likewise.
8888         (vect_create_data_ref_ptr): Likewise.
8889         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
8890         (vect_can_advance_ivs_p): Likewise.
8891         (vect_update_ivs_after_vectorizer): Likewise.
8892         (vect_gen_prolog_loop_niters): Likewise.
8893         (vect_prepare_for_masked_peels): Likewise.
8894         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8895         (vect_determine_vectorization_factor): Likewise.
8896         (vect_is_simple_iv_evolution): Likewise.
8897         (vect_analyze_scalar_cycles_1): Likewise.
8898         (vect_analyze_loop_operations): Likewise.
8899         (report_vect_op): Likewise.
8900         (vect_is_slp_reduction): Likewise.
8901         (check_reduction_path): Likewise.
8902         (vect_is_simple_reduction): Likewise.
8903         (vect_create_epilog_for_reduction): Likewise.
8904         (vect_finalize_reduction:): Likewise.
8905         (vectorizable_induction): Likewise.
8906         (vect_transform_loop_stmt): Likewise.
8907         (vect_transform_loop): Likewise.
8908         (optimize_mask_stores): Likewise.
8909         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
8910         (vect_split_statement): Likewise.
8911         (vect_recog_over_widening_pattern): Likewise.
8912         (vect_recog_average_pattern): Likewise.
8913         (vect_determine_min_output_precision_1): Likewise.
8914         (vect_determine_precisions_from_range): Likewise.
8915         (vect_determine_precisions_from_users): Likewise.
8916         (vect_mark_pattern_stmts): Likewise.
8917         (vect_pattern_recog_1): Likewise.
8918         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8919         (vect_record_max_nunits): Likewise.
8920         (vect_build_slp_tree_1): Likewise.
8921         (vect_build_slp_tree_2): Likewise.
8922         (vect_print_slp_tree): Likewise.
8923         (vect_analyze_slp_instance): Likewise.
8924         (vect_detect_hybrid_slp_stmts): Likewise.
8925         (vect_detect_hybrid_slp_1): Likewise.
8926         (vect_slp_analyze_operations): Likewise.
8927         (vect_slp_analyze_bb_1): Likewise.
8928         (vect_transform_slp_perm_load): Likewise.
8929         (vect_schedule_slp_instance): Likewise.
8930         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8931         (vect_mark_stmts_to_be_vectorized): Likewise.
8932         (vect_init_vector_1): Likewise.
8933         (vect_get_vec_def_for_operand): Likewise.
8934         (vect_finish_stmt_generation_1): Likewise.
8935         (vect_check_load_store_mask): Likewise.
8936         (vectorizable_call): Likewise.
8937         (vectorizable_conversion): Likewise.
8938         (vectorizable_operation): Likewise.
8939         (vectorizable_load): Likewise.
8940         (vect_analyze_stmt): Likewise.
8941         (vect_is_simple_use): Likewise.
8942         (vect_get_vector_types_for_stmt): Likewise.
8943         (vect_get_mask_type_for_stmt): Likewise.
8944         * tree-vectorizer.c (increase_alignment): Likewise.
8946 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
8948         * doc/rtl.texi: Adjust vec_select description.
8949         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
8950         non-constant selectors.
8952 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
8954         * config/aarch64/aarch64-protos.h
8955         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
8956         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
8957         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
8958         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
8959         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
8960         (AARCH64_FL_PROFILE): Move index so flags are ordered.
8961         (AARCH64_ISA_RCPC8_4): New flag.
8962         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
8963         to aarch64_offset_9bit_signed_unscaled_p.
8964         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
8965         and use stlur.
8966         * config/aarch64/constraints.md (Ust): New constraint.
8967         * config/aarch64/predicates.md.
8968         (aarch64_9bit_offset_memory_operand): New predicate.
8969         (aarch64_rcpc_memory_operand): New predicate.
8971 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
8973         PR rtl-optimization/87361
8974         * rtlanal.c (nonzero_bits1): Revert accidental change.
8976 2018-09-19  Richard Biener  <rguenther@suse.de>
8978         PR tree-optimization/87349
8979         PR tree-optimization/87342
8980         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
8982 2018-09-18  Marek Polacek  <polacek@redhat.com>
8984         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
8985         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
8987 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
8989         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
8991 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
8993         PR rtl-optimization/86882
8994         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
8996 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
8998         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
8999         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
9001 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
9003         PR other/87353
9004         * doc/invoke.texi (Link Options): Fix formatting and grammar.
9006 2018-09-18  Richard Biener  <rguenther@suse.de>
9008         PR middle-end/63155
9009         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
9010         (compute_samebase_partition_bases): Likewise.
9011         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
9012         (gimple_can_coalesce_p): Simplify.
9014 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
9016         Handle a library implementation of ffs calling __builtin_ffs.
9017         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
9018         (mmix_init_libfuncs): New function: make __builtin_ffs expand
9019         to __ffsdi2.
9021 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
9023         * diagnostic-show-locus.c (class layout_range): Add field
9024         "m_original_idx".
9025         (layout_range::layout_range): Add "original_idx" param and use it
9026         to initialize new field.
9027         (make_range): Use 0 for original_idx.
9028         (layout::layout): Pass in index to calls to
9029         maybe_add_location_range.
9030         (layout::maybe_add_location_range): Add param "original_idx" and
9031         pass it on to layout_range.
9032         (layout::print_any_labels): Pass on range->m_original_idx to
9033         get_text call.
9034         (gcc_rich_location::add_location_if_nearby): Use 0 for
9035         original_idx.
9036         * gcc-rich-location.h (text_range_label::get_text): Update for new
9037         param.
9038         (range_label_for_type_mismatch::get_text): Likewise.
9040 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
9042         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
9044 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
9046         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
9047         format_string_diagnostic_t.
9048         (fmtwarn_n): Likewise.
9049         * substring-locations.c
9050         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
9051         (format_warning_n_va): Convert to...
9052         (format_string_diagnostic_t::emit_warning_n_va): ...this.
9053         (format_warning_va): Convert to...
9054         (format_string_diagnostic_t::emit_warning_va): ...this.
9055         (format_warning_at_substring): Convert to...
9056         (format_string_diagnostic_t::emit_warning): ...this.
9057         (format_warning_at_substring_n): Convert to...
9058         (format_string_diagnostic_t::emit_warning_n): ...this.
9059         * substring-locations.h (class format_string_diagnostic_t): New
9060         class.
9061         (format_warning_va): Convert to
9062         format_string_diagnostic_t::emit_warning_va.
9063         (format_warning_n_va): Convert to
9064         format_string_diagnostic_t::emit_warning_n_va.
9065         (format_warning_at_substring): Convert to
9066         format_string_diagnostic_t::emit_warning.
9067         (format_warning_at_substring_n): Convert to
9068         format_string_diagnostic_t::emit_warning_n.
9070 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
9071             Bernd Schmidt <bernds_cb1@t-online.de>
9073         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
9074         SImode args.
9076 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
9078         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
9079         operand 0 predicate to nonimmediate operand.
9080         (rint<mode>2_frndint): Remove insn pattern.
9081         (rint<mode>2): Change operand 1 predicate to general_operand.
9082         Extend operand 1 to XFmode and generate rintxf2 insn.
9083         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
9084         Do not use X87MODEF mode macro.
9085         (frndintxf2_<rounding>_i387): Rename from
9086         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
9087         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
9088         to XFmode and generate significandxf3 insn.
9090 2018-09-17  Richard Biener  <rguenther@suse.de>
9092         PR tree-optimization/87328
9093         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
9094         visiting unexecutable backedges when not iterating.
9095         (do_rpo_vn): Mark all edges not executable even when not
9096         iterating.
9098 2018-09-17  Martin Jambor  <mjambor@suse.cz>
9100         PR c/63886
9101         * doc/invoke.texi (Warning Options): Likewise.
9103 2018-09-17  Richard Biener  <rguenther@suse.de>
9105         PR tree-optimization/87301
9106         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
9107         clean EH info from leftover copy assignments.
9109 2018-09-17  Martin Liska  <mliska@suse.cz>
9111         PR gcov-profile/85871
9112         * gcov.c (output_intermediate_file): Fix out of bounds
9113         access.
9115 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
9117         * config/arc/arc.c: Object attributes for core4 not reflected
9118         correctly.
9119         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
9120         core3).
9122 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
9124         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
9126 2018-09-17  Martin Liska  <mliska@suse.cz>
9128         * doc/gcov.texi: Document new option --use-hotness-colors.
9129         * gcov.c (struct source_info): Declare new field.
9130         (source_info::source_info): Set default for maximum_count.
9131         (print_usage): Add new -q option.
9132         (process_args): Process it.
9133         (accumulate_line_info): Save src->maximum_count.
9134         (output_line_beginning): Make color line number if
9135         flag_use_hotness_colors is set.
9136         (output_line_details): Pass default argument value.
9137         (output_lines): Pass src->maximum_count.
9139 2018-09-17  Martin Liska  <mliska@suse.cz>
9141         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
9142         Use processor_names table.
9143         * config/i386/i386.c (ix86_default_align): Use
9144         processor_cost_table for alignment values.
9145         (ix86_option_override_internal): Use processor_names.
9146         (ix86_function_specific_print): Likewise.
9147         * config/i386/i386.h (struct processor_costs):
9148         Add alignment values.
9149         (struct ptt): Remove and replace with const char *.
9150         * config/i386/x86-tune-costs.h (struct processor_costs):
9151         Declare default alignments for all costs.
9153 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
9155         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
9156         symbolics or VR_VARYING ranges for ABS_EXPR.
9157         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
9158         when range will wrap.
9160 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
9162         PR middle-end/86864
9163         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
9164         before and after a JUMP_TABLE_DATA.
9166 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
9168         PR middle-end/87188
9169         * dojump.c (do_compare_and_jump): Canonicalize function pointers
9170         when one operand is a function pointer.  Use POINTER_TYPE_P and
9171         FUNC_OR_METHOD_TYPE_P.
9172         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
9173         * fold-const.c (build_range_check): Likewise.
9174         * match.pd (simple_comparison): Likewise.
9176 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
9178         PR c/82967
9179         * spellcheck.c (get_edit_distance_cutoff): New function.
9180         (selftest::test_edit_distance_unit_test_oneway): Rename to...
9181         (selftest::test_get_edit_distance_one_way): ...this.
9182         (selftest::test_get_edit_distance_unit): Rename to...
9183         (selftest::test_get_edit_distance_both_ways): ...this.
9184         (selftest::test_edit_distances): Move tests to this new function,
9185         and test some more pairs of strings.  Update for above renaming.
9186         (selftest::get_old_cutoff): New function.
9187         (selftest::test_get_edit_distance_cutoff): New function.
9188         (selftest::assert_suggested_for): New function.
9189         (ASSERT_SUGGESTED_FOR): New macro.
9190         (selftest::assert_not_suggested_for): New function.
9191         (ASSERT_NOT_SUGGESTED_FOR): New macro.
9192         (selftest::test_suggestions): New function.
9193         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
9194         tests to selftest::test_edit_distances and call it.  Add calls to
9195         selftest::test_get_edit_distance_cutoff and
9196         selftest::test_suggestions.
9197         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
9198         (best_match::consider): Replace hard-coded cutoff calculation with
9199         a call to...
9200         (best_match::get_cutoff): New declaration.
9201         (best_match::get_best_meaningful_candidate): Likewise.
9203 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9205         * builtins.c (fold_builtin_strlen): Remove TODO comment.
9207 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9209         revert:
9210         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9212         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9213         terminated string literal.
9215 2018-09-14  Martin Sebor  <msebor@redhat.com>
9217         * builtins.c (unterminated_array): Handle ARRAY_REF.
9218         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
9219         * builtins.h (unterminated_array): Declare extern.
9220         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
9221         arrays.
9222         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
9223         calls.
9225 2018-09-14  Martin Sebor  <msebor@redhat.com>
9226             Jeff Law  <law@redhat.com>
9228         * builtins.c (unterminated_array): New.
9229         (expand_builtin_strcpy): Adjust.
9230         (expand_builtin_strcpy_args): Detect unterminated arrays.
9231         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
9232         unterminated arrays.
9233         * gimple-fold.h (get_maxval_strlen): Add argument.
9234         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
9236         * gimple-fold.c (get_range_strlen): Add argument.
9237         (get_maxval_strlen): Adjust.
9238         * gimple-fold.h (get_range_strlen): Add argument.
9240 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
9242         * config/i386/movdirintrin.h: Fix copyright year.
9244 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
9246         * reg-stack.c: Include regs.h.
9247         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
9248         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
9249         FIRST_STACK_REG, not DFmode.
9250         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
9251         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
9252         (change stack): Default register mode to the reg_raw_mode of
9253         FIRST_STACK_REG, not DFmode.
9254         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
9255         (*swapxf): Rename from swapxf.
9257 2018-09-14  Carl Love  <cel@us.ibm.com>
9259         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
9260         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
9262 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
9264         PR target/87224
9265         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
9266         alternatives.
9268 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
9270         PR target/85628
9271         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
9273 2018-09-14  Jason Merrill  <jason@redhat.com>
9275         Fix --enable-gather-detailed-mem-stats.
9276         * hash-table.c (hash_table_usage): Change from variable to function.
9277         * hash-table.h: Adjust.
9278         * Makefile.in: Add missing dependencies on hash-table.h.
9280 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9282         PR tree-optimization/87259
9283         PR lto/87283
9284         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
9285         execute_cse_reciprocals_1 has tried transforming.
9287 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
9289         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
9290         VR_VARYING for PLUS/MINUS_EXPR.
9292 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9294         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
9295         formatting.
9297 2018-09-14  Richard Biener  <rguenther@suse.de>
9299         PR middle-end/63155
9300         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
9301         bits for the merged partition.
9303 2018-09-13  Martin Sebor  <msebor@redhat.com>
9304             Bernd Edlinger  <bernd.edlinger@hotmail.de>
9306         * builtins.h (c_srlen): Add argument.
9307         * builtins.c (warn_string_no_nul): New function.
9308         (c_strlen): Add argument and use it.  Update recursive calls.
9309         Pass DECL argument to string_constant to get info on non
9310         terminated strings.  Update *NONSTR as needed.
9311         (fold_builtin_strlen): Add argument to calls to c_strlen.
9312         Warn for unterminated arrays.
9313         (warn_string_no_null): Add prototype.
9314         * expr.c (string_constant): Update arguments.  Update recursive
9315         calls appropriately.  Detect missing NUL terminator and outermost
9316         declaration its missing in.
9317         Improve checks for arrays with nonzero lower bound or elements
9318         that are not a single byte.  Simplify offset computation.
9319         Simplify checks for non-NUL terminated strings.
9320         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
9321         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
9323 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9325         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
9326         correctly.
9327         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
9328         argument.  Fix range checks.
9329         * fold-const.h (c_getstr): Adjust protoype.
9330         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
9331         string is constant but contains no NUL byte.
9333         * expr.c (string_constant): Adjust function comment.
9334         Remove bogus check for zero termination.
9336         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
9338         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
9339         (get_constant_size): Don't make STRING_CSTs larger than they are.
9340         (check_string_literal): New check function for STRING_CSTs.
9341         (output_constant): Use it.
9343 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
9345         PR target/86812
9346         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
9348 2018-09-13  Richard Biener  <rguenther@suse.de>
9350         PR tree-optimization/87263
9351         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
9352         (struct unwind_state): Add max_rpo field.
9353         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
9354         Compute max_rpo, the max RPO number a block can be backwards reached
9355         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
9356         separating it from the iterating mode.
9358 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
9360         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
9361         (rfs_decision): New scheduling decision.
9363 2018-09-13  Richard Biener  <rguenther@suse.de>
9365         PR bootstrap/87134
9366         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
9367         (vn_nary_op_insert_pieces_predicated): Do not write useless
9368         valid_dominated_by_p entry outside of the allocated storage.
9370 2018-09-13  Omar Sandoval  <osandov@osandov.com>
9371             Tom de Vries  <tdevries@suse.de>
9373         PR debug/86985
9374         * dwarf2out.c (is_c): New function.
9375         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
9377 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
9379         PR target/85628
9380         * config/aarch64/aarch64.md (*aarch64_bfxil):
9381         Define.
9382         * config/aarch64/constraints.md (Ulc): Define.
9383         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
9384         Define.
9385         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
9386         New function.
9388 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
9390         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
9391         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
9392         aarch64_layout_frame call.
9393         (aarch64_expand_epilogue): Likewise.
9394         (aarch64_initial_elimination_offset): Likewise.
9395         (aarch64_get_separate_components): Likewise.
9396         (aarch64_use_return_insn_p): Likewise.
9397         (aarch64_layout_frame): Remove unneeded check.
9399 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
9401         * configure.ac: Only append
9402         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
9403         gcc_config_arguments if it was never reconfigured or last reconfigure
9404         was with different arguments.
9405         * configure: Regenerated.
9407 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
9408             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9410         PR middle-end/87290
9411         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
9412         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
9414 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
9416         PR tree-optimization/87287
9417         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
9418         X % C == 0 to X % (unsigned) C == 0 optimization to ...
9419         * match.pd (X % C == 0): ... here.  New optimization.
9421 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
9423         PR middle-end/82853
9424         * expr.h (maybe_optimize_mod_cmp): Declare.
9425         * expr.c (mod_inv): New function.
9426         (maybe_optimize_mod_cmp): New function.
9427         (do_store_flag): Use it.
9428         * cfgexpand.c (expand_gimple_cond): Likewise.
9430 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
9431             Julian Brown  <julian@codesourcery.com>
9433         PR middle-end/86336
9434         * gimplify.c (gimplify_scan_omp_clauses): Set
9435         target_firstprivatize_array_bases in OpenACC parallel and kernels
9436         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
9437         OpenACC data regions.
9439 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
9441         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
9442         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
9444 2018-09-12  Richard Biener  <rguenther@suse.de>
9446         PR tree-optimization/87280
9447         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
9448         edge but unreachable target.
9449         (do_rpo_vn): For conservatively handling a PHI only mark
9450         the backedge executable but not the block reachable.
9452 2018-09-12  Richard Biener  <rguenther@suse.de>
9454         PR tree-optimization/87266
9455         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
9456         visited blocks.
9458 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
9460         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
9461         constants.
9462         ("trunc<BFP:mode><DFP_ALL:mode>2")
9463         ("trunc<DFP_ALL:mode><BFP:mode>2")
9464         ("extend<BFP:mode><DFP_ALL:mode>2")
9465         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
9466         according to the target operand type.
9468 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
9469             Andreas Krebbel  <krebbel@linux.ibm.com>
9471         PR tree-optimization/86844
9472         * gimple-ssa-store-merging.c
9473         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
9474         there are any overlapping stores in between them, make sure they are
9475         also coalesced or we give up completely.
9477 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
9479         PR middle-end/87248
9480         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
9481         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
9483 2018-09-12  Tom de Vries  <tdevries@suse.de>
9485         * common.opt (gdescribe-dies): Add option.
9486         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
9487         attribute for artifical and nameless decls.
9488         (dwarf2out_register_external_die): Add description attribute to
9489         external reference die.
9490         (add_desc_attribute): New functions.
9491         (gen_subprogram_die): Add description attribute to
9492         DW_TAG_call_site_parameter.
9493         * tree-pretty-print.c (print_generic_expr_to_str): New function.
9494         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
9495         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
9496         -gno-describe-dies.
9497         (@item -gdescribe-dies): Add.
9499 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
9501         * tree-vrp.c (vrp_shift_undefined_p): Remove.
9502         (extract_range_from_binary_expr_1: Call
9503         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
9504         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
9505         depend on sign.
9507 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
9509         * gimple-ssa-warn-alloca.c
9510         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
9511         field for ALLOCA_BOUND_*_LARGE.
9513 2018-09-11  Nathan Sidwell  <nathan@acm.org>
9515         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
9517 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
9519         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
9520         for clobbers.  Remove obsolete comment.
9522 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
9524         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
9525         mpxchk, mpxld and mpxst types.
9526         (define_attr length_immediate): Remove all processing of mpx types.
9527         (define_attr prefix_0f): Ditto.
9528         (define_attr memory): Ditto.
9530 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
9532         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
9533         (log<mode>2): Change operand 1 predicate to general_operand.
9534         Extend operand 1 to XFmode and generate logxf3 insn.
9535         (log10<mode>2): Change operand 1 predicate to general_operand.
9536         Extend operand 1 to XFmode and generate log10xf3 insn.
9537         (log2<mode>2): Change operand 1 predicate to general_operand.
9538         Extend operand 1 to XFmode and generate log2xf3 insn.
9539         (fyl2xp1_extend<mode>xf3_i387): Remove.
9540         (log1p<mode>2): Change operand 1 predicate to general_operand.
9541         Extend operand 1 to XFmode and generate log1pxf3 insn.
9542         (fxtract_extend<mode>xf3_i387): Remove.
9543         (logb<mode>2): Change operand 1 predicate to general_operand.
9544         Extend operand 1 to XFmode and generate logbxf3 insn.
9545         (ilogb<mode>2): Change operand 1 predicate to general_operand.
9546         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
9547         (significand<mode>2): Change operand 1 predicate to general_operand.
9548         Extend operand 1 to XFmode and generate significandxf3 insn.
9550 2018-09-11  Nathan Sidwell  <nathan@acm.org>
9552         * gcc.c (perror_with_name, pfatal_with_name): Delete.
9553         (load_specs): Use fatal_error.
9554         (DELETE_IF_ORDINARY, process_command): Use error.
9555         (execute, run_attempt): Use fatal_error.
9557 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
9559         * diagnostic-core.h (sorry_at): New prototype.
9560         * diagnostic.c (sorry_at): New function.
9562 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
9564         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
9565         by zero as VR_UNDEFINED.
9567 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
9569         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
9570         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
9571         (<sincos>mode2): New expander.
9572         (sincos_extend<mode>xf3_i387): Remove insn pattern.
9573         (sincos -> sin, cos splitters): Remove splitter patterns.
9574         (sincos<mode>3): Change operand 2 predicate to general_operand.
9575         Extend operand 2 to XFmode and generate sincosxf3 insn.
9576         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
9577         Change operand 3 predicate to const1_operand.
9578         (fptan_extend<mode>xf4_i387): Remove insn pattern.
9579         (tanxf2): Update operands in the call to fptanxf4_i387.
9580         (tan<mode>2): Change operand 1 predicate to general_operand.
9581         Extend operand 1 to XFmode and generate tanxf3 insn.
9582         (atan2xf3): Rename from *fpatanxf3_i387.
9583         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
9584         (atan2xf3): Remove expander.
9585         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
9586         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
9587         (atan<mode>2): Change operand 1 predicate to general_operand.
9588         Extend operand 1 to XFmode and generate atanxf3 insn.
9590 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
9592         * config/i386/i386.md (x87/SSE constant load splitter): Use
9593         memory_operand instead of nonimmediate_operand for input operand
9594         predicate.
9596 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
9598         * config/i386/i386.md (float partial SSE register stall splitter): Move
9599         splitter near its instruction pattern.
9600         (float_extend partial SSE register stall splitter): Ditto.
9601         (float_truncate partial SSE register stall splitter): Ditto.
9603 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
9605         PR target/86794
9606         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9607         to speculation_safe_value_not_needed.
9609         PR target/85666
9610         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
9611         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
9612         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
9613         leaf_function_p, instead use has_hard_reg_initial_val.
9615 2018-09-09  Nathan Sidwell  <nathan@acm.org>
9617         * gcc.h (pfatal_with_name): Don't declare here.
9618         * gcc.c (pfatal_with_name): Make static.
9620 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
9622         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
9623         earlyclobber.
9625 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
9627         PR rtl-optimization/85458
9628         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
9629         priority hook to reduce the priority of EXPR.
9631 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
9633         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
9634         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
9635         with X87_ENABLE_FLOAT.  Remove preparation code.
9636         (*float<SWI48:mode><MODEF:mode>2): Rename from
9637         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
9638         math using "enabled" attribute.
9639         (*floatdi<MODEF:mode>2_i387): Rename from
9640         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
9641         enable for 32bit targets only.
9642         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
9643         splitter.
9644         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
9645         as operand 1 predicate.  Rewrite as define_insn_and_split.
9646         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
9648 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
9650         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
9651         to fallthru to FLOAT case.
9653 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
9655         PR target/86731
9656         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
9657         around folding of vec_sl to handle out of range shift values.
9659 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
9661         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
9662         Update callers to gen_fix_trunc<mode>_i387_fisttp
9663         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
9664         nonimmediate_operand.
9665         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
9666         and corresponding splitters.
9667         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
9668         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
9669         (fix_truncdi_i387_with_temp): Remove insn pattern
9670         and corresponding splitters.
9671         (fix_trunc<mode>_i387): Change operand 0 predicate to
9672         nonimmediate_operand.
9673         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
9674         and corresponding splitters.
9675         (*fistdi2_1): Remove.
9676         (fistdi2): Ditto.
9677         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
9678         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
9679         (*fist<mode>2_1): Remove.
9680         (fist<mode>2): Ditto.
9681         (fist<mode>2_with_temp): Remove insn pattern and corresponding
9682         splitters.
9683         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
9684         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
9685         (fistdi2_<rounding>): Change operand 0 predicate to
9686         nonimmediate_operand.
9687         (fistdi2_<rounding>_with_temp): Remove insn pattern
9688         and corresponding splitters.
9689         (fist<mode>2_<rounding>): Change operand 0 predicate to
9690         nonimmediate_operand.
9691         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
9692         and corresponding splitters.
9694         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
9696 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9698         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
9699         the init value.
9701 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
9703         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
9704         early gimple folding of vec_splat().
9705         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
9706         * gimple-fold.h: Add an extern define for tree_vec_extract().
9708 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
9710         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
9711         wrappers around TREE_TYPE comparisons.
9713 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
9715         PR target/80080
9716         * config/s390/predicates.md: Add nonsym_memory_operand.
9717         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
9718         contains a SYMBOL_REF, load it into an intermediate pseudo.
9719         (s390_emit_compare_and_swap): Legitimize operand.
9720         * config/s390/s390.md: Use the new nonsym_memory_operand
9721         with UNSPECV_CAS patterns.
9723 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
9725         PR target/80080
9726         * config/s390/s390-passes.def: New file.
9727         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
9728         declaration.
9729         (make_pass_s390_early_mach): Add declaration.
9730         * config/s390/s390.c (make_pass_s390_early_mach):
9731         (s390_option_override): Remove dynamic registration.
9732         * config/s390/t-s390: Add s390-passes.def.
9734 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
9736         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9737         Remove UNSPEC_LTREL_BASE check.
9738         (annotate_constant_pool_refs): Likewise.
9739         (find_constant_pool_ref): Likewise.
9740         (find_ltrel_base): Removed.
9741         (replace_ltrel_base): Removed.
9742         (s390_mainpool_finish): Remove replace_ltrel_base call.
9743         (s390_chunkify_start): Remove pending LTREL_BASE logic.
9744         (s390_chunkify_finish): Remove replace_ltrel_base call.
9745         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
9747 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
9749         PR target/86779
9750         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9751         to speculation_safe_value_not_needed.
9753 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
9754             Bernd Schmidt  <bernds_cb1@t-online.de>
9756         * config/nvptx/nvptx-opts.h: New file.
9757         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
9758         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
9759         (ASM_SPEC): Define.
9760         (TARGET_SM35): New macro.
9761         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
9762         correct predicate.
9763         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
9764         values.
9765         (misa=): New option.
9766         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
9768 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
9770         * config/i386/i386.md (truncdfsf2): Remove expander.
9771         (truncdfsf2_with_temp): Ditto.
9772         (truncxf<mode>2): Ditto.
9773         (*truncdfsf_fast_mixed): Remove insn pattern.
9774         (*truncdfsf_fast_i387): Ditto.
9775         (*truncdfsf_mixed): Ditto.
9776         (*truncdfsf_i387): Ditto.
9777         (*truncdfsf2_i387_1): Ditto.
9778         (*truncxfsf2_mixed): Ditto.
9779         (*truncxfdf2_mixed): Ditto.
9780         (*truncxf<mode>2_i387_noop): Ditto. Update callers
9781         to call gen_truncxf<mode>2 instead.
9782         (*truncxf<mode>2_i387): Remove.
9783         (reg->reg splitters): Remove splitter pattern.
9784         (reg->mem splitters): Ditto.
9786         (truncdfsf2): New insn pattern.
9787         (truncxf<mode>2): Ditto.
9789 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9791         * tree-ssa-math-opts.c (is_mult_by): New function.
9792         (is_square_of): Use the above.
9793         (optimize_recip_sqrt): New function.
9794         (pass_cse_reciprocals::execute): Use the above.
9796 2018-09-05  Richard Biener  <rguenther@suse.de>
9798         PR bootstrap/87134
9799         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
9800         to zero-init the emplaced vec.
9802 2018-09-05  Martin Liska  <mliska@suse.cz>
9804         PR tree-optimization/87205
9805         * tree-switch-conversion.c (pass_lower_switch::execute):
9806         Group cases for switch statements.
9808 2018-09-05  Richard Biener  <rguenther@suse.de>
9810         PR tree-optimization/87217
9811         * tree-ssa-sccvn.c (vuse_valueize): New.
9812         (vn_reference_lookup_pieces): Use it.
9813         (vn_reference_lookup): Likewise.
9815 2018-09-05  Nathan Sidwell  <nathan@acm.org>
9817         PR c++/87137
9818         * stor-layout.c (place_field): Scan forwards to check last
9819         bitfield when ms_bitfield_placement is in effect.
9821 2018-09-05  Richard Biener  <rguenther@suse.de>
9823         PR bootstrap/87225
9824         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
9825         return.
9827 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9828             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
9830         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
9831         * config.gcc (extra_objs): Build it.
9832         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
9833         Likewise.
9834         * config/aarch64/aarch64-passes.def
9835         (pass_tag_collision_avoidance): New pass.
9836         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
9837         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
9838         (aarch64_classify_address): Remove static qualifier.
9839         (aarch64_address_info, aarch64_address_type): Move to...
9840         * config/aarch64/aarch64-protos.h: ... here.
9841         (make_pass_tag_collision_avoidance): New function.
9842         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
9843         New tuning flag.
9845 2018-09-05  Martin Liska  <mliska@suse.cz>
9847         * doc/gcov.texi: Update documentation of humar
9848         readable mode.
9849         * gcov.c (format_count): Print one decimal place, it provides
9850         more fine number of situations like '1G' vs. '1.4G'.
9852 2018-09-05  Martin Liska  <mliska@suse.cz>
9854         PR target/87164
9855         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
9856         * optc-gen.awk: Allow 'Var' for Deprecated options in order
9857         to generate a MASK value.
9859 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
9861         PR debug/86593
9862         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
9863         if frame pointer isn't used.
9864         (compute_frame_pointer_to_fb_displacement): Likewise.
9866 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
9868         PR target/87198
9869         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
9870         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
9871         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
9872         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
9873         and OPTION_MASK_ISA_XSAVEC_UNSET.
9875 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
9877         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
9878         XOR operations in NAND case.
9880 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
9882         * wide-int-range.cc (wide_int_range_convert): New.
9883         * wide-int-range.h (wide_int_range_convert): New.
9884         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
9885         code into wide_int_range_convert.
9886         (extract_range_into_wide_ints): Do not munge anti range constants
9887         into the entire domain.  Just return the range back.
9889 2018-09-04  Martin Liska  <mliska@suse.cz>
9891         * genmatch.c (output_line_directive): Add new argument
9892         fnargs.
9893         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
9895 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
9897         * doc/invoke.texi (Option Summary): Add whitespace.
9899         * doc/invoke.texi (Option Summary): Add -Waligned-new.
9901 2018-09-04  Richard Biener  <rguenther@suse.de>
9903         PR tree-optimization/87211
9904         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
9905         backedge value we're supposed to treat as VARYING also number
9906         the PHI to VARYING in case it got a different value-number already.
9908 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
9910         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
9911         (extract_range_from_binary_expr_1): Do not call
9912         vrp_can_optimize_bit_op.
9913         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
9914         static.
9915         (wide_int_range_get_mask_and_bounds): New.
9916         (wide_int_range_optimize_bit_op): New.
9917         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
9918         (wide_int_range_bit_and): Same.
9919         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
9920         (wide_int_range_optimize_bit_op): New.
9921         (wide_int_range_get_mask_and_bounds): New.
9923 2018-09-04  Richard Biener  <rguenther@suse.de>
9925         PR tree-optimization/87176
9926         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
9927         variable.  When value-numbering a virtual PHI node make sure
9928         to not value-number to the backedge value.
9930 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
9932         * doc/extend.texi (Long Long, Hex Floats): Document support for
9933         long long and hex floats in more recent versions of ISO C++.
9935 2018-09-03  Richard Biener  <rguenther@suse.de>
9937         PR tree-optimization/87177
9938         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
9939         cleanup.
9941 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9943         * bb-reorder.c (edge_order): Convert to C-qsort-style
9944         tri-state comparator.
9945         (reorder_basic_blocks_simple): Change std::stable_sort to
9946         gcc_stablesort.
9948 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9950         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
9951         tri-state comparator.
9952         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
9954 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9956         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
9957         (mergesort): ... here as maximum count for using netsort.
9958         (gcc_qsort): Set nlim to 3 if stable sort is requested.
9959         (gcc_stablesort): New.
9960         * system.h (gcc_stablesort): Declare.
9962 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
9964         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
9965         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
9966         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
9968 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
9970         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
9971         lxsdx and stxsdx alternatives.
9972         (*mov<mode>_hardfloat64): Ditto.
9973         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
9975 2018-09-03  Richard Biener  <rguenther@suse.de>
9977         PR tree-optimization/87200
9978         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
9979         simplify result.
9981 2018-09-03  Martin Liska  <mliska@suse.cz>
9983         PR tree-optimization/87201
9984         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
9985         Fix parenthesis in an expression.
9987 2018-09-03  Richard Biener  <rguenther@suse.de>
9989         PR tree-optimization/87197
9990         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
9991         visited.  CSE the VN_INFO hashtable lookup.
9993         PR tree-optimization/87169
9994         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
9995         iterating make sure there's no extra backedges from irreducible
9996         regions feeding the header.  Mark the destination block
9997         executable.
9999 2018-09-03  Martin Liska  <mliska@suse.cz>
10001         PR driver/83193
10002         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
10003         * common/common-targhooks.c (default_get_valid_option_values):
10004         New function.
10005         * common/common-targhooks.h (default_get_valid_option_values):
10006         Likewise.
10007         * common/config/i386/i386-common.c: Move processor_target_table
10008         from i386.c.
10009         (ix86_get_valid_option_values): New function.
10010         (TARGET_GET_VALID_OPTION_VALUES): New macro.
10011         * config/i386/i386.c (struct ptt): Move to i386-common.c.
10012         (PTA_*): Move all defined masks into i386-common.c.
10013         (ix86_function_specific_restore): Use new processor_cost_table.
10014         * config/i386/i386.h (struct ptt): Moved from i386.c.
10015         (struct pta): Likewise.
10016         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
10017         * doc/tm.texi.in: Likewise.
10018         * opt-suggestions.c (option_proposer::suggest_option):
10019         Pass prefix to build_option_suggestions.
10020         (option_proposer::get_completions): Likewise.
10021         (option_proposer::build_option_suggestions): Use the new target
10022         hook.
10023         * opts.c (struct option_help_tuple): New struct.
10024         (print_filtered_help): Use the new target hook.
10026 2018-09-03  Martin Liska  <mliska@suse.cz>
10028         PR middle-end/59521
10029         * predict.c (set_even_probabilities): Add likely_edges
10030         argument and handle cases where we have precisely one
10031         likely edge.
10032         (combine_predictions_for_bb): Catch also likely_edges.
10033         (tree_predict_by_opcode): Handle gswitch statements.
10034         * tree-cfg.h (find_case_label_for_value): New declaration.
10035         (find_taken_edge_switch_expr): Likewise.
10036         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
10037         Find pivot in decision tree based on probabily, not by number of
10038         nodes.
10040 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
10042         * doc/standards.texi (Standards): Update Objective-C reference.
10044 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10046         * doc/install.texi (Prerequisites): Update link for MPC.
10048 2018-09-01  Michael Matz  <matz@suse.de>
10050         PR tree-optimization/87074
10051         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
10052         PHIs for outer-loop uses.
10054 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10056         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
10057         * doc/invoke.texi (C Dialect Options): Ditto.
10059 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
10061         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
10063 2018-08-31  Richard Biener  <rguenther@suse.de>
10065         PR tree-optimization/87168
10066         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
10067         (rpo_elim::eliminate_avail): When OP was not visited it must
10068         be available.
10070 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
10072         * tree-vrp.c (copy_value_range): Convert param "from" from
10073         "value_range *" to "const value_range *".
10074         (range_is_null): Likewise for param "vr".
10075         (range_int_cst_p): Likewise.
10076         (range_int_cst_singleton_p): Likewise.
10077         (symbolic_range_p): Likewise.
10078         (value_ranges_intersect_p): Likewise for both params.
10079         (value_range_nonnegative_p): Likewise for param "vr".
10080         (value_range_constant_singleton): Likewise.
10081         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
10082         (extract_range_into_wide_ints): Likewise for param "vr".
10083         (extract_range_from_multiplicative_op): Likewise for params "vr0"
10084         and "vr1".
10085         (vrp_can_optimize_bit_op): Likewise.
10086         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
10087         "vr1_".
10088         (extract_range_from_unary_expr): Likewise.
10089         (debug_value_range): Likewise for param "vr".
10090         (value_range::dump): Add "const" qualifier.
10091         (vrp_prop::check_array_ref): Convert local "vr" from
10092         "value_range *" to "const value_range *".
10093         (vrp_prop::check_mem_ref): Likewise.
10094         (vrp_prop::visit_stmt): Likewise for local "old_vr".
10095         (vrp_intersect_ranges_1): Likewise for param "vr_1".
10096         (vrp_intersect_ranges): Likewise.
10097         (simplify_stmt_for_jump_threading): Likewise for local "vr".
10098         (vrp_prop::vrp_finalize): Likewise.
10099         * tree-vrp.h (value_range::dump): Add "const" qualifier.
10100         (vrp_intersect_ranges): Add "const" qualifier to params as above.
10101         (extract_range_from_unary_expr): Likewise.
10102         (value_range_constant_singleton): Likewise.
10103         (symbolic_range_p): Likewise.
10104         (copy_value_range): Likewise.
10105         (extract_range_from_binary_expr_1): Likewise.
10106         (range_int_cst_p): Likewise.
10107         (vrp_set_zero_nonzero_bits): Likewise.
10108         (range_int_cst_singleton_p): Likewise.
10110 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
10112         * config/aarch64/arm_neon.h (vabsd_s64): New.
10113         (vnegd_s64): Likewise.
10115 2018-08-31  Martin Jambor  <mjambor@suse.cz>
10117         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
10119 2018-08-31  Martin Liska  <mliska@suse.cz>
10121         * ipa-icf.c (sem_item::add_type): Use
10122         sem_item::m_type_hash_cache.
10123         * ipa-icf.h: Move the cache from sem_item_optimizer
10124         to sem_item.
10126 2018-08-31  Nathan Sidwell  <nathan@acm.org>
10128         * doc/extend.texi (Backwards Compatibility): Remove implicit
10129         extern C leeway of () being (...).
10131 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10133         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
10135 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
10137         PR target/86684
10138         PR target/87149
10139         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
10141 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
10143         PR middle-end/87138
10144         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
10145         gen_int_mode.  Formatting fixes.
10147 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
10149         * target.def (custom_function_descriptors): Improve documentation.
10150         * doc/tm.texi.in (Trampolines): Expand discussion of function
10151         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
10152         beginning of the section.
10153         * doc/tm.texi: Regenerated.
10155 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
10157         * cfg.h (class auto_edge_flag): Spell out the template-id of the
10158         base class in the initializer list.  This is a workaround for
10159         building with older GCC.
10160         (class auto_bb_flag): Likewise.
10162 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
10164         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
10165         (altivec_vcmpequ<VI_char>_p): Remove star.
10166         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
10167         vector load modes.
10168         (expand_strncmp_vec_sequence): New function.
10169         (emit_final_str_compare_vec): New function.
10170         (expand_strn_compare): Add support for vector strncmp.
10171         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
10172         length specification to bytes.
10173         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
10174         (vcmpnezb_p): New pattern.
10175         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
10176         for option -mstring-compare-inline-limit.
10178 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
10180         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
10181         (PTA_SKYLAKE): Add PTA_AES.
10182         (PTA_GOLDMONT): Likewise.
10184 2018-08-29  Jan Hubicka  <jh@suse.cz>
10186         PR lto/86517
10187         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
10188         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
10190 2018-08-29  Jan Hubicka  <jh@suse.cz>
10192         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
10193         TYPE_STUB_DECL.
10194         (hash_tree): Do not visit TYPE_STUB_DECL.
10195         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
10196         stream TYPE_STUB_DECL.
10197         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
10198         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
10199         after free_lang_data.
10200         (type_in_anonymous_namespace_p): Likewise.
10202 2018-08-29  Jan Hubicka  <jh@suse.cz>
10204         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
10205         comment that it has to be even number.
10206         (class sreal): Change m_sig type to int32_t.
10207         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
10208         int64_t for temporary calculations.
10209         (sreal_verify_basics): Drop one bit from minimum and maximum.
10211 2018-08-30  Richard Biener  <rguenther@suse.de>
10213         PR tree-optimization/87147
10214         * tree-ssa-sccvn.c (SSA_VISITED): New function.
10215         (visit_phi): When the degenerate result is from the backedge and
10216         we didn't visit its definition yet drop to VARYING.
10217         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
10219 2018-08-29  Jan Hubicka  <jh@suse.cz>
10221         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
10222         DECL_VINDEX.
10223         (hash_tree): Likewise.
10225 2018-08-29  Jan Hubicka  <jh@suse.cz>
10227         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
10228         and TYPE_NEXT_REF_TO.
10230 2018-08-29  Jan Hubicka  <jh@suse.cz>
10232         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
10233         comment that it has to be even number.
10234         (class sreal): Change m_sig type to int32_t.
10235         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
10236         int64_t for temporary calculations.
10237         (sreal_verify_basics): Drop one bit from minimum and maximum.
10239 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
10241         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
10243 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
10245         PR middle-end/86995
10246         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
10247         if to_add is negative.
10249 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10251         PR middle-end/87053
10252         * builtins.c (c_strlen): Improve range checks.
10254 2018-08-29  Martin Sebor  <msebor@redhat.com>
10255             Jeff Law  <law@redhat.com>
10257         PR tree-optimization/86714
10258         PR tree-optimization/86711
10259         * builtins.c (c_strlen): Add arguments to call to string_constant.
10260         * expr.c (string_constant): Add argument.  Detect missing nul
10261         terminator and outermost declaration it's missing in.
10262         * expr.h (string_constant): Add argument.
10263         * fold-const.c (read_from_constant_string): Add arguments to call to
10264         string_constant.
10265         (c_getstr): Likewise.
10266         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
10267         to string_constant.
10268         * tree-ssa-strlen.c (get_stridx): Likewise.
10270 2018-08-29  Jan Hubicka  <jh@suse.cz>
10272         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
10273         Do not stream DECL_VINDEX.
10274         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
10275         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
10276         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
10278 2018-08-29  Richard Biener  <rguenther@suse.de>
10280         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
10281         virtual operands that are not default defs to honor region
10282         boundaries.
10283         (rpo_vn_valueize): Remove ineffective code here.
10285 2018-08-29  Richard Biener  <rguenther@suse.de>
10287         PR tree-optimization/87132
10288         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
10289         when skipping defs reachable over backedges.
10291 2018-08-29  Richard Biener  <rguenther@suse.de>
10293         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
10294         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
10295         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
10296         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
10297         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
10298         (vn_reference_lookup_3): Remove use of const_parms.
10299         (free_rpo_vn): Do not free const_parms.
10300         (do_rpo_vn): Do not call init_const_parms.
10301         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
10302         SSA_NAME_POINTS_TO_READONLY_MEMORY.
10303         (call_may_clobber_ref_p_1): Likewise.
10305 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
10307         PR other/86726
10308         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
10309         (-O): Ditto.
10310         (-ftree-scev-cprop): Document.
10312 2018-08-29  Jan Hubicka  <jh@suse.cz>
10314         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
10315         parameters.
10316         (sreal constructor): Update.
10317         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
10318         sreal:operator/): Update.
10320 2018-08-29  Martin Liska  <mliska@suse.cz>
10322         * tree-switch-conversion.c (switch_conversion::expand):
10323         Strenghten assumption about gswitch statements.
10325 2018-08-29  Richard Biener  <rguenther@suse.de>
10327         PR tree-optimization/87117
10328         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
10329         re-value-number released SSA VDEFs.
10331 2018-08-29  Richard Biener  <rguenther@suse.de>
10333         PR tree-optimization/87126
10334         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
10336 2018-08-28  Jim Wilson  <jimw@sifive.com>
10338         * config/riscv/pic.md: Rewrite.
10339         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
10340         invalid address.
10341         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
10342         (SOFTF, default_load, softload, softstore): New.
10344 2018-08-28  Jeff Law  <law@redhat.com>
10346         * fold-const.c (fold_binary_loc): Remove recently added assert.
10348 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
10350         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
10351         to OP parmeter of generated function.
10353 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
10355         PR tree-optimization/87009
10356         * match.pd: Add boolean optimizations.
10358 2018-08-28  Martin Sebor  <msebor@redhat.com>
10360         PR middle-end/86631
10361         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
10362         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
10363         (pass_walloca::gate): Use it.
10364         (alloca_call_type): Same.
10365         (pass_walloca::execute): Same.
10366         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
10368 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
10370         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
10371         GCC_VERSION for usage of "__gcc_dump_printf__" format from
10372         >= 3005 to >= 9000.
10374 2018-08-28  Richard Biener  <rguenther@suse.de>
10376         PR tree-optimization/87124
10377         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
10378         constants before looking up avail.
10380 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
10382         PR middle-end/87099
10383         * calls.c (maybe_warn_nonstring_arg): Punt early if
10384         warn_stringop_overflow is zero.  Don't call get_range_strlen
10385         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
10386         Swap comparison operands to have constants on rhs.  Only use
10387         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
10388         increment lenrng[0].
10390 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
10392         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
10393         use of tree_to_shwi.  Remove duplicated test for the size being
10394         a whole number of bytes.
10396 2018-08-28  Richard Biener  <rguenther@suse.de>
10398         PR tree-optimization/87117
10399         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
10400         Handle removed stmt without LHS (GIMPLE_NOP).
10402 2018-08-28  Richard Biener  <rguenther@suse.de>
10404         PR tree-optimization/87117
10405         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
10406         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
10408 2018-08-28  Richard Biener  <rguenther@suse.de>
10410         PR tree-optimization/87117
10411         * tree-ssa-pre.c (compute_avail): Do not make expressions
10412         with predicated values available.
10413         (get_expr_value_id): Assert we do not run into predicated value
10414         expressions.
10416 2018-08-28  Richard Biener  <rguenther@suse.de>
10418         PR tree-optimization/87117
10419         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
10420         get virtual operands.
10421         (get_expr_operands): Handle STRING_CST like other decls.
10423 2018-08-28  Martin Liska  <mliska@suse.cz>
10425         * tree.h: Update documentation of fndecl_built_in_p
10426         functions.
10429 2018-08-27  Jeff Law  <law@redhat.com>
10430         PR tree-optimization/87110
10431         * tree-ssa-dse.c (compute_trims): Handle non-constant
10432         TYPE_SIZE_UNIT.
10434 2018-08-27  Martin Sebor  <msebor@redhat.com>
10436         PR tree-optimization/86914
10437         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
10439 2018-08-27  Martin Sebor  <msebor@redhat.com>
10441         PR tree-optimization/87112
10442         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
10443         the type of the bound argument.
10445 2018-08-27  Jeff Law  <law@redhat.com>
10447         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
10448         type does not have a TYPE_SIZE_UNIT.
10450 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
10452         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
10453         with include of backend.h.
10455 2018-08-27  Richard Biener  <rguenther@suse.de>
10457         PR tree-optimization/86927
10458         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
10459         use const cond reduction code.
10461 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
10463         PR tree-optimization/85758
10464         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
10466 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
10468         PR c++/87091
10469         * diagnostic-show-locus.c (class layout_range): Update for
10470         conversion of show_caret_p to a tri-state.
10471         (layout_range::layout_range): Likewise.
10472         (make_range): Likewise.
10473         (layout::maybe_add_location_range): Likewise.
10474         (layout::should_print_annotation_line_p): Don't show annotation
10475         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
10476         (layout::get_state_at_point): Update for conversion of
10477         show_caret_p to a tri-state.  Bail out early for
10478         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
10479         underlining or source colorization.
10480         (gcc_rich_location::add_location_if_nearby): Update for conversion
10481         of show_caret_p to a tri-state.
10482         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
10483         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
10484         Likewise.
10485         (selftest::test_one_liner_labels): Likewise.
10486         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
10487         conversion of show_caret_p to a tri-state.
10488         * pretty-print.c (text_info::set_location): Likewise.
10489         * pretty-print.h (text_info::set_location): Likewise.
10490         * substring-locations.c (format_warning_n_va): Likewise.
10491         * tree-diagnostic.c (default_tree_printer): Likewise.
10492         * tree-pretty-print.c (newline_and_indent): Likewise.
10494 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
10496         PR c++/87091
10497         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
10498         line above for line-insertion fix-it hints.
10499         (selftest::test_fixit_insert_containing_newline): Update the
10500         expected results, and add a test with line-numbering enabled.
10502 2018-08-27  Martin Liska  <mliska@suse.cz>
10504         PR sanitizer/86962
10505         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
10506         params with DECL_HAS_VALUE_EXPR_P.
10508 2018-08-27  Martin Liska  <mliska@suse.cz>
10510         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
10511         selected expansion strategy.
10513 2018-08-27  Martin Liska  <mliska@suse.cz>
10515         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
10516         * builtins.c (is_builtin_fn): Likewise.
10517         * attribs.c (diag_attr_exclusions): Use new function
10518         fndecl_built_in_p and remove check for FUNCTION_DECL if
10519         possible.
10520         (builtin_mathfn_code): Likewise.
10521         (fold_builtin_expect): Likewise.
10522         (fold_call_expr): Likewise.
10523         (fold_builtin_call_array): Likewise.
10524         (fold_call_stmt): Likewise.
10525         (set_builtin_user_assembler_name): Likewise.
10526         (is_simple_builtin): Likewise.
10527         * calls.c (gimple_alloca_call_p): Likewise.
10528         (maybe_warn_nonstring_arg): Likewise.
10529         * cfgexpand.c (expand_call_stmt): Likewise.
10530         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
10531         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
10532         (cgraph_node::verify_node): Likewise.
10533         * cgraphclones.c (build_function_decl_skip_args): Likewise.
10534         (cgraph_node::create_clone): Likewise.
10535         * config/arm/arm.c (arm_insert_attributes): Likewise.
10536         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
10537         * dse.c (scan_insn): Likewise.
10538         * expr.c (expand_expr_real_1): Likewise.
10539         * fold-const.c (operand_equal_p): Likewise.
10540         (fold_binary_loc): Likewise.
10541         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
10542         * gimple-low.c (lower_stmt): Likewise.
10543         * gimple-pretty-print.c (dump_gimple_call): Likewise.
10544         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10545         Likewise.
10546         * gimple.c (gimple_build_call_from_tree): Likewise.
10547         (gimple_call_builtin_p): Likewise.
10548         (gimple_call_combined_fn): Likewise.
10549         * gimplify.c (gimplify_call_expr): Likewise.
10550         (gimple_boolify): Likewise.
10551         (gimplify_modify_expr): Likewise.
10552         (gimplify_addr_expr): Likewise.
10553         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
10554         * ipa-cp.c (determine_versionability): Likewise.
10555         * ipa-fnsummary.c (compute_fn_summary): Likewise.
10556         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
10557         * ipa-split.c (visit_bb): Likewise.
10558         (split_function): Likewise.
10559         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10560         * lto-cgraph.c (input_node): Likewise.
10561         * lto-streamer-out.c (write_symbol): Likewise.
10562         * omp-low.c (setjmp_or_longjmp_p): Likewise.
10563         (lower_omp_1): Likewise.
10564         * predict.c (strip_predict_hints): Likewise.
10565         * print-tree.c (print_node): Likewise.
10566         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
10567         * trans-mem.c (is_tm_irrevocable): Likewise.
10568         (is_tm_load): Likewise.
10569         (is_tm_simple_load): Likewise.
10570         (is_tm_store): Likewise.
10571         (is_tm_simple_store): Likewise.
10572         (is_tm_abort): Likewise.
10573         (tm_region_init_1): Likewise.
10574         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
10575         * tree-cfg.c (verify_gimple_call): Likewise.
10576         (move_stmt_r): Likewise.
10577         (stmt_can_terminate_bb_p): Likewise.
10578         * tree-eh.c (lower_eh_constructs_2): Likewise.
10579         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
10580         * tree-inline.c (remap_gimple_stmt): Likewise.
10581         (copy_bb): Likewise.
10582         (estimate_num_insns): Likewise.
10583         (fold_marked_statements): Likewise.
10584         * tree-sra.c (scan_function): Likewise.
10585         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
10586         (optimize_stack_restore): Likewise.
10587         (pass_fold_builtins::execute): Likewise.
10588         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
10589         (mark_all_reaching_defs_necessary_1): Likewise.
10590         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
10591         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
10592         (pass_forwprop::execute): Likewise.
10593         * tree-ssa-loop-im.c (stmt_cost): Likewise.
10594         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
10595         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
10596         * tree-ssa-strlen.c (get_string_length): Likewise.
10597         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
10598         (find_func_aliases_for_call): Likewise.
10599         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
10600         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
10601         * tree-tailcall.c (find_tail_calls): Likewise.
10602         * tree.c (need_assembler_name_p): Likewise.
10603         (free_lang_data_in_decl): Likewise.
10604         (get_call_combined_fn): Likewise.
10605         * ubsan.c (is_ubsan_builtin_p): Likewise.
10606         * varasm.c (incorporeal_function_p): Likewise.
10607         * tree.h (DECL_BUILT_IN): Remove and replace with
10608         fndecl_built_in_p.
10609         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
10610         (fndecl_built_in_p): New.
10612 2018-08-27  Martin Liska  <mliska@suse.cz>
10614         PR tree-optimization/86847
10615         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
10616         Dump also subtree probability.
10617         (switch_decision_tree::do_jump_if_equal): New function.
10618         (switch_decision_tree::emit_case_nodes): Handle special
10619         situations in balanced tree that can be emitted much simpler.
10620         Fix calculation of probabilities that happen in tree expansion.
10621         * tree-switch-conversion.h (struct cluster): Add
10622         is_single_value_p.
10623         (struct simple_cluster): Likewise.
10624         (struct case_tree_node): Add new function has_child.
10625         (do_jump_if_equal): New.
10627 2018-08-27  Martin Liska  <mliska@suse.cz>
10629         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
10630         Add new argument to bit_test_cluster constructor.
10631         (bit_test_cluster::emit): Set bits really number of values
10632         handlel by a test.
10633         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
10634         probability argument.
10635         * tree-switch-conversion.h (struct bit_test_cluster):
10636         Add m_handles_entire_switch.
10638 2018-08-27  Martin Liska  <mliska@suse.cz>
10640         PR tree-optimization/86702
10641         * tree-switch-conversion.c (jump_table_cluster::emit):
10642         Make probabilities even for values in jump table
10643         according to number of cases handled.
10644         (switch_decision_tree::compute_cases_per_edge): Pass
10645         argument to reset_out_edges_aux function.
10646         (switch_decision_tree::analyze_switch_statement): Likewise.
10647         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
10648         Make it static.
10650 2018-08-27  Martin Liska  <mliska@suse.cz>
10652         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
10653         cfun argument explicitly.
10654         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
10655         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
10656         function gimple_switch_default_bb.
10657         (convert_switch_statements):
10658         (expand_builtins):
10659         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
10660         * stmt.c (label_to_block_fn): Use label_to_block and pass
10661         cfun argument explicitly and use gimple_switch_label_bb.
10662         (expand_case): Likewise.
10663         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
10664         cfun argument explicitly. Likewise.
10665         (make_edges_bb): Likewise.
10666         (make_cond_expr_edges): Likewise.
10667         (get_cases_for_edge): Likewise.
10668         (make_gimple_switch_edges): Likewise.
10669         (label_to_block_fn): Likewise.
10670         (label_to_block): Likewise.
10671         (make_goto_expr_edges): Likewise.
10672         (make_gimple_asm_edges): Likewise.
10673         (main_block_label): Likewise.
10674         (group_case_labels_stmt): Likewise.
10675         (find_taken_edge_computed_goto): Likewise.
10676         (find_taken_edge_switch_expr): Likewise.
10677         (gimple_verify_flow_info): Likewise.
10678         (gimple_redirect_edge_and_branch): Likewise.
10679         (gimple_switch_label_bb): New function.
10680         (gimple_switch_default_bb): Likewise.
10681         (gimple_switch_edge): Likewise.
10682         (gimple_switch_default_edge): Likewise.
10683         * tree-cfg.h (label_to_block_fn): Remove and replace ...
10684         (label_to_block): ... with this.
10685         (gimple_switch_label_bb): New.
10686         (gimple_switch_default_bb): Likewise.
10687         (gimple_switch_edge): Likewise.
10688         (gimple_switch_default_edge): Likewise.
10689         * tree-cfgcleanup.c (convert_single_case_switch): Use
10690         new gimple functions and pass new argument to label_to_block.
10691         (cleanup_control_flow_bb):
10692         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
10693         cfun argument explicitly.
10694         (make_eh_edges): Likewise.
10695         (redirect_eh_dispatch_edge): Likewise.
10696         (lower_resx): Likewise.
10697         (lower_eh_dispatch): Likewise.
10698         (maybe_remove_unreachable_handlers): Likewise.
10699         (unsplit_eh): Likewise.
10700         (cleanup_empty_eh): Likewise.
10701         (verify_eh_edges): Likewise.
10702         (verify_eh_dispatch_edge): Likewise.
10703         * tree-ssa-dom.c (record_edge_info): Likewise.
10704         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
10705         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
10706         (thread_through_normal_block): Likewise.
10707         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
10708         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
10709         * tree-switch-conversion.c (switch_conversion::collect): Use new
10710         gimple functions.
10711         (switch_conversion::check_final_bb): Likewise.
10712         (switch_conversion::gather_default_values): Pass new argument
10713         to label_to_block.
10714         (switch_conversion::build_constructors): Likewise.
10715         (switch_decision_tree::compute_cases_per_edge): Use new
10716         gimple_switch_edge function.
10717         (switch_decision_tree::analyze_switch_statement): Pass new argument
10718         to label_to_block.
10719         (switch_decision_tree::try_switch_expansion): Use
10720         gimple_switch_default_edge.
10721         * tree-vrp.c (find_switch_asserts): Pass new argument
10722         to label_to_block.
10723         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
10724         (vr_values::simplify_switch_using_ranges): Likewise.
10726 2018-08-27  Richard Biener  <rguenther@suse.de>
10728         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
10729         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
10731         * tree-ssa-sccvn.h (struct vn_pval): New structure.
10732         (struct vn_nary_op_s): Add unwind_to member.  Add
10733         predicated_values flag and put result into a union together
10734         with a linked list of vn_pval.
10735         (struct vn_ssa_aux): Add name member to make maintaining
10736         a map of SSA name to vn_ssa_aux possible.  Remove no longer
10737         needed info, dfsnum, low, visited, on_sccstack, use_processed
10738         and range_info_anti_range_p members.
10739         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
10740         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
10741         New functions.
10742         (vn_valueize): New global.
10743         (vn_context_bb): Likewise.
10744         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
10745         VN_INFO_PTR_INFO): Remove.
10746         * tree-ssa-sccvn.c: ... (rewrite)
10747         (pass_fre::execute): For -O2+ initialize loops and run
10748         RPO VN in optimistic mode (iterating).  For -O1 and -Og
10749         run RPO VN in non-optimistic mode.
10750         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
10751         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
10752         * doc/invoke.texi (sccvn-max-scc-size): Remove.
10753         (rpo-vn-max-loop-depth): Document.
10754         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
10755         when valuezing the VUSE signals we walked out of the region.
10756         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
10757         (phi_translate): Set VN context block to use for availability
10758         lookup.
10759         (compute_avail): Likewise.
10760         (pre_valueize): New function.
10761         (pass_pre::execute): Adjust to the RPO VN API.
10763         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
10764         (propagate_constants_for_unrolling): Remove.
10765         (tree_unroll_loops_completely): Perform value-numbering
10766         on the unrolled bodies loop parent.
10768 2018-08-27  Richard Biener  <rguenther@suse.de>
10770         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
10771         for partial antic compute.
10773 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
10775         PR rtl-optimization/87065
10776         * combine.c (simplify_if_then_else): Formatting fix.
10777         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
10778         check.
10779         (known_cond): Don't return const_true_rtx for vector modes.  Use
10780         CONST0_RTX instead of const0_rtx.  Formatting fixes.
10782 2018-08-27  Martin Liska  <mliska@suse.cz>
10784         PR gcov-profile/87069
10785         * gcov.c (process_file): Record files already processed
10786         and warn about a file being processed multiple times.
10788 2018-08-27  Martin Liska  <mliska@suse.cz>
10790         PR driver/83193
10791         * config/aarch64/aarch64.c (aarch64_override_options_internal):
10792         Set default values for x_aarch64_*_string strings.
10793         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
10794         prefix.  For -mabi do not print '=ABI' in help and use
10795         <option_value> format for -msve-vector-bits and -moverride
10796         options.
10798 2018-08-26  Jeff Law  <law@redhat.com>
10800         * config/mips/frame-header-opt.c: Include "backend.h" rather than
10801         "cfg.h"
10803 2018-08-26  Marek Polacek  <polacek@redhat.com>
10805         PR c++/87029, Implement -Wredundant-move.
10806         * doc/invoke.texi: Document -Wredundant-move.
10808 2018-08-25  Martin Sebor  <msebor@redhat.com>
10810         PR tree-optimization/87059
10811         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
10812         to the same type as the other.
10813         * fold-const.c (fold_binary_loc): Assert expectation.
10815 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
10817         * config/darwin.c (machopic_legitimize_pic_address): Clean up
10818         extraneous parentheses, dead code section and formatting.
10820 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
10822         PR c++/87091
10823         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
10824         wide enough for jumps in the line-numbering to be visible.
10825         (layout::print_gap_in_line_numbering): New member function.
10826         (layout::calculate_line_spans): When using line numbering, merge
10827         line spans that are only 1 line apart.
10828         (diagnostic_show_locus): When printing line numbers, show gaps in
10829         line numbering directly, rather than printing headers.
10830         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
10831         line-numbering with multiple line spans.
10832         (selftest::test_fixit_insert_containing_newline_2): Add test of
10833         line-numbering, in which the spans are close enough to be merged.
10835 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
10837         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
10838         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
10839         * tree-vrp.c (range_is_nonnull): Remove.
10840         (range_includes_zero_p): Accept value_range instead of min/max.
10841         (extract_range_from_binary_expr_1): Do not early bail on
10842         POINTER_PLUS_EXPR.
10843         Use range_includes_zero_p instead of range_is_nonnull.
10844         (extract_range_from_unary_expr): Use range_includes_zero_p instead
10845         of range_is_nonnull.
10846         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
10847         special case VR_ANTI_RANGE.
10848         (vrp_finalize): Same.
10849         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
10850         instead of min/max.
10851         (range_is_nonnull): Remove.
10852         * vr-values.c (vrp_stmt_computes_nonzero): Use
10853         range_includes_zero_p instead of range_is_nonnull.
10854         (extract_range_basic): Pass value_range to range_includes_zero_p
10855         instead of range_is_nonnull.
10857 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
10859         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
10860         * emit-rtl.h (rtl_data): Remove return_bnd.
10861         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
10862         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
10863         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
10864         (POINTER_BOUNDS_MODE): Remove definition.
10865         (make_pointer_bounds_mode): Remove.
10866         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
10867         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
10868         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
10869         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
10870         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
10871         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
10872         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
10874         * config/i386/i386-modes.def (BND32, BND64): Remove.
10875         * config/i386/i386.c (dbx_register_map): Remove bound registers.
10876         (dbx64_register_map): Ditto.
10877         (svr4_dbx_register_map): Ditto.
10878         (indirect_thunk_bnd_needed): Remove.
10879         (indirect_thunks_bnd_used): Ditto.
10880         (indirect_return_bnd_needed): Ditto.
10881         (indirect_return_via_cx_bnd): Ditto.
10882         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
10883         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
10884         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
10885         (output_indirect_thunk_function): Remove handling of
10886         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
10887         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
10888         (ix86_save_reg): Remove handling of crtl->return_bnd.
10889         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
10890         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
10891         and UNSPEC_BNDLX_ADDR.
10892         (ix86_output_indirect_branch_via_reg): Remove handling of
10893         indirect_thunk_prefix_bnd.
10894         (ix86_output_indirect_branch_via_push): Ditto.
10895         (ix86_output_function_return): Ditto.
10896         (ix86_output_indirect_function_return): Ditto.
10897         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
10898         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
10899         (CALL_USED_REGISTERS): Ditto.
10900         (REG_ALLOC_ORDER): Update for removal of bound registers.
10901         (HI_REGISTER_NAMES): Ditto.
10902         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
10903         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
10904         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
10905         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
10906         (FIRST_PSEUDO_REG): Update.
10907         (BND): Remove mode iterator.
10908         * config/i386/predicates.md (bnd_mem_operator): Remove.
10910 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
10912         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
10913         vectors.
10915 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
10917         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
10918         the case in which the permute needs only a single element and
10919         repeats for every vector of the result.  Extend that case to
10920         handle variable-length vectors.
10921         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10923 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
10925         PR debug/79342
10926         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
10927         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
10929 2018-08-24  Richard Biener  <rguenther@suse.de>
10931         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
10932         bb_flags_allocated members.
10933         (auto_flag): New RAII class for allocating flags.
10934         (auto_edge_flag): New RAII class for allocating edge flags.
10935         (auto_bb_flag): New RAII class for allocating bb flags.
10936         * cfgloop.c (verify_loop_structure): Allocate temporary edge
10937         flag dynamically.
10938         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
10939         in favor of temporarily allocated BB flag.
10940         * hsa-brig.c: Re-order includes.
10941         * hsa-dump.c: Likewise.
10942         * hsa-regalloc.c: Likewise.
10943         * print-rtl.c: Likewise.
10944         * profile-count.c: Likewise.
10946 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
10948         PR target/86989
10949         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
10950         the TOC register.
10952 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
10954         PR 87073/bootstrap
10955         * wide-int-range.cc (wide_int_range_div): Do not ignore result
10956         from wide_int_range_multiplicative_op.
10958 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10960         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
10961         "permutaion".
10963 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
10965         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
10966         to 'expanded'.
10968 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
10970         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
10971         full GENERIC expression used for replacement.
10973 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
10975         * tree-vrp.c (abs_extent_range): Remove.
10976         (extract_range_into_wide_ints): Pass wide ints by reference.
10977         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
10978         Pass wide ints by reference in all calls to
10979         extract_range_into_wide_ints.
10980         * wide-int-range.cc (wide_int_range_div): New.
10981         * wide-int-range.h (wide_int_range_div): New.
10982         (wide_int_range_includes_zero_p): New.
10983         (wide_int_range_zero_p): New.
10985 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
10987         * config/aarch64/aarch64.md (arches): New enum.
10988         (arch): New enum attr.
10989         (arch_enabled): New attr.
10990         (enabled): Now uses arch_enabled only.
10991         (simd, sve, fp16): Removed attribute.
10992         (fp): Attr now defined in terms of 'arch'.
10993         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
10994         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
10995         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
10996         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
10997         attributes into 'arch'.
10998         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
10999         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
11000         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
11001         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
11002         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
11003         'simd' attribute into 'arch'.
11004         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
11005         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
11006         Convert use of 'fp' attribute to 'arch'.
11007         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
11008         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
11009         into 'arch'.
11010         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
11011         (different modes) Merge 'fp' and 'simd' into 'arch'.
11012         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
11013         'simd' into 'arch'.
11015 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
11017         PR rtl-optimization/87026
11018         * expmed.c (canonicalize_comparison): If we can no longer create
11019         pseudoregisters, don't.
11021 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
11023         PR target/86951
11024         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
11025         prototype.
11026         * config/arm/arm.c (speculation_barrier_libfunc): New static
11027         variable.
11028         (arm_init_libfuncs): Initialize it.
11029         (arm_emit_speculation_barrier): New function.
11030         * config/arm/arm.md (speculation_barrier): Call
11031         arm_emit_speculation_barrier for architectures that do not have
11032         DSB or ISB.
11033         (speculation_barrier_insn): Only match on Armv7 or later.
11035 2018-08-23  Richard Biener  <rguenther@suse.de>
11037         PR middle-end/87024
11038         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
11039         calls.
11041 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
11043         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
11044         of single-vector TBLs.
11045         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
11046         one input is given.
11048 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
11050         PR target/85910
11051         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
11052         aarch64_evpc_tbl guard.
11054 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11056         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
11057         behaviour.
11059 2018-08-22  Martin Sebor  <msebor@redhat.com>
11061         PR middle-end/87052
11062         * tree-pretty-print.c (pretty_print_string): Add argument.
11063         (dump_generic_node): Call to pretty_print_string with string size.
11065 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
11067         PR rtl-optimization/86771
11068         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
11069         of two SETs into those two SETs, one to be placed at i2, if that SETs
11070         destination is modified between i2 and i3.
11072 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11074         PR tree-optimization/86725
11075         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
11076         function.
11077         (vect_analyze_scalar_cycles_1): Check it.
11079 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11081         PR tree-optimization/86725
11082         * tree-vect-loop.c (vect_is_simple_reduction): When treating
11083         an outer loop phi as a double reduction, make sure that the
11084         single user of the phi result is an inner loop phi.
11086 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11088         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
11089         grouped stores with gaps to a strided group.
11091 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
11093         * tree-vect-stmts.c (get_group_load_store_type)
11094         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
11095         first statement in a group.
11097 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11099         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
11100         the sequence used in gcc/gcc.c.
11102 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11104         PR other/704
11105         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
11106         building it.
11108 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11110         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
11111         Darwin10-specific unwinder-shim.
11112         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
11113         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
11114         New to cater for Darwin10 Rosetta.
11116 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11118         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
11119         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
11121 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
11123         PR bootstrap/81033
11124         PR target/81733
11125         PR target/52795
11126         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
11127         (dwarf2out_switch_text_section): Generate a local label for the
11128         second function sub-section and apply it as the second FDE start
11129         label.
11130         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
11131         second sub-section start.
11133 2018-08-22  Richard Biener  <rguenther@suse.de>
11135         PR tree-optimization/86988
11136         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
11138 2018-08-22  Richard Biener  <rguenther@suse.de>
11140         PR tree-optimization/86945
11141         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
11143 2018-08-22  Alexandre Oliva <oliva@adacore.com>
11145         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
11146         a comment about how uses of r2 for .sdata2 come about.
11148 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
11150         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
11152 2018-08-21  Marek Polacek  <polacek@redhat.com>
11154         PR c++/86981, Implement -Wpessimizing-move.
11155         * doc/invoke.texi: Document -Wpessimizing-move.
11157 2018-08-21  Jan Hubicka  <jh@suse.cz>
11159         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
11160         * tree.h (is_redundant_typedef): Remove.
11161         * dwarf2out.c (is_redundant_typedef): Turn into static function.
11163 2018-08-21  Jan Hubicka  <jh@suse.cz>
11165         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
11166         when possible.
11168 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
11170         * expmed.c (extract_low_bits): Reject invalid subregs early.
11172 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11174         PR middle-end/86121
11175         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
11176         behaviour.
11178 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11180         * config/vxworks.h: Guard vxworks_asm_out_constructor and
11181         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
11182         * config/vxworks.c: Likewise.
11184 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11186         * config/vxworks.c: Set targetm.have_ctors_dtors
11187         if HAVE_INITFINI_ARRAY_SUPPORT.
11188         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
11189         if HAVE_INITFINI_ARRAY_SUPPORT.
11191 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11193         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
11194         default search path for VxWorks < 7.
11196 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11198         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
11199         (get_format_string): Refer to c_getstr.
11201 2018-08-21  Tom de Vries  <tdevries@suse.de>
11203         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
11204         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
11205         (debuginfo_early_stop): Declare.
11206         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
11207         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
11208         (debuginfo_early_stop): New function.
11209         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
11210         and debuginfo_early_stop.
11211         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
11212         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
11213         (general_init): Call debuginfo_early_init.
11214         (finalize): Call debuginfo_fini.
11215         (do_compile): Call debuginfo_init.
11216         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
11217         -fdump-early-debug.
11218         (@item -fdump-debug, @item -fdump-earlydebug): Add.
11220 2018-08-21  Tom de Vries  <tdevries@suse.de>
11222         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
11223         flag_dump_noaddr and flag_dump_unnumbered.
11225 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
11227         * wide-int-range.cc (wide_int_range_abs): New.
11228         (wide_int_range_order_set): Rename from wide_int_range_min_max.
11229         * wide-int-range.h (wide_int_range_abs): New.
11230         (wide_int_range_min_max): New.
11231         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
11232         case to call wide_int_range_abs.
11233         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
11234         (extract_range_from_abs_expr): Delete.
11236 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
11238         PR target/87033
11239         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
11240         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
11241         for indexed loads.
11243 2018-08-20  Nathan Sidwell  <nathan@acm.org>
11244             Jeff Law <law@redhat.com>
11246         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
11247         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
11249 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
11251         PR other/84889
11252         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
11253         (decl_attributes): Likewise.
11254         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
11255         instance.
11256         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11257         * diagnostic-core.h (class auto_diagnostic_group): New class.
11258         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
11259         (diagnostic_report_diagnostic): Handle the first diagnostics within
11260         a group.
11261         (emit_diagnostic): Add auto_diagnostic_group instance.
11262         (inform): Likewise.
11263         (inform_n): Likewise.
11264         (warning): Likewise.
11265         (warning_at): Likewise.
11266         (warning_n): Likewise.
11267         (pedwarn): Likewise.
11268         (permerror): Likewise.
11269         (error): Likewise.
11270         (error_n): Likewise.
11271         (error_at): Likewise.
11272         (sorry): Likewise.
11273         (fatal_error): Likewise.
11274         (internal_error): Likewise.
11275         (internal_error_no_backtrace): Likewise.
11276         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
11277         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
11278         * diagnostic.h (struct diagnostic_context): Add fields
11279         "diagnostic_group_nesting_depth",
11280         "diagnostic_group_emission_count", "begin_group_cb",
11281         "end_group_cb".
11282         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
11283         Add auto_diagnostic_group instance(s).
11284         (find_explicit_erroneous_behavior): Likewise.
11285         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
11286         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
11287         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
11288         (gimplify_va_arg_expr): Likewise.
11289         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
11290         (HSA_SORRY_AT): Likewise.
11291         * ipa-devirt.c (compare_virtual_tables): Likewise.
11292         (warn_odr): Likewise.
11293         * multiple_target.c (expand_target_clones): Likewise.
11294         * opts-common.c (cmdline_handle_error): Likewise.
11295         * reginfo.c (globalize_reg): Likewise.
11296         * substring-locations.c (format_warning_n_va): Likewise.
11297         * tree-inline.c (expand_call_inline): Likewise.
11298         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
11299         * tree-ssa-loop-niter.c
11300         (do_warn_aggressive_loop_optimizations): Likewise.
11301         * tree-ssa-uninit.c (warn_uninit): Likewise.
11302         * tree.c (warn_deprecated_use): Likewise.
11304 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
11306         PR target/87014
11307         * config/i386/i386.md (eh_return): Always update EH return
11308         address in word_mode.
11310 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
11312         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
11313         TARGET_SPLIT_COMPLEX_ARG is defined.
11315 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11317         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
11319 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11321         PR target/86984
11322         * expr.c (expand_assignment): Assert that bitpos is positive.
11323         (store_field): Likewise
11324         (expand_expr_real_1): Make sure that bitpos is positive.
11325         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
11326         integer overflow.
11328 2018-08-20  Nathan Sidwell  <nathan@acm.org>
11330         * Makefile.in (CPP_ID_DATA_H): Delete.
11331         (CPP_INTERNAL_H): Don't add it.
11332         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
11333         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
11335 2018-08-20  Richard Biener  <rguenther@suse.de>
11337         PR tree-optimization/78655
11338         * tree-vrp.c (extract_range_from_binary_expr_1): Make
11339         pointer + offset nonnull if either operand is nonnull work.
11341 2018-08-20  Tom de Vries  <tdevries@suse.de>
11343         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
11344         unless the referenced die describes the added property using
11345         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
11346         Otherwise, add a DW_AT_location to the referenced die.
11348 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
11350         PR target/86994
11351         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
11352         register_operand when calling ix86_set_reg_reg_cost.
11353         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
11354         Set *total to 0 for operands that satisfy x86_64_immediate_operand
11355         predicate and to 1 otherwise.
11357 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
11359         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
11360         emit a diagnostic that it is not supported and reset the option.
11361         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
11362         supported and consume the option.  (ASM_FINAL_SPEC): New.
11364 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
11366         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
11367         a sentence.
11369 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
11371         C-SKY port: Documentation
11373         * doc/extend.texi (C-SKY Function Attributes): New section.
11374         * doc/invoke.texi (Option Summary): Add C-SKY options.
11375         (C-SKY Options): New section.
11376         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
11378 2018-08-17  Jojo  <jijie_rong@c-sky.com>
11379             Huibin Wang  <huibin_wang@c-sky.com>
11380             Sandra Loosemore  <sandra@codesourcery.com>
11381             Chung-Lin Tang  <cltang@codesourcery.com>
11383         C-SKY port: Backend implementation
11385         * config/csky/*: New.
11386         * common/config/csky/*: New.
11388 2018-08-17  Jojo  <jijie_rong@c-sky.com>
11389             Huibin Wang  <huibin_wang@c-sky.com>
11390             Sandra Loosemore  <sandra@codesourcery.com>
11391             Chung-Lin Tang  <cltang@codesourcery.com>
11392             Andrew Jenner  <andrew@codesourcery.com>
11394         C-SKY port: Configury
11396         * config.gcc (csky-*-*): New.
11397         * configure.ac: Add csky to targets for dwarf2 debug_line support.
11398         * configure: Regenerated.
11400 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
11402         * dump-context.h: Include "dumpfile.h".
11403         (dump_context::dump_printf_va): Convert final param from va_list
11404         to va_list *.  Convert from ATTRIBUTE_PRINTF to
11405         ATTRIBUTE_GCC_DUMP_PRINTF.
11406         (dump_context::dump_printf_loc_va): Likewise.
11407         * dumpfile.c: Include "stringpool.h".
11408         (make_item_for_dump_printf_va): Delete.
11409         (make_item_for_dump_printf): Delete.
11410         (class dump_pretty_printer): New class.
11411         (dump_pretty_printer::dump_pretty_printer): New ctor.
11412         (dump_pretty_printer::emit_items): New member function.
11413         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
11414         function.
11415         (dump_pretty_printer::emit_item): New member function.
11416         (dump_pretty_printer::stash_item): New member function.
11417         (dump_pretty_printer::format_decoder_cb): New member function.
11418         (dump_pretty_printer::decode_format): New member function.
11419         (dump_context::dump_printf_va): Reimplement in terms of
11420         dump_pretty_printer.
11421         (dump_context::dump_printf_loc_va): Convert final param from va_list
11422         to va_list *.
11423         (dump_context::begin_scope): Reimplement call to
11424         make_item_for_dump_printf.
11425         (dump_printf): Update for change to dump_printf_va.
11426         (dump_printf_loc): Likewise.
11427         (selftest::test_capture_of_dump_calls): Convert "stmt" from
11428         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
11429         with %T, %E, and %G.
11430         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
11431         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
11432         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
11433         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
11434         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
11435         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
11436         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
11437         within a dump_printf_loc call to "%wu".
11438         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
11439         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
11440         missing space after "=".
11441         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
11442         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
11443         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
11444         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
11445         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
11446         duplicate "vectorized" from message.
11448 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
11450         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
11451         polyNxK_t element's TYPE_STRING_FLAG.
11453 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
11455         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
11456         (they were unnamed before).  Fix comments.
11458 2018-08-17  Nathan Sidwell  <nathan@acm.org>
11460         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
11462 2018-08-17  Richard Biener  <rguenther@suse.de>
11464         PR tree-optimization/86841
11465         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
11467 2018-08-17  Martin Liska  <mliska@suse.cz>
11469         * common.opt: Remove Warn, Init and Report for options with
11470         Ignore/Deprecated flag. Warning is done automatically for
11471         Deprecated flags.
11472         * config/i386/i386.opt: Likewise.
11473         * config/ia64/ia64.opt: Likewise.
11474         * config/rs6000/rs6000.opt: Likewise.
11475         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
11476         Remove usage of flag_check_pointer_bounds.
11477         * lto-wrapper.c (merge_and_complain): Do not handle
11478         OPT_fcheck_pointer_bounds.
11479         (append_compiler_options): Likewise.
11480         * opt-functions.awk: Do not handle Deprecated.
11481         * optc-gen.awk: Check that Var, Report and Init are not
11482         used for an option with Ignore/Deprecated flag.
11483         * opts-common.c (decode_cmdline_option): Do not report
11484         CL_ERR_DEPRECATED.
11485         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
11486         options.
11487         * opts.h (struct cl_option): Remove cl_deprecated flag.
11488         (CL_ERR_DEPRECATED): Remove error enum value.
11490 2018-08-17  Richard Biener  <rguenther@suse.de>
11492         PR middle-end/86505
11493         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
11494         across a va-arg-pack using call adjust its return value accordingly.
11496 2018-08-16  Martin Sebor  <msebor@redhat.com>
11498         PR tree-optimization/86853
11499         * gimple-ssa-sprintf.c (struct format_result): Rename member.
11500         (struct fmtresult): Add member and initialize it in ctors.
11501         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
11502         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
11503         (format_directive): Set POSUNDER4K when MAYFAIL is set.
11504         (parse_directive): Handle %C same as %c.
11505         (sprintf_dom_walker::compute_format_length): Adjust.
11506         (is_call_safe): Adjust.
11508 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11510         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
11511         for determining how to count the elements.
11512         * builtins.h (c_strlen): Adjust prototype.
11513         * expr.c (string_constant): Add new parameter mem_size.
11514         Set *mem_size appropriately.
11515         * expr.h (string_constant): Adjust protoype.
11516         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
11517         * gimple-fold.h (get_range_strlen): Adjust prototype.
11518         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
11519         (format_string): Call get_string_length with eltsize.
11521 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
11523         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
11524         to emit the span, rather than setting it as the prefix.
11526 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
11528         * diagnostic-show-locus.c (layout::start_annotation_line): Add
11529         "margin_char" parameter, defaulting to space.  Use it in place
11530         of pp_space for the initial part of the margin.
11531         (layout::print_leading_fixits): Use '+' when filling the margin
11532         of line-insertion fix-it hints.
11534 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11536         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
11537         Delete.
11539 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11541         * config/rs6000/altivec.md: Don't set length attribute to the default
11542         value.
11543         * config/rs6000/darwin.md: Ditto.
11544         * config/rs6000/dfp.md: Ditto.
11545         * config/rs6000/htm.md: Ditto.
11546         * config/rs6000/rs6000.md: Ditto.
11547         * config/rs6000/sync.md: Ditto.
11548         * config/rs6000/vsx.md: Ditto.
11550 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11552         * config/rs6000/altivec.md: Don't set length attribute to the default
11553         value, for branch instructions.
11554         * config/rs6000/darwin.md: Ditto.
11555         * config/rs6000/rs6000.md: Ditto.
11557 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
11559         * config/rs6000/rs6000.md (length): Always define as const_int 4.
11560         (unnamed conditional branch define_insn): Set length to 4 or 8
11561         depending on offset.
11562         (<bd>_<mode>): Similar, for alternative 0.
11563         (<bd>tf_<mode>): Ditto.
11565 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
11567         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
11569 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
11571         * doc/rtl.texi: Replace old RTX class names with new names.
11574 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
11576         * expmed.h (canonicalize_comparison): New declaration.
11577         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
11578         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
11579         * optabs.c (prepare_cmp_insn): Likewise.
11580         * rtl.h (unsigned_condition_p): New function which checks if a
11581         comparison operator is unsigned.
11583 2018-08-16  Nathan Sidwell  <nathan@acm.org>
11585         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
11586         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
11588 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
11590         PR target/84711
11591         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
11592         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
11593         (mov<mov>): ..this and enable unconditionally.
11595 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
11597         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
11599 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
11601         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
11602         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
11603         "Common" with "Target".
11605 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
11607         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
11608         * doc/invoke.texi (mmitigate-rop): Remove.
11609         * config/i386/i386.c: Do not include "regrename.h".
11610         (ix86_rop_should_change_byte_p, reg_encoded_number)
11611         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
11612         Remove.
11613         (ix86_reorg): Remove call to ix86_mitigate_rop.
11614         * config/i386/i386.md (attr "modrm_class"): Remove.
11615         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
11616         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
11617         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
11619 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
11621         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
11622         allow folding of mergeh() and mergel() for the float and double types.
11623         (fold_mergehl_helper): Rework to handle building a permute tree
11624         for float vectors.
11626 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
11628         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
11629         for TARGET_SSE.
11631 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
11633         * common.opt (fdiagnostics-show-labels): New option.
11634         * diagnostic-show-locus.c (class layout_range): Add field
11635         "m_label".
11636         (class layout): Add field "m_show_labels_p".
11637         (layout_range::layout_range): Add param "label" and use it to
11638         initialize m_label.
11639         (make_range): Pass in NULL for new "label" param of layout_range's
11640         ctor.
11641         (layout::layout): Initialize m_show_labels_p.
11642         (layout::maybe_add_location_range): Pass in loc_range->m_label
11643         when constructing layout_range instances.
11644         (struct line_label): New struct.
11645         (layout::print_any_labels): New member function.
11646         (layout::print_line): Call it if label-printing is enabled.
11647         (selftest::test_one_liner_labels): New test.
11648         (selftest::test_diagnostic_show_locus_one_liner): Call it.
11649         * diagnostic.c (diagnostic_initialize): Initialize
11650         context->show_labels_p.
11651         * diagnostic.h (struct diagnostic_context): Add field
11652         "show_labels_p".
11653         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
11654         -fno-diagnostics-show-labels.
11655         * dwarf2out.c (gen_producer_string): Add
11656         OPT_fdiagnostics_show_labels to the ignored options.
11657         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
11658         param.
11659         (gcc_rich_location::maybe_add_expr): Likewise.
11660         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
11661         label" param, defaulting to NULL.
11662         (gcc_rich_location::add_expr): Add "label" param.
11663         (gcc_rich_location::maybe_add_expr): Likewise.
11664         (class text_range_label): New class.
11665         (class range_label_for_type_mismatch): New class.
11666         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
11667         of format_warning_va.
11668         (fmtwarn_n): Likewise for new params of format_warning_n_va.
11669         * lto-wrapper.c (merge_and_complain): Add
11670         OPT_fdiagnostics_show_labels to the "pick one setting" options.
11671         (append_compiler_options): Likewise to the dropped options.
11672         (append_diag_options): Likewise to the passed-on options.
11673         * opts.c (common_handle_option): Handle the new option.
11674         * selftest-diagnostic.c
11675         (test_diagnostic_context::test_diagnostic_context): Enable
11676         show_labels_p.
11677         * substring-locations.c: Include "gcc-rich-location.h".
11678         (format_warning_n_va): Add "fmt_label" and "param_label" params
11679         and use them as appropriate.
11680         (format_warning_va): Add "fmt_label" and "param_label" params,
11681         passing them on to format_warning_n_va.
11682         (format_warning_at_substring): Likewise.
11683         (format_warning_at_substring_n): Likewise.
11684         * substring-locations.h (format_warning_va): Add "fmt_label" and
11685         "param_label" params.
11686         (format_warning_n_va): Likewise.
11687         (format_warning_at_substring): Likewise.
11688         (format_warning_at_substring_n): Likewise.
11689         * toplev.c (general_init): Initialize global_dc->show_labels_p.
11691 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
11693         PR testsuite/86519
11694         * builtins.c (expand_builtin_memcmp): Do not expand the call
11695         when overflow is detected.
11697 2018-08-15  Martin Sebor  <msebor@redhat.com>
11699         PR tree-optimization/71625
11700         * config/aarch64/aarch64-builtins.c
11701         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
11703 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
11705         * config/s390/s390.c (s390_reorg): Remove loop.
11707 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
11709         * config/darwin.c
11710          (darwin_function_switched_text_sections): Delete.
11711         * gcc/config/darwin.h
11712          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
11714 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
11716         PR target/81685
11717         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
11718         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
11719         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
11721 2018-08-15  Martin Liska  <mliska@suse.cz>
11723         PR tree-optimization/86925
11724         * predict.c (expr_expected_value_1): When taking
11725         later predictor, assign also probability.
11726         Use fold_build2_initializer_loc in order to fold
11727         the expression in -frounding-math.
11729 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
11731         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
11732         patterns.
11733         (expand_vec_perm_1): Try the new method.
11735 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
11737         PR target/86547
11738         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
11739         Check whether lra_live_max_point is 0 before dividing.
11741 2018-08-14  Martin Sebor  <msebor@redhat.com>
11743         PR tree-optimization/86650
11744         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
11745         (vrp_prop::check_mem_ref): Same.
11747 2018-08-13  Liu Hao <lh_mouse@126.com>
11749         * pretty-print.c (eat_esc_sequence): Swap the foreground and
11750         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
11751         and clear it thereafter, as it only works for DBCS.
11753 2018-08-13  Liu Hao <lh_mouse@126.com>
11755         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
11756         handle returned by _get_osf_handle().
11758 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
11760         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
11761         for folding vec_perm.
11763 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
11765         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
11766         Add support for gimple-folding of vec_pack() and vec_unpack()
11767         intrinsics.
11769 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
11771         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
11772         vec_xst variants to the list.
11773         (rs6000_gimple_fold_builtin): Add support for folding unaligned
11774         vector loads and stores.
11776 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
11778         * config.gcc (rs6000-ibm-aix4.x): Delete.
11779         (rs6000-ibm-aix5.1): Delete.
11780         (rs6000-ibm-aix5.2): Delete.
11781         (rs6000-ibm-aix5.3): Delete.
11782         * config/rs6000/aix43.h: Delete.
11783         * config/rs6000/aix51.h: Delete.
11784         * config/rs6000/aix52.h: Delete.
11785         * config/rs6000/t-aix43: Delete.
11787 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
11789         * config/s390/s390.c (s390_decompose_constant_pool_ref):
11790         New function.
11791         (s390_decompose_address): Factor out constant pool ref
11792         decomposition.
11794 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11796         * config/nds32/nds32-predicates.c
11797         (nds32_can_use_bclr_p): Change return type as bool.
11798         (nds32_can_use_bset_p): Ditto.
11799         (nds32_can_use_btgl_p): Ditto.
11800         (nds32_can_use_bitci_p): Ditto.
11801         * config/nds32/nds32-protos.h
11802         (nds32_can_use_bclr_p): Change declaration.
11803         (nds32_can_use_bset_p): Ditto.
11804         (nds32_can_use_btgl_p): Ditto.
11805         (nds32_can_use_bitci_p): Ditto.
11807 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11809         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
11810         Support -msched-prolog-epilog option.
11811         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
11813 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11815         * common/config/nds32/nds32-common.c
11816         (nds32_option_optimization_table): Enalbe -malways-align.
11818 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
11820         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
11821         extra_headers.
11822         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
11823         OPT_misr_secure_ case.
11824         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
11825         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
11826         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
11827         secure attribute.
11828         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
11829         (nds32_isr_info): New field security_level.
11830         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
11831         * config/nds32/nds32.md (return_internal): Consider critical attribute.
11832         * config/nds32/nds32.opt (misr-secure): New option.
11833         * config/nds32/nds32_init.inc: New file.
11834         * config/nds32/nds32_isr.h: New file.
11836 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
11838         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
11839         Update comment for atomic instructions.
11840         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
11841         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
11842         Remove.
11843         (atomic_loaddi): Revise fence expansion to only emit fence prior to
11844         load for __ATOMIC_SEQ_CST model.
11845         (atomic_loaddi_1): Remove float register target.
11846         (atomic_storedi): Handle CONST_INT values.
11847         (atomic_storedi_1): Remove float register source.  Add special case
11848         for zero value.
11849         (memory_barrier): New expander and insn.
11851 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
11853         PR tree-optimization/86835
11854         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
11855         new_stmt after def_gsi, make sure to insert new_square_stmt after
11856         that stmt, not 2 stmts before it.
11858 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
11860         PR target/82418
11861         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
11862         instead of SWI48.
11864 2018-08-10  Martin Liska  <mliska@suse.cz>
11866         PR target/83610
11867         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
11868         function type.
11869         * builtins.c (expand_builtin_expect_with_probability):
11870         New function.
11871         (expand_builtin_expect_with_probability): New function.
11872         (build_builtin_expect_predicate): Add new argumnet probability
11873         for BUILT_IN_EXPECT_WITH_PROBABILITY.
11874         (fold_builtin_expect):
11875         (fold_builtin_2):
11876         (fold_builtin_3):
11877         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
11878         * builtins.h (fold_builtin_expect): Set new argument.
11879         * doc/extend.texi: Document __builtin_expect_with_probability.
11880         * doc/invoke.texi: Likewise.
11881         * gimple-fold.c (gimple_fold_call): Pass new argument.
11882         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
11883         also BUILT_IN_EXPECT_WITH_PROBABILITY.
11884         * predict.c (get_predictor_value): New function.
11885         (expr_expected_value): Add new argument probability. Assume
11886         that predictor and probability are always non-null.
11887         (expr_expected_value_1): Likewise.  For __builtin_expect and
11888         __builtin_expect_with_probability set probability.  Handle
11889         combination in binary expressions.
11890         (tree_predict_by_opcode): Simplify code by simply calling
11891         get_predictor_value.
11892         (pass_strip_predict_hints::execute): Add handling of
11893         BUILT_IN_EXPECT_WITH_PROBABILITY.
11894         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
11895         new predictor.
11896         * tree.h (DECL_BUILT_IN_P): New function.
11898 2018-08-10  Martin Liska  <mliska@suse.cz>
11900         PR tree-optimization/85799
11901         * passes.def: Add argument for pass_strip_predict_hints.
11902         * predict.c (class pass_strip_predict_hints): Add new argument
11903         early_p.
11904         (strip_predictor_early): New function.
11905         (pass_strip_predict_hints::execute): Call the function to
11906         strip predictors.
11907         (strip_predict_hints): New function.
11908         * predict.def: Fix comment.
11910 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
11912         * Makefile.in: Clarify which tm.texi to copy over to assert the
11913         right to grant a GFDL license for all.
11915 2018-08-09  Jeff Law  <law@redhat.com>
11917         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
11918         unused variable.
11920 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
11922         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
11923         prototype.
11925 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
11927         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
11928         reductions for variable-length vectors.
11930 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
11932         PR other/84889
11933         * common.opt (fdiagnostics-show-line-numbers): New option.
11934         * diagnostic-show-locus.c (class layout): Add fields
11935         "m_show_line_numbers_p" and "m_linenum_width";
11936         (num_digits): New function.
11937         (test_num_digits): New function.
11938         (layout::layout): Initialize new fields.  Update m_x_offset
11939         logic to handle any left margin.
11940         (layout::print_source_line): Print line number when requested.
11941         (layout::start_annotation_line): New member function.
11942         (layout::print_annotation_line): Call it.
11943         (layout::print_leading_fixits): Likewise.
11944         (layout::print_trailing_fixits): Likewise.  Update calls to
11945         move_to_column for new parameter.
11946         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
11947         it to potentially call start_annotation_line.
11948         (layout::show_ruler): Call start_annotation_line.
11949         (selftest::test_line_numbers_multiline_range): New selftest.
11950         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
11951         and selftest::test_line_numbers_multiline_range.
11952         * diagnostic.c (diagnostic_initialize): Initialize
11953         show_line_numbers_p.
11954         * diagnostic.h (struct diagnostic_context): Add field
11955         "show_line_numbers_p".
11956         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
11957         -fno-diagnostics-show-line-numbers.
11958         * dwarf2out.c (gen_producer_string): Add
11959         OPT_fdiagnostics_show_line_numbers to the ignored options.
11960         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
11961         one setting" options.
11962         (append_compiler_options): Likewise to the dropped options.
11963         (append_diag_options): Likewise to the passed-on options.
11964         * opts.c (common_handle_option): Handle the new option.
11965         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
11967 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11969         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
11970         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
11971         third argument of type "const signed char" to descriptions of
11972         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
11973         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
11974         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
11975         __builtin_bcdsub_ov functions.
11977 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
11979         PR tree-optimization/86858
11980         * tree-vect-loop.c (vect_is_simple_reduction): Restore
11981         flow_bb_inside_loop_p calls.
11983 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
11985         PR tree-optimization/86871
11986         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
11987         instead of gimple_assign_lhs.
11989 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
11991         PR target/86887
11992         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
11993         register constraint to operand 0.
11994         (add<mode>3_carryinC): Likewise.
11995         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
11997 2018-08-09  Martin Liska  <mliska@suse.cz>
11999         PR c/86895
12000         * common.opt: Remove extra line.
12002 2018-08-09  Martin Liska  <mliska@suse.cz>
12004         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
12005         at the end of a line, make first letter capital and end up
12006         a sentence with a dot.
12007         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
12008         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
12009         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
12010         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
12011         (PARAM_MAX_ISL_OPERATIONS): Likewise.
12012         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
12013         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
12014         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
12015         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
12016         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
12017         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
12018         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
12019         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
12020         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
12021         (PARAM_TREE_REASSOC_WIDTH): Likewise.
12022         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
12023         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
12024         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
12026 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
12028         PR target/84332
12029         * config/s390/s390.c (s390_option_override_internal): Reduce the
12030         stack-clash-protection-probe-interval param if it would be too big
12031         for z900.
12033 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
12035         PR target/46179
12036         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
12037         * config/m68k/m68k.c (handle_move_double): Don't call
12038         m68k_final_prescan_insn.
12039         (m68k_adjust_decorated_operand): Renamed from
12040         m68k_final_prescan_insn, remove first and third operand and
12041         simplify.
12042         (print_operand): Call it.
12043         (print_operand_address): Call it.
12045 2018-08-08  Nathan Sidwell  <nathan@acm.org>
12047         * diagnostic.c (diagnostic_report_current_module): Use
12048         linemap_included_from & linemap_included_from_linemap.
12050 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
12052         * config/aarch64/aarch64-cores.def: Add phecda core.
12053         * config/aarch64/aarch64-tune.md: Regenerate.
12054         * doc/invoke.texi: Add phecda core.
12056 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
12058         PR target/85295
12059         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
12060         definitions.
12061         * config/s390/s390.md ("movti"): Add more alternatives for
12062         constant to GPR copies.
12064 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
12066         * config/s390/s390.c: Fix whitespace damage throughout the file.
12067         * config/s390/s390.h: Likewise.
12068         * config/s390/tpf.h: Likewise.
12070 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
12072         * config/s390/s390.c (s390_loadrelative_operand_p):
12073         Remove TARGET_CPU_ZARCH usages.
12074         (s390_rtx_costs): Likewise.
12075         (s390_legitimate_constant_p): Likewise.
12076         (s390_cannot_force_const_mem): Likewise.
12077         (legitimate_reload_constant_p): Likewise.
12078         (s390_preferred_reload_class): Likewise.
12079         (legitimize_pic_address): Likewise.
12080         (legitimize_tls_address): Likewise.
12081         (s390_split_branches): Removed.
12082         (s390_add_execute): Removed.
12083         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
12084         (s390_mainpool_start): Likewise.
12085         (s390_mainpool_finish): Likewise.
12086         (s390_mainpool_cancel): Removed.
12087         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
12088         (s390_chunkify_cancel): Likewise.
12089         (s390_return_addr_rtx): Likewise.
12090         (s390_register_info): Remove split_branches_pending_p uages.
12091         (s390_optimize_register_info): Likewise.
12092         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
12093         split_branches_pending_p usages.
12094         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
12095         (s390_load_got): Likewise.
12096         (s390_expand_split_stack_prologue): Likewise.
12097         (output_asm_nops): Likewise.
12098         (s390_function_profiler): Likewise.
12099         (s390_emit_call): Likewise.
12100         (s390_conditional_register_usage): Likewise.
12101         (s390_optimize_prologue): Likewise.
12102         (s390_reorg): Remove TARGET_CPU_ZARCH and
12103         split_branches_pending_p usages.
12104         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
12105         usages.
12106         (s390_output_indirect_thunk_function): Likewise.
12107         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
12108         (TARGET_CPU_ZARCH_P): Removed.
12109         (struct machine_function): Remove split_branches_pending_p.
12110         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
12112 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
12114         * common/config/s390/s390-common.c (processor_flags_table):
12115         Remove flags.
12116         * config.gcc: Remove with_arch/with_tune support.
12117         * config/s390/2064.md: Remove cpu attribute comparisons.
12118         * config/s390/driver-native.c (s390_host_detect_local_cpu):
12119         Remove MTN.
12120         * config/s390/linux.h (ASM_SPEC):
12121         Remove -march support.
12122         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
12123         Use a table to get an arch level.
12124         * config/s390/s390-opts.h (enum processor_type):
12125         Remove enum values.
12126         * config/s390/s390.c
12127         (processor_table): Remove entries, add arch_level values.
12128         (s390_issue_rate): Remove cases.
12129         (s390_option_override): Adjust
12130         s390_option_override_internal() call.
12131         (s390_option_override_internal): Remove deprecation warning.
12132         (s390_valid_target_attribute_tree): Adjust
12133         s390_option_override_internal() call.
12134         * config/s390/s390.h (struct s390_processor):
12135         Share with s390-c.c, add arch_level field.
12136         * config/s390/s390.md:
12137         Remove occurrences in cpu attribute.
12138         * config/s390/s390.opt: Remove -march/-mtune support.
12139         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
12140         * doc/invoke.texi: Remove deprecation warning.
12142 2018-08-08  Luis Machado  <luis.machado@linaro.org>
12144         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
12145         global.
12146         (qdf24xx_tunings): Set vector cost structure to
12147         qdf24xx_vector_cost.
12149         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
12150         <register_sextend>: Set to 3.
12152 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
12154         PR target/86838
12155         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
12156         * config/aarch64/aarch64-simd.md
12157         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
12158         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
12159         (aarch64_frecpx<mode>): ...this new pattern.
12160         * config/aarch64/aarch64-simd-builtins.def: Remove comment
12161         about aarch64_frecp<FRECP:frecp_suffix><mode>.
12163 2018-08-07  Martin Liska  <mliska@suse.cz>
12165         PR middle-end/83023
12166         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
12167         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
12168         * predict.def (PRED_MALLOC_NONNULL): New predictor.
12169         * doc/extend.texi: Document that malloc attribute adds
12170         hit to compiler.
12172 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
12174         PR target/86785
12175         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12176         Define to speculation_safe_value_not_needed.
12178 2018-08-06  Jeff Law  <law@redhat.com>
12180         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
12181         the vr_values instance to cprop_into_stmt.
12182         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
12183         (cprop_operand): Also query EVRP to determine if OP is a constant.
12185 2018-08-06  Nathan Sidwell  <nathan@acm.org>
12187         * diagnostic.c (diagnostic_report_current_module): Reroll
12188         included-at loop.  Translate text.
12190 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
12192         * function-tests.c (selftest::test_expansion_to_rtl): Call
12193         free_after_compilation.
12195 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12197         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
12199 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
12201         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
12202         loops with memory block operations from getting unrolled.
12204 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
12206         PR target/86807
12207         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12208         Define to speculation_safe_value_not_needed.
12210 2018-08-06  Jeff Law  <law@redhat.com>
12212         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
12213         assert.
12215 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12217         PR target/86662
12218         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
12219         with all enabled __intN types.
12221         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
12223 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12225         * alias.c (record_set): Check for clobber high.
12226         * cfgexpand.c (expand_gimple_stmt): Likewise.
12227         * combine-stack-adj.c (single_set_for_csa): Likewise.
12228         * combine.c (find_single_use_1): Likewise.
12229         (set_nonzero_bits_and_sign_copies): Likewise.
12230         (get_combine_src_dest): Likewise.
12231         (is_parallel_of_n_reg_sets): Likewise.
12232         (try_combine): Likewise.
12233         (record_dead_and_set_regs_1): Likewise.
12234         (reg_dead_at_p_1): Likewise.
12235         (reg_dead_at_p): Likewise.
12236         * dce.c (deletable_insn_p): Likewise.
12237         (mark_nonreg_stores_1): Likewise.
12238         (mark_nonreg_stores_2): Likewise.
12239         * df-scan.c (df_find_hard_reg_defs): Likewise.
12240         (df_uses_record): Likewise.
12241         (df_get_call_refs): Likewise.
12242         * dwarf2out.c (mem_loc_descriptor): Likewise.
12243         * haifa-sched.c (haifa_classify_rtx): Likewise.
12244         * ira-build.c (create_insn_allocnos): Likewise.
12245         * ira-costs.c (scan_one_insn): Likewise.
12246         * ira.c (equiv_init_movable_p): Likewise.
12247         (rtx_moveable_p): Likewise.
12248         (interesting_dest_for_shprep): Likewise.
12249         * jump.c (mark_jump_label_1): Likewise.
12250         * postreload-gcse.c (record_opr_changes): Likewise.
12251         * postreload.c (reload_cse_simplify): Likewise.
12252         (struct reg_use): Add source expr.
12253         (reload_combine): Check for clobber high.
12254         (reload_combine_note_use): Likewise.
12255         (reload_cse_move2add): Likewise.
12256         (move2add_note_store): Likewise.
12257         * print-rtl.c (print_pattern): Likewise.
12258         * recog.c (decode_asm_operands): Likewise.
12259         (store_data_bypass_p): Likewise.
12260         (if_test_bypass_p): Likewise.
12261         * regcprop.c (kill_clobbered_value): Likewise.
12262         (kill_set_value): Likewise.
12263         * reginfo.c (reg_scan_mark_refs): Likewise.
12264         * reload1.c (maybe_fix_stack_asms): Likewise.
12265         (eliminate_regs_1): Likewise.
12266         (elimination_effects): Likewise.
12267         (mark_not_eliminable): Likewise.
12268         (scan_paradoxical_subregs): Likewise.
12269         (forget_old_reloads_1): Likewise.
12270         * reorg.c (find_end_label): Likewise.
12271         (try_merge_delay_insns): Likewise.
12272         (redundant_insn): Likewise.
12273         (own_thread_p): Likewise.
12274         (fill_simple_delay_slots): Likewise.
12275         (fill_slots_from_thread): Likewise.
12276         (dbr_schedule): Likewise.
12277         * resource.c (update_live_status): Likewise.
12278         (mark_referenced_resources): Likewise.
12279         (mark_set_resources): Likewise.
12280         * rtl.c (copy_rtx): Likewise.
12281         * rtlanal.c (reg_referenced_p): Likewise.
12282         (single_set_2): Likewise.
12283         (noop_move_p): Likewise.
12284         (note_stores): Likewise.
12285         * sched-deps.c (sched_analyze_reg): Likewise.
12286         (sched_analyze_insn): Likewise.
12288 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12290         * cse.c (invalidate_reg): New function extracted from...
12291         (invalidate): ...here.
12292         (canonicalize_insn): Check for clobber high.
12293         (invalidate_from_clobbers): invalidate clobber highs.
12294         (invalidate_from_sets_and_clobbers): Likewise.
12295         (count_reg_usage): Check for clobber high.
12296         (insn_live_p): Likewise.
12297         * cselib.c (cselib_expand_value_rtx_1):Likewise.
12298         (cselib_invalidate_regno): Check for clobber in setter.
12299         (cselib_invalidate_rtx): Pass through setter.
12300         (cselib_invalidate_rtx_note_stores):
12301         (cselib_process_insn): Check for clobber high.
12302         * cselib.h (cselib_invalidate_rtx): Add operand.
12304 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12306         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
12307         (mark_not_eliminable): Likewise.
12308         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
12309         * lra-lives.c (process_bb_lives): Check for clobber high.
12310         * lra.c (new_insn_reg): Remember clobber highs.
12311         (collect_non_operand_hard_regs): Check for clobber high.
12312         (lra_set_insn_recog_data): Likewise.
12313         (add_regs_to_insn_regno_info): Likewise.
12314         (lra_update_insn_regno_info): Likewise.
12316 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12318         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
12319         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
12321 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12323         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
12324         (copy_insn_1): Likewise.
12325         (gen_hard_reg_clobber_high): New gen function.
12326         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
12327         * genemit.c (gen_exp): Likewise.
12328         (gen_emit_seq): Pass through info.
12329         (gen_insn): Check for CLOBBER_HIGH.
12330         (gen_expand): Pass through info.
12331         (gen_split): Likewise.
12332         (output_add_clobbers): Likewise.
12333         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
12334         (remove_clobbers): Likewise.
12335         * rtl.h (gen_hard_reg_clobber_high): New declaration.
12337 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
12339         * doc/rtl.texi (clobber_high): Add.
12340         (parallel): Add in clobber high
12341         * rtl.c (rtl_check_failed_code3): Add function.
12342         * rtl.def (CLOBBER_HIGH): Add expression.
12343         * rtl.h (RTL_CHECKC3): Add macro.
12344         (rtl_check_failed_code3): Add declaration.
12345         (XC3EXP): Add macro.
12347 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
12349         PR target/86386
12350         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
12351         cfun->machine->max_used_stack_alignment if needed.
12353 2018-08-04  Martin Sebor  <msebor@redhat.com>
12355         PR tree-optimization/86571
12356         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
12357         NaN output to 4.
12359 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
12361         PR target/86799
12362         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12363         Define.
12365 2018-08-03  Jeff Law  <law@redhat.com>
12367         PR target/86795
12368         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12369         Define to speculation_safe_value_not_needed.
12371 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
12373         * doc/gcov.texi (-x): Remove duplicate "to".
12374         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
12375         (-Wif-not-aligned): Remove duplicate "is".
12376         (-flto): Remove duplicate "the".
12377         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
12378         duplicate "v5.00.b".
12379         (MSP430 Options): Remove duplicate "and" from the description
12380         of "-mgprel-sec=regexp".
12381         (x86 Options): Remove duplicate copies of "vmldLog102" and
12382         vmlsLog104 from description of "-mveclibabi=type".
12384 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
12386         * internal-fn.h (first_commutative_argument): Declare.
12387         * internal-fn.c (first_commutative_argument): New function.
12388         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
12389         restrictions for pattern statements.  Use first_commutative_argument
12390         to look for commutative operands in calls to internal functions.
12392 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
12394         * Makefile.in (wide-int-range.o): New.
12395         * tree-vrp.c: Move all the wide_int_* functions to...
12396         * wide-int-range.cc: ...here.
12397         * tree-vrp.h: Move all the wide_int_* prototypes to...
12398         * wide-int-range.h: ...here.
12400 2018-08-03  Tom de Vries  <tdevries@suse.de>
12402         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
12403         UI_NONE.
12404         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
12405         * except.c (output_function_exception_table): Do early exit if
12406         targetm_common.except_unwind_info (&global_options) == UI_NONE.
12408 2018-08-03  Martin Liska  <mliska@suse.cz>
12410         * predict.c (dump_prediction): Change to 2 digits
12411         in fraction part.
12413 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
12415         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
12416         neon_dup_q to...
12417         (falkor_am_1_gtov_gtov): ... a new insn reservation.
12419 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
12421         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
12422         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
12423         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
12424         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
12425         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
12427 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
12429         * diagnostic-show-locus.c (diagnostic_show_locus): Use
12430         pp_take_prefix when saving the existing prefix.
12431         * diagnostic.c (diagnostic_append_note): Likewise.
12432         * langhooks.c (lhd_print_error_function): Likewise.
12433         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
12434         param's type.  Free the existing prefix.
12435         (pp_take_prefix): New function.
12436         (pretty_printer::pretty_printer): Drop the prefix parameter.
12437         Rename the length parameter to match the comment.
12438         (pretty_printer::~pretty_printer): Free the prefix.
12439         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
12440         parameter.
12441         (struct pretty_printer): Drop the "const" from "prefix" field's
12442         type and clarify memory management.
12443         (pp_set_prefix): Drop the "const" from the 2nd param.
12444         (pp_take_prefix): New decl.
12446 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
12448         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
12449         for word_mode_ok here instead of passing as argument.
12450         (expand_block_compare): Change select_block_compare_mode() call.
12451         (expand_strncmp_gpr_sequence): New function.
12452         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
12454 2018-08-02  Jeff Law  <law@redhat.com>
12456         PR target/86790
12457         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12458         Define to speculation_safe_value_not_needed.
12460         PR target/86784
12461         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12462         Define to speculation_safe_value_not_needed.
12464 2018-08-02  Tom de Vries  <tdevries@suse.de>
12466         PR target/86660
12467         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
12468         function.  Return UI_TARGET unconditionally.
12469         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
12470         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
12472 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
12474         * genemit.c (print_overload_test): Fix typo.
12476 2018-08-02  Richard Biener  <rguenther@suse.de>
12478         PR tree-optimization/86816
12479         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
12480         which checks for value availability before querying it.
12481         (gvn_uses_equal): Use it.
12482         (same_succ_hash): Likewise.
12483         (gimple_equal_p): Likewise.
12485 2018-08-02  Nick Clifton  <nickc@redhat.com>
12487         PR target/86813
12488         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12489         Define to speculation_safe_value_not_needed.
12491         PR target/86810
12492         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12493         Define to speculation_safe_value_not_needed.
12495         PR target/86810
12496         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12497         Define to speculation_safe_value_not_needed.
12499         PR target/86803
12500         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12501         Define to speculation_safe_value_not_needed.
12503         PR target/86797
12504         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12505         Define to speculation_safe_value_not_needed.
12507         PR target/86791
12508         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12509         Define to speculation_safe_value_not_needed.
12511         PR target/86789
12512         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12513         Define to speculation_safe_value_not_needed.
12515         PR target/86787
12516         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
12517         Define to speculation_safe_value_not_needed.
12519         PR target/86782
12520         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
12521         speculation_safe_value_not_needed.
12523         PR target/86781
12524         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
12525         to speculation_safe_value_not_needed.
12527 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
12529         * doc/md.texi: Expand the documentation of instruction names
12530         to mention port-local uses.  Document '@' in pattern names.
12531         * read-md.h (overloaded_instance, overloaded_name): New structs.
12532         (mapping): Declare.
12533         (md_reader::handle_overloaded_name): New member function.
12534         (md_reader::get_overloads): Likewise.
12535         (md_reader::m_first_overload): New member variable.
12536         (md_reader::m_next_overload_ptr): Likewise.
12537         (md_reader::m_overloads_htab): Likewise.
12538         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
12539         m_next_overload_ptr and m_overloads_htab.
12540         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
12541         (get_mode_token, get_code_token, get_int_token): New functions.
12542         (map_attr_string): Add an optional argument that passes back
12543         the associated iterator.
12544         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
12545         (md_reader::handle_overloaded_name, add_overload_instance): New
12546         functions.
12547         (apply_iterators): Handle '@' names.  Report an error if '@'
12548         is used without iterators.
12549         (initialize_iterators): Initialize the new iterator_group fields.
12550         * genopinit.c (handle_overloaded_code_for)
12551         (handle_overloaded_gen): New functions.
12552         (main): Use them to print declarations of maybe_code_for_* and
12553         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
12554         * genemit.c (print_overload_arguments, print_overload_test)
12555         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
12556         (main): Use it to print definitions of maybe_code_for_* and
12557         maybe_gen_* functions.
12558         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
12559         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
12560         instead of explicit mode checks.
12561         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
12562         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
12563         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
12564         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
12565         (aarch64_expand_compare_and_swap): Likewise
12566         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
12567         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
12568         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
12569         (aarch64_constant_pool_reload_icode): Delete.
12570         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
12571         instead of aarch64_constant_pool_reload_icode.  Use
12572         code_for_aarch64_reload_mov instead of explicit mode checks.
12573         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
12574         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
12575         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
12576         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
12577         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
12578         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
12579         (aarch64_atomic_load_op_code): Delete.
12580         (aarch64_emit_atomic_load_op): Likewise.
12581         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
12582         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
12583         instead of aarch64_emit_atomic_load_op.
12584         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
12585         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
12586         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
12587         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
12588         character before the pattern name.
12589         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
12590         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
12591         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
12592         (aarch64_frecps<mode>): Likewise.
12593         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
12594         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
12595         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
12596         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
12597         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
12599 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
12601         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
12602         Allow HFmode constants if TARGET_FP_F16INST.
12604 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
12606         PR target/86014
12607         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
12608         No longer check last store for clobber of address register.
12610 2018-08-02  Martin Liska  <mliska@suse.cz>
12612         PR gcov-profile/86817
12613         * gcov.c (process_all_functions): New function.
12614         (main): Call it.
12615         (process_file): Move functions processing to
12616         process_all_functions.
12618 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
12620         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
12621         "const" to the "gimple *" and "rtx_insn *" parameters.
12622         * dumpfile.h (dump_user_location_t::dump_user_location_t):
12623         Likewise.
12624         (dump_location_t::dump_location_t): Likewise.
12626 2018-08-01  Martin Sebor  <msebor@redhat.com>
12628         PR tree-optimization/86650
12629         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
12630         rather than a "gcall *".  Directly pass the data of interest
12631         to percent_K_format, rather than building a temporary CALL_EXPR
12632         to hold it.
12633         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
12634         (gimple_fold_builtin_strncat): Adjust.
12635         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
12636         gcall* argument with gimple*.
12637         * gimple-ssa-warn-restrict.c (check_call): Same.
12638         (wrestrict_dom_walker::before_dom_children): Same.
12639         (builtin_access::builtin_access): Same.
12640         (check_bounds_or_overlap): Same
12641         (maybe_diag_overlap): Same.
12642         (maybe_diag_offset_bounds): Same.
12643         * tree-diagnostic.c (default_tree_printer): Move usage of
12644         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
12645         to this callsite.
12646         * tree-pretty-print.c (percent_K_format): Add argument.
12647         * tree-pretty-print.h: Add argument.
12648         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
12649         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
12650         (maybe_diag_stxncpy_trunc): Same.
12651         (handle_builtin_stxncpy): Same.
12652         (handle_builtin_strcat): Same.
12654 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12656         * match.pd: Optimise pointer range checks.
12658 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12660         PR tree-optimization/86758
12661         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
12662         to remove pattern statements.
12664 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12666         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
12667         result of dfs_enumerate_from when constructing stmt_vec_infos,
12668         instead of additionally calling get_loop_body.
12670 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12672         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
12673         parameter.
12674         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
12675         When creating an iv, assert that the step is not known to be zero.
12676         (vect_setup_realignment): Update call accordingly.
12677         * tree-vect-stmts.c (vectorizable_store): Likewise.
12678         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
12680 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12682         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
12683         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
12684         (vectorizable_reduction): Likewise.
12685         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12686         (vect_detect_hybrid_slp_stmts): Likewise.
12687         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12689 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
12691         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
12692         (wide_int_set_zero_nonzero_bits): ...this.
12693         (zero_nonzero_bits_from_vr): Rename to...
12694         (vrp_set_zero_nonzero_bits): ...this.
12695         (extract_range_from_multiplicative_op_1): Abstract wide int
12696         code...
12697         (wide_int_range_multiplicative_op): ...here.
12698         (extract_range_from_binary_expr_1): Extract wide int binary
12699         operations into their own functions.
12700         (wide_int_range_lshift): New.
12701         (wide_int_range_can_optimize_bit_op): New.
12702         (wide_int_range_shift_undefined_p): New.
12703         (wide_int_range_bit_xor): New.
12704         (wide_int_range_bit_ior): New.
12705         (wide_int_range_bit_and): New.
12706         (wide_int_range_trunc_mod): New.
12707         (extract_range_into_wide_ints): New.
12708         (vrp_shift_undefined_p): New.
12709         (extract_range_from_multiplicative_op): New.
12710         (vrp_can_optimize_bit_op): New.
12711         * tree-vrp.h (value_range::dump): New.
12712         (wide_int_range_multiplicative_op): New.
12713         (wide_int_range_lshift):New.
12714         (wide_int_range_shift_undefined_p): New.
12715         (wide_int_range_bit_xor): New.
12716         (wide_int_range_bit_ior): New.
12717         (wide_int_range_bit_and): New.
12718         (wide_int_range_trunc_mod): New.
12719         (zero_nonzero_bits_from_bounds): Rename to...
12720         (wide_int_set_zero_nonzero_bits): ...this.
12721         (zero_nonzero_bits_from_vr): Rename to...
12722         (vrp_set_zero_nonzero_bits): ...this.
12723         (range_easy_mask_min_max): Rename to...
12724         (wide_int_range_can_optimize_bit_op): this.
12725         * vr-values.c (simplify_bit_ops_using_ranges): Rename
12726         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
12728 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12730         * tree-vectorizer.h (vect_orig_stmt): New function.
12731         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
12732         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
12733         (vect_create_epilog_for_reduction): Likewise.
12734         (vectorizable_live_operation): Likewise.
12735         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
12736         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
12737         * tree-vect-stmts.c (vectorizable_call): Likewise.
12738         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
12740 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12742         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
12743         argument.
12744         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
12745         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
12746         (vect_transform_loop): Likewise.
12747         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
12749 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12751         * tree-vectorizer.h (vect_schedule_slp): Return void.
12752         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
12753         (vect_schedule_slp): Likewise.
12755 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12757         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
12758         argument.
12759         (vect_transform_loop): Update calls accordingly.  Schedule SLP
12760         instances before the main loop, if any exist.
12762 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
12764         PR tree-optimization/86749
12765         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
12766         If the lhs is used in a COND_EXPR, check that it is being used
12767         as the "then" or "else" value.
12769 2018-08-01  Tom de Vries  <tdevries@suse.de>
12771         PR target/86800
12772         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
12773         speculation_safe_value_not_needed.
12775 2018-08-01  Richard Biener  <rguenther@suse.de>
12777         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
12778         as base and offset.
12780 2018-08-01  Martin Liska  <mliska@suse.cz>
12782         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
12783         format how successful transformation is dumped.
12784         (gimple_mod_pow2_value_transform): Likewise.
12785         (gimple_mod_subtract_transform): Likewise.
12786         (gimple_stringops_transform): Likewise.
12788 2018-08-01  Martin Liska  <mliska@suse.cz>
12790         PR value-prof/35543
12791         * value-prof.c (interesting_stringop_to_profile_p):
12792         Simplify the code and add BUILT_IN_MEMMOVE.
12793         (gimple_stringops_transform): Likewise.
12795 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
12797         * config/aarch64/aarch64-simd.md
12798         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
12799         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
12800         use GPI iterator instead of SI mode.
12802 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
12804         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
12805         rs6000_speculation_barrier.
12806         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
12807         new barrier pattern name.
12809 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
12811         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
12812         (speculation_barrier): New insn.
12814 2018-08-01  Richard Biener  <rguenther@suse.de>
12816         PR bootstrap/86724
12817         * graphite.h: Include isl/id.h and isl/space.h to allow build
12818         with ISL 0.20.
12820 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
12822         PR target/86651
12823         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12824         mode for COFF targets.
12825         * defaults.h (TARGET_COFF): Define.
12826         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
12827         TARGET_COFF): Define.
12828         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
12829         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
12830         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
12832 2018-07-31  Alexandre Oliva <oliva@adacore.com>
12834         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
12835         * gimple-streamer-out.c (output_bb): Save it.
12836         * lto-streamer-in.c (input_struct_function_base): Restore
12837         instance discriminator if available.  Create map on demand.
12838         * lto-streamer-out.c (output_struct_function_base): Save it if
12839         available.
12840         * final.c (decl_to_instance_map): Document LTO strategy.
12842 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
12843             Olivier Hainque  <hainque@adacore.com>
12845         * debug.h (decl_to_instance_map_t): New type.
12846         (decl_to_instance_map): Declare.
12847         (maybe_create_decl_to_instance_map): New inline function.
12848         * final.c (bb_discriminator, last_bb_discriminator): New statics,
12849         to track basic block discriminators.
12850         (final_start_function_1): Initialize them.
12851         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
12852         bb_discriminator.
12853         (decl_to_instance_map): New variable.
12854         (map_decl_to_instance, maybe_set_discriminator): New functions.
12855         (notice_source_line): Set discriminator.
12857 2018-07-31  Ian Lance Taylor  <iant@golang.org>
12859         * targhooks.c (default_have_speculation_safe_value): Add
12860         ATTRIBUTE_UNUSED.
12862 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
12864         * dump-context.h: Include "pretty-print.h".
12865         (dump_context::refresh_dumps_are_enabled): New decl.
12866         (dump_context::emit_item): New decl.
12867         (class dump_context): Add fields "m_test_pp" and
12868         "m_test_pp_flags".
12869         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
12870         (temp_dump_context::get_dumped_text): New decl.
12871         (class temp_dump_context): Add field "m_pp".
12872         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
12873         (dump_context::refresh_dumps_are_enabled): ...and add a test for
12874         m_test_pp.
12875         (set_dump_file): Update for above change.
12876         (set_alt_dump_file): Likewise.
12877         (dump_loc): New overload, taking a pretty_printer *.
12878         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
12879         to any test pretty-printer.
12880         (make_item_for_dump_gimple_stmt): New function, adapted from
12881         optinfo::add_gimple_stmt.
12882         (dump_context::dump_gimple_stmt): Call it, and use the result,
12883         eliminating the direct usage of dump_file and alt_dump_file in
12884         favor of indirectly using them via emit_item.
12885         (make_item_for_dump_gimple_expr): New function, adapted from
12886         optinfo::add_gimple_expr.
12887         (dump_context::dump_gimple_expr): Call it, and use the result,
12888         eliminating the direct usage of dump_file and alt_dump_file in
12889         favor of indirectly using them via emit_item.
12890         (make_item_for_dump_generic_expr): New function, adapted from
12891         optinfo::add_tree.
12892         (dump_context::dump_generic_expr): Call it, and use the result,
12893         eliminating the direct usage of dump_file and alt_dump_file in
12894         favor of indirectly using them via emit_item.
12895         (make_item_for_dump_printf_va): New function, adapted from
12896         optinfo::add_printf_va.
12897         (make_item_for_dump_printf): New function.
12898         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
12899         and use the result, eliminating the direct usage of dump_file and
12900         alt_dump_file in favor of indirectly using them via emit_item.
12901         (make_item_for_dump_dec): New function.
12902         (dump_context::dump_dec): Call it, and use the result,
12903         eliminating the direct usage of dump_file and alt_dump_file in
12904         favor of indirectly using them via emit_item.
12905         (make_item_for_dump_symtab_node): New function, adapted from
12906         optinfo::add_symtab_node.
12907         (dump_context::dump_symtab_node): Call it, and use the result,
12908         eliminating the direct usage of dump_file and alt_dump_file in
12909         favor of indirectly using them via emit_item.
12910         (dump_context::begin_scope): Reimplement, avoiding direct usage
12911         of dump_file and alt_dump_file in favor of indirectly using them
12912         via emit_item.
12913         (dump_context::emit_item): New member function.
12914         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
12915         Set up test pretty-printer on the underlying context.  Call
12916         refresh_dumps_are_enabled.
12917         (temp_dump_context::~temp_dump_context): Call
12918         refresh_dumps_are_enabled.
12919         (temp_dump_context::get_dumped_text): New member function.
12920         (selftest::verify_dumped_text): New function.
12921         (ASSERT_DUMPED_TEXT_EQ): New macro.
12922         (selftest::test_capture_of_dump_calls): Run all tests twice, with
12923         and then without optinfo enabled.  Add uses of
12924         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
12925         * dumpfile.h: Update comment for the dump_* API.
12926         * optinfo-emit-json.cc
12927         (selftest::test_building_json_from_dump_calls): Update for new
12928         param for temp_dump_context ctor.
12929         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
12930         and "m_owned" field.
12931         (optinfo_item::~optinfo_item): Likewise.
12932         (optinfo::add_item): New member function.
12933         (optinfo::emit): Update comment.
12934         (optinfo::add_string): Delete.
12935         (optinfo::add_printf): Delete.
12936         (optinfo::add_printf_va): Delete.
12937         (optinfo::add_gimple_stmt): Delete.
12938         (optinfo::add_gimple_expr): Delete.
12939         (optinfo::add_tree): Delete.
12940         (optinfo::add_symtab_node): Delete.
12941         (optinfo::add_dec): Delete.
12942         * optinfo.h (class dump_context): New forward decl.
12943         (optinfo::add_item): New decl.
12944         (optinfo::add_string): Delete.
12945         (optinfo::add_printf): Delete.
12946         (optinfo::add_printf_va): Delete.
12947         (optinfo::add_gimple_stmt): Delete.
12948         (optinfo::add_gimple_expr): Delete.
12949         (optinfo::add_tree): Delete.
12950         (optinfo::add_symtab_node): Delete.
12951         (optinfo::add_dec): Delete.
12952         (optinfo::add_poly_int): Delete.
12953         (optinfo_item::optinfo_item): Remove "owned" param.
12954         (class optinfo_item): Remove field "m_owned".
12956 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12958         PR middle-end/86705
12959         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
12960         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
12961         requested variable alignment.
12962         (expand_one_ssa_partition): Likewise.
12963         (expand_one_var): Likewise.
12965 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12967         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
12968         to speculation_safe_value_not_needed.
12970 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12972         * targhooks.h (speculation_safe_value_not_needed): New prototype.
12973         * targhooks.c (speculation_safe_value_not_needed): New function.
12974         * target.def (have_speculation_safe_value): Update documentation.
12975         * doc/tm.texi: Regenerated.
12977 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12979         * config/aarch64/iterators.md (ALLI_TI): New iterator.
12980         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
12981         expand.
12982         (despeculate_copy<ALLI:mode>_insn): New insn.
12983         (despeculate_copyti_insn): New insn.
12984         (despeculate_simple<ALLI:mode>): New insn
12985         (despeculate_simpleti): New insn.
12986         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
12987         function.
12988         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
12989         aarch64_speculation_safe_value.
12990         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
12992 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
12994         * config/aarch64/aarch64-speculation.cc: New file.
12995         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
12996         before pass_reorder_blocks.
12997         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
12998         prototype.
12999         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
13000         X14 and X15 when tracking speculation.
13001         * config/aarch64/aarch64.md (register name constants): Add
13002         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
13003         (unspec): Add UNSPEC_SPECULATION_TRACKER.
13004         (speculation_barrier): New insn attribute.
13005         (cmp<mode>): Allow SP in comparisons.
13006         (speculation_tracker): New insn.
13007         (speculation_barrier): Add speculation_barrier attribute.
13008         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
13009         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
13010         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
13012 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13014         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
13015         aarch64_track_speculation is true.
13016         (tb<optab><mode>1): Likewise.
13017         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
13018         generate CB[N]Z when tracking speculation.
13019         (aarch64_split_compare_and_swap): Likewise.
13020         (aarch64_split_atomic_op): Likewise.
13022 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13024         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
13026 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13028         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
13029         (speculation_barrier): New insn.
13031 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13033         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
13034         * config/arm/arm.md (speculation_barrier): New expand.
13035         (speculation_barrier_insn): New pattern.
13037 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
13039         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
13040         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
13041         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
13042         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
13043         list.
13044         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
13045         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
13046         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
13047         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
13048         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
13049         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
13050         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
13051         * builtins.c (expand_speculation_safe_value): New function.
13052         (expand_builtin): Call it.
13053         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
13054         * doc/extend.texi: Document __builtin_speculation_safe_value.
13055         * doc/md.texi: Document "speculation_barrier" pattern.
13056         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
13057         TARGET_HAVE_SPECULATION_SAFE_VALUE.
13058         * doc/tm.texi: Regenerated.
13059         * target.def (have_speculation_safe_value, speculation_safe_value): New
13060         hooks.
13061         * targhooks.c (default_have_speculation_safe_value): New function.
13062         (default_speculation_safe_value): New function.
13063         * targhooks.h (default_have_speculation_safe_value): Add prototype.
13064         (default_speculation_safe_value): Add prototype.
13066 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
13068         * dump-context.h (dump_context::dump_loc): New decl.
13069         * dumpfile.c (dump_context::dump_loc): New member function.
13070         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
13071         and dump_gimple_stmt.
13072         (dump_context::dump_gimple_expr_loc): Likewise, using
13073         dump_gimple_expr.
13074         (dump_context::dump_generic_expr_loc): Likewise, using
13075         dump_generic_expr.
13076         (dump_context::dump_printf_loc_va): Likewise, using
13077         dump_printf_va.
13078         (dump_context::begin_scope): Explicitly using the global function
13079         "dump_loc", rather than the member function.
13081 2018-07-31  Martin Sebor  <msebor@redhat.com>
13083         PR tree-optimization/86741
13084         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
13086 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
13088         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
13089         depend on whether prefetch instructions will be emitted or not.
13090         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
13091         will be emitted or not.
13092         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
13093         (TARGET_SETMEM_PFD): New macros.
13095 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13097         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
13098         (NULL_STMT_VEC_INFO): Delete.
13099         (stmt_vec_info::operator*): Likewise.
13100         (stmt_vec_info::operator gimple *): Likewise.
13101         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
13102         of NULL_STMT_VEC_INFO.
13103         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13104         (vect_reassociating_reduction_p): Likewise.
13105         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
13106         (vectorizable_store): Likewise.
13107         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
13108         (vec_info::free_stmt_vec_infos): Likewise.
13110 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13112         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
13113         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
13115 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13117         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
13118         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
13119         (vec_info::free_stmt_vec_info): New private member functions.
13120         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
13121         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
13122         * tree-parloops.c (gather_scalar_reductions): Remove calls to
13123         set_stmt_vec_info_vec and free_stmt_vec_infos.
13124         * tree-vect-loop.c (_loop_vec_info): Remove call to
13125         set_stmt_vec_info_vec.
13126         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
13127         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
13128         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
13129         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
13130         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
13131         assignments in {vec_info::,}new_stmt_vec_info that are redundant
13132         with the clearing in the xcalloc.
13133         (stmt_vec_info_vec): Delete.
13134         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
13135         (vectorize_loops): Likewise.
13136         (vec_info::~vec_info): Remove argument from call to
13137         free_stmt_vec_infos.
13138         (vec_info::add_stmt): Remove vinfo argument from call to
13139         new_stmt_vec_info.
13141 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13143         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
13144         rather than a gimple stmt.
13145         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
13146         information for pattern statements when passed the original
13147         statement; instead wait to be passed the pattern statement itself.
13148         Don't call set_vinfo_for_stmt here.
13149         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
13150         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
13151         stmt_vec_infos here.
13152         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
13153         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
13154         stmt_vec_infos entry.
13156 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13158         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
13159         * tree-vectorizer.c (vec_info::replace_stmt): New function.
13160         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
13161         * tree-vect-stmts.c (vectorizable_call): Likewise.
13162         (vectorizable_simd_clone_call): Likewise.
13164 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13166         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
13167         * tree-vectorizer.c (vec_info::remove_stmt): New function.
13168         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
13169         * tree-vect-loop.c (vect_transform_loop): Likewise.
13170         * tree-vect-slp.c (vect_schedule_slp): Likewise.
13171         * tree-vect-stmts.c (vect_remove_stores): Likewise.
13173 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13175         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
13176         (vect_dr_stmt): Delete.
13177         * tree-vectorizer.c (vec_info::lookup_dr): New function.
13178         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
13179         of DR_VECT_AUX.
13180         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13181         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
13182         (vect_verify_datarefs_alignment, vect_peeling_supportable)
13183         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
13184         (vect_analyze_data_refs): Likewise.
13185         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
13186         argument.
13187         (vect_find_same_alignment_drs): Likewise.
13188         (vect_slp_analyze_node_dependences): Update calls accordingly.
13189         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
13190         instead of DR_VECT_AUX.
13191         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
13192         of a vector data references.  Use vec_info::lookup_dr instead of
13193         DR_VECT_AUX.
13194         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
13195         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
13196         instead of DR_VECT_AUX.
13198 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13200         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
13201         dr_vec_info.
13202         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
13203         accordingly.
13204         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13205         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13206         (vect_gen_prolog_loop_niters): Likewise.
13208 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13210         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
13211         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
13212         (vect_known_alignment_in_bytes, vect_dr_behavior)
13213         (vect_get_scalar_dr_size): Take references as dr_vec_infos
13214         instead of data_references.  Update calls to other routines for
13215         which the same change has been made.
13216         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
13217         dr_vec_infos instead of stmt_vec_infos.
13218         (vect_analyze_data_ref_dependence): Update call accordingly.
13219         (vect_slp_analyze_data_ref_dependence)
13220         (vect_record_base_alignments): Use DR_VECT_AUX.
13221         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
13222         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
13223         (vector_alignment_reachable_p, vect_get_data_access_cost)
13224         (vect_peeling_supportable, vect_analyze_group_access_1)
13225         (vect_analyze_group_access, vect_analyze_data_ref_access)
13226         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
13227         (vect_compile_time_alias, vect_small_gap_p)
13228         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
13229         (vect_supportable_dr_alignment): Take references as dr_vec_infos
13230         instead of data_references.  Update calls to other routines for
13231         which the same change has been made.
13232         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
13233         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
13234         (vect_slp_analyze_and_verify_node_alignment)
13235         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
13236         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
13237         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
13238         above changes.
13239         (_vect_peel_info::dr): Replace with...
13240         (_vect_peel_info::dr_info): ...this new field.
13241         (vect_peeling_hash_get_most_frequent)
13242         (vect_peeling_hash_choose_best_peeling): Update accordingly.
13243         (vect_peeling_hash_get_lowest_cost):
13244         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
13245         routines for which the same change has been made.
13246         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
13247         data_reference.
13248         * tree-vect-loop-manip.c (get_misalign_in_elems)
13249         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
13250         above changes.
13251         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13252         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
13253         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
13254         (get_group_load_store_type, get_negative_load_store_type)
13255         (vect_get_data_ptr_increment, vectorizable_store)
13256         (vectorizable_load): Likewise.
13257         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
13258         Update calls to other routines for which the same change has been made.
13260 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13262         * tree-vectorizer.h (vec_info::move_dr): New member function.
13263         (dataref_aux): Rename to...
13264         (dr_vec_info): ...this and add "dr" and "stmt" fields.
13265         (_stmt_vec_info::dr_aux): Update accordingly.
13266         (_stmt_vec_info::data_ref_info): Delete.
13267         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
13268         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
13269         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
13270         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
13271         of data_ref.
13272         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
13273         (STMT_VINFO_DR_INFO): New macro.
13274         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
13275         (set_dr_misalignment): Update after rename of dataref_aux.
13276         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
13277         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
13278         initialization of STMT_VINFO_DATA_REF.
13279         * tree-vectorizer.c (vec_info::move_dr): New function.
13280         * tree-vect-patterns.c (vect_recog_bool_pattern)
13281         (vect_recog_mask_conversion_pattern)
13282         (vect_recog_gather_scatter_pattern): Use it.
13283         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
13284         the "dr" and "stmt" fields of dr_vec_info instead of
13285         STMT_VINFO_DATA_REF.
13287 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13289         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
13290         (is_pattern_stmt_p): Use it.
13291         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
13292         on pattern statements.
13294 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13296         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
13297         original stmt as a stmt_vec_info rather than a gimple stmt.
13298         (vect_pattern_recog_1): Take the statement directly as a
13299         stmt_vec_info, rather than via a gimple_stmt_iterator.
13300         Update call to vect_mark_pattern_stmts.
13301         (vect_pattern_recog): Update calls accordingly.
13303 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13305         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
13306         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
13307         a vect_def_type for the first argument.
13308         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
13309         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
13310         operand if it isn't defined by a vectorized statement.
13311         (vect_build_gather_load_calls): Remove the mask_dt argument and
13312         update calls to vect_get_vec_def_for_stmt_copy.
13313         (vectorizable_bswap): Likewise the dt argument.
13314         (vectorizable_call): Update calls to vectorizable_bswap and
13315         vect_get_vec_def_for_stmt_copy.
13316         (vectorizable_simd_clone_call, vectorizable_assignment)
13317         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
13318         (vectorizable_comparison): Update calls to
13319         vect_get_vec_def_for_stmt_copy.
13320         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
13321         vect_is_simple_use.
13322         (vect_get_loop_based_defs): Remove dt argument and update call
13323         to vect_get_vec_def_for_stmt_copy.
13324         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
13325         and vect_get_vec_def_for_stmt_copy.
13326         (vectorizable_load): Update calls to vect_build_gather_load_calls
13327         and vect_get_vec_def_for_stmt_copy.
13328         * tree-vect-loop.c (vect_create_epilog_for_reduction)
13329         (vectorizable_reduction, vectorizable_live_operation): Update calls
13330         to vect_get_vec_def_for_stmt_copy.
13332 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13334         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
13335         and gimple stmt arguments with a stmt_vec_info.
13336         (vect_record_base_alignments): Update calls accordingly.
13337         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
13338         and gimple stmt arguments with a stmt_vec_info.
13339         (vect_build_slp_tree_1): Remove vinfo argument and update call
13340         to vect_record_max_nunits.
13341         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
13342         and vect_record_max_nunits.
13344 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13346         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
13347         file and take a stmt_vec_info instead of a gimple stmt.
13348         (supportable_widening_operation, vect_finish_replace_stmt)
13349         (vect_finish_stmt_generation, vect_get_store_rhs)
13350         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
13351         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
13352         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
13353         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
13354         (vect_create_data_ref_ptr, bump_vector_ptr)
13355         (vect_permute_store_chain, vect_setup_realignment)
13356         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
13357         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
13358         (vectorizable_reduction, vectorizable_induction)
13359         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
13360         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
13361         than gimple stmts as arguments.
13362         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
13363         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
13364         (can_group_stmts_p, vect_check_gather_scatter)
13365         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
13366         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
13367         (vect_permute_load_chain, vect_shift_permute_load_chain)
13368         (vect_transform_grouped_load)
13369         (vect_record_grouped_load_vectors): Likewise.
13370         * tree-vect-loop.c (vect_fixup_reduc_chain)
13371         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
13372         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
13373         (vectorizable_reduction, vectorizable_induction)
13374         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
13375         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
13376         (vect_get_load_store_mask): Likewise.
13377         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
13378         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
13379         * tree-vect-stmts.c (vect_mark_relevant)
13380         (is_simple_and_all_uses_invariant)
13381         (exist_non_indexing_operands_for_use_p, process_use)
13382         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
13383         (vect_get_vec_def_for_operand, vect_get_vec_defs)
13384         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
13385         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
13386         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
13387         (get_negative_load_store_type, get_load_store_type)
13388         (vect_check_load_store_mask, vect_check_store_rhs)
13389         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
13390         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
13391         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
13392         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13393         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
13394         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
13395         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
13396         (supportable_widening_operation): Likewise.
13398 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13400         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
13401         a stmt_vec_info instead of a gcall.
13402         (vect_check_gather_scatter): Update call accordingly.
13403         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
13404         of a gphi.
13405         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
13406         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
13407         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
13408         instead of a gimple stmt.
13409         (vect_transform_loop): Update calls accordingly.
13410         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
13411         stmt_vec_infos instead of gimple stmts.
13412         (vect_analyze_slp_instance): Update use accordingly.
13413         * tree-vect-stmts.c (read_vector_array, write_vector_array)
13414         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
13415         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
13416         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
13417         (vect_gen_widened_results_half, vect_get_loop_based_defs)
13418         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
13419         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
13420         down to subroutines.
13422 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13424         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
13425         of the worklist from a vector of gimple stmts to a vector of
13426         stmt_vec_infos.
13427         * tree-vect-stmts.c (vect_mark_relevant, process_use)
13428         (vect_mark_stmts_to_be_vectorized): Likewise
13430 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13432         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
13433         statement before passing it to vect_analyze_stmt.
13434         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
13435         the chain of phi vector definitions.  Track the exit phi via its
13436         stmt_vec_info.
13437         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
13438         STMT_VINFO_REDUC_DEF.
13439         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
13440         stmt_vec_infos to handle the statement chains.
13441         (vect_get_slp_defs): Record the first statement in the node
13442         using a stmt_vec_info.
13443         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
13444         statements here and pass their stmt_vec_info down to subroutines.
13445         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
13446         down to vect_finish_stmt_generation.
13447         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
13448         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
13449         stmt_vec_infos to subroutines.
13450         (vect_remove_stores): Use stmt_vec_infos to handle the statement
13451         chains.
13453 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13455         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
13456         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
13457         (vect_permute_store_chain, vect_setup_realignment)
13458         (vect_permute_load_chain, vect_shift_permute_load_chain)
13459         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
13460         stmts internally, and when passing values to other vectorizer routines.
13461         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
13462         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
13463         (vect_analyze_loop_operations, get_initial_def_for_reduction)
13464         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
13465         (vectorizable_reduction, vectorizable_induction)
13466         (vectorizable_live_operation, vect_transform_loop_stmt)
13467         (vect_transform_loop): Likewise.
13468         * tree-vect-patterns.c (vect_reassociating_reduction_p)
13469         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
13470         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
13471         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13472         (vect_slp_analyze_node_operations_1): Likewise.
13473         * tree-vect-stmts.c (vect_mark_relevant, process_use)
13474         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
13475         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
13476         (vect_finish_stmt_generation_1, get_group_load_store_type)
13477         (get_load_store_type, vect_build_gather_load_calls)
13478         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
13479         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
13480         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13481         (vectorizable_store, vectorizable_load, vectorizable_condition)
13482         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
13483         (supportable_widening_operation): Likewise.
13484         (vect_get_vector_types_for_stmt): Likewise.
13485         * tree-vectorizer.h (vect_dr_behavior): Likewise.
13487 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13489         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13490         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
13491         (vect_permute_store_chain, vect_permute_load_chain)
13492         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
13493         repeated stmt_vec_info lookups.
13494         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
13495         (vect_update_ivs_after_vectorizer): Likewise.
13496         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
13497         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
13498         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
13499         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13500         (vect_bb_slp_scalar_cost): Likewise.
13501         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
13503 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13505         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
13506         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
13507         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
13508         to get gassigns and gcalls, rather than operating on generc gimple
13509         stmts.
13510         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
13511         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
13512         (vectorizable_load, vect_analyze_stmt): Likewise.
13513         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
13515 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13517         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
13518         return stmt_vec_infos rather than gimple stmts.  Do not accept
13519         null arguments.
13520         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
13521         of a gimple stmt.
13522         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
13523         Update use of get_later_stmt.
13524         (vect_get_constant_vectors): Update call accordingly.
13525         (vect_schedule_slp_instance): Likewise
13526         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
13527         (vect_slp_analyze_instance_dependence): Likewise.
13528         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
13530 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13532         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
13533         (stmt_info_for_cost::stmt_info): ...this new field.
13534         (add_stmt_costs): Update accordingly.
13535         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
13536         (vect_get_known_peeling_cost): Likewise.
13537         (vect_estimate_min_profitable_iters): Likewise.
13538         * tree-vect-stmts.c (record_stmt_cost): Likewise.
13540 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13542         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
13543         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
13544         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
13545         accordingly.
13546         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
13548 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13550         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
13551         a gimple stmt to a stmt_vec_info.
13552         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
13554 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13556         * tree-vectorizer.h (vec_info::grouped_stores): Change from
13557         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
13558         (_loop_vec_info::reduction_chains): Likewise.
13559         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
13560         accordingly.
13561         * tree-vect-slp.c (vect_analyze_slp): Likewise.
13563 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13565         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
13566         a gimple stmt to a stmt_vec_info.
13567         (_stmt_vec_info::next_element): Likewise.
13568         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
13569         (vect_slp_analyze_and_verify_node_alignment)
13570         (vect_analyze_group_access_1, vect_analyze_group_access)
13571         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
13572         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
13573         (vect_supportable_dr_alignment): Update accordingly.
13574         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
13575         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
13576         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
13577         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13578         * tree-vect-slp.c (vect_build_slp_tree_1)
13579         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
13580         (vect_split_slp_store_group, vect_analyze_slp_instance)
13581         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
13582         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
13583         (get_group_load_store_type, get_load_store_type)
13584         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
13585         (vect_transform_stmt, vect_remove_stores): Likewise.
13587 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13589         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
13590         than a gimple stmt.
13591         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13592         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
13593         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
13594         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
13595         (vector_alignment_reachable_p, vect_get_data_access_cost)
13596         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
13597         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
13598         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
13599         (vect_analyze_group_access_1, vect_analyze_group_access)
13600         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
13601         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
13602         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
13603         result of vect_dr_stmt and use the stmt_vec_info instead of
13604         the associated gimple stmt.
13605         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13606         (vect_gen_prolog_loop_niters): Likewise.
13607         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13609 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13611         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
13612         to a vec<stmt_vec_info>.
13613         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
13614         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
13615         vec<stmt_vec_info>.
13616         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
13617         to a vec<stmt_vec_info>.
13618         (bst_traits::value_type, bst_traits::value_type): Likewise.
13619         (bst_traits::hash): Update accordingly.
13620         (vect_get_and_check_slp_defs): Change the stmts parameter from
13621         a vec<gimple *> to a vec<stmt_vec_info>.
13622         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
13623         (vect_build_slp_tree): Likewise.
13624         (vect_build_slp_tree_2): Likewise.  Update uses of
13625         SLP_TREE_SCALAR_STMTS.
13626         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
13627         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
13628         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
13629         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
13630         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
13631         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
13632         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
13633         (vect_get_constant_vectors, vect_get_slp_defs)
13634         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
13635         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
13636         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
13637         instead of gimple stmts.
13638         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
13639         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
13640         (vect_slp_analyze_instance_dependence): Update uses of
13641         SLP_TREE_SCALAR_STMTS.
13642         (vect_slp_analyze_and_verify_node_alignment): Likewise.
13643         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
13644         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
13645         (get_initial_defs_for_reduction): Likewise.
13646         (vect_create_epilog_for_reduction): Likewise.
13647         (vectorize_fold_left_reduction): Likewise.
13648         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
13649         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
13650         (can_vectorize_live_stmts): Likewise.
13652 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13654         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
13655         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
13656         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
13657         than gimple stmts.
13658         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
13659         of a gimple stmt.
13660         (gather_scalar_reductions): Update after above interface changes.
13661         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
13662         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
13663         than gimple stmts.
13664         (vect_force_simple_reduction): Likewise.
13665         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
13666         LOOP_VINFO_REDUCTIONS.
13667         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13669 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13671         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
13672         a gimple stmt to a stmt_vec_info.
13673         * tree-vect-loop.c (vect_active_double_reduction_p)
13674         (vect_force_simple_reduction, vectorizable_reduction): Update
13675         accordingly.
13677 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13679         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
13680         vec<gimple *> to a vec<stmt_vec_info>.
13681         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
13682         the reduction_phis argument from a vec<gimple *> to a
13683         vec<stmt_vec_info>.
13684         (vectorizable_reduction): Likewise the phis local variable that
13685         is passed to vect_create_epilog_for_reduction.  Update for new type
13686         of SLP_TREE_VEC_STMTS.
13687         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
13688         (vectorizable_live_operation): Likewise.
13689         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
13690         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
13692 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13694         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
13695         a gimple stmt to a stmt_vec_info.
13696         (vectorizable_condition, vectorizable_live_operation)
13697         (vectorizable_reduction, vectorizable_induction): Pass back the
13698         vectorized statement as a stmt_vec_info.
13699         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
13700         use of STMT_VINFO_VEC_STMT.
13701         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
13702         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
13703         as stmt_vec_infos rather than gimple stmts.
13704         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
13705         to a stmt_vec_info.
13706         (vectorizable_live_operation): Likewise.
13707         (vectorizable_reduction, vectorizable_induction): Likewise,
13708         updating use of STMT_VINFO_VEC_STMT.
13709         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
13710         of STMT_VINFO_VEC_STMT.
13711         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
13712         (vectorizable_simd_clone_call, vectorizable_conversion)
13713         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13714         (vectorizable_store, vectorizable_load, vectorizable_condition)
13715         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
13716         from a gimple stmt to a stmt_vec_info.
13717         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
13718         pointer to a stmt_vec_info to the vectorizable_* routines.
13720 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13722         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
13723         a gimple stmt to a stmt_vec_info.
13724         (is_pattern_stmt_p): Update accordingly.
13725         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
13726         (vect_record_grouped_load_vectors): Likewise.
13727         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
13728         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
13729         (vect_model_reduction_cost): Likewise.
13730         (vect_create_epilog_for_reduction): Likewise.
13731         (vectorizable_reduction, vectorizable_induction): Likewise.
13732         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13733         Return the stmt_vec_info for the pattern statement.
13734         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
13735         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
13736         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
13737         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
13738         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
13739         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
13740         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
13741         (free_stmt_vec_info, vect_is_simple_use): Likewise.
13743 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13745         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
13746         (vect_finish_stmt_generation): Likewise.
13747         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
13748         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
13749         (vect_build_gather_load_calls): Use the return value of the above
13750         functions instead of a separate call to vinfo_for_stmt.  Use narrow
13751         scopes for the input gimple stmt and wider scopes for the associated
13752         stmt_vec_info.  Use vec_info::lookup_def when setting these
13753         stmt_vec_infos from an SSA_NAME definition.
13754         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
13755         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
13756         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
13757         (vectorizable_store, vectorizable_load, vectorizable_condition)
13758         (vectorizable_comparison): Likewise.
13759         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
13760         (vectorizable_reduction): Likewise.
13762 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13764         * tree-vectorizer.h (vect_is_simple_use): Add an optional
13765         stmt_vec_info * parameter before the optional gimple **.
13766         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13767         (process_use, vect_get_vec_def_for_operand_1): Update callers.
13768         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
13769         * tree-vect-loop.c (vectorizable_reduction): Likewise.
13770         (vectorizable_live_operation): Likewise.
13771         * tree-vect-patterns.c (type_conversion_p): Likewise.
13772         (vect_look_through_possible_promotion): Likewise.
13773         (vect_recog_rotate_pattern): Likewise.
13774         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13778         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
13779         a typedef to a wrapper class.
13780         (NULL_STMT_VEC_INFO): New macro.
13781         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
13782         (stmt_vec_info::operator*): New function.
13783         (stmt_vec_info::operator gimple *): Likewise.
13784         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
13785         (add_stmt_costs): Likewise.
13786         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
13787         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
13788         (vect_get_known_peeling_cost): Likewise.
13789         (vect_estimate_min_profitable_iters): Likewise.
13790         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13791         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
13792         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
13793         (vectorizable_store, free_stmt_vec_infos): Likewise.
13794         (new_stmt_vec_info): Change return type of xcalloc to
13795         _stmt_vec_info *.
13797 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13799         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
13800         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
13801         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
13802         a single_imm_use-based sequence.
13803         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
13805 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13807         * tree-vectorizer.h (vec_info::lookup_def): Declare.
13808         * tree-vectorizer.c (vec_info::lookup_def): New function.
13809         * tree-vect-patterns.c (vect_get_internal_def): Use it.
13810         (vect_widened_op_tree): Likewise.
13811         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13812         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
13813         (vectorizable_reduction): Likewise.
13814         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
13815         of a gimple *.
13816         (vect_is_slp_reduction): Update calls accordingly.  Use
13817         vec_info::lookup_def.
13818         (vect_is_simple_reduction): Likewise
13819         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
13821 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13823         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
13824         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
13825         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
13826         of vinfo_for_stmt.
13827         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
13828         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
13829         (vect_update_vf_for_slp, vect_analyze_loop_operations)
13830         (vect_is_slp_reduction, vectorizable_induction)
13831         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
13832         * tree-vect-patterns.c (vect_init_pattern_stmt):
13833         (vect_determine_min_output_precision_1, vect_determine_precisions)
13834         (vect_pattern_recog): Likewise.
13835         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
13836         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
13837         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
13838         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
13839         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
13840         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
13841         info field from a loop to a loop_vec_info.
13843 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13845         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
13846         (vec_info::add_stmt): Declare.
13847         * tree-vectorizer.c (vec_info::add_stmt): New function.
13848         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
13849         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
13850         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
13851         (vectorizable_induction): Likewise.
13852         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
13853         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
13854         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
13855         (vectorizable_load): Likewise.
13856         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13857         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
13858         (vect_recog_gather_scatter_pattern): Likewise.
13859         (append_pattern_def_seq): Likewise.  Remove a check that is
13860         performed by add_stmt itself.
13862 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13864         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
13865         which make_ssa_name was called with new_stmt before new_stmt
13866         had been created.
13868 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13870         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
13871         split out from...
13872         (vect_is_slp_reduction): ...here...
13873         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
13874         that are already known to be false.
13876 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13878         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
13879         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
13880         STMT_VINFO_NUM_SLP_USES when it's true.
13881         (vect_free_slp_instance): Add a final_p parameter and pass it to
13882         vect_free_slp_tree.
13883         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
13884         (vect_analyze_slp_instance): Likewise.
13885         (vect_slp_analyze_operations): Likewise.
13886         (vect_slp_analyze_bb_1): Likewise.
13887         * tree-vectorizer.c (vec_info): Likewise.
13888         * tree-vect-loop.c (vect_transform_loop): Likewise.
13890 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13892         * tree-vect-loop.c (vectorizable_reduction): Assert that the
13893         function is not called for second and subsequent members of
13894         a reduction group.
13896 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
13898         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
13899         cases for nested loops from here to ...
13900         (vect_create_epilog_for_reduction): ...here.  Only call
13901         vect_is_simple_use for inner-loop reductions.
13903 2018-07-31  Martin Liska  <mliska@suse.cz>
13905         PR gcov-profile/85338
13906         PR gcov-profile/85350
13907         PR gcov-profile/85372
13908         * profile.c (struct location_triplet): New.
13909         (struct location_triplet_hash): Likewise.
13910         (output_location): Do not output a BB that
13911         is already recorded for a line.
13912         (branch_prob): Use streamed_locations.
13914 2018-07-31  Martin Liska  <mliska@suse.cz>
13916         PR gcov-profile/85370
13917         * coverage.c (coverage_begin_function): Do not mark target
13918         clones as artificial functions.
13920 2018-07-31  Martin Liska  <mliska@suse.cz>
13922         PR gcov-profile/83813
13923         PR gcov-profile/84758
13924         PR gcov-profile/85217
13925         PR gcov-profile/85332
13926         * profile.c (branch_prob): Do not record GOTO expressions
13927         for GIMPLE statements which locations are already streamed.
13929 2018-07-31  Olivier Hainque  <hainque@adacore.com>
13931         * gcc.c (handle_spec_function): Accept a soft_matched_part
13932         argument, as do_spec_1.  Pass it down to ...
13933         (eval_spec_function): Accept a soft_matched_part argument,
13934         and pass it down to ...
13935         (do_spec_2): Accept a soft_matched_part argument, and pass
13936         it down to do_spec_1.
13937         (do_spec_1): Pass soft_matched_part to handle_spec_function.
13938         (handle_braces): Update call to handle_spec_function.
13939         (driver::set_up_specs): Update calls to do_spec_2.
13940         (compare_debug_dump_opt_spec_function): Likewise.
13941         (compare_debug_self_opt_spec_function): Likewise.
13943 2018-07-31  Olivier Hainque  <hainque@adacore.com>
13945         * common.opt (nolibc): New option.
13946         * doc/invoke.texi (Link Options): Document it.
13947         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
13948         * config/alpha/linux.h: Likewise.
13949         * config/arc/elf.h: Likewise.
13950         * config/arm/uclinux-elf.h: Likewise.
13951         * config/arm/unknown-elf.h: Likewise.
13952         * config/avr/avrlibc.h: Likewise.
13953         * config/bfin/bfin.h: Likewise.
13954         * config/bfin/linux.h: Likewise.
13955         * config/bfin/uclinux.h: Likewise.
13956         * config/darwin.h: Likewise.
13957         * config/darwin10.h: Likewise.
13958         * config/darwin12.h: Likewise.
13959         * config/gnu-user.h: Likewise.
13960         * config/lm32/uclinux-elf.h: Likewise.
13961         * config/pa/pa-hpux11.h: Likewise.
13962         * config/pa/pa64-hpux.h: Likewise.
13963         * config/sparc/sparc.h: Likewise.
13965 2018-07-31  Olivier Hainque  <hainque@adacore.com>
13967         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
13968         undefined variables.
13970 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
13972         PR target/86640
13973         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
13974         instead of GEN_INT.
13976 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13978         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
13979         terminated string literal.
13981 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
13983         PR rtl-optimization/85160
13984         * combine.c (is_just_move): New function.
13985         (try_combine): Allow combining two instructions into two if neither of
13986         the original instructions was a move.
13988 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
13990         PR target/86673
13991         * doc/extend.texi (Global Register Variables): Discourage use of type
13992         qualifiers.
13993         (Local Register Variables): Likewise.
13995 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
13997         PR tree-optimization/86506
13998         * hwint.h (ceil_log2): Resync with hwint.c implementation.
14000 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
14002         PR target/86547
14003         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
14004         hard_regno, make sure no insn between `from` and `to` clobbers it.
14006 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
14007             Tom de Vries  <tdevries@suse.de>
14009         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
14010         (PTX_DEFAULT_RUNTIME_DIM): ... this.
14011         (nvptx_goacc_validate_dims): Set default worker and gang dims to
14012         PTX_DEFAULT_RUNTIME_DIM.
14013         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
14015 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
14017         * config/pa/pa.c (pa_output_addr_vec): Align address table.
14018         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
14019         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
14021 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
14023         * config/rs6000/constraints.md (wG constraint): Delete, no longer
14024         used.
14025         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
14026         predicate to reflect toc fusion has been deleted.
14027         (toc_fusion_mem_raw): Delete, no longer used.
14028         (toc_fusion_mem_wrapped): Likewise.
14029         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
14030         fusion mask bit.
14031         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
14032         Delete, no longer used.
14033         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
14034         meant to be used for toc fusion.
14035         (rs6000_debug_print_mode): Delete toc fusion debugging.
14036         (rs6000_debug_reg_global): Likewise.
14037         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
14038         fusion and secondary reload support that were never used.
14039         (rs6000_option_override_internal): Delete TOC fusion, that was only
14040         partially defined, and it did not work unless you also used the
14041         -mcmodel= switch.
14042         (rs6000_legitimate_address_p): Delete TOC fusion support.
14043         (rs6000_opt_masks): Likewise.
14044         (fusion_wrap_memory_address): Delete function, no longer used.
14045         (fusion_split_address); Delete TOC fusion support.
14046         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
14047         longer used with toc fusion being deleted.
14048         (TARGET_TOC_FUSION_FP): Likewise.
14049         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
14050         UNSPEC.
14051         (toc fusion spliter): Delete TOC fusion support.
14052         (toc_fusionload_<mode>): Likewise.
14053         (toc_fusionload_di): Likewise.
14054         (fusion_gpr_load_<mode>): Delete generator function, this insn no
14055         longer needs to be named.  Rename predicate to delete TOC fusion.
14056         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
14057         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
14058         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
14059         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
14060         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
14062 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14064         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
14065         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
14066         __int128 in built-in function prototypes.
14067         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
14068         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
14070 2018-07-27  Martin Sebor  <msebor@redhat.com>
14072         PR tree-optimization/86696
14073         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
14074         types, including enums.
14075         (handle_char_store): Be prepared for the above function to fail.
14077 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
14079         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
14080         when optimization level is lower than 2 or optimize for size.
14082 2018-07-26  Martin Sebor  <msebor@redhat.com>
14084         PR tree-optimization/86043
14085         PR tree-optimization/86042
14086         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
14087         (get_string_cst_length): Rename...
14088         (get_min_string_length): ...to this.  Add argument.
14089         (handle_char_store): Extend to handle multi-character stores by
14090         MEM_REF.
14091         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
14092         * tree.h (initializer_zerop): Add argument.
14094 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
14096         PR middle-end/86660
14097         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
14098         declare target to variables if they have always,{to,from,tofrom} map
14099         kinds.
14101 2018-07-26  Martin Liska  <mliska@suse.cz>
14103         PR lto/86548
14104         * lto-wrapper.c: Add linker_output as prefix
14105         for ltrans_output_file.
14107 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
14109         PR rtl-optimization/85805
14110         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
14111         value for hard registers if that was written in the same mode.
14113 2018-07-26  Martin Liska  <mliska@suse.cz>
14115         PR gcov-profile/86536
14116         * gcov.c (format_gcov): Use printf format %.*f directly
14117         and do not handle special values.
14119 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14121         * common/config/arc/arc-common.c (arc_option_optimization_table):
14122         Update default optimizations for size.
14124 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14126         * config/arc/arc.md (movsf_insn): Add short instruction selection.
14127         * config/arc/constraints.md (CfZ): New constraint.
14128         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
14129         (subsf3_fpu): Likewise.
14130         (cmpsf_fpu): Likewise.
14131         (cmpsf_fpu_uneq): Likewise.
14133 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14135         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
14136         accesses as well.
14137         (arc_is_uncached_mem_p): uncached applies to both the variable and
14138         the pointer.
14140 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
14142         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
14143         register names.
14145 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
14147         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
14148         field "m_scopes" from vec to auto_vec.
14150 2018-07-25  Martin Liska  <mliska@suse.cz>
14152         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
14153         return type.
14155 2018-07-25  Richard Biener  <rguenther@suse.de>
14157         PR debug/86654
14158         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
14159         special wrt context_die late.
14160         (gen_subprogram_die): Re-use DIEs in local scope.
14162 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
14164         PR tree-optimization/86644
14165         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
14167 2018-07-25  Martin Liska  <mliska@suse.cz>
14169         PR middle-end/86645
14170         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
14171         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
14173 2018-07-25  Martin Liska  <mliska@suse.cz>
14175         PR sanitizer/79635
14176         * params.def: Explain ASan abbreviation and provide
14177         a documentation link.
14179 2018-07-24  Martin Sebor  <msebor@redhat.com>
14181         PR tree-optimization/86622
14182         PR tree-optimization/86532
14183         * builtins.h (string_length): Declare.
14184         * builtins.c (c_strlen): Correct handling of non-constant offsets.
14185         (check_access): Be prepared for non-constant length ranges.
14186         (string_length): Make extern.
14187         * expr.c (string_constant): Only handle the minor non-constant
14188         array index.  Use string_constant to compute the length of
14189         a generic string constant.
14191 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
14193         PR tree-optimization/86618
14194         * tree-vect-stmts.c (vectorizable_call): Don't take the address
14195         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
14197 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
14199         PR tree-optimization/86636
14200         * json.cc (json::object::set): Fix comment.  Add assertions.
14201         (json::array::append): Move here from json.h.  Add comment and an
14202         assertion.
14203         (json::string::string): Likewise.
14204         * json.h (json::array::append): Move to json.cc.
14205         (json::string::string): Likewise.
14206         * optinfo-emit-json.cc
14207         (optrecord_json_writer::impl_location_to_json): Assert that we
14208         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
14209         wrapper around it.  Expand the location once, rather than three
14210         times.
14211         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
14212         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
14213         wrappers.
14214         (optrecord_json_writer::optinfo_to_json): Likewise, in four
14215         places.  Fix some overlong lines.
14217 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
14219         * config/aarch64/aarch64-simd.md
14220         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
14221         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
14222         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
14223         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
14224         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
14225         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
14226         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
14227         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
14228         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
14230 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
14232         PR middle-end/86627
14233         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
14234         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
14235         and abs_d == d, do the power of two handling if profitable.
14237 2018-07-24  Richard Biener  <rguenther@suse.de>
14239         * match.pd: Add BIT_FIELD_REF canonicalizations.
14241 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14243         PR c/86617
14244         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
14246 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14248         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
14249         terminated STRING_CST object.
14251 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14253         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
14255 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
14257         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
14258         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
14259         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
14260         the elements into a register.
14261         (rs6000_split_v4si_init_di_reg): Delete.
14262         (rs6000_split_v4si_init): Delete.
14263         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
14264         (vsx_init_v4si): Rewrite as a define_expand.
14266 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
14268         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
14269         zero_extend argument from memory): New.
14271 2018-07-22  Martin Sebor  <msebor@redhat.com>
14273         PR bootstrap/86621
14274         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
14275         diagnosing calls with unknown arguments unless -Walloca-larger-than
14276         is restricted to less than PTRDIFF_MAX bytes.
14278 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
14280         * doc/gcov.texi (Invoking Gcov): Editorial changes.
14282 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
14284         * pretty-print.c (text_info::set_location): Remove redundant
14285         "line_table" parameter from call to rich_location::set_range.
14287 2018-07-20  Martin Sebor  <msebor@redhat.com>
14289         PR middle-end/82063
14290         * builtins.c (expand_builtin_alloca): Adjust.
14291         * calls.c (alloc_max_size): Simplify.
14292         * cgraphunit.c (cgraph_node::expand): Adjust.
14293         * common.opt (larger_than_size, warn_frame_larger_than): Remove
14294         variables.
14295         (frame_larger_than_size): Same.
14296         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
14297         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
14298         Initialize.
14299         * doc/invoke.texi (GCC Command Options): Document option arguments.
14300         Explain byte-size arguments and suffixes.
14301         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
14302         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
14303         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
14304         * doc/options.texi (UInteger): Expand.
14305         (Host_Wide_Int, ByteSize): Document new properties.
14306         * final.c (final_start_function_1): Include sizes in an error message.
14307         * function.c (frame_offset_overflow): Same.
14308         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
14309         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
14310         Diagnose unbounded alloca calls only for limits of less than
14311         PTRDIFF_MAX.
14312         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
14313         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
14314         for alloca(0).
14315         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
14316         only for limits of less than PTRDIFF_MAX.
14317         * langhooks-def.h (lhd_handle_option): Change function argument
14318         to HOST_WIDE_INT.
14319         * langhooks.c (lhd_handle_option): Same.
14320         * langhooks.h (handle_option): Same.
14321         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
14322         ByteSize flags.
14323         (var_type, var_type_struct): Same.
14324         (var_set): Handle ByteSize flag.
14325         * optc-gen.awk: Add comments to output to ease debugging.  Make
14326         use of HOST_WIDE_INT where appropriate.
14327         * opts-gen-save.awk:  Use %lx to format unsigned long.
14328         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
14329         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
14330         arguments.  Parse bytes-size suffixes.
14331         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
14332         (enum_value_to_arg): Same.
14333         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
14334         (handle_option): Adjust.
14335         (generate_option): Change function argument to HOST_WIDE_INT.
14336         (cmdline_handle_error): Adjust.
14337         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
14338         (set_option): Change function argument to HOST_WIDE_INT.
14339         (option_enabled): Handle cl_host_wide_int.
14340         (get_option_state): Handle CLVC_SIZE.
14341         (control_warning_option): Same.
14342         * opts.c (common_handle_option): Change function argument to
14343         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
14344         OPT_Wvla_larger_than_.
14345         * opts.h (enum cl_var_type): Add an enumerator.
14346         * stor-layout.c (layout_decl): Print a more meaningful warning.
14347         * toplev.c (output_stack_usage): Adjust.
14349 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
14351         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
14352         call to inline_expand_builtin_string_cmp.
14353         (expand_builtin_strcmp): Likewise.
14354         (expand_builtin_strncmp): Likewise.
14355         (inline_string_cmp): Delete the last parameter, change char_type_node
14356         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
14357         two operands.
14358         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
14359         the inlining expansion on target where the type of the call has same or
14360         narrower precision than unsigned char.
14362 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
14364         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
14365         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
14366         * common.opt (fsave-optimization-record): New option.
14367         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
14368         * doc/invoke.texi (-fsave-optimization-record): New option.
14369         * dumpfile.c: Include "optinfo-emit-json.h".
14370         (struct kv_pair): Move to coretypes.h.
14371         (optgroup_options): Make non-static.
14372         (dump_context::end_scope): Call
14373         optimization_records_maybe_pop_dump_scope.
14374         * dumpfile.h (optgroup_options): New decl.
14375         * json.cc: New file.
14376         * json.h: New file.
14377         * optinfo-emit-json.cc: New file.
14378         * optinfo-emit-json.h: New file.
14379         * optinfo.cc: Include "optinfo-emit-json.h".
14380         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
14381         (optinfo_enabled_p): Check optimization_records_enabled_p.
14382         (optinfo_wants_inlining_info_p): Likewise.
14383         * optinfo.h: Update comment.
14384         * profile-count.c (profile_quality_as_string): New function.
14385         * profile-count.h (profile_quality_as_string): New decl.
14386         (profile_count::quality): New accessor.
14387         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
14388         and optinfo_emit_json_cc_tests.
14389         * selftest.h (selftest::json_cc_tests): New decl.
14390         (selftest::optinfo_emit_json_cc_tests): New decl.
14391         * toplev.c: Include "optinfo-emit-json.h".
14392         (compile_file): Call optimization_records_finish.
14393         (do_compile): Call optimization_records_start.
14394         * tree-ssa-live.c: Include optinfo.h.
14395         (remove_unused_scope_block_p): Retain inlining information if
14396         optinfo_wants_inlining_info_p returns true.
14398 2018-07-20  Richard Biener  <rguenther@suse.de>
14400         PR debug/86585
14401         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
14402         to cover -flto-partition=none.
14404 2018-07-20  Martin Liska  <mliska@suse.cz>
14406         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
14407         (get_decl_source_range): Remove unused function.
14409 2018-07-20  Richard Biener  <rguenther@suse.de>
14411         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
14412         (struct vn_phi_s): Likewise.
14413         (struct vn_reference_s): Likewise.
14414         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
14415         for searching the slot of an entry known to be in the hash itself.
14416         (vn_phi_hasher::equal): Likewise.
14417         (vn_reference_hasher::equal): Likewise.
14418         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
14419         globals.
14420         (optimistic_info, current_info): Remove, keeping only valid_info.
14421         (vn_reference_lookup_1): Remove fallback lookup.
14422         (vn_reference_lookup_2): Likewise.
14423         (vn_nary_op_lookup_1): Likewise.
14424         (vn_phi_lookup): Likewise.
14425         (vn_nary_build_or_lookup_1): Make sure to not chain the built
14426         hash element.
14427         (vn_reference_insert): Adjust, chain the inserted hash element
14428         at last_inserted_ref.
14429         (vn_reference_insert_pieces): Likewise.
14430         (visit_reference_op_call): Likewise.
14431         (vn_nary_op_insert_into): Chain the inserted hash element at
14432         last_inserted_nary.
14433         (vn_nary_op_insert_pieces): Adjust.
14434         (vn_nary_op_insert): Likewise.
14435         (vn_nary_op_insert_stmt): Likewise.
14436         (vn_phi_insert): Adjust, chain the inserted hash element at
14437         last_inserted_phi.
14438         (process_scc): Remove clearing and copying the optimistic
14439         table.  Instead remove elements inserted during an optimistic
14440         iteration from the single table we maintain.
14441         (init_scc_vn): Adjust.
14442         (free_scc_vn): Likewise.
14443         (sccvn_dom_walker::record_cond): Likewise.
14444         (sccvn_dom_walker::after_dom_children): Likewise.
14446 2018-07-19  Martin Sebor  <msebor@redhat.com>
14448         PR tree-optimization/84047
14449         PR tree-optimization/83776
14450         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
14451         (check_array_bounds): Call it.
14453 2018-07-19  Martin Sebor  <msebor@redhat.com>
14455         * align.h (align_flags): Use member initialization.
14457 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
14459         * Makefile.in (OBJS): Add optinfo.o.
14460         * coretypes.h (class symtab_node): New forward decl.
14461         (struct cgraph_node): New forward decl.
14462         (class varpool_node): New forward decl.
14463         * dump-context.h: New file.
14464         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
14465         "tree-pass.h".
14466         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
14467         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
14468         (set_alt_dump_file): Likewise.
14469         (dump_context::~dump_context): New dtor.
14470         (dump_gimple_stmt): Move implementation to...
14471         (dump_context::dump_gimple_stmt): ...this new member function.
14472         Add the stmt to any pending optinfo, creating one if need be.
14473         (dump_gimple_stmt_loc): Move implementation to...
14474         (dump_context::dump_gimple_stmt_loc): ...this new member function.
14475         Start a new optinfo and add the stmt to it.
14476         (dump_gimple_expr): Move implementation to...
14477         (dump_context::dump_gimple_expr): ...this new member function.
14478         Add the stmt to any pending optinfo, creating one if need be.
14479         (dump_gimple_expr_loc): Move implementation to...
14480         (dump_context::dump_gimple_expr_loc): ...this new member function.
14481         Start a new optinfo and add the stmt to it.
14482         (dump_generic_expr): Move implementation to...
14483         (dump_context::dump_generic_expr): ...this new member function.
14484         Add the tree to any pending optinfo, creating one if need be.
14485         (dump_generic_expr_loc): Move implementation to...
14486         (dump_context::dump_generic_expr_loc): ...this new member
14487         function.  Add the tree to any pending optinfo, creating one if
14488         need be.
14489         (dump_printf): Move implementation to...
14490         (dump_context::dump_printf_va): ...this new member function.  Add
14491         the text to any pending optinfo, creating one if need be.
14492         (dump_printf_loc): Move implementation to...
14493         (dump_context::dump_printf_loc_va): ...this new member function.
14494         Start a new optinfo and add the stmt to it.
14495         (dump_dec): Move implementation to...
14496         (dump_context::dump_dec): ...this new member function.  Add the
14497         value to any pending optinfo, creating one if need be.
14498         (dump_context::dump_symtab_node): New member function.
14499         (dump_context::get_scope_depth): New member function.
14500         (dump_context::begin_scope): New member function.
14501         (dump_context::end_scope): New member function.
14502         (dump_context::ensure_pending_optinfo): New member function.
14503         (dump_context::begin_next_optinfo): New member function.
14504         (dump_context::end_any_optinfo): New member function.
14505         (dump_context::s_current): New global.
14506         (dump_context::s_default): New global.
14507         (dump_scope_depth): Delete global.
14508         (dumpfile_ensure_any_optinfo_are_flushed): New function.
14509         (dump_symtab_node): New function.
14510         (get_dump_scope_depth): Reimplement in terms of dump_context.
14511         (dump_begin_scope): Likewise.
14512         (dump_end_scope): Likewise.
14513         (selftest::temp_dump_context::temp_dump_context): New ctor.
14514         (selftest::temp_dump_context::~temp_dump_context): New dtor.
14515         (selftest::verify_item): New function.
14516         (ASSERT_IS_TEXT): New macro.
14517         (ASSERT_IS_TREE): New macro.
14518         (ASSERT_IS_GIMPLE): New macro.
14519         (selftest::test_capture_of_dump_calls): New test.
14520         (selftest::dumpfile_c_tests): Call it.
14521         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
14522         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
14523         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
14524         descriptive comment.
14525         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
14526         (dump_node, dump_bb): Move these unrelated decls.
14527         (class dump_manager): Add leading comment.
14528         * optinfo.cc: New file.
14529         * optinfo.h: New file.
14531 2018-07-19  Michael Collison  <michael.collison@arm.com>
14532             Richard Henderson <rth@redhat.com>
14534         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
14535         (subti): Handle op1 zero.
14536         (subvti4, usub4ti4): New.
14537         (*sub<GPI>3_compare1_imm): New.
14538         (sub<GPI>3_carryinCV): New.
14539         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
14540         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
14542 2018-07-19  Michael Collison  <michael.collison@arm.com>
14543             Richard Henderson <rth@redhat.com>
14545         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
14546         (addti3): Create simpler code if low part is already known to be 0.
14547         (addvti4, uaddvti4): New.
14548         (*add<GPI>3_compareC_cconly_imm): New.
14549         (*add<GPI>3_compareC_cconly): New.
14550         (*add<GPI>3_compareC_imm): New.
14551         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
14552         handle constants within this pattern..
14553         (*add<GPI>3_compareV_cconly_imm): New.
14554         (*add<GPI>3_compareV_cconly): New.
14555         (*add<GPI>3_compareV_imm): New.
14556         (add<GPI>3_compareV): New.
14557         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
14558         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
14559         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
14560         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
14561         with 'comparison' operator.
14562         (*add<GPI>3_compareV_cconly_imm): Ditto.
14563         (*add<GPI>3_compareV_cconly): Ditto.
14564         (*add<GPI>3_compareV_imm): Ditto.
14565         (add<GPI>3_compareV): Ditto.
14566         (add<mode>3_carryinC): Ditto.
14567         (*add<mode>3_carryinC_zero): Ditto.
14568         (*add<mode>3_carryinC): Ditto.
14569         (add<mode>3_carryinV): Ditto.
14570         (*add<mode>3_carryinV_zero): Ditto.
14571         (*add<mode>3_carryinV): Ditto.
14573 2018-07-19  Michael Collison  <michael.collison@arm.com>
14574             Richard Henderson <rth@redhat.com>
14576         * config/aarch64/aarch64-modes.def (CC_V): New.
14577         * config/aarch64/aarch64-protos.h
14578         (aarch64_addti_scratch_regs): Declare
14579         (aarch64_subvti_scratch_regs): Declare.
14580         (aarch64_expand_subvti): Declare.
14581         (aarch64_gen_unlikely_cbranch): Declare
14582         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
14583         for signed overflow using CC_Vmode.
14584         (aarch64_get_condition_code_1): Handle CC_Vmode.
14585         (aarch64_gen_unlikely_cbranch): New function.
14586         (aarch64_addti_scratch_regs): New function.
14587         (aarch64_subvti_scratch_regs): New function.
14588         (aarch64_expand_subvti): New function.
14590 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14592         * config/aarch64/aarch64-option-extensions.def: New entry for profile
14593         extension.
14594         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
14595         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
14596         extension.
14598 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14600         PR target/83009
14601         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14602         address check not strict.
14604 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14606         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
14607         Umq with Umn.
14608         (store_pair_lanes<mode>): Likewise.
14609         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
14610         enum value 'ADDR_QUERY_LDP_STP_N'.
14611         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
14612         (aarch64_print_address_internal): Add declaration.
14613         (aarch64_print_ldpstp_address): Remove.
14614         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
14615         (aarch64_print_operand): Change printing of 'y'.
14616         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
14617         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
14618         'true' rather than '1'.
14619         * config/aarch64/constraints.md (Uml): Likewise.
14620         (Uml): Rename to Umn.
14621         (Umq): Remove.
14623 2018-07-19  Richard Biener  <rguenther@suse.de>
14625         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
14626         a trailing array.
14627         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
14628         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
14629         (vn_reference_hasher): Likewise.
14630         (struct vn_tables_s): Remove obstack and alloc-pool members.
14631         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
14632         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
14633         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
14634         (vn_reference_insert_pieces): Likewise.
14635         (alloc_vn_nary_op_noinit): Adjust.
14636         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
14637         (vn_phi_eq): Adjust.
14638         (shared_lookup_phiargs): Remove.
14639         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
14640         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
14641         (visit_reference_op_call): Likewise.
14642         (copy_nary, copy_phi, copy_reference): Remove.
14643         (process_scc): Rewind the obstack when iterating.  Do not
14644         copy the elements to valid_info but just move them from one
14645         hashtable to the other.
14646         (allocate_vn_table): Adjust.
14647         (free_vn_table): Likewise.
14648         (init_scc_vn): Likewise.
14649         (free_scc_vn): Likewise.
14651 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
14653         PR target/86560
14654         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
14655         indirect_return as function type attribute.
14656         (ix86_attribute_table): Change indirect_return to function
14657         type attribute.
14658         * doc/extend.texi: Update indirect_return attribute.
14660 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
14662         * wide-int.h (widest2_int): New.
14663         * gimple-fold.c (arith_overflowed_p): Use it.
14664         * tree.h (widest2_int_cst): New.
14665         * tree-vrp.c (wide_int_binop_overflow): Rename from
14666         vrp_int_const_binop.
14667         Rewrite to work on trees.
14668         (extract_range_from_multiplicative_op_1): Abstract code to...
14669         (wide_int_range_min_max): ...here.
14670         (wide_int_range_cross_product): ...and here.
14671         (extract_range_from_binary_expr_1): Abstract overflow code to...
14672         (wide_int_range_mult_wrapping): ...here.
14673         * tree-vrp.h (wide_int_range_cross_product): New.
14674         (wide_int_range_mult_wrapping): New.
14676 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
14677             Julia Koval  <julia.koval@intel.com>
14679         * config/i386/x86-tune-costs.h (skylake_memcpy,
14680         skylake_memset): Replace rep_prefix with unrolling for size 512.
14682 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
14684         PR middle-end/86544
14685         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
14686         comparision with EQ_EXPR in last stmt.
14688 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14690         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
14691         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
14692         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
14693         previously known as "PowerPC AltiVec Built-in Functions".  Move
14694         some material to new subsubsections "PowerPC AltiVec Built-in
14695         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
14696         ISA 2.07".
14697         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
14698         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
14699         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
14700         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
14702 2018-07-18  Richard Biener  <rguenther@suse.de>
14704         PR tree-optimization/86557
14705         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
14706         EXACT_DIV_EXPR.
14708 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
14710         * config/s390/s390.c (s390_function_profiler): Generate CFI.
14712 2018-07-17  Jeff Law  <law@redhat.com>
14714         * config/arm/arm.c (get_label_padding): Update for recent
14715         changes to label_to_alignment.
14717         PR tree-optimization/86010
14718         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
14720         * config/mips/mips.c (vr4130_align_insns): Update for recent
14721         changes to label_to_alignment.
14723         * config/frv/frv.c (frv_label_align): Update for recent changes
14724         to label_to_alignment.
14726         * config/nios2/nios2.c (nios2_label_align): Update for recent
14727         changes which dropped ALIGN_LABELS_LOG.
14729 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
14731         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
14732         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
14734 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
14736         * config/arc/arc.c (arc_label_align): Use align_labels instead of
14737         deprecated align_labels_log.
14739 2018-07-17  Richard Biener  <rguenther@suse.de>
14741         PR lto/86456
14742         * dwarf2out.c (init_sections_and_labels): Always generate
14743         a debug_line_str_section for early LTO debug.
14744         (dwarf2out_finish): Reset debug_line_str_hash output early.
14745         Bump counter for extra dwarf5 .debug_loc labels to not conflict
14746         with fat LTO part.
14747         (dwarf2out_early_finish): Output debug_line_str.
14749 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
14751         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
14752         index register on z196 or later.
14754 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
14756         * config/s390/s390.c (s390_default_align): Set default function
14757         alignment to 16.
14758         (s390_override_options_after_change): Call s390_default align.
14759         (s390_option_override_internal): Call s390_default_align.
14760         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
14762 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
14764         PR middle-end/86542
14765         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
14766         to _looptemp_ clauses, other than the first two.
14768 2018-07-17  Martin Liska  <mliska@suse.cz>
14770         * opts.c: Do not enable OPT_falign_* for -Os.
14772 2018-07-17  Martin Liska  <mliska@suse.cz>
14774         * align.h (MAX_CODE_ALIGN): New.
14775         (MAX_CODE_ALIGN_VALUE): New.
14776         * common/config/i386/i386-common.c (ix86_handle_option):
14777         (MAX_CODE_ALIGN): Moved to align.h.
14778         * final.c (MAX_CODE_ALIGN): Likewise.
14779         * opts.c (parse_and_check_align_values):
14780         (MAX_CODE_ALIGN): Likewise.
14781         (MAX_CODE_ALIGN_VALUE): Likewise.
14783 2018-07-17  Martin Liska  <mliska@suse.cz>
14785         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
14786         in order to fulfil coding style.
14787         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
14788         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14789         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
14790         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
14791         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
14792         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
14793         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
14794         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
14795         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14797 2018-07-17  Martin Liska  <mliska@suse.cz>
14799         * align.h: New file.
14800         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
14801         directly.
14802         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
14803         align_flags of label_to_alignment.
14804         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
14805         align_flags class.
14806         * config/m68k/m68k.c: Do not use removed align_labels_value and
14807         align_loops_value.
14808         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
14809         (LOOP_ALIGN): Likewise.
14810         (LABEL_ALIGN): Likewise.
14811         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
14812         Remove not used macro.
14813         (rs6000_loop_align): Change return type to align_flags.
14814         (rs6000_loop_align_max_skip): Remove.
14815         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
14816         Change return type to align_flags.
14817         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
14818         Remove not used macro.
14819         (rs6000_loop_align):  Change return type to align_flags.
14820         (rs6000_loop_align_max_skip): Remove.
14821         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
14822         * config/rx/rx-protos.h (rx_align_for_label): Make it
14823         static function.
14824         * config/rx/rx.c (rx_align_for_label): Change return type
14825         to align_flags.
14826         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
14827         macro definitions.
14828         into align_flags class.
14829         (LABEL_ALIGN): Likewise.
14830         (LOOP_ALIGN): Likewise.
14831         * config/s390/s390.c (s390_label_align): Use align_flags
14832         class member.
14833         (s390_asm_output_function_label): Likewise.
14834         * config/sh/sh.c (sh_override_options_after_change):
14835         Use align_flags class directly without macros.
14836         (find_barrier): Likewise.
14837         (barrier_align): Likewise.
14838         (sh_loop_align): Likewise.
14839         * config/spu/spu.c (spu_option_override):
14840         Use align_flags_tuple::get_value instead of removed macros.
14841         (spu_sched_init): Likewise.
14842         * config/spu/spu.h (GTY): Likewise.
14843         * config/visium/visium.c (visium_option_override):
14844         Set "8" as default secondary alignment.
14845         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
14846         in order to guarantee secondary alignment of 8.
14847         * coretypes.h: Include align.h header file.
14848         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
14849         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
14850         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
14851         * doc/tm.texi.in: Likewise.
14852         * final.c (struct label_alignment): Remove not used structure.
14853         (LABEL_ALIGN): Change type to align_flags.
14854         (LOOP_ALIGN): Likewise.
14855         (JUMP_ALIGN): Likewise.
14856         (default_loop_align_max_skip): Remove.
14857         (default_label_align_max_skip): Likewise.
14858         (default_jump_align_max_skip): Likewise.
14859         (default_label_align_after_barrier_max_skip):
14860         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
14861         (LABEL_TO_MAX_SKIP): Remove.
14862         (label_to_alignment): Return align_flags type instead of integer.
14863         (label_to_max_skip): Remove.
14864         (align_fuzz): Use align_flags type.
14865         (compute_alignments): Use align_flags type and use align_flags::max
14866         to combine multiple alignments.
14867         (grow_label_align): Grow vec instead of C array.
14868         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
14869         (shorten_branches):  Use align_flags type and use align_flags::max
14870         to combine multiple alignments.
14871         (final_scan_insn_1): Remove usage of secondary alignment that comes
14872         from label alignment, but instead use proper secondary alignment
14873         which is computed in grow_label_align.
14874         * flags.h (struct align_flags_tuple): Move to align.h.
14875         (struct align_flags): Likewise.
14876         (state_align_loops): Rename to align_loops.
14877         (state_align_jumps): Rename to align_jumps.
14878         (state_align_labels): Rename to align_labels.
14879         (state_align_functions): Rename to align_functions.
14880         (align_loops_log): Remove.
14881         (align_jumps_log): Remove.
14882         (align_labels_log): Remove.
14883         (align_functions_log): Remove.
14884         (align_loops_max_skip): Remove.
14885         (align_jumps_max_skip): Remove.
14886         (align_labels_max_skip): Remove.
14887         (align_functions_max_skip): Remove.
14888         (align_loops_value): Remove.
14889         (align_jumps_value): Remove.
14890         (align_labels_value): Remove.
14891         (align_functions_value): Remove.
14892         * output.h (label_to_alignment): Change return type to align_flags.
14893         (label_to_max_skip): Remove.
14894         * target.def: Remove loop_align_max_skip, label_align_max_skip,
14895         jump_align_max_skip macros.
14896         * targhooks.h (default_loop_align_max_skip): Remove.
14897         (default_label_align_max_skip): Likewise.
14898         (default_jump_align_max_skip): Likewise.
14899         (default_label_align_after_barrier_max_skip): Remove.
14900         * toplev.c (read_log_maxskip): Use ::normalize function.
14901         (parse_N_M): Remove not used argument and also call ::normalize.
14902         (parse_alignment_opts): Do not pass unused arguments.
14903         * varasm.c (assemble_start_function): Use directly align_functions
14904         instead of removed macros.
14905         * system.h: Do not poison removed macros.
14907 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
14909         PR middle-end/86539
14910         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
14911         and cond temporaries don't have reference type if iterator has
14912         pointer type.  For init use &for_pre_body instead of pre_p if
14913         for_pre_body is non-empty.
14915 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
14917         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
14918         double-double modes to SFmode directly directly.
14919         (trunc<mode>sf2_fprs): Delete.
14921 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
14923         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
14924         for conversions between IFmode and the decimal floating point modes.
14925         (init_float128_ieee): Use the correct names for conversions between
14926         KFmode and the decimal floating point modes.
14928 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
14930         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
14931         for the conversions between TDmode and IFmode.
14932         (init_float128_ieee): Use more correct names for the conversions
14933         between TDmode and KFmode.
14935 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
14937         PR tree-optimization/86526
14938         * builtins.c (expand_builtin_memcmp): Formatting fixes.
14939         (inline_expand_builtin_string_cmp): Likewise.
14940         (inline_string_cmp): Likewise.  Use c_readstr instead of
14941         builtin_memcpy_read_str.  Add unit_mode temporary.
14943 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14945         PR middle-end/86528
14946         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
14947         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
14949 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14951         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
14952         Alphabetize prototypes of built-in functions, separating out
14953         built-in functions that are listed in this section but should be
14954         described elsewhere.
14956 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
14958         PR target/86511
14959         * expmed.c (emit_store_flag): Do not emit setcc followed by a
14960         conditional move when trapping comparison was split to a
14961         non-trapping one (and vice versa).
14963 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
14965         * config/s390/s390.c (s390_function_profiler): Generate nops
14966         instead of profiler call sequences.
14967         * config/s390/s390.opt: Add the new option.
14969 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
14971         * config/s390/s390.c (s390_function_profiler): Generate
14972         __mcount_loc section.
14973         * config/s390/s390.opt: Add the new option.
14975 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
14977         * common.opt: Add the new warning.
14978         * config/s390/s390.c (s390_function_profiler): Emit "brasl
14979         %r0,__fentry__" when -mfentry is specified.
14980         (s390_option_override_internal): Disallow -mfentry for 31-bit
14981         CPUs.
14982         * config/s390/s390.opt: Add the new option.
14984 2018-07-16  Richard Biener  <rguenther@suse.de>
14986         PR lto/86523
14987         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
14988         for function-local FUNCTION_DECL and RESULT_DECL immediately.
14990 2018-07-16  Martin Liska  <mliska@suse.cz>
14992         PR ipa/86529
14993         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
14994         to ::get_create.
14996 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
14998         * config/arc/arcHS.md: Update ARCHS scheduling rules.
15000 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
15002         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
15003         for ARCHS4x.
15004         * config/arc/arc-cpus.def (hs4x): New cpu.
15005         (hs4xd): Likewise.
15006         * config/arc/arc-tables.opt: Regenerate.
15007         * config/arc/arc.c (arc_sched_issue_rate): New function.
15008         (TARGET_SCHED_ISSUE_RATE): Define.
15009         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
15010         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
15011         fpu_cvt.
15012         (attr tune): Add ARCHS4x tune values.
15013         (attr tune_dspmpy): Define.
15014         (*tst): Correct instruction type.
15015         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
15016         * config/arc/arcHS4x.md: New file.
15017         * config/arc/fpu.md: Update instruction type attributes.
15018         * config/arc/t-multilib: Regenerate.
15020 2018-07-16  Tom de Vries  <tdevries@suse.de>
15022         PR debug/86455
15023         * var-tracking.c (vt_initialize): Fix pre_dec handling.
15025 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
15027         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
15028         early clobber.
15030 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
15032         PR tree-optimization/86514
15033         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
15034         conversion to a boolean type from a type with greater precision.
15036 2018-07-16  Tom de Vries  <tdevries@suse.de>
15038         * var-tracking.c (vt_initialize): Print adjusted insn slim if
15039         dump_flags request TDF_SLIM.
15041 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
15043         * fold-const.c (int_const_binop_1): Abstract...
15044         (wide_int_binop): ...wide int code here.
15045         (poly_int_binop): ...poly int code here.
15046         Abstract the rest of int_const_binop_1 into int_const_binop.
15047         * fold-const.h (wide_int_binop): New.
15048         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
15049         Remove useless PLUS/MINUS_EXPR case.
15050         (zero_nonzero_bits_from_vr): Move wide int code...
15051         (zero_nonzero_bits_from_bounds): ...here.
15052         (extract_range_from_binary_expr_1): Move mask optimization code...
15053         (range_easy_mask_min_max): ...here.
15054         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
15055         (range_easy_mask_min_max): New.
15057 2018-07-15  Jeff Law  <law@redhat.com>
15059         PR target/85993
15060         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
15061         block.
15063 2018-07-14  Jim Wilson  <jimw@sifive.com>
15065         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
15067 2018-07-14  Paul Koning  <ni1d@arrl.net>
15069         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
15071 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
15073         * lto-streamer-out.c (copy_function_or_variable): Dump info about
15074         copying section.
15076 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
15077             Steve Munroe  <munroesj52@gmail.com>
15079         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
15080         (_mm_andnot_si128): Likewise.
15081         (_mm_or_si128): Likewise.
15082         (_mm_xor_si128): Likewise.
15084 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
15086         PR middle-end/78809
15087         * builtins.c (expand_builtin_memcmp): Inline the calls first
15088         when result_eq is false.
15089         (expand_builtin_strcmp): Inline the calls first.
15090         (expand_builtin_strncmp): Likewise.
15091         (inline_string_cmp): New routine. Expand a string compare
15092         call by using a sequence of char comparison.
15093         (inline_expand_builtin_string_cmp): New routine. Inline expansion
15094         a call to str(n)cmp/memcmp.
15095         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
15096         New option.
15097         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
15099 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
15101         * config/arm/driver-arm.c: Include arm-native.h.
15102         (host_detect_local_cpu): Use auto-generated data tables.
15103         (vendors, arm_cpu_table): Delete.  Move part information to ...
15104         * config/arm/arm-cpus.in: ... here.
15105         * config/arm/parsecpu.awk (gen_native): New function.
15106         (vendor, part): New CPU fields.
15107         (END): Add support for building the native CPU detection tables.
15108         * config/arm/t-arm (arm-native.h): Add build rule.
15109         (driver-arm.o): Add dependency on arm-native.h.
15111 2018-07-13  Richard Biener  <rguenther@suse.de>
15113         PR middle-end/85974
15114         * match.pd (addr1 - addr2): Allow either of the operand to
15115         have a conversion.
15117 2018-07-13  Tom de Vries  <tdevries@suse.de>
15119         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
15120         in remap_ssa_name.
15122 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
15124         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
15125         arrays instead of numbered variables.
15127 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
15129         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
15130         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
15132 2018-07-13  Richard Biener  <rguenther@suse.de>
15134         PR debug/86452
15135         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
15136         instead of get_context_die.
15138 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
15139             Richard Biener  <rguenther@suse.de>
15141         PR middle-end/86489
15142         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
15143         that the loop latch destination where phi is defined.
15145 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
15147         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
15148         (riscv_expand_epilogue): Add assertion to check interrupt mode.
15149         (riscv_set_current_function): Extract getting interrupt type to new
15150         function.
15151         (riscv_get_interrupt_type): New function.
15152         (riscv_merge_decl_attributes): New function, checking interrupt type is
15153         same.
15154         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
15156 2018-07-12  Paul Koning  <ni1d@arrl.net>
15158         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
15159         directive.
15161 2018-07-12  Paul Koning  <ni1d@arrl.net>
15163         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
15164         zero reference, add doloop_end instead.
15165         * doc/md.texi (decrement_and_branch_until_zero): Remove.
15166         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
15167         detail for doloop_end.
15169 2018-07-12  Martin Sebor  <msebor@redhat.com>
15171         PR c/86453
15172         * attribs.c (decl_attributes): Reject conflicting attributes before
15173         calling attribute handlers.
15175 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
15177         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
15178          parameter.
15179         (gcc::dump_manager::get_dump_file_name): likewise.
15180         (dump_begin): Likewise.
15181         * dumpfile.h (dump_begin): Update prototype.
15182         (gcc::dump_manager::get_dump_file_name,
15183         gcc::dump_manager::get_dump_file_name): Update prototype.
15185 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15187         * internal-fn.h (vectorizable_internal_fn_p): New function.
15188         * tree-vect-slp.c (compatible_calls_p): Likewise.
15189         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
15190         to internal functions.
15191         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
15193 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15195         * fold-const.h (inverse_conditions_p): Declare.
15196         * fold-const.c (inverse_conditions_p): New function.
15197         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
15198         that test the inverse condition of a conditional internal function.
15199         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
15200         * internal-fn.c (internal_fn_mask_index): Handle conditional
15201         internal functions.
15202         (vectorized_internal_fn_supported_p): New function.
15203         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
15204         (any_pred_load_store): Replace with...
15205         (need_to_predicate): ...this new variable.
15206         (redundant_ssa_names): New variable.
15207         (ifcvt_can_use_mask_load_store): Move initial checks to...
15208         (ifcvt_can_predicate): ...this new function.  Handle tree codes
15209         for which a conditional internal function exists.
15210         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
15211         instead of ifcvt_can_use_mask_load_store.  Update after variable
15212         name change.
15213         (predicate_load_or_store): New function, split out from
15214         predicate_mem_writes.
15215         (check_redundant_cond_expr): New function.
15216         (value_available_p): Likewise.
15217         (predicate_rhs_code): Likewise.
15218         (predicate_mem_writes): Rename to...
15219         (predicate_statements): ...this.  Use predicate_load_or_store
15220         and predicate_rhs_code.
15221         (combine_blocks, tree_if_conversion): Update after above name changes.
15222         (ifcvt_local_dce): Handle redundant_ssa_names.
15223         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
15224         general conditional functions.
15225         * tree-vect-stmts.c (vectorizable_call): Likewise.
15227 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15228             Alan Hayward  <alan.hayward@arm.com>
15229             David Sherwood  <david.sherwood@arm.com>
15231         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
15232         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
15233         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
15234         plus and minus and convert them into IFN_COND_FMA-based sequences.
15235         (convert_mult_to_fma): Handle conditional plus and minus.
15237 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15239         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
15240         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
15241         (cond_fnms_optab): New optabs.
15242         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
15243         internal functions.
15244         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
15245         * internal-fn.h (get_conditional_internal_fn): Declare.
15246         (get_unconditional_internal_fn): Likewise.
15247         * internal-fn.c (cond_ternary_direct): New macro.
15248         (expand_cond_ternary_optab_fn): Likewise.
15249         (direct_cond_ternary_optab_supported_p): Likewise.
15250         (FOR_EACH_COND_FN_PAIR): Likewise.
15251         (get_conditional_internal_fn): New function.
15252         (get_unconditional_internal_fn): Likewise.
15253         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
15254         (gimple_match_op::gimple_match_op): Add a new overload for 5
15255         operands.
15256         (gimple_match_op::set_op): Likewise.
15257         (gimple_resimplify5): Declare.
15258         * genmatch.c (decision_tree::gen): Generate simplifications for
15259         5 operands.
15260         * gimple-match-head.c (gimple_simplify): Define an overload for
15261         5 operands.  Handle calls with 5 arguments in the top-level overload.
15262         (convert_conditional_op): Handle conversions from unconditional
15263         internal functions to conditional ones.
15264         (gimple_resimplify5): New function.
15265         (build_call_internal): Pass a fifth operand.
15266         (maybe_push_res_to_seq): Likewise.
15267         (try_conditional_simplification): Try converting conditional
15268         internal functions to unconditional internal functions.
15269         Handle 3-operand unconditional forms.
15270         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
15271         Define ternary equivalents of the current rules for binary conditional
15272         internal functions.
15273         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
15274         ternary operations.
15275         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
15276         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
15277         (optab): Handle them.
15278         (SVE_COND_FP_TERNARY): New int iterator.
15279         (sve_fmla_op, sve_fmad_op): New int attributes.
15280         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
15281         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
15282         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
15284 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
15286         * target.def (preferred_else_value): New target hook.
15287         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
15288         * doc/tm.texi: Regenerate.
15289         * targhooks.h (default_preferred_else_value): Declare.
15290         * targhooks.c (default_preferred_else_value): New function.
15291         * internal-fn.h (conditional_internal_fn_code): Declare.
15292         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
15293         (get_conditional_internal_fn): Use it.
15294         (conditional_internal_fn_code): New function.
15295         * gimple-match.h (gimple_match_cond): New struct.
15296         (gimple_match_op): Add a cond member function.
15297         (gimple_match_op::gimple_match_op): Update all forms to take a
15298         gimple_match_cond.
15299         * genmatch.c (expr::gen_transform): Use the same condition as res_op
15300         for the suboperation, but don't specify a particular else_value.
15301         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
15302         (visit_nary_op, visit_reference_op_load): Pass
15303         gimple_match_cond::UNCOND to the gimple_match_op constructor.
15304         * gimple-match-head.c: Include tree-eh.h
15305         (convert_conditional_op): New function.
15306         (maybe_resimplify_conditional_op): Likewise.
15307         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
15308         (gimple_resimplify2): Likewise.
15309         (gimple_resimplify3): Likewise.
15310         (gimple_resimplify4): Likewise.
15311         (maybe_push_res_to_seq): Return null for conditional operations.
15312         (try_conditional_simplification): New function.
15313         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
15314         constructor.
15315         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
15316         IFN_COND_* call.
15317         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
15318         function.
15319         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
15321 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
15323         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15324         DECL_FCONTEXT
15325         (hash_tree): Do not hash DECL_FCONTEXT
15326         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
15327         Do not stream DECL_FCONTEXT.
15328         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
15329         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
15331 2018-07-12  Richard Biener  <rguenther@suse.de>
15333         PR debug/86462
15334         * dwarf2out.c (gen_block_die): Only output blocks when they have
15335         at least one !DECL_IGNORED_P variable.
15337 2018-07-12  Richard Biener  <rguenther@suse.de>
15339         PR target/84829
15340         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
15341         Remove -mieee-fp handling.
15343 2018-07-12  Richard Biener  <rguenther@suse.de>
15345         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
15346         left-over from last patch.
15348 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
15350         PR tree-optimization/86492
15351         * gimple-ssa-store-merging.c
15352         (imm_store_chain_info::coalesce_immediate_stores): Call
15353         check_no_overlap even for the merge_overlapping case.  Formatting fix.
15355 2018-07-12  Richard Biener  <rguenther@suse.de>
15357         PR middle-end/86479
15358         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
15359         move possibly trapping operations into the conditional.
15361 2018-07-12  Richard Biener  <rguenther@suse.de>
15363         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
15364         (vn_lookup_simplify_result): Remove recursion limit applied
15365         here.
15366         (vn_nary_build_or_lookup_1): Adjust.
15367         (try_to_simplify): Likewise.
15368         * gimple-match-head.c (gimple_resimplify1): Instead apply one
15369         here.
15370         (gimple_resimplify2): Likewise.
15371         (gimple_resimplify3): Likewise.
15372         (gimple_resimplify4): Likewise.
15374 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
15376         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
15377         Use __mmask64 type instead of __mmask8 for __M argument.
15378         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
15379         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
15380         __U argument.
15381         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
15382         __mmask16 for __M argument.
15383         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
15384         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
15385         to __mmask16 instead of __mmask8.
15386         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
15387         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
15388         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
15389         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
15390         instead of __mmask16 for __U argument.
15391         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
15392         __mmask16 instead of __mmask8 for __U argument.
15393         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
15394         __U argument.
15395         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
15396         __mmask16.
15397         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
15398         argument.
15399         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
15400         __U argument.
15401         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
15402         __mmask16.
15403         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
15404         of __mmask16.
15405         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
15406         __U argument.
15407         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
15408         __U argument.
15409         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
15410         __U argument.
15411         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
15412         __U argument.
15413         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
15414         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
15415         return type as well as __M argument type and all casts from __mmask8
15416         to __mmask32.
15417         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
15418         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
15419         return type as well as __M argument type and all casts from __mmask8
15420         to __mmask16.
15421         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
15422         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
15423         return type as well as __M argument type and all casts from __mmask8
15424         to __mmask32.
15425         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
15426         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
15427         return type as well as __M argument type and all casts from __mmask8
15428         to __mmask16.
15429         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
15430         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
15431         __mmask16.
15433 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
15435         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
15436         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
15437         for __U argument.
15439 2018-07-11  Paul Koning  <ni1d@arrl.net>
15441         * doc/md.texi (define_subst): Document how multiple occurrences of
15442         the same argument in the replacement pattern are handled.
15444 2018-07-11  Paul Koning  <ni1d@arrl.net>
15446         * doc/extend.texi (Common Variable Attributes): Move "mode" into
15447         alphabetical order.
15448         (Common Type Attributes): Add "mode" attribute.
15450 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
15452         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
15453         stream DECL_ORIGINAL_TYPE.
15454         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
15455         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
15456         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
15457         Do not walk original type.
15458         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
15459         external decls.
15460         (write_ts_decl_non_common_tree_pointers): Do not stream
15461         DECL_ORIGINAL_TYPE
15462         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
15463         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
15465 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
15467         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
15468         thread twice from the same starting edge.
15470 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
15472         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
15473         * gimple.c (gimple_call_nonnull_result_p): ...here...
15474         (gimple_call_nonnull_arg): ...and here.
15475         * gimple.h (gimple_call_nonnull_result_p): New.
15476         (gimple_call_nonnull_arg): New.
15478 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
15480         * config/arm/arm-cpus.in: Move information from fpu field of each
15481         cpu definition to the isa field.
15482         * config/arm/parsecpu.awk (fpu): Delete match rule.
15483         (gen_comm_data): Don't add bits from the CPU's FPU entry.
15485 2018-07-11  Richard Biener  <rguenther@suse.de>
15487         PR debug/86457
15488         * dwarf2out.c (init_sections_and_labels): Use
15489         output_asm_line_debug_info consistently.
15490         (dwarf2out_early_finish): Likewise.
15491         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
15492         type units.
15494 2018-07-11  Richard Biener  <rguenther@suse.de>
15496         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
15497         Rework father_bb setting in a way to avoid propagating constants
15498         multiple times on a loop body.
15500 2018-07-10  Mark Wielaard  <mark@klomp.org>
15502         PR debug/86459
15503         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
15505 2018-07-10  Richard Biener  <rguenther@suse.de>
15507         * hash-map.h (hash_map::iterator::operator*): Return
15508         references to key and value.
15510 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
15512         PR c++/86443
15513         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
15514         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
15515         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
15516         (gimplify_omp_for): For composite loops, move outer
15517         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
15518         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
15519         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
15520         TREE_LIST for both the original class iterator and the "last" helper
15521         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
15522         loop, remember has_decl_expr from outer composite loops for the
15523         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
15525 2018-07-09  Martin Sebor  <msebor@redhat.com>
15527         PR middle-end/77357
15528         PR middle-end/86428
15529         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
15530         accessing implicitly initialized array elements.
15531         * expr.c (string_constant): Handle string initializers of
15532         character arrays within aggregates.
15533         * gimple-fold.c (fold_array_ctor_reference): Add argument.
15534         Store element offset.  As a special case, handle zero size.
15535         (fold_nonarray_ctor_reference): Same.
15536         (fold_ctor_reference): Add argument.  Store subobject offset.
15537         * gimple-fold.h (fold_ctor_reference): Add argument.
15539 2018-07-09  Paul Koning  <ni1d@arrl.net>
15541         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
15542         (pdp11_insn_cost): New function.
15543         (pdp11_md_asm_adjust): New function.
15544         (TARGET_INVALID_WITHIN_DOLOOP): Define.
15545         (pdp11_rtx_costs): Update to match machine better.
15546         (output_addr_const_pdp11): Correct format mismatch warnings.
15547         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
15548         * config/pdp11/pdp11.md: General change to add base_cost and/or
15549         length attributes for use by new pdp11_insn_cost function.
15550         (MIN_BRANCH): Correct definition.
15551         (MIN_SOB): Ditto.
15552         (doloop_end): Use standard pattern name for looping pattern.
15553         (doloop_end_nocc): New.
15554         (movsf): Add another constraint alternative.
15555         (zero_extendqihi2): Add constraint alternatives for not in place
15556         extend.
15557         (zero_extendhisi2): Remove.
15558         (shift patterns): Add CC handling variants.
15559         (bswaphi2): New.
15560         (bswapsi2): New.
15561         (rothi3): New.
15562         (define_peephole2): New peephole to recognize mov that sets CC for
15563         subsequent test.
15565 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15567         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
15568         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
15569         wi::add.
15571 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
15573         PR c/86420
15574         * real.c (real_nextafter): Return true if result is denormal.
15576 2018-07-09  Martin Liska  <mliska@suse.cz>
15578         * common.opt: Add back wrongly removed attribute.
15580 2018-07-09  Richard Biener  <rguenther@suse.de>
15582         PR debug/86413
15583         * dwarf2out.c (gen_block_die): For an early generated DIE
15584         always output high/low PC attributes.
15586 2018-07-09  Tom de Vries  <tdevries@suse.de>
15588         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
15589         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
15590         onto VAR_DECL with abstract origin.
15592 2018-07-07  Jim Wilson  <jimw@sifive.com>
15594         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
15596 2018-07-07  Tom de Vries  <tdevries@suse.de>
15598         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
15600 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
15602         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
15603         overflow_type.
15604         (combine_bound): Use wide-int overflow calculation instead of
15605         rolling our own.
15606         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
15607         overflow_type.
15608         * fold-const.c (int_const_binop_2): Same.
15609         (extract_muldiv_1): Same.
15610         (fold_div_compare): Same.
15611         (fold_abs_const): Same.
15612         * match.pd: Same.
15613         * poly-int.h (add): Same.
15614         (sub): Same.
15615         (neg): Same.
15616         (mul): Same.
15617         * predict.c (predict_iv_comparison): Same.
15618         * profile-count.c (slow_safe_scale_64bit): Same.
15619         * simplify-rtx.c (simplify_const_binary_operation): Same.
15620         * tree-chrec.c (tree_fold_binomial): Same.
15621         * tree-data-ref.c (split_constant_offset_1): Same.
15622         * tree-if-conv.c (idx_within_array_bound): Same.
15623         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
15624         * tree-ssa-phiopt.c (minmax_replacement): Same.
15625         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
15626         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
15627         * vr-values.c (vr_values::adjust_range_with_scev): Same.
15628         * wide-int.cc (wi::add_large): Same.
15629         (wi::mul_internal): Same.
15630         (wi::sub_large): Same.
15631         (wi::divmod_internal): Same.
15632         * wide-int.h: Change overflow type to overflow_type for neg, add,
15633         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
15634         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
15635         mul_internal, divmod_internal.
15636         (overflow_type): New enum.
15637         (accumulate_overflow): New.
15639 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
15641         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
15642         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
15644 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
15646         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
15647         argument is checked for zero before entering loop, avoid checking again.
15649 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
15651         * gimplify.h (generic_expr_could_trap_p): Set as global function.
15652         * gimplify.h (generic_expr_could_trap_p): Likwise.
15653         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
15655 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
15657         PR tree-optimization/86401
15658         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
15659         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
15660         (fold_bit_and_mask): ... here.  New helper function for match.pd.
15661         * fold-const.h (fold_bit_and_mask): Declare.
15662         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
15664 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
15666         PR target/86324
15667         * target.def (translate_mode_attribute): New hook.
15668         * targhooks.h (default_translate_mode_attribute): Declare.
15669         * targhooks.c (default_translate_mode_attribute): New function.
15670         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
15671         * doc/tm.texi: Regenerate.
15672         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
15673         (rs6000_translate_mode_attribute): New function.
15675 2018-07-06  Paul Koning  <ni1d@arrl.net>
15677         * doc/md.texi (define_split): Document DONE and FAIL.
15678         (define_peephole2): Ditto.
15680 2018-07-05  Jeff Law  <law@redhat.com>
15682         PR tree-optimization/86010
15683         * tree-ssa-dse.c (compute_trims): More aggressively trim at
15684         both the head and tail of mem* and str* calls.
15686 2018-07-05  Jim Wilson  <jimw@sifive.com>
15688         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
15690 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
15692         * config/aarch64/aarch64-simd.md: correct flags text for
15693         MIN_EXPR replacement.
15695 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
15697         * configure: Regenerated.
15699 2018-07-05  Carl Love  <cel@us.ibm.com>
15701         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
15702         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
15703         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
15704         VSX_BUILTIN_DOUBLEL_V4SF.
15706 2018-07-05  Martin Sebor  <msebor@redhat.com>
15708         PR c++/86400
15709         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
15710         than its domain to compute its the upper bound of a char array.
15712 2018-07-05  Nathan Sidwell  <nathan@acm.org>
15714         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
15715         * doc/cpp.texi: Update comment.
15716         * doc/tm.texi: Rebuilt.
15717         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
15718         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
15719         * doc/extend.texi (Backwards Compatibility): Clarify it is system
15720         headers affected by extern "C".
15721         * system.h: Poison NO_IMPLICIT_EXTERN_C.
15722         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
15723         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
15724         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
15725         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
15726         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
15727         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
15728         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
15729         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
15730         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
15731         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
15732         config/sparc/sp64-elf.h, config/spu/spu.h,
15733         config/stormy16/stormy16.h, config/v850/v850.h,
15734         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
15735         define NO_IMPLICIT_EXTERN_C.
15736         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
15738 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
15740         PR target/84711
15741         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
15742         instead of GET_MODE_SIZE when comparing Units.
15744 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
15746         PR target/84711
15747         * rtlanal.c (set_noop_p): Constrain on mode change,
15748         include hard-reg-set.h
15750 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
15752         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
15754 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
15756         Revert
15757         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
15759         PR sanitizer/84250
15760         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
15761         libasan.
15762         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
15764 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
15766         PR sanitizer/84250
15767         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
15768         libasan.
15769         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
15771 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
15773         PR middle-end/86380
15774         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
15776 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
15778         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
15779         neg_*_op* variables.
15781 2018-07-04  Martin Liska  <mliska@suse.cz>
15783         * tree-switch-conversion.c: Define
15784         max_ratio_for_speed and max_ratio_for_size constants.
15786 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
15787             Martin Liska  <mliska@suse.cz>
15789         PR middle-end/66240
15790         PR target/45996
15791         PR c/84100
15792         * common.opt: Rename align options with 'str_' prefix.
15793         * common/config/i386/i386-common.c (set_malign_value): New
15794         function.
15795         (ix86_handle_option): Use it to set -falign-* options/
15796         * config/aarch64/aarch64-protos.h (struct tune_params): Change
15797         type from int to string.
15798         * config/aarch64/aarch64.c: Update default values from int
15799         to string.
15800         * config/alpha/alpha.c (alpha_override_options_after_change):
15801         Likewise.
15802         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
15803         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15804         max skip conditionally.
15805         * config/i386/freebsd.h (SUBALIGN_LOG): New.
15806         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15807         max skip conditionally.
15808         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15809         max skip conditionally.
15810         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
15811         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15812         max skip conditionally.
15813         * config/i386/i386.c (struct ptt): Change type from int to
15814         string.
15815         (ix86_default_align): Set default values.
15816         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
15817         max skip conditionally.
15818         * config/i386/iamcu.h (SUBALIGN_LOG): New.
15819         (ASM_OUTPUT_MAX_SKIP_ALIGN):
15820         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
15821         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15822         max skip conditionally.
15823         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
15824         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
15825         * config/i386/x86-64.h (SUBALIGN_LOG): New.
15826         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15827         max skip conditionally.
15828         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15829         * config/ia64/ia64.c (ia64_option_override): Set default values
15830         for alignment options.
15831         * config/m68k/m68k.c: Handle new str_align_* options.
15832         * config/mips/mips.c (mips_set_compression_mode): Change
15833         type of constants.
15834         (mips_option_override): Set default values for options.
15835         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
15836         Likewise.
15837         * config/rs6000/rs6000.c (rs6000_option_override_internal):
15838         Likewise.
15839         * config/rx/rx.c (rx_option_override): Likewise.
15840         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
15841         (LABEL_ALIGN): Use align_labels_log.
15842         (LOOP_ALIGN): Use align_loops_align.
15843         * config/s390/s390.c (s390_asm_output_function_label): Use new
15844         macros.
15845         * config/sh/sh.c (sh_override_options_after_change):
15846         Change type of constants.
15847         * config/spu/spu.c (spu_sched_init): Likewise.
15848         * config/sparc/sparc.c (sparc_option_override): Set default
15849         values for options.
15850         * config/visium/visium.c (visium_option_override): Likewise.
15851         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
15852         emit p2align format with last argument if it's not needed.
15853         * doc/invoke.texi: Document extended format of -falign-*.
15854         * final.c: Use align_labels alignment.
15855         * flags.h (struct target_flag_state): Change type to use
15856         align_flags.
15857         (struct align_flags_tuple): New.
15858         (struct align_flags): Likewise.
15859         (align_loops_log): Redefine macro to use new types.
15860         (align_loops_max_skip): Redefine macro to use new types.
15861         (align_jumps_log): Redefine macro to use new types.
15862         (align_jumps_max_skip): Redefine macro to use new types.
15863         (align_labels_log): Redefine macro to use new types.
15864         (align_labels_max_skip): Redefine macro to use new types.
15865         (align_functions_log): Redefine macro to use new types.
15866         (align_loops): Redefine macro to use new types.
15867         (align_jumps): Redefine macro to use new types.
15868         (align_labels): Redefine macro to use new types.
15869         (align_functions): Redefine macro to use new types.
15870         (align_functions_max_skip): Redefine macro to use new types.
15871         (align_loops_value): New macro.
15872         (align_jumps_value): New macro.
15873         (align_labels_value): New macro.
15874         (align_functions_value): New macro.
15875         * function.c (invoke_set_current_function_hook): Propagate
15876         alignment values from flags to global variables default in
15877         topleev.h.
15878         * ipa-icf.c (sem_function::equals_wpa): Use
15879         cl_optimization_option_eq instead of memcmp.
15880         * lto-streamer.h (cl_optimization_stream_out): Support streaming
15881         of string types.
15882         (cl_optimization_stream_in): Likewise.
15883         * optc-save-gen.awk: Support strings in cl_optimization.
15884         * opth-gen.awk: Likewise.
15885         * opts.c (finish_options): Remove error checking of invalid
15886         value ranges.
15887         (MAX_CODE_ALIGN): Remove.
15888         (MAX_CODE_ALIGN_VALUE): Likewise.
15889         (parse_and_check_align_values): New function.
15890         (check_alignment_argument): Likewise.
15891         (common_handle_option): Use check_alignment_argument.
15892         * opts.h (parse_and_check_align_values): Declare.
15893         * toplev.c (init_alignments): Remove.
15894         (read_log_maxskip): New.
15895         (parse_N_M): Likewise.
15896         (parse_alignment_opts): Likewise.
15897         (backend_init_target): Remove usage of init_alignments.
15898         * toplev.h (parse_alignment_opts): Declare.
15899         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
15900         argument.
15901         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
15902         * tree.c (cl_option_hasher::equal): New.
15903         * varasm.c: Use new global macros.
15905 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
15907         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
15908         Use a simpler align directive also if MAXSKIP = ALIGN-1.
15909         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15910         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15911         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15912         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15913         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
15914         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
15915         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
15916         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15917         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15918         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15919         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15921 2018-07-04  Martin Liska  <mliska@suse.cz>
15922             Jonathan Wakely  <jwakely@redhat.com>
15924         * coverage.c: Use correct type.
15925         * doc/invoke.texi: Language correction.
15927 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
15929         PR target/85620
15930         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
15931         ENDBRANCH for non-tail call which may return via indirect branch.
15932         * doc/extend.texi: Document indirect_return attribute.
15934 2018-07-03  Martin Sebor  <msebor@redhat.com>
15936         PR tree-optimization/86274
15937         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
15938         precondition.
15939         (format_floating): Correct handling of infinities and NaNs.
15941 2018-07-03  Martin Sebor  <msebor@redhat.com>
15943         * print-tree.c (print_real_cst): New function.
15944         (print_node_brief): Call it.
15945         (print_node): Ditto.
15947 2018-07-03  Jeff Law  <law@redhat.com>
15949         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
15950         into a single pattern.
15952         * config/h8300/h8300.md (ors code_iterator): New.
15953         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
15954         a single pattern and single splitter.
15955         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
15956         (iorqi3_1, xorqi3_1): Likewise.
15957         (iorqi3, xorqi3 expanders): Similarly.
15959         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
15960         (movmd_internal) into a single pattern using the P mode iterator.
15961         (movmd splitters): Similarly.
15962         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
15963         (movsd splitters): Similarly.
15965         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
15966         ADDB, ADDW and ADDL into a single ADD attribute which selects the
15967         right table based on the size of the operand.
15968         * config/h8300/h8300.md (length_table): Corresponding changes. All
15969         references to "addb", "addw" and "addl" changed to "add".
15970         (btst patterns): Merge two variants into a single pattern.
15971         (tstqi, tsthi): Likewise.
15972         (addhi3_incdec, addsi3_incdec): Likewise.
15973         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
15974         (mulhi3, mulsi3): Likewise.
15975         (udivhi3, udivsi3): Likewise.
15976         (divhi3, divsi3): Likewise.
15977         (andorqi3, andorhi3, andorsi3): Likewise.
15979 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
15981         PR target/85694
15982         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
15983         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
15985 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
15987         PR tree-optimization/85694
15988         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
15989         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
15990         UNSPEC_URHADD.
15991         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
15992         (<u>avg<mode>3_ceil): New patterns.
15994 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
15996         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
15997         scan-tree-dump directive.
15998         * gcc.dg/vect/slp-perm-2.c: Likewise.
15999         * gcc.dg/vect/slp-perm-3.c: Likewise.
16000         * gcc.dg/vect/slp-perm-5.c: Likewise.
16001         * gcc.dg/vect/slp-perm-6.c: Likewise.
16002         * gcc.dg/vect/slp-perm-7.c: Likewise.
16003         * gcc.dg/vect/slp-perm-8.c: Likewise.
16005 2018-07-03  Marek Polacek  <polacek@redhat.com>
16007         PR middle-end/86202
16008         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
16010 2018-07-03  Richard Biener  <rguenther@suse.de>
16012         PR ipa/86389
16013         * tree-ssa-structalias.c (find_func_clobbers): Properly
16014         handle indirect calls.
16016 2018-07-03  Jeff Law  <law@redhat.com>
16018         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
16019         (shifts): New code iterator.
16020         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
16021         expander.  Fix HImode handling on H8/SX.
16022         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
16023         (subqi3, subhi3, subsi3 expanders): Likewise.
16024         (andqi3, andhi3, andsi3 expanders): Likewise.
16025         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
16026         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
16027         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
16028         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
16029         (zero_extendqihi2, zero_extendqisi2): Likewise.
16030         (extendqihi2, extendqisi2): Likewise.
16031         (rotlqi3, rotlhi3, rotlsi3): Likewise.
16032         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
16033         (rotlqi3_1, rotlhi3_1): Likewise.
16034         (logicalhi3_sn, logicalsi3_sn): Likewise.
16035         (logicalhi3, logicalsi3): Likewise.
16037 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16039         * tree-vect-patterns.c (vect_recog_rotate_pattern)
16040         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
16041         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
16042         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
16043         type to append_pattern_def_seq instead of creating a stmt_vec_info
16044         directly.
16045         (build_mask_conversion): Likewise.  Remove vinfo argument.
16046         (vect_add_conversion_to_patterm): Likewise, renaming to...
16047         (vect_add_conversion_to_pattern): ...this.
16048         (vect_recog_mask_conversion_pattern): Update call to
16049         build_mask_conversion.  Pass the vector type to
16050         append_pattern_def_seq here too.
16051         (vect_recog_gather_scatter_pattern): Update call to
16052         vect_add_conversion_to_pattern.
16054 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16056         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
16057         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16058         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
16059         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
16060         STMT_VINFO_PATTERN_DEF_SEQ to null here.
16061         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
16062         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
16063         append_pattern_def_seq instead of new_pattern_def_seq.
16064         (vect_recog_divmod_pattern): Do both of the above.
16065         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
16066         is null.
16068 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16070         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
16071         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
16072         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
16073         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
16074         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
16075         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
16076         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
16077         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
16078         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
16079         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
16080         parameter with a single stmt_vec_info.
16081         (vect_recog_func_ptr): Likewise.
16082         (vect_recog_gather_scatter_pattern): Likewise, folding in...
16083         (vect_try_gather_scatter_pattern): ...this.
16084         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
16085         the stmt_vec_info of the statement to be matched.  Don't clear
16086         STMT_VINFO_RELATED_STMT.
16087         (vect_pattern_recog): Update call accordingly.
16089 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16091         PR tree-optimization/85694
16092         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
16093         (uavgM3_ceil): Document new optabs.
16094         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
16095         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
16096         functions.
16097         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
16098         (savg_ceil_optab): New optabs.
16099         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
16100         (vect_vect_recog_func_ptrs): Add it.
16101         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
16102         constant directly from the associated lhs.
16104 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16106         * tree-vect-patterns.c (vect_split_statement): New function.
16107         (vect_convert_input): Use it to try to split an existing cast.
16109 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16111         * poly-int.h (print_hex): New function.
16112         * dumpfile.h (dump_dec, dump_hex): Declare.
16113         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
16114         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
16115         min_input_precision, operation_precision and operation_sign.
16116         * tree-vect-patterns.c (vect_get_range_info): New function.
16117         (vect_same_loop_or_bb_p, vect_single_imm_use)
16118         (vect_operation_fits_smaller_type): Delete.
16119         (vect_look_through_possible_promotion): Add an optional
16120         single_use_p parameter.
16121         (vect_recog_over_widening_pattern): Rewrite to use new
16122         stmt_vec_info infomration.  Handle one operation at a time.
16123         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
16124         (vect_truncatable_operation_p, vect_set_operation_type)
16125         (vect_set_min_input_precision): New functions.
16126         (vect_determine_min_output_precision_1): Likewise.
16127         (vect_determine_min_output_precision): Likewise.
16128         (vect_determine_precisions_from_range): Likewise.
16129         (vect_determine_precisions_from_users): Likewise.
16130         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
16131         (vect_vect_recog_func_ptrs): Put over_widening first.
16132         Add cast_forwprop.
16133         (vect_pattern_recog): Call vect_determine_precisions.
16135 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
16137         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
16138         statements that have been replaced by further pattern statements.
16139         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
16141 2018-07-03  Richard Biener  <rguenther@suse.de>
16143         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
16144         always set *dt.  Dump vectype in vectype overload.
16145         * dumpfile.h (dump_gimple_expr): New function.
16146         (dump_gimple_expr_loc): Likewise.
16147         * dumpfile.c (dump_gimple_expr): New function.
16148         (dump_gimple_expr_loc): Likewise.
16150 2018-07-02  Jeff Law  <law@redhat.com>
16152         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
16153         the H8/300, H8/300H and H8/S variants into a single pattern.
16154         (movhi_h8300, movqi_h8300hs): Similarly.
16155         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
16156         (QHI mode iterator): New.
16158         * config/h8300/h8300.md: Remove trailing whitespace.
16160 2018-07-02  Jim Wilson  <jimw@sifive.com>
16162         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
16163         instead of emit_insn for interrupt returns.
16164         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
16165         (riscv_sret, riscv_uret): Likewise.
16167 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
16169         * pretty-print.c (selftest::test_pp_format): Move save and restore
16170         of quotes to class auto_fix_quotes, and add an instance.
16171         * selftest.c: Include "intl.h".
16172         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
16173         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
16174         * selftest.h (selftest::auto_fix_quotes): New class.
16176 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
16178         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
16179         (aarch64_sve_prepare_conditional_op): Remove.
16180         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
16181         Allow aarch64_simd_reg_or_zero as select operand; remove
16182         the aarch64_sve_prepare_conditional_op call.
16183         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
16184         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
16185         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
16186         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
16187         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
16188         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
16189         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
16190         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
16191         and a splitters to match all of the *_any patterns.
16192         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
16194         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
16195         (SVE_COND_FP_BINARY_REV): Remove.
16196         (sve_int_op_rev, sve_fp_op_rev): New.
16197         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
16198         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
16199         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
16200         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
16201         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
16202         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
16203         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
16204         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
16205         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
16207         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
16208         Remove match_dup 1 from the inner unspec.
16209         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
16211         * config/aarch64/aarch64.md (movprfx): New attr.
16212         (length): Default movprfx to 8.
16213         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
16214         (*madd<SVE_I>, *msub<SVE_I): Likewise.
16215         (*<su>mul<SVE_I>3_highpart): Likewise.
16216         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
16217         (*v<ASHIFT><SVE_I>3): Likewise.
16218         (*<su><MAXMIN><SVE_I>3): Likewise.
16219         (*<su><MAXMIN><SVE_F>3): Likewise.
16220         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
16221         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
16222         (*div<SVE_F>4): Likewise.
16224 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
16226         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
16227         in dump string.
16229 2018-07-02  Richard Biener  <rguenther@suse.de>
16231         PR tree-optimization/86363
16232         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
16233         memset argument refers to a non-variable address.
16235 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
16237         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
16238         {PLUS,MINUS}_EXPR code to...
16239         (adjust_symbolic_bound): ...here,
16240         (combine_bound): ...here,
16241         (set_value_range_with_overflow): ...and here.
16243 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
16245         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
16246         code...
16247         (extract_range_from_abs_expr): ...here.
16249 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
16251         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
16252         -fno-omit-frame-pointer when not optimizing.
16254 2018-07-02  Martin Liska  <mliska@suse.cz>
16256         PR ipa/86279
16257         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
16258         (propagate_nothrow): Likewise.
16260 2018-07-02  Martin Liska  <mliska@suse.cz>
16262         PR ipa/86323
16263         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
16265 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
16267         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
16268         function in r262149, changing "loc" param from source_location to
16269         const dump_location_t &.
16270         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
16271         declaration, as above.
16273 2018-07-01  Paul Koning  <ni1d@arrl.net>
16275         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
16276         -munit-asm, -mgnu-asm, -mdec-asm.
16277         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
16278         (pdp11_output_labelref): New.
16279         (pdp11_output_def): New.
16280         (pdp11_output_addr_vec_elt): New.
16281         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
16282         %# and %@ format codes.
16283         (pdp11_option_override): New.
16284         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
16285         (pdp11_output_ident): New.
16286         (pdp11_asm_named_section): New.
16287         (pdp11_asm_init_sections): New.
16288         (pdp11_file_start): New.
16289         (pdp11_file_end): New.
16290         (output_ascii): Use .ascii/.asciz for -mdec-asm.
16291         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
16292         %o, like %c but octal.
16293         (pdp11_option_override): New.
16294         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
16295         -mdec-asm.
16296         (DATA_SECTION_ASM_OP): Ditto.
16297         (READONLY_DATA_SECTION_ASM_OP): New.
16298         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
16299         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
16300         (ASM_OUTPUT_LABELREF): Ditto.
16301         (ASM_OUTPUT_DEF): Ditto.
16302         (ASM_OUTPUT_EXTERNAL): New.
16303         (ASM_OUTPUT_SOURCE_FILENAME): New.
16304         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
16305         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
16306         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
16307         %# and %@ format codes.
16308         * config/pdp11/pdp11.opt (mgnu-asm): New.
16309         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
16310         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
16311         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
16313 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
16315         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
16316         dereferencing path[] beyond its length.
16317         (debug_path): New.
16318         (debug_all_paths): New.
16319         (rewire_first_differing_edge): New.
16320         (adjust_paths_after_duplication): New.
16321         (duplicate_thread_path): Call adjust_paths_after_duplication.
16322         Add new argument.
16323         (thread_through_all_blocks): Add new argument to
16324         duplicate_thread_path.
16326 2018-06-30  Jim Wilson  <jimw@sifive.com>
16328         * config/riscv/predicates.md (p2m1_shift_operand): New.
16329         (high_mask_shift_operand): New.
16330         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
16331         pattern using p2m1_shift_operand.
16332         (lshsi3_zero_extend_3+2): New combiner pattern using
16333         high_mask_shift_operand.
16335 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16337         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
16338         split out from...
16339         (vect_recog_rotate_pattern): ...here.
16340         (vect_convert_input): Try to insert casts of invariants in the
16341         preheader.
16342         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
16343         preheader to be empty.
16345 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16347         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
16348         vector type.  If given, install it in the new statement's
16349         STMT_VINFO_VECTYPE.
16350         (vect_element_precision): New function.
16351         (vect_unpromoted_value): New struct.
16352         (vect_unpromoted_value::vect_unpromoted_value): New function.
16353         (vect_unpromoted_value::set_op): Likewise.
16354         (vect_look_through_possible_promotion): Likewise.
16355         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
16356         (vect_widened_op_tree, vect_convert_input): Likewise.
16357         (vect_convert_inputs, vect_convert_output): Likewise.
16358         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
16359         to handle the optional cast of the multiplication result and
16360         vect_widened_op_tree to detect the widened multiplication itself.
16361         Do not require the input and output of promotion casts to have
16362         the same sign, but base the signedness of the operation on the
16363         input rather than the result.  If the pattern includes two
16364         promotions, check that those promotions have the same sign.
16365         Do not restrict the MULT_EXPR handling to a double-width result;
16366         handle quadruple-width results and wider.  Use vect_convert_inputs
16367         to convert the inputs to the common type.
16368         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
16369         to handle the optional cast of the ABS result.  Also allow a sign
16370         change or a sign extension between the ABS and MINUS.
16371         Use vect_widened_op_tree to detect the widened subtraction and use
16372         vect_convert_inputs to convert the inputs to the common type.
16373         (vect_handle_widen_op_by_const): Delete.
16374         (vect_recog_widen_op_pattern): New function.
16375         (vect_recog_widen_mult_pattern): Use it.
16376         (vect_recog_widen_shift_pattern): Likewise.
16377         (vect_recog_widen_sum_pattern): Use
16378         vect_look_through_possible_promotion to handle the promoted
16379         PLUS_EXPR operand.
16381 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16383         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
16384         the containing gimple_seq *.
16385         * gimple-iterator.h (gsi_for_stmt): Declare it.
16386         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
16387         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
16388         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
16389         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
16390         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
16391         checks.
16392         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
16393         split out from...
16394         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
16395         statement being replaced is part of an existing pattern
16396         definition sequence, inserting the new pattern statements before
16397         the original one.
16398         (vect_pattern_recog_1): Don't return a bool.  If the statement
16399         is already part of a pattern, instead apply pattern matching
16400         to the pattern definition statements.  Don't clear the
16401         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
16402         (vect_pattern_recog): Don't break after the first match;
16403         continue processing the pattern definition statements instead.
16404         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
16406 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16408         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16409         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16410         (vect_recog_widen_sum_pattern): Use it.
16412 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16414         * tree-vect-loop.c (vectorizable_reduction): Assert that the
16415         phi is not a pattern statement and has not been replaced by
16416         a pattern statement.
16417         * tree-vect-patterns.c (type_conversion_p): Don't check
16418         STMT_VINFO_IN_PATTERN_P.
16419         (vect_recog_vector_vector_shift_pattern): Likewise.
16420         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
16421         the pattern statement rather than the original statement; check
16422         directly for a WIDEN_MULT_EXPR here.
16423         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
16424         vect_is_simple_use to return the pattern statement rather
16425         than the original statement; use is_pattern_stmt_p to check
16426         for such a pattern statement.
16427         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
16428         to return the pattern statement rather than the original statement;
16429         don't do the same transformation here.
16430         (vect_is_simple_use): If the defining statement has been replaced
16431         by a pattern statement, return the pattern statement instead.
16432         Remove the corresponding (local) transformation from the vectype
16433         overload.
16435 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
16437         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
16438         end and default to null.
16439         * tree-vect-loop.c (vect_create_epilog_for_reduction)
16440         (vectorizable_reduction): Update calls accordingly, dropping the
16441         gimple ** argument if the passed-back statement isn't needed.
16442         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
16443         (vect_recog_rotate_pattern): Likewise.
16444         (vect_recog_mask_conversion_pattern): Likewise.
16445         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
16446         (vect_mask_constant_operand_p): Likewise.
16447         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
16448         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
16449         (get_group_load_store_type, get_load_store_type): Likewise.
16450         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
16451         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
16452         (vectorizable_conversion, vectorizable_assignment): Likewise.
16453         (vectorizable_shift, vectorizable_operation): Likewise.
16454         (vectorizable_store, vect_is_simple_cond): Likewise.
16455         (vectorizable_condition, vectorizable_comparison): Likewise.
16456         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
16457         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
16458         and move it to the end.  Cope with null def_stmt_outs.
16460 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16462         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
16464 2018-06-29  Jeff Law  <law@redhat.com>
16466         * config/v850/v850.c (v850_legitimate_address_p): Handle large
16467         displacements for TARGET_V850E2V3 and newer.
16468         (TARGET_LRA_P): Remove.  Defaults to LRA now.
16469         * config/v850/v850.md (sign23byte_load): Remove.
16470         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
16471         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
16473 2018-06-29  Martin Liska  <mliska@suse.cz>
16475         PR lto/85759
16476         * coverage.c (coverage_init): Mangle full path name.
16477         * doc/invoke.texi: Document the change.
16478         * gcov-io.c (mangle_path): New.
16479         * gcov-io.h (mangle_path): Likewise.
16480         * gcov.c (mangle_name): Use mangle_path for path mangling.
16482 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16484         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
16485         if starting source register is not even.
16487 2018-06-29  Martin Liska  <mliska@suse.cz>
16489         PR tree-optimization/86263
16490         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
16491         Make edge redirection.
16493 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
16495         * dumpfile.c (dump_loc): Add indentation based on scope depth.
16496         (dump_scope_depth): New variable.
16497         (get_dump_scope_depth): New function.
16498         (dump_begin_scope): New function.
16499         (dump_end_scope): New function.
16500         * dumpfile.h (get_dump_scope_depth): New declaration.
16501         (dump_begin_scope): New declaration.
16502         (dump_end_scope): New declaration.
16503         (class auto_dump_scope): New class.
16504         (AUTO_DUMP_SCOPE): New macro.
16505         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
16506         AUTO_DUMP_SCOPE.
16508 2018-06-29  Richard Biener  <rguenther@suse.de>
16510         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
16511         compute_all_dependences succeeds.
16512         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
16513         exceed --param loop-max-datarefs-for-datadeps.
16515 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
16517         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
16519 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
16521         PR target/86348
16522         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
16523         alternative 0 in preferred_for_speed attribute.
16525 2018-06-28  Paul Koning  <ni1d@arrl.net>
16527         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
16528         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
16529         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
16530         * config/pdp11/pdp11.md: Correct "length" attribute calculation
16531         for shift insn patterns.
16533 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
16535         * cgraph.c (cgraph_node::get_body): Replace assignments to
16536         "dump_file" with calls to set_dump_file.
16537         * dumpfile.c (alt_dump_file): Make static, and group with...
16538         (alt_flags): ...this definition.
16539         (dumps_are_enabled): New variable.
16540         (refresh_dumps_are_enabled): New function.
16541         (set_dump_file): New function.
16542         (set_alt_dump_file): New function.
16543         (gcc::dump_manager::dump_start): Replace assignments to
16544         "dump_file" and "alt_dump_file" with calls to set_dump_file and
16545         set_alt_dump_file.
16546         (gcc::dump_manager::dump_finish): Likewise.
16547         * dumpfile.h (alt_dump_file): Delete decl.
16548         (dumps_are_enabled): New variable decl.
16549         (set_dump_file): New function decl.
16550         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
16551         global.
16552         * tree-nested.c (lower_nested_functions): Replace assignments to
16553         "dump_file" with calls to set_dump_file.
16555 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
16557         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
16558         goto_locus of each outgoing edge of each basic block.
16560 2018-06-28  Richard Biener  <rguenther@suse.de>
16562         * dwarf2out.c (decl_scope_table): Remove.
16563         (push_decl_scope): Likewise.
16564         (pop_decl_scope): Likewise.
16565         (gen_type_die_for_member): Do not call push/pop_decl_scope.
16566         (gen_struct_or_union_type_die): Likewise.
16567         (gen_tagged_type_die): Likewise.
16568         (dwarf2out_init): Do not initialize decl_scope_table.
16569         (dwarf2out_c_finalize): Do not free it.
16571 2018-06-28  Richard Biener  <rguenther@suse.de>
16573         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
16574         deciding whether to not re-use a DIE.
16576 2018-06-28  Richard Biener  <rguenther@suse.de>
16578         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
16579         DW_AT_abstract_origin attribute.
16581 2018-06-28  Martin Liska  <mliska@suse.cz>
16583         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
16584         Use newly introduced constants.
16585         * tree-switch-conversion.h (struct jump_table_cluster):
16586         Define max_ratio_for_size and max_ratio_for_speed.
16588 2018-06-28  Martin Liska  <mliska@suse.cz>
16590         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
16591         Add new checking assert to catch invalid state.
16592         (jump_table_cluster::can_be_handled): Handle single case
16593         clusters.
16594         (jump_table_cluster::is_beneficial): Bail out for such case.
16595         (bit_test_cluster::find_bit_tests):
16596         Add new checking assert to catch invalid state.
16597         (bit_test_cluster::can_be_handled): Handle single case
16598         clusters.
16599         (bit_test_cluster::is_beneficial): Bail out for such case.
16600         (switch_decision_tree::analyze_switch_statement):
16601         Fix comment.
16603 2018-06-28  Martin Liska  <mliska@suse.cz>
16605         * common.opt: Introduce -completion option.
16606         * gcc.c (driver_handle_option): Handle it.
16607         (driver::main): Print completions if completion
16608         is set.
16609         * opt-suggestions.c (option_proposer::get_completions):
16610         New function.
16611         (option_proposer::suggest_completion): Likewise.
16612         (option_proposer::find_param_completions): Likewise.
16613         (verify_autocompletions): Likewise.
16614         (test_completion_valid_options): Likewise.
16615         (test_completion_valid_params): Likewise.
16616         (in_completion_p): Likewise.
16617         (empty_completion_p): Likewise.
16618         (test_completion_partial_match): Likewise.
16619         (test_completion_garbage): Likewise.
16620         (opt_proposer_c_tests): Likewise.
16621         * opt-suggestions.h: Declare new functions.
16622         * opts.c (common_handle_option): Handle OPT__completion_.
16623         * selftest-run-tests.c (selftest::run_tests): Add
16624         opt_proposer_c_tests.
16625         * selftest.c (assert_str_startswith): New.
16626         * selftest.h (assert_str_startswith): Likewise.
16627         (opt_proposer_c_tests): New.
16628         (ASSERT_STR_STARTSWITH): Likewise.
16630 2018-06-28  Martin Liska  <mliska@suse.cz>
16632         * Makefile.in: Add opt-suggestions.o.
16633         * gcc-main.c: Include opt-suggestions.h.
16634         * gcc.c (driver::driver): Likewise.
16635         (driver::~driver): Remove m_option_suggestions.
16636         (driver::build_option_suggestions): Moved to option_proposer.
16637         (driver::suggest_option): Likewise.
16638         (driver::handle_unrecognized_options): Use option_proposer.
16639         * gcc.h (class driver): Add new memver m_option_proposer.
16640         * opt-suggestions.c: New file.
16641         * opt-suggestions.h: New file.
16643 2018-06-28  Martin Liska  <mliska@suse.cz>
16645         * vec.h (class auto_string_vec): New (moved from auto_argvec).
16646         (auto_string_vec::~auto_string_vec): Likewise.
16648 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
16650         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
16651         prevent_decl_creation_for_types fields up and add reset_location field.
16652         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
16653         statement if id->reset_location is true.
16654         (copy_edges_for_bb): Do not set goto_locus on the new edges if
16655         id->reset_location is true.
16656         (copy_phis_for_bb): Force input_location on the arguments if
16657         id->reset_location is true.
16658         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
16659         is set on the function to be inlined.
16661 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
16663         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
16665 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
16667         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
16668         registers for Pmode.
16669         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
16670         hard registers for the clobbered pseudo.
16672 2018-06-27  Paul Koning  <ni1d@arrl.net>
16674         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
16675         mutually exclusive options.
16676         * config/pdp11/constraints.md (h): New constraint.
16677         (O): Update definition to match shift code generation.
16678         (D): New constraint.
16679         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
16680         (CCFP): Remove.
16681         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
16682         function.
16683         (output_jump): Change arguments.
16684         (pdp11_fixed_cc_regs): New function.
16685         (pdp11_cc_mode): Ditto.
16686         (pdp11_expand_shift): Ditto.
16687         (pdp11_assemble_shift): Ditto.
16688         (pdp11_small_shift): Ditto.
16689         (pdp11_branch_cost): Remove.
16690         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
16691         from output.
16692         (pdp11_register_move_cost): Update for CC registers.
16693         (pdp11_rtx_costs): Add case for LSHIFTRT.
16694         (pdp11_output_jump): Add CCNZ mode conditional branches.
16695         (notice_update_cc_on_set): Remove.
16696         (pdp11_cc_mode): New function.
16697         (simple_memory_operand): Correct pre/post decrement case.
16698         (no_side_effect_operand): New function.
16699         (pdp11_regno_reg_class): Add CC_REGS class.
16700         (pdp11_fixed_cc_regs): New function.
16701         (pdp11_small_shift): New function.
16702         (pdp11_expand_shift): New function to expand shift insns.
16703         (pdp11_assemble_shift): New function to output shifts.
16704         (pdp11_branch_cost): Remove.
16705         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
16706         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
16707         (WCHAR_TYPE): Ditto.
16708         (PTRDIFF_TYPE): Ditto.
16709         (ADJUST_INSN_LENGTH): New macro.
16710         (FIXED_REGISTERS): Add CC registers.
16711         (CALL_USED_REGISTERS): Ditto.
16712         (reg_class): Ditto.
16713         (REG_CLASS_NAMES): Ditto.
16714         (REG_CLASS_CONTENTS): Ditto.
16715         (SELECT_CC_MODE): Use new function.
16716         (TARGET_FLAGS_REGNUM): New macro.
16717         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
16718         (cc0_reg_rtx): Remove.
16719         (CC_STATUS_MDEP): Remove.
16720         (CC_STATUS_MDEFP_INIT): Remove.
16721         (CC_IN_FPU): Remove.
16722         (NOTICE_UPDATE_CC): Remove.
16723         (REGISTER_NAMES): Add CC registers.
16724         (BRANCH_COST): Change to constant 1.
16725         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
16726         handling.
16727         * config/pdp11/pdp11.opt (mbcopy): Remove.
16728         (mbcopy-builtin): Remove.
16729         (mbranch-cheap): Remove.
16730         (mbranch-expensive): Remove.
16731         * config/pdp11/predicates.md (expand_shift_operand): Update to
16732         match shift code generation.
16733         (ccnz_operator): New predicate.
16734         * doc/invoke.texi (PDP-11 Options): Remove deleted options
16735         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
16736         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
16737         exclusive options.
16738         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
16739         description of O constraint.
16741 2018-06-27  Jeff Law  <law@redhat.com>
16742             Austin Law  <austinklaw@gmail.com>
16744         * config/v850/v850.md (addsi3_set_flags): New pattern.
16745         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
16746         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
16747         (zero_extendhisi2_v850_set_flags): Likewise.
16748         (zero_extendqisi2_v850_set_flags): Likewise.
16749         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
16750         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
16751         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
16753         * config/v850/v850-protos.h (notice_update_cc): Remove.
16754         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
16755         (v850_print_operand): Handle 'D' and "d".
16756         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
16757         Add handling of arithmetic/logical operations compared against zero.
16758         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
16759         Do not look at v850_compare_op, instead get mode from last argument.
16760         (v850_gen_compare): Remove
16761         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
16762         after reload for prologue insns.
16763         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
16764         patterns.
16765         (construct_save_jarl): Likewise.
16766         (TARGET_FLAGS_REGNUM): Define.
16767         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
16768         (NOTICE_UPDATE_CC): Remove.
16769         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
16770         than cc0.  Conditionalize on reload_completed.
16771         (cmpsi_insn, setfcc_insn): Likewise.
16772         (tst1 splitter): Turn into define_and_split which sets the flags
16773         after reload.
16774         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
16775         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
16776         (cstoresf4, cstoredf4): Clobber the flags.
16777         (cmpsi, cmpsf, cmpdf): Remove expanders.
16778         (setf_insn): Remove pattern.
16779         (addsi3): Turn into define_and_split which clobbers the flags after
16780         reload and a suitable pattern (addsi3_clobber_flags) for use after
16781         reload.
16782         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
16783         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
16784         (ashrsi3, ashrsi3_v850e2): Likewise.
16785         (bins): Clobber the flags.
16786         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
16787         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
16788         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
16789         (call_value_internal_short, call_value_internal_long): Likewise.
16790         (callt_save_interrupt, callt_return_interrupt): Likewise.
16791         (save_interrupt, return_interrupt): Likewise.
16792         (callt_save_all_interrupt, save_all_interrupt): Likewise.
16793         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
16794         (restore_all_interrupt, _restore_all_interrupt): Likewise.
16795         (All FP comparisons): Only allow after reload has completed.
16796         (trfsr): Likewise.
16797         (divh, divhu): Tweak output template.
16798         (branch_z_normal, branch_z_invert): Remove
16799         (branch_nz_normal, branch_nz_invert): Likewise.
16800         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
16802         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
16803         * config/v850/v850.c (notice_update_cc): Remove.
16804         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
16805         (CC_NO_CARRY): Likewise.
16806         (NOTICE_UPDATE_CC): Define to nothing.
16807         * config/v850/v850.md: Remove block comment on cc0 handling
16808         Remove "cc" attribute from all patterns.  Remove cc_status handling
16809         from all patterns.  Minor formatting fixes.
16811 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16813         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
16814         (cortex-a76.cortex-a55): Likewise.
16815         * config/aarch64/aarch64-tune.md: Regenerate.
16816         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
16817         cortex-a76.cortex-a55.
16819 2018-06-27  Jeff Law  <law@redhat.com>
16821         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
16822         (MULTILIB_DIRNAMES): Similarly.
16824 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
16826         * gimple.h (gimple_return_retbnd): Delete.
16827         (gimple_return_set_retbnd): Likewise.
16828         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
16829         gimple_return_set_retbnd.
16830         * gimple-pretty-print.c (dump_gimple_return): Remove call to
16831         gimple_return_retbnd and adjust.
16832         * tree-inline.h (struct copy_body_data): Remove retbnd field.
16833         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
16834         Explicitly return NULL in a couple more cases.  Move assertion
16835         on debug statements and remove unreachable code.
16836         (reset_debug_binding): Do not test id->retbnd.
16837         (expand_call_inline): Do not set it.
16839 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
16841         * configure.ac: Add --disable-gcov option.
16842         * configure: Regenerate.
16843         * Makefile.in: Honour @enable_gcov@.
16844         * doc/install.texi: Document --disable-gcov.
16846 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16848         * config/arm/arm-cpus.in (cortex-a76): New entry.
16849         (cortex-a76.cortex-a55): Likewise.
16850         * config/arm/arm-tables.opt: Regenerate.
16851         * config/arm/arm-tune.md: Likewise.
16852         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
16853         * doc/invoke.texi (ARM Options): Document cortex-a76 and
16854         cortex-a76.cortex-a55.
16856 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
16858         PR target/85769
16859         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
16861 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
16863         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
16864         comment.
16865         (EPILOGUE_USES): Likewise.
16867 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
16869         * tree-inline.c (remap_location): New function extracted from...
16870         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
16871         (copy_phis_for_bb): ...here.  Call remap_location.
16872         (copy_cfg_body): Adjust call to copy_edges_for_bb.
16874 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
16876         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
16877         unaligned vsx for 16B memset.
16879 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
16881         PR target/86285
16882         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
16883         ieee128_float_type_node to long_double_type_node unless
16884         TARGET_LONG_DOUBLE_128 is set.
16886 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
16888         * cfgloop.c (get_loop_location): Convert return type from
16889         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
16890         by implicit construction from rtx_insn *, and using
16891         dump_user_location_t::from_function_decl for the fallback case.
16892         * cfgloop.h (get_loop_location): Convert return type from
16893         location_t to dump_user_location_t.
16894         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
16895         dump_printf_loc to pass in a dump_location_t rather than a
16896         location_t, via the gimple stmt.
16897         * coverage.c (get_coverage_counts): Update calls to
16898         dump_printf_loc to pass in dump_location_t rather than a
16899         location_t.
16900         * doc/optinfo.texi (Dump types): Convert example of
16901         dump_printf_loc from taking "locus" to taking "insn".  Update
16902         description of the "_loc" calls to cover dump_location_t.
16903         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
16904         "selftest.h".
16905         (dump_user_location_t::dump_user_location_t): New constructors,
16906         from gimple *stmt and rtx_insn *.
16907         (dump_user_location_t::from_function_decl): New function.
16908         (dump_loc): Make static.
16909         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
16910         const dump_location_t &.
16911         (dump_generic_expr_loc): Delete.
16912         (dump_printf_loc): Convert param "loc" from location_t to
16913         const dump_location_t &.
16914         (selftest::test_impl_location): New function.
16915         (selftest::dumpfile_c_tests): New function.
16916         * dumpfile.h: Include "profile-count.h".
16917         (class dump_user_location_t): New class.
16918         (struct dump_impl_location_t): New struct.
16919         (class dump_location_t): New class.
16920         (dump_printf_loc): Convert 2nd param from source_location to
16921         const dump_location_t &.
16922         (dump_generic_expr_loc): Delete.
16923         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
16924         const dump_location_t &.
16925         * gimple-fold.c (fold_gimple_assign): Update call to
16926         dump_printf_loc to pass in a dump_location_t rather than a
16927         location_t, via the gimple stmt.
16928         (gimple_fold_call): Likewise.
16929         * gimple-loop-interchange.cc
16930         (loop_cand::analyze_iloop_reduction_var): Update for change to
16931         check_reduction_path.
16932         (tree_loop_interchange::interchange): Update for change to
16933         find_loop_location.
16934         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
16935         change in return-type of find_loop_location.
16936         (graphite_regenerate_ast_isl): Likewise.
16937         * graphite-optimize-isl.c (optimize_isl): Likewise.
16938         * graphite.c (graphite_transform_loops): Likewise.
16939         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
16940         pass in a dump_location_t rather than a location_t, via the
16941         gimple stmt.
16942         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16943         * ipa.c (walk_polymorphic_call_targets): Likewise.
16944         * loop-unroll.c (report_unroll): Convert "locus" param from
16945         location_t to dump_location_t.
16946         (decide_unrolling): Update for change to get_loop_location's
16947         return type.
16948         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
16949         location_t to dump_user_location_t.
16950         (grid_find_single_omp_among_assignments_1): Updates calls to
16951         dump_printf_loc to pass in a dump_location_t rather than a
16952         location_t, via the gimple stmt.
16953         (grid_parallel_clauses_gridifiable): Convert "tloc" from
16954         location_t to dump_location_t.  Updates calls to dump_printf_loc
16955         to pass in a dump_location_t rather than a location_t, via the
16956         gimple stmt.
16957         (grid_inner_loop_gridifiable_p): Likewise.
16958         (grid_dist_follows_simple_pattern): Likewise.
16959         (grid_gfor_follows_tiling_pattern): Likewise.
16960         (grid_target_follows_gridifiable_pattern): Likewise.
16961         (grid_attempt_target_gridification): Convert initialization
16962         of local "grid" from memset to zero-initialization; FIXME: does
16963         this require C++11?  Update call to dump_printf_loc to pass in a
16964         optinfo_location rather than a location_t, via the gimple stmt.
16965         * profile.c (read_profile_edge_counts): Updates call to
16966         dump_printf_loc to pass in a dump_location_t rather than a
16967         location_t
16968         (compute_branch_probabilities): Likewise.
16969         * selftest-run-tests.c (selftest::run_tests): Call
16970         dumpfile_c_tests.
16971         * selftest.h (dumpfile_c_tests): New decl.
16972         * tree-loop-distribution.c (pass_loop_distribution::execute):
16973         Update for change in return type of find_loop_location.
16974         * tree-parloops.c (parallelize_loops): Likewise.
16975         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
16976         "locus" from location_t to dump_user_location_t.
16977         (canonicalize_loop_induction_variables): Likewise.
16978         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
16979         for change in return type of find_loop_location.
16980         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
16981         to dump_printf_loc to pass in a dump_location_t rather than a
16982         location_t, via the stmt.
16983         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
16984         Likewise.
16985         * tree-vect-loop-manip.c (find_loop_location): Convert return
16986         type from source_location to dump_user_location_t.
16987         (vect_do_peeling): Update for above change.
16988         (vect_loop_versioning): Update for change in type of
16989         vect_location.
16990         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
16991         from location_t to dump_user_location_t.
16992         (vect_estimate_min_profitable_iters): Update for change in type
16993         of vect_location.
16994         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
16995         location_t to dump_location_t.
16996         (vect_slp_bb): Update for change in type of vect_location.
16997         * tree-vectorizer.c (vect_location): Convert from source_location
16998         to dump_user_location_t.
16999         (try_vectorize_loop_1): Update for change in vect_location's type.
17000         (vectorize_loops): Likewise.
17001         (increase_alignment): Likewise.
17002         * tree-vectorizer.h (vect_location): Convert from source_location
17003         to dump_user_location_t.
17004         (find_loop_location): Convert return type from source_location to
17005         dump_user_location_t.
17006         (check_reduction_path): Convert 1st param from location_t to
17007         dump_user_location_t.
17008         * value-prof.c (check_counter): Update call to dump_printf_loc to
17009         pass in a dump_user_location_t rather than a location_t; update
17010         call to error_at for change in type of "locus".
17011         (check_ic_target): Update call to dump_printf_loc to
17012         pass in a dump_user_location_t rather than a location_t, via the
17013         call_stmt.
17015 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
17017         * config/s390/s390.h (enum processor_flags): Do not use
17018         default tune parameter when -march was specified.
17020 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
17022         PR target/86314
17023         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
17024         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
17025         operands.
17027 2018-06-26  Richard Biener  <rguenther@suse.de>
17029         PR tree-optimization/86287
17030         PR bootstrap/86316
17031         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
17032         (vect_analyze_loop): Initialize n_stmts.
17034 2018-06-26  Richard Biener  <rguenther@suse.de>
17036         PR middle-end/86271
17037         * fold-const.c (fold_convertible_p): Pointer extension
17038         isn't valid.
17040 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
17042         PR debug/86064
17043         * dwarf2out.c (loc_list_has_views): Adjust comments.
17044         (dw_loc_list): Split single cross-partition range with
17045         nonzero locview.
17047 2018-06-25  Jeff Law  <law@redhat.com>
17049         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
17050         on -mbig-switch by default.
17052         * config/v850/predicates.md (const_float_1_operand): Fix match_code
17053         test.
17054         (const_float_0_operand): Remove unused predicate.
17055         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
17056         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
17057         (recipsf2): New expander.  Original pattern now called
17058         (recipsf2_insn).
17059         (recipdf2, recipdf2_insn): Similarly.
17060         (rsqrtsf2, rsqrtsf2_insn): Similarly
17061         (rsqrtdf2, rsqrtdf2_insn): Similarly
17063 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
17065         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
17066         Simplify logic for FreeBSD (twice).
17068 2018-06-25  Martin Sebor  <msebor@redhat.com>
17070         PR tree-optimization/86204
17071         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
17072         a strnlen result if it's less than the length of the string.
17074 2018-06-25  Martin Sebor  <msebor@redhat.com>
17076         PR tree-optimization/85700
17077         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
17078         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
17079         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
17081 2018-06-25  Martin Sebor  <msebor@redhat.com>
17083         * doc/extend.texi (Zero-length arrays): Update and clarify.
17085 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
17087         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
17088         added IEEE/IBM long double multilib support on PowerPC little
17089         endian Linux systems.
17090         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
17091         (MULTILIB_DEFAULTS): Likewise.
17092         * config/rs6000/rs6000.c (rs6000_option_override_internal):
17093         Likewise.
17094         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
17095         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
17096         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
17098 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
17100         PR middle-end/86311
17101         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
17102         (REORDER_45): Likewise.
17104 2018-06-25  Jeff Law  <law@redhat.com>
17106         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
17107         dividend to 32 bits.  Adjust length.
17108         (udivmodhi4): Cleanup output template.  Fix length.
17110 2018-06-25  Carl Love  <cel@us.ibm.com>
17112         * config/rs6000/vsx.md: Change word selector to prefered location.
17114 2018-06-25  Richard Biener  <rguenther@suse.de>
17116         PR tree-optimization/86304
17117         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
17118         epilogue-if-converted loops as well.
17120 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
17122         * lto-section-out.c (lto_begin_section): Do not print section
17123         name for noaddr and unnumbered dumps.
17125 2018-06-25  Richard Biener  <rguenther@suse.de>
17127         * tree-vectorizer.h (struct vec_info_shared): New structure
17128         with parts split out from struct vec_info and loop_nest from
17129         struct _loop_vec_info.
17130         (struct vec_info): Adjust accordingly.
17131         (struct _loop_vec_info): Likewise.
17132         (LOOP_VINFO_LOOP_NEST): Adjust.
17133         (LOOP_VINFO_DATAREFS): Likewise.
17134         (LOOP_VINFO_DDRS): Likewise.
17135         (struct _bb_vec_info): Likewise.
17136         (BB_VINFO_DATAREFS): Likewise.
17137         (BB_VINFO_DDRS): Likewise.
17138         (struct _stmt_vec_info): Add dr_aux member.
17139         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
17140         (DR_MISALIGNMENT_UNINITIALIZED): New.
17141         (set_dr_misalignment): Adjust.
17142         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
17143         (vect_analyze_loop): Adjust prototype.
17144         (vect_analyze_loop_form): Likewise.
17145         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
17146         Compute dependences lazily.
17147         (vect_record_base_alignments): Use shared datarefs/ddrs.
17148         (vect_verify_datarefs_alignment): Likewise.
17149         (vect_analyze_data_refs_alignment): Likewise.
17150         (vect_analyze_data_ref_accesses): Likewise.
17151         (vect_analyze_data_refs): Likewise.
17152         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
17153         constructor parameter for shared part.
17154         (vect_analyze_loop_form): Pass in shared part and adjust.
17155         (vect_analyze_loop_2): Pass in storage for the number of
17156         stmts.  Move loop nest finding to the caller.  Compute
17157         datarefs lazily.
17158         (vect_analyze_loop): Pass in shared part.
17159         (vect_transform_loop): Verify shared datarefs are unchanged.
17160         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
17161         constructor parameter for shared part.
17162         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
17163         (vect_slp_bb): Verify shared datarefs are unchanged before
17164         transform.
17165         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
17166         change.
17167         (new_stmt_vec_info): Initialize DR_AUX misalignment to
17168         DR_MISALIGNMENT_UNINITIALIZED.
17169         * tree-vectorizer.c (vec_info::vec_info): Add constructor
17170         parameter for shared part.
17171         (vec_info::~vec_info): Adjust.
17172         (vec_info_shared::vec_info_shared): New.
17173         (vec_info_shared::~vec_info_shared): Likewise.
17174         (vec_info_shared::save_datarefs): Likewise.
17175         (vec_info_shared::check_datarefs): Likewise.
17176         (try_vectorize_loop_1): Construct shared part live for analyses
17177         of a single loop for multiple vector sizes.
17178         * tree-parloops.c (gather_scalar_reductions): Adjust.
17180 2018-06-25  Richard Biener  <rguenther@suse.de>
17182         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
17183         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
17184         (vect_analyze_data_refs): Remove similar code from here and
17185         simplify accordingly.
17187 2018-06-25  Richard Biener  <rguenther@suse.de>
17189         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
17190         for reverse storage order accesses rather than asserting
17191         they cannot happen here.
17193 2018-06-25  Tom de Vries  <tdevries@suse.de>
17195         PR debug/86257
17196         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
17197         Use data16 instead of .byte for insn prefix.
17199 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
17201         PR C++/86082
17202         * parser.c (make_char_string_pack): Pass this literal chars
17203         through cpp_interpret_string.
17204         (cp_parser_userdef_numeric_literal): Check the result of
17205         make_char_string_pack.
17207 2018-06-24  Maya Rashish  <coypu@sdf.org>
17209         * ginclude/stddef.h: Simplify conditions around avoiding
17210         re-definition of __size_t.
17212 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17214         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
17215         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
17217 2018-06-22  Maya Rashish  <coypu@sdf.org>
17219         * doc/invoke.texi (mno-fancy-math-387): Update for changes
17220         made to OpenBSD and NetBSD through the years.
17222 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17224         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
17225         behavior of vec_pack (vector double, vector double) to match
17226         behavior of vec_float2 (vector double, vector double).
17228 2018-06-22  Olivier Hainque  <hainque@adacore.com>
17230         * gimplify.c (gimplify_function_tree): Prevent creation
17231         of a trampoline for the address of the current function
17232         passed to entry/exit instrumentation hooks.
17234 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
17236         PR target/86222
17237         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
17238         correctly.
17240 2018-06-22  Martin Liska  <mliska@suse.cz>
17242         PR tree-optimization/86263
17243         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17244         Bail out if is_enabled is false.
17245         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
17246         New declaration.
17247         (jump_table_cluster::is_enabled): New function.
17249 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17251         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
17252         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
17253         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
17254         (lto_input_ts_binfo_tree_pointers): Likewise.
17255         * tree-streamer-out.c (streamer_write_tree_bitfields,
17256         write_ts_binfo_tree_pointers): Likewise.
17257         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
17259 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
17261         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
17263 2018-06-22  Martin Liska  <mliska@suse.cz>
17265         * symbol-summary.h (get): Make it pure and inline move
17266         functionality from ::get function.
17267         (get): Remove and inline into ::get and ::get_create.
17268         (get_create): Move code from ::get function.
17270 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17272         PR target/85994
17273         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
17274         -x assembler-with-cpp.
17276 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17278         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
17279         _FILE_OFFSET_BITS=64 for C++.
17281 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
17283         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
17284         conversion insn that shows up when pr85657-3.c is compiled using
17285         IEEE 128-bit long double.
17286         (neg<mode>2_internal): Use the correct mode to check whether the
17287         mode is IBM extended.
17288         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
17289         multiply and divide external functions from being created more
17290         than once.
17292 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17294         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
17295         functions.
17296         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
17297         the edge can be forwarded.
17298         (cfg_layout_merge_blocks): Likewise.
17300 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17302         * except.c (finish_eh_generation): Commit edge insertions only after
17303         the EH edges have been redirected from post-landing to landing pads.
17305 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17307         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
17308         create_tmp_var_for to create the FRAME decl.
17309         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
17311 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
17313         * tree-inline.c (copy_edges_for_bb): Minor tweak.
17314         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
17315         debug statement when resetting its value.
17316         (expand_call_inline): Copy the locus of the call onto the assignment
17317         of the return value, if any.  Use local variable in more cases.
17319 2018-06-21  Martin Liska  <mliska@suse.cz>
17321         * ipa-pure-const.c (propagate_nothrow): Use
17322         funct_state_summaries->get.
17323         (dump_malloc_lattice): Likewise.
17324         (propagate_malloc): Likewise.
17326 2018-06-21  Richard Biener  <rguenther@suse.de>
17328         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
17329         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
17330         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
17331         comment.
17332         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
17333         BLOCK_ABSTRACT_ORIGIN unconditionally.
17335 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
17337         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
17338         deleting it.
17339         * ipa-reference.c (ipa_reference_c_finalize): Delete
17340         ipa_ref_opt_sum_summaries and set it to NULL.
17342 2018-06-21  Tom de Vries  <tdevries@suse.de>
17344         PR tree-optimization/85859
17345         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
17346         test with comment from bb_no_side_effects_p.
17348 2018-06-21  Richard Biener  <rguenther@suse.de>
17350         PR tree-optimization/86232
17351         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
17352         max for constant niter.
17354 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17356         * config/aarch64/aarch64-simd.md
17357         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
17359 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17361         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
17362         Make opernads of the unspec commutative.
17364 2018-06-21  Richard Biener  <rguenther@suse.de>
17366         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
17367         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
17368         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
17369         (vect_analyze_data_ref_dependence): Re-order checks to deal with
17370         NULL DR_STEP.
17371         (vect_record_base_alignments): Do not record base alignment
17372         for gathers or scatters.
17373         (vect_compute_data_ref_alignment): Drop return value that is always
17374         true.  Bail out early for gathers or scatters.
17375         (vect_enhance_data_refs_alignment): Bail out early for gathers
17376         or scatters.
17377         (vect_find_same_alignment_drs): Likewise.
17378         (vect_analyze_data_refs_alignment): Remove dead code.
17379         (vect_slp_analyze_and_verify_node_alignment): Likewise.
17380         (vect_analyze_data_refs): For possible gathers or scatters do
17381         not create an alternate DR, just check their possible validity
17382         and mark them.  Adjust DECL_NONALIASED handling to not rely
17383         on DR_BASE_ADDRESS.
17384         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
17385         update inits of gathers or scatters.
17386         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
17387         Also copy gather/scatter flag to pattern vinfo.
17389 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17391         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
17392         behavior of vec_packsu (vector unsigned long long, vector unsigned
17393         long long) to match behavior of vec_packs with same signature.
17395 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
17396             Thomas Schwinge <thomas@codesourcery.com>
17397             Cesar Philippidis  <cesar@codesourcery.com>
17399         * gimplify.c (gimplify_scan_omp_clauses): Add support for
17400         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
17401         (gimplify_adjust_omp_clauses): Likewise.
17402         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
17403         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
17404         (gimplify_omp_target_update): Update handling of acc update and
17405         enter/exit data.
17406         * omp-low.c (install_var_field): Remove unused parameter
17407         base_pointers_restrict.
17408         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
17409         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
17410         FINALIZE}
17411         (omp_target_base_pointers_restrict_p): Delete.
17412         (scan_omp_target): Update call to scan_sharing_clauses.
17413         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
17414         FINALIZE}.
17415         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
17416         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
17417         (convert_local_omp_clauses): Likewise.
17418         * tree-pretty-print.c (dump_omp_clause): Likewise.
17419         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
17420         FINALIZE}.
17421         (omp_clause_code_name): Likewise.
17423 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
17425         PR debug/86194
17426         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
17427         be narrowed.
17429         PR tree-optimization/86231
17430         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
17431         anti-range don't overwrite *vr0min before using it to compute *vr0max.
17433 2018-06-20  Tom de Vries  <tdevries@suse.de>
17435         PR tree-optimization/86097
17436         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
17437         iv type if signedness of iv type is not the same as that of *nit.
17439 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
17441         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
17442         EDGE_EH edges, verify they are all EDGE_EH.
17444 2018-06-20  Maya Rashish  <coypu@sdf.org>
17446         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
17448 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17450         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
17451         * config/aarch64/aarch64.c (xgene1_tunings): Add
17452         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
17453         (aarch64_mode_valid_for_sched_fusion_p):
17454         Allow 16-byte modes.
17455         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
17456         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
17457         128-bit modes.
17458         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
17459         New pattern.
17460         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
17461         * config/aarch64/iterators.md (VQ2): New mode iterator.
17463 2018-06-20  Martin Liska  <mliska@suse.cz>
17465         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
17466         Change default ratio from 10 to 8.
17468 2018-06-20  Martin Liska  <mliska@suse.cz>
17470         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17471         New.
17472         (bit_test_cluster::find_bit_tests): Likewise.
17473         (switch_decision_tree::analyze_switch_statement): Find clusters.
17474         * tree-switch-conversion.h (struct jump_table_cluster): Document
17475         hierarchy.
17477 2018-06-20  Martin Liska  <mliska@suse.cz>
17479         * tree-switch-conversion.c (switch_conversion::collect):
17480         Record m_uniq property.
17481         (switch_conversion::expand): Bail out for special conditions.
17482         (group_cluster::~group_cluster): New.
17483         (group_cluster::group_cluster): Likewise.
17484         (group_cluster::dump): Likewise.
17485         (jump_table_cluster::emit): New.
17486         (switch_decision_tree::fix_phi_operands_for_edges): New.
17487         (struct case_node): Remove struct.
17488         (jump_table_cluster::can_be_handled): New.
17489         (case_values_threshold): Moved to header.
17490         (reset_out_edges_aux): Likewise.
17491         (jump_table_cluster::is_beneficial): New.
17492         (bit_test_cluster::can_be_handled): Likewise.
17493         (add_case_node): Remove.
17494         (bit_test_cluster::is_beneficial): New.
17495         (case_bit_test::cmp): New.
17496         (bit_test_cluster::emit): New.
17497         (expand_switch_as_decision_tree_p): Remove.
17498         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
17499         (fix_phi_operands_for_edge): Likewise.
17500         (switch_decision_tree::analyze_switch_statement): New.
17501         (compute_cases_per_edge): Move ...
17502         (switch_decision_tree::compute_cases_per_edge): ... here.
17503         (try_switch_expansion): Likewise.
17504         (switch_decision_tree::try_switch_expansion): Likewise.
17505         (record_phi_operand_mapping): Likewise.
17506         (switch_decision_tree::record_phi_operand_mapping): Likewise.
17507         (emit_case_decision_tree): Likewise.
17508         (switch_decision_tree::emit): Likewise.
17509         (balance_case_nodes): Likewise.
17510         (switch_decision_tree::balance_case_nodes): Likewise.
17511         (dump_case_nodes): Likewise.
17512         (switch_decision_tree::dump_case_nodes): Likewise.
17513         (emit_jump): Likewise.
17514         (switch_decision_tree::emit_jump): Likewise.
17515         (emit_cmp_and_jump_insns): Likewise.
17516         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
17517         (emit_case_nodes): Likewise.
17518         (switch_decision_tree::emit_case_nodes): Likewise.
17519         (conditional_probability): Remove.
17520         * tree-switch-conversion.h (enum cluster_type): New.
17521         (PRINT_CASE): New.
17522         (struct cluster): Likewise.
17523         (cluster::cluster): Likewise.
17524         (struct simple_cluster): Likewise.
17525         (simple_cluster::simple_cluster): Likewise.
17526         (struct group_cluster): Likewise.
17527         (struct jump_table_cluster): Likewise.
17528         (struct bit_test_cluster): Likewise.
17529         (struct min_cluster_item): Likewise.
17530         (struct case_tree_node): Likewise.
17531         (case_tree_node::case_tree_node): Likewise.
17532         (jump_table_cluster::case_values_threshold): Likewise.
17533         (struct case_bit_test): Likewise.
17534         (struct switch_decision_tree): Likewise.
17535         (struct switch_conversion): Likewise.
17536         (switch_decision_tree::reset_out_edges_aux): Likewise.
17538 2018-06-20  Martin Liska  <mliska@suse.cz>
17540         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
17541         (hoist_edge_and_branch_if_true): Likewise.
17542         (expand_switch_using_bit_tests_p): Likewise.
17543         (struct case_bit_test): Likewise.
17544         (case_bit_test_cmp): Likewise.
17545         (emit_case_bit_tests): Likewise.
17546         (switch_conversion::switch_conversion): New class.
17547         (struct switch_conv_info): Remove old struct.
17548         (collect_switch_conv_info): More to ...
17549         (switch_conversion::collect): ... this.
17550         (check_range): Likewise.
17551         (switch_conversion::check_range): Likewise.
17552         (check_all_empty_except_final): Likewise.
17553         (switch_conversion::check_all_empty_except_final): Likewise.
17554         (check_final_bb): Likewise.
17555         (switch_conversion::check_final_bb): Likewise.
17556         (create_temp_arrays): Likewise.
17557         (switch_conversion::create_temp_arrays): Likewise.
17558         (free_temp_arrays): Likewise.
17559         (gather_default_values): Likewise.
17560         (switch_conversion::gather_default_values): Likewise.
17561         (build_constructors): Likewise.
17562         (switch_conversion::build_constructors): Likewise.
17563         (constructor_contains_same_values_p): Likewise.
17564         (switch_conversion::contains_same_values_p): Likewise.
17565         (array_value_type): Likewise.
17566         (switch_conversion::array_value_type): Likewise.
17567         (build_one_array): Likewise.
17568         (switch_conversion::build_one_array): Likewise.
17569         (build_arrays): Likewise.
17570         (switch_conversion::build_arrays): Likewise.
17571         (gen_def_assigns): Likewise.
17572         (switch_conversion::gen_def_assigns): Likewise.
17573         (prune_bbs): Likewise.
17574         (switch_conversion::prune_bbs): Likewise.
17575         (fix_phi_nodes): Likewise.
17576         (switch_conversion::fix_phi_nodes): Likewise.
17577         (gen_inbound_check): Likewise.
17578         (switch_conversion::gen_inbound_check): Likewise.
17579         (process_switch): Use the newly created class.
17580         (switch_conversion::expand): New.
17581         (switch_conversion::~switch_conversion): New.
17582         * tree-switch-conversion.h: New file.
17584 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17586         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
17587         tree-vect-patterns.c.
17588         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
17589         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
17590         (vect_recog_sad_pattern): Likewise.
17591         (vect_recog_widen_sum_pattern): Likewise.
17592         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
17593         (vect_recog_widen_shift_pattern): Remove the type_in argument.
17594         (vect_recog_rotate_pattern): Likewise.
17595         (vect_recog_mult_pattern): Likewise.
17596         (vect_recog_vector_vector_shift_pattern): Likewise.
17597         (vect_recog_divmod_pattern): Likewise.
17598         (vect_recog_mixed_size_cond_pattern): Likewise.
17599         (vect_recog_bool_pattern): Likewise.
17600         (vect_recog_mask_conversion_pattern): Likewise.
17601         (vect_try_gather_scatter_pattern): Likewise.
17602         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
17603         (vect_recog_over_widening_pattern): Likewise.
17604         (vect_recog_gather_scatter_pattern): Likewise.
17605         (vect_recog_func_ptr): Move from tree-vectorizer.h
17606         (vect_vect_recog_func_ptrs): Move further down the file.
17607         (vect_recog_func): Likewise.  Remove the third argument.
17608         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
17609         (vect_pattern_recog_1): Expect the pattern function to do any
17610         necessary target tests.  Also expect it to provide a vector type.
17611         Remove the type_in handling.
17613 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17615         * tree-vect-patterns.c (vect_pattern_detected): New function.
17616         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
17617         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
17618         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
17619         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
17620         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
17621         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
17622         (vect_recog_mask_conversion_pattern)
17623         (vect_try_gather_scatter_pattern): Likewise.
17625 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17627         * tree-vect-patterns.c (vect_get_internal_def): New function.
17628         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17629         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
17630         (search_type_for_mask_1): Use it.
17632 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17634         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
17635         redundant WIDEN_SUM_EXPR handling.
17636         (vect_recog_sad_pattern): Likewise.
17638 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17640         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
17641         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
17642         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
17643         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
17644         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
17646 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17648         * tree-vect-stmts.c (vectorizable_call): Make sure that we
17649         use the stmt_vec_info of the original bb statement for the
17650         new zero assignment, even if the call is part of a pattern.
17652 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17654         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
17655         that the sequence is attached to the original statement rather
17656         than the pattern statement.
17657         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
17658         PATTERN_DEF_SEQ from the original statement rather than
17659         the main pattern statement.
17660         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
17661         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
17662         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
17664 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
17666         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
17667         definition statements before the early exit for statements that aren't
17668         live or relevant.
17669         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
17670         split out from...
17671         (vect_transform_loop): ...here.  Process pattern definition
17672         statements without first checking whether the main pattern
17673         statement is live or relevant.
17675 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
17677         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
17678         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
17680 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
17682         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
17683         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
17684         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
17685         (expand_block_compare): Change select_block_compare_mode call.
17686         (expand_strncmp_align_check): Use new functions, fix comment.
17687         (emit_final_str_compare_gpr): New function.
17688         (expand_strn_compare): Refactor and clean up code.
17689         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
17691 2018-06-19  Tony Reix  <tony.reix@atos.com>
17692             Damien Bergamini  <damien.bergamini@atos.com>
17693             David Edelsohn  <dje.gcc@gmail.com>
17695         * collect2.c (static_obj): New variable.
17696         (static_libs): New variable.
17697         (is_in_list): Uncomment declaration.
17698         (main): Track AIX libraries linked statically.
17699         (is_in_list): Uncomment definition.
17700         (scan_prog_file): Don't add AIX shared libraries initializer
17701         to constructor list if linking statically.
17703 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
17705         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
17706         constant.
17707         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
17709 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
17711         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
17712         blocks.
17714 2018-06-19  Martin Liska  <mliska@suse.cz>
17716         * config/i386/i386.c (ix86_can_inline_p): Do not use
17717         ipa_fn_summaries::get_create.
17718         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
17719         get.
17720         (devirtualization_time_bonus): Likewise.
17721         (ipcp_propagate_stage): Likewise.
17722         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
17723         (edge_set_predicate): Likewise.
17724         (evaluate_conditions_for_known_args): Likewise.
17725         (evaluate_properties_for_edge): Likewise.
17726         (ipa_call_summary::reset): Tranform to ...
17727         (ipa_call_summary::~ipa_call_summary): ... this.
17728         (ipa_fn_summary::reset): Transform to ...
17729         (ipa_fn_summary::~ipa_fn_summary): ... this.
17730         (ipa_fn_summary_t::remove): Rename to ...
17731         (ipa_fn_summary_t::remove_callees): ... this.
17732         (ipa_fn_summary_t::duplicate): Use placement new
17733         instead of memory copy.
17734         (ipa_call_summary_t::duplicate): Likewise.
17735         (ipa_call_summary_t::remove): Remove.
17736         (dump_ipa_call_summary): Change get_create to get.
17737         (ipa_dump_fn_summary): Dump only when summary exists.
17738         (analyze_function_body): Use symbol_summary::get instead
17739         of get_create.
17740         (compute_fn_summary): Likewise.
17741         (estimate_edge_devirt_benefit): Likewise.
17742         (estimate_edge_size_and_time): Likewise.
17743         (inline_update_callee_summaries): Likewise.
17744         (remap_edge_change_prob): Likewise.
17745         (remap_edge_summaries): Likewise.
17746         (ipa_merge_fn_summary_after_inlining): Likewise.
17747         (write_ipa_call_summary): Likewise.
17748         (ipa_fn_summary_write): Likewise.
17749         (ipa_free_fn_summary): Likewise.
17750         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
17751         (struct ipa_call_summary): Likewise.
17752         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
17753         of get_create.
17754         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
17755         (estimate_size_after_inlining): Likewise.
17756         (estimate_growth): Likewise.
17757         (growth_likely_positive): Likewise.
17758         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
17759         (inline_call): Likewise.
17760         * ipa-inline.c (caller_growth_limits): Likewise.
17761         (can_inline_edge_p): Likewise.
17762         (can_inline_edge_by_limits_p): Likewise.
17763         (compute_uninlined_call_time): Likewise.
17764         (compute_inlined_call_time): Likewise.
17765         (want_inline_small_function_p): Likewise.
17766         (edge_badness): Likewise.
17767         (update_caller_keys): Likewise.
17768         (update_callee_keys): Likewise.
17769         (inline_small_functions): Likewise.
17770         (inline_to_all_callers_1): Likewise.
17771         (dump_overall_stats): Likewise.
17772         (early_inline_small_functions): Likewise.
17773         (early_inliner): Likewise.
17774         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
17775         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
17776         * ipa-pure-const.c (malloc_candidate_p): Likewise.
17777         * ipa-split.c (execute_split_functions): Likewise.
17778         * symbol-summary.h: Likewise.
17779         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
17781 2018-06-19  Richard Biener  <rguenther@suse.de>
17783         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
17784         (vectorize_loops): ... here.  Fix dbgcnt handling.
17785         (try_vectorize_loop): Wrap try_vectorize_loop_1.
17787 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
17789         PR target/86197
17790         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
17791         ieee128 argument takes up only one (vector) register, not two (floating
17792         point) registers.
17794 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
17796         * gimplify.c (gimplify_init_constructor): Really never clear for an
17797         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
17799 2018-06-19  Richard Biener  <rguenther@suse.de>
17801         PR tree-optimization/86179
17802         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
17803         after failed recognition.
17805 2018-06-18  Martin Sebor  <msebor@redhat.com>
17807         PR middle-end/85602
17808         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
17809         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
17810         Handle integer subtraction.
17811         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
17812         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
17814 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
17816         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
17817         param from rtx to rtx_insn *.
17818         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
17819         param.
17820         (frv_ifcvt_modify_insn): Likwise.
17821         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
17822         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
17823         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
17824         as_a <rtx_insn *> cast to local "unprotected_region" once
17825         it's been established that it's not NULL or pc_rtx.
17826         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
17827         param "sethi" from rtx to rtx_insn *.
17828         (nds32_group_float_insns): Likewise for param "insn".
17829         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
17830         param.
17831         (vax_output_int_subtract): Likewise.
17832         * config/vax/vax.c (vax_output_int_add): Likewise for param
17833         "insn".
17834         (vax_output_int_subtract): Likewise.
17835         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
17836         (emit_pattern_after): Likewise for param "after".
17837         (emit_insn_after): Likewise.
17838         (emit_jump_insn_after): Likewise.
17839         (emit_call_insn_after): Likewise.
17840         (emit_debug_insn_after): Likewise.
17841         (emit_pattern_before): Likewise for param "before".
17842         (emit_insn_before): Likewise.
17843         (emit_jump_insn_before): Likewise.
17844         * final.c (get_insn_template): Likewise for param "insn", removing
17845         a cast.
17846         * output.h (get_insn_template): Likewise for 2nd param.
17847         * rtl.h (emit_insn_before): Likewise.
17848         (emit_jump_insn_before): Likewise.
17849         (emit_debug_insn_before_noloc): Likewise.
17850         (emit_insn_after): Likewise.
17851         (emit_jump_insn_after): Likewise.
17852         (emit_call_insn_after): Likewise.
17853         (emit_debug_insn_after): Likewise.
17854         (set_insn_deleted): Likewise for param.
17856 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
17858         PR target/85358
17859         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
17860         floating point modes, so that IFmode is numerically greater than
17861         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
17862         to declare the ordering.  This prevents IFmode from being
17863         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
17864         machine.  Include rs6000-modes.h to share the fractional values
17865         between genmodes* and the rest of the compiler.
17866         (IFmode): Likewise.
17867         (KFmode): Likewise.
17868         (TFmode): Likewise.
17869         * config/rs6000/rs6000-modes.h: New file.
17870         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
17871         meaning of rs6000_long_double_size so that 126..128 selects an
17872         appropriate 128-bit floating point type.
17873         (rs6000_option_override_internal): Likewise.
17874         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
17875         (TARGET_LONG_DOUBLE_128): Change the meaning of
17876         rs6000_long_double_size so that 126..128 selects an appropriate
17877         128-bit floating point type.
17878         (LONG_DOUBLE_TYPE_SIZE): Update comment.
17879         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
17880         source and destination to match the standard usage.
17881         (truncifkf2): Likewise.
17882         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
17883         ISA 2.07 to use an explicit clobber, instead of passing in a
17884         temporary.
17885         (copysign<mode>3_soft): Likewise.
17887 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
17889         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
17890         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
17891         (vect_slp_analyze_instance_dependence): Likewise.
17892         (vect_enhance_data_refs_alignment): Likewise.
17893         (vect_analyze_data_refs_alignment): Likewise.
17894         (vect_slp_analyze_and_verify_instance_alignment
17895         (vect_analyze_data_ref_accesses): Likewise.
17896         (vect_prune_runtime_alias_test_list): Likewise.
17897         (vect_analyze_data_refs): Likewise.
17898         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
17899         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17900         (vect_analyze_scalar_cycles_1): Likewise.
17901         (vect_get_loop_niters): Likewise.
17902         (vect_analyze_loop_form_1): Likewise.
17903         (vect_update_vf_for_slp): Likewise.
17904         (vect_analyze_loop_operations): Likewise.
17905         (vect_analyze_loop): Likewise.
17906         (vectorizable_induction): Likewise.
17907         (vect_transform_loop): Likewise.
17908         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
17909         * tree-vect-slp.c (vect_analyze_slp): Likewise.
17910         (vect_make_slp_decision): Likewise.
17911         (vect_detect_hybrid_slp): Likewise.
17912         (vect_slp_analyze_operations): Likewise.
17913         (vect_slp_bb): Likewise.
17914         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
17915         (vectorizable_bswap): Likewise.
17916         (vectorizable_call): Likewise.
17917         (vectorizable_simd_clone_call): Likewise.
17918         (vectorizable_conversion): Likewise.
17919         (vectorizable_assignment): Likewise.
17920         (vectorizable_shift): Likewise.
17921         (vectorizable_operation): Likewise.
17922         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
17924 2018-06-18  Martin Sebor  <msebor@redhat.com>
17926         PR tree-optimization/81384
17927         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
17928         * builtins.c (expand_builtin_strnlen): New function.
17929         (expand_builtin): Call it.
17930         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
17931         * builtins.def (BUILT_IN_STRNLEN): New.
17932         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
17933         Warn for bounds in excess of maximum object size.
17934         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
17935         single-value ranges.  Handle strnlen.
17936         (handle_builtin_strlen): Handle strnlen.
17937         (strlen_check_and_optimize_stmt): Same.
17938         * doc/extend.texi (Other Builtins): Document strnlen.
17940 2018-06-18  Maya Rashish  <coypu@sdf.org>
17942         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
17943         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
17944         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
17946         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
17947         here to ...
17948         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
17950 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17952         * tree.c (escaped_string::escape): Replace cast to char * by
17953         const_cast<char *> (unescaped).
17955 2018-06-18  Nick Clifton  <nickc@redhat.com>
17957         PR 84195
17958         * tree.c (escaped_string): New class.  Converts an unescaped
17959         string into its escaped equivalent.
17960         (warn_deprecated_use): Use the new class to convert the
17961         deprecation message, if present.
17962         (test_escaped_strings): New self test.
17963         (test_c_tests): Add test_escaped_strings.
17964         * doc/extend.texi (deprecated): Add a note that the
17965         deprecation message is affected by the -fmessage-length
17966         option, and that control characters will be escaped.
17967         (#pragma GCC error): Document this pragma.
17968         (#pragma GCC warning): Likewise.
17969         * doc/invoke.texi (-fmessage-length): Document this option's
17970         effect on the #warning and #error preprocessor directives and
17971         the deprecated attribute.
17973 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
17975         * tree.c (decl_value_expr_lookup): Revert latest change.
17976         (decl_value_expr_insert): Likewise.
17978 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
17980         * gimplify.c (nonlocal_vlas): Delete.
17981         (nonlocal_vla_vars): Likewise.
17982         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
17983         referenced VLAs.
17984         (gimplify_body): Do not create and destroy nonlocal_vlas.
17985         * tree-nested.c: Include diagnostic.h.
17986         (use_pointer_in_frame): Tweak.
17987         (lookup_field_for_decl): Add assertion and declare the transformation.
17988         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
17989         internal error when the reference is in a wrong context.  Do not
17990         create a debug decl by default.
17991         (note_nonlocal_block_vlas): Delete.
17992         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
17993         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
17994         create a debug decl by default.
17995         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
17996         call is in a wrong context.
17997         (fixup_vla_decls): New function.
17998         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
17999         debug variables were created.
18000         * tree.c (decl_value_expr_lookup): Add checking assertion.
18001         (decl_value_expr_insert): Likewise.
18003 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
18005         PR middle-end/82479
18006         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
18007         * tree-scalar-evolution.c (interpret_expr): Likewise.
18008         (expression_expensive_p): Likewise.
18009         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
18010         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
18011         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
18012         (ssa_defined_by_minus_one_stmt_p): New.
18014 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
18016         PR middle-end/64946
18017         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
18018         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
18019         * dojump.c (do_jump): Likewise.
18020         * expr.c (expand_expr_real_2): Check operand type's sign.
18021         * fold-const.c (const_unop): Handle ABSU_EXPR.
18022         (fold_abs_const): Likewise.
18023         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
18024         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
18025         (strip_sign_op_1): Likesise.
18026         * match.pd: Add new pattern to generate ABSU_EXPR.
18027         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
18028         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18029         * tree-eh.c (operation_could_trap_helper_p): Likewise.
18030         * tree-inline.c (estimate_operator_cost): Likewise.
18031         * tree-pretty-print.c (dump_generic_node): Likewise.
18032         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
18033         * tree.def (ABSU_EXPR): New.
18035 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
18037         PR middle-end/86095
18038         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
18039         documented as preserved for backward compatibility only.
18040         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
18042         PR rtl-optimization/86108
18043         * bb-reorder.c (create_forwarder_block): Renamed to ...
18044         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
18045         jump from new landing pad to the second part.
18046         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
18047         Adjust callers.
18049 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
18051         PR middle-end/85878
18052         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
18053         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
18054         Only call store_expr for halves if the mode is the same.
18056         PR middle-end/86123
18057         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
18058         Fix up comment formatting.
18060 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18062         * typed-splay-tree.h (typed_splay_tree::remove): New function.
18063         (typed_splay_tree::closure,
18064         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
18065         (typed_splay_tree::typed_splay_tree,
18066         typed_splay_tree::operator =): Declared private.
18067         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
18068         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
18069         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
18070         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
18071         typed_splay_tree::splay_tree_splay,
18072         typed_splay_tree::splay_tree_foreach_helper,
18073         typed_splay_tree::splay_tree_insert,
18074         typed_splay_tree::splay_tree_remove,
18075         typed_splay_tree::splay_tree_lookup,
18076         typed_splay_tree::splay_tree_predecessor,
18077         typed_splay_tree::splay_tree_successor,
18078         typed_splay_tree::splay_tree_min,
18079         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
18080         (typed_splay_tree::root, typed_splay_tree::comp,
18081         typed_splay_tree::delete_key,
18082         typed_splay_tree::delete_value): New data members.
18083         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
18084         typed_splay_tree::remove.
18086 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
18088         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
18089         -mginv and -mno-ginv to the assembler.
18090         * config/mips/mips.opt (-mcrc): New option.
18091         (-mginv): Likewise.
18092         * doc/invoke.text (-mcrc): Document.
18093         (-mginv): Likewise.
18095 2018-06-15  Nick Clifton  <nickc@redhat.com>
18097         PR 84195
18098         * tree.c (escaped_string): New class.  Converts an unescaped
18099         string into its escaped equivalent.
18100         (warn_deprecated_use): Use the new class to convert the
18101         deprecation message, if present.
18102         (test_escaped_strings): New self test.
18103         (test_c_tests): Add test_escaped_strings.
18104         * doc/extend.texi (deprecated): Add a note that the
18105         deprecation message is affected by the -fmessage-length
18106         option, and that control characters will be escaped.
18107         (#pragma GCC error): Document this pragma.
18108         (#pragma GCC warning): Likewise.
18109         * doc/invoke.texi (-fmessage-length): Document this option's
18110         effect on the #warning and #error preprocessor directives and
18111         the deprecated attribute.
18113 2018-06-15  Richard Biener  <rguenther@suse.de>
18115         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
18116         here, also noting vector size used.
18117         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
18118         size used in MSG_OPTIMIZED_LOCATIONS dump.
18119         (pass_slp_vectorize::execute): Adjust.
18121 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
18123         PR target/85968
18124         * config/arc/arc.c (arc_return_address_register): Fix
18125         if-condition.
18127 2018-06-15  Richard Biener  <rguenther@suse.de>
18129         PR middle-end/86159
18130         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
18131         leave useless conversion stripping to force_gimple_operand_gsi.
18132         (gimplify_build2): Likewise.
18133         (gimplify_build1): Likewise.
18135 2018-06-15  Richard Biener  <rguenther@suse.de>
18137         PR middle-end/86076
18138         * tree-cfg.c (move_stmt_op): unshare invariant addresses
18139         before adjusting their block.
18141 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
18143         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
18144         multilibs for *-*-rtems*.
18145         * config/riscv/t-rtems: New file.
18147 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
18149         PR middle-end/86122
18150         * match.pd ((A +- CST1) +- CST2): Punt if last resort
18151         unsigned_type_for returns NULL.
18153         PR target/85945
18154         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
18155         subregs of multi-word pseudos unless the float mode has word size.
18157 2018-06-14  Richard Biener  <rguenther@suse.de>
18159         PR middle-end/86139
18160         * tree-vect-generic.c (build_word_mode_vector_type): Remove
18161         duplicate and harmful type_hash_canon.
18162         * tree.c (type_hash_canon): Assert we didn't find ourselves.
18164 2018-06-14  Richard Biener  <rguenther@suse.de>
18166         PR ipa/86124
18167         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
18168         NULL cgraph_node.
18170 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
18172         * config/rtems.h (STDINT_LONG32): Define.
18174 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
18175             Prachi Godbole  <prachi.godbole@imgtec.com>
18177         * config/mips/mips-cpus.def: Define P6600.
18178         * config/mips/mips-tables.opt: Regenerate.
18179         * config/mips/mips.c (mips_ucbranch_type): New enum.
18180         (mips_rtx_cost_data): Add support for P6600.
18181         (mips_issue_rate): Likewise.
18182         (mips_multipass_dfa_lookahead): Likewise.
18183         (mips_avoid_hazard): Likewise.
18184         (mips_reorg_process_insns): Likewise.
18185         (mips_classify_branch_p6600): New function.
18186         * config/mips/mips.h (TUNE_P6600): New define.
18187         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
18188         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
18189         * config/mips/mips.md: Include p6600.md.
18190         (processor): Add p6600.
18191         * config/mips/p6600.md: New file.
18192         * doc/invoke.texi: Add p6600 to supported architectures.
18194 2018-06-13  Martin Sebor  <msebor@redhat.com>
18196         PR tree-optimization/86114
18197         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
18198         of integer types.
18199         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
18201 2018-06-13  Richard Biener  <rguenther@suse.de>
18203         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
18204         Properly set vector type of the intermediate stmt.
18205         * tree-vect-stmts.c (vectorizable_operation): The destination
18206         var always has vectype_out type.
18208 2018-06-13  Jeff Law  <law@redhat.com>
18210         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
18211         integer 0 for argument to print_rtl_with_bb.
18212         (rl78_reorg): Likewise.
18214 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
18216         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
18217         from rtx to rtx_insn *.
18218         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
18219         "label".
18220         (add_sched_insns_for_speculation): Likewise for local "target",
18221         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
18222         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
18223         from rtx_insn ** to rtx_code_label **.
18224         (reorg_emit_nops): Likewise.
18225         (c6x_reorg): Likewise for local "call_labels".
18226         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
18227         rtx to rtx_insn *.
18228         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
18229         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
18230         the loops over LABEL_REFS.
18231         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
18232         braf_label.
18233         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
18234         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
18235         (split_branches): Strengthen local "olabel" from rtx to
18236         rtx_insn *, adding a safe_as_a cast.
18237         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
18238         to "rtx_insn *".
18239         (add_insn_after): Likewise for first two params.
18240         (add_insn_before): Likewise.
18241         (remove_insn): Likewise for param.
18242         (emit_pattern_before_noloc): Likewise for second and third params.
18243         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
18244         (emit_call_insn_before_noloc): Likewise.
18245         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
18246         to "rtx_insn *".
18247         (emit_barrier_before): Likewise.
18248         (emit_label_before): Strengthen "label" param from "rtx" to
18249         "rtx_code_label *".  Strengthen "before" param from "rtx" to
18250         "rtx_insn *".
18251         (emit_insn_after_1): Strengthen "after" param from "rtx" to
18252         "rtx_insn *".
18253         (emit_pattern_after_noloc): Likewise.
18254         (emit_insn_after_noloc): Likewise.
18255         (emit_jump_insn_after_noloc): Likewise.
18256         (emit_call_insn_after_noloc): Likewise.
18257         (emit_debug_insn_after_noloc): Likewise.
18258         (emit_barrier_after): Likewise.
18259         (emit_label_after): Likewise for both params.
18260         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
18261         "loc" param from "int" to "location_t".
18262         (emit_insn_after_setloc): Likewise.
18263         (emit_jump_insn_after_setloc): Likewise.
18264         (emit_call_insn_after_setloc): Likewise.
18265         (emit_debug_insn_after_setloc): Likewise.
18266         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
18267         "loc" param from "int" to "location_t".
18268         (emit_pattern_before): Convert NULL_RTX to NULL.
18269         (emit_insn_before_setloc): Convert "loc" param from "int" to
18270         "location_t".
18271         (emit_jump_insn_before_setloc): Likewise.
18272         (emit_call_insn_before_setloc): Likewise.
18273         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
18274         rtx_insn *.  Convert "loc" param from "int" to "location_t".
18275         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
18276         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
18277         Convert 3rd param from "int" to "location_t".
18278         (emit_barrier_before, emit_barrier_after, next_real_insn):
18279         Strengthen param from rtx to rtx_insn *.
18280         (emit_label_before): Strengthen 1st param from "rtx" to
18281         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
18282         "rtx_insn *".
18283         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
18284         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
18285         Strengthen 2nd param from "rtx" to "rtx_insn *".
18286         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
18287         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
18288         Likewise. Convert 3rd param from "int" to "location_t".
18289         (emit_label_after): Strengthen 1st param from "rtx" to
18290         "rtx_code_label *".
18291         (next_real_insn, remove_insn): Strengthen param from "rtx" to
18292         "rtx_insn *".
18293         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
18294         from "rtx" to "rtx_insn *".
18296 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
18298         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
18299         bodies streamed in with -Q.
18300         * dumpfile.c (dump_files): Add lto-stream-out dump file.
18301         * dumpfile.h (tree_dump_index): Add lto_stream_out.
18302         * gimple-streamer-out.c: Include gimple-pretty-print.h
18303         (output_bb): Dump stmts streamed.
18304         * lto-section-out.c: Include print-tree.h
18305         (lto_begin_section): Dump sections created.
18306         (lto_output_decl_index): Dump decl encoded.
18307         * lto-streamer-out.c: Include print-tree.h
18308         (create_output_block): Dump output block created.
18309         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
18310         (output_function): Dump function output.
18311         (output_constructor): Dump constructor streamed.
18312         (write_global_stream): Output indexes encoded.
18313         (produce_asm_for_decls): Dump streams encoded.
18314         * lto-streamer.c (streamer_dump_file): New global var.
18315         * lto-streamer.h (streamer_dump_file): Declare.
18316         * passes.c (ipa_write_summaries): Initialize streamer dump.
18317         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
18318         in.
18320 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
18322         PR target/86048
18323         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
18324         offsets for register save directives.  Emit a second batch of save
18325         directives, if need be, when the function accesses prior frames.
18327 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18329         * config/arc/fpu.md (fmasf4): Force operand to register.
18330         (fnmasf4): Likewise.
18332 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18334         * config/arc/arc-protos.h (arc_pad_return): Remove.
18335         * config/arc/arc.c (machine_function): Remove force_short_suffix
18336         and size_reason.
18337         (arc_print_operand): Adjust printing of '&'.
18338         (arc_verify_short): Remove conditional printing of short suffix.
18339         (arc_final_prescan_insn): Remove reference to size_reason.
18340         (pad_return): New function.
18341         (arc_reorg): Call pad_return.
18342         (arc_pad_return): Remove.
18343         (arc_init_machine_status): Remove reference to force_short_suffix.
18344         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
18345         (attr length): When attribute iscompact is true force to 2
18346         regardless; in the case of maybe check if we want to force the
18347         instruction to have 4 bytes length.
18348         (nopv): Change it to generate 4 byte long nop as well.
18349         (blockage): New pattern.
18350         (simple_return): Remove call to arc_pad_return.
18351         (p_return_i): Likewise.
18353 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18355         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
18357 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18359         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
18360         ARC cores.
18362 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18364         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
18365         for ARC700 and ARCv2.
18367 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
18369         PR target/86076
18370         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
18371         operands[2] instead of operands[1].
18374 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
18376         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
18377         case, check whether the outer register overlaps an unallocatable
18378         register, not just whether it fits the required class.
18380 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
18382         * poly-int.h (can_div_trunc_p): Add new overload in which all values
18383         are poly_ints.
18384         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
18385         (memrefs_conflict_p): Likewise.
18386         (init_alias_analysis): Likewise.
18387         * cfgexpand.c (expand_debug_expr): Likewise.
18388         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
18389         * cse.c (fold_rtx): Likewise.
18390         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
18391         * expr.c (emit_block_move_hints): Likewise.
18392         (clear_storage_hints, push_block, emit_push_insn): Likewise.
18393         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
18394         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
18395         (emit_group_store): Likewise.
18396         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
18397         to read the PRE/POST_MODIFY increment.
18398         * calls.c (store_one_arg): Use strip_offset.
18399         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
18400         poly_int_rtx_p.
18401         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
18402         by a VEC_SELECT.
18403         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
18404         (simplify_binary_operation_1): Extend CONST_INT handling to
18405         poly_int_rtx_p.
18406         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
18407         than a HOST_WIDE_INT.
18408         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
18409         poly_int64.
18410         (adjust_mems, add_stores): Update accodingly.
18411         (vt_canonicalize_addr): Track polynomial offsets.
18412         (emit_note_insn_var_location): Likewise.
18413         (vt_add_function_parameter): Likewise.
18414         (vt_initialize): Likewise.
18416 2018-06-12  Jeff Law  <law@redhat.com>
18418         * config.gcc (alpha*-*-freebsd*): Remove.
18419         * config/alpha/freebsd.h: Remove.
18421 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
18423         PR other/69968
18424         * spellcheck-tree.c (levenshtein_distance): Rename to...
18425         (get_edit_distance): ...this, and update for underlying renaming.
18426         * spellcheck-tree.h (levenshtein_distance): Rename to...
18427         (get_edit_distance): ...this.
18428         * spellcheck.c (levenshtein_distance): Rename to...
18429         (get_edit_distance): ...this.  Convert from Levenshtein distance
18430         to Damerau-Levenshtein distance by supporting transpositions of
18431         adjacent characters.  Rename "v1" to "v_next" and "v0" to
18432         "v_one_ago".
18433         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
18434         (selftest::test_edit_distance_unit_test_oneway): ...this, and
18435         update for underlying renaming.
18436         (selftest::levenshtein_distance_unit_test): Rename to...
18437         (selftest::test_get_edit_distance_unit): ...this, and update for
18438         underlying renaming.
18439         (selftest::test_find_closest_string): Add example from PR 69968
18440         where transposition helps
18441         (selftest::test_metric_conditions): Update for renaming.
18442         (selftest::test_metric_conditions): Likewise.
18443         (selftest::spellcheck_c_tests): Likewise.
18444         * spellcheck.h (levenshtein_distance): Rename both overloads to...
18445         (get_edit_distance): ...this.
18446         (best_match::consider): Update for renaming.
18448 2018-06-12  Martin Sebor  <msebor@redhat.com>
18450         PR tree-optimization/85259
18451         * builtins.c (compute_objsize): Handle constant offsets.
18452         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
18453         true iff a warning has been issued.
18454         * gimple.h (gimple_nonartificial_location): New function.
18455         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
18456         gimple_nonartificial_location and handle -Wno-system-headers.
18457         (handle_builtin_stxncpy): Same.
18459 2018-06-12  Martin Sebor  <msebor@redhat.com>
18461         PR c/85931
18462         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
18464 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
18466         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
18467         BUILTIN_VEC_XST entries for pointer to double and long long.
18469 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
18471         PR target/85990
18472         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
18473         Update comments.
18474         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
18475         Likewise.
18477 2018-06-12  Martin Liska  <mliska@suse.cz>
18479         * doc/options.texi: Document IntegerRange.
18481 2018-06-12  Martin Liska  <mliska@suse.cz>
18483         * config/i386/i386.opt: Make MPX-related options as Deprecated.
18484         * opt-functions.awk: Handle Deprecated flag.
18485         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
18486         and report error.
18487         (read_cmdline_option): Report warning for a deprecated option.
18488         * opts.h (struct cl_option): Add new field cl_deprecated.
18489         (CL_ERR_DEPRECATED): New.
18491 2018-06-12  Martin Liska  <mliska@suse.cz>
18493         * doc/options.texi: Document Deprecated option flag.
18495 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18497         * config/arc/arc-arch.h (arc_extras): New enum.
18498         (arc_cpu_t):Add field extra.
18499         (arc_cpu_types): Consider the extras.
18500         * config/arc/arc-cpus.def: Add extras info.
18501         * config/arc/arc-opts.h (processor_type): Consider extra field.
18502         * config/arc/arc.c (arc_override_options): Handle extra field.
18504 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
18506         * config/arc/arc-arch.h: Update ARC_OPTX macro.
18507         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
18508         field.
18509         * config/arc/arc.c (arc_init): Update pic warning.
18510         (irq_range): Update irq range parsing warnings.
18511         (arc_override_options): Update various warning messages.
18512         (arc_handle_aux_attribute): Likewise.
18514 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
18516         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
18518 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18520         * doc/sourcebuild.texi: Document usage of line number 0 in verify
18521         compiler messages directives.
18523 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
18525         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
18526         * config/mips/mips-tables.opt: Regenerate.
18527         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
18528         mips64r6.
18529         * doc/invoke.texi: Document -march=i6500.
18531 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
18533         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
18534         (i6400_gpmul): Add cpu_unit.
18535         (i6400_gpdiv): Likewise.
18536         (i6400_msa_add_d): Update reservations.
18537         (i6400_msa_int_add) Likewise.
18538         (i6400_msa_short_logic3) Likewise.
18539         (i6400_msa_short_logic2) Likewise.
18540         (i6400_msa_short_logic) Likewise.
18541         (i6400_msa_move) Likewise.
18542         (i6400_msa_cmp) Likewise.
18543         (i6400_msa_short_float2) Likewise.
18544         (i6400_msa_div_d) Likewise.
18545         (i6400_msa_long_logic1) Likewise.
18546         (i6400_msa_long_logic2) Likewise.
18547         (i6400_msa_mult) Likewise.
18548         (i6400_msa_long_float2) Likewise.
18549         (i6400_msa_long_float4) Likewise.
18550         (i6400_msa_long_float5) Likewise.
18551         (i6400_msa_long_float8) Likewise.
18552         (i6400_fpu_fadd): Include frint type.
18553         (i6400_fpu_store): New define_insn_reservation.
18554         (i6400_fpu_load): Likewise.
18555         (i6400_fpu_move): Likewise.
18556         (i6400_fpu_fcmp): Likewise.
18557         (i6400_fpu_fmadd): Likewise.
18558         (i6400_int_mult): Include imul3nc type and update reservation.
18559         (i6400_int_div): Include idiv3 type and update reservation.
18560         (i6400_int_load): Update to check type not move_type.
18561         (i6400_int_store): Likewise.
18562         (i6400_int_prefetch): Set zero latency.
18564 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
18566         * gcc.c: Document new %@{...} sequence.
18567         (LINK_COMMAND_SPEC): Use it for the -L switches.
18568         (cpp_unique_options): Use it for the -I switches.
18569         (at_file_argbuf): New global variable.
18570         (in_at_file): Likewise.
18571         (alloc_args): Create at_file_argbuf.
18572         (clear_args): Truncate at_file_argbuf.
18573         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
18574         (open_at_file): New function.
18575         (close_at_file): Likewise.
18576         (create_at_file): Delete.
18577         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
18578         <'o'>: Likewise.
18579         <'@'>: New case.
18580         (validate_switches_from_spec): Deal with %@{...} sequence.
18581         (validate_switches): Likewise.
18582         (driver::finalize): Call clear_args.
18584 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
18586         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
18588 2018-06-11  Martin Sebor  <msebor@redhat.com>
18590         * doc/invoke.texi (-Wall): List -Wc++17-compat.
18591         (Wno-class-memaccess): Add @opindex.
18592         (Wno-templates, Wno-multiple-inheritance): Same.
18593         (Wno-virtual-inheritance, Wno-namespaces): Same.
18594         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
18595         (Wno-format-overflow, Wno-format-truncation): Same.
18596         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
18597         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
18598         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
18599         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
18600         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
18601         (Wno-misspelled-isr): Same.
18603 2018-06-11  Martin Sebor  <msebor@redhat.com>
18605         * PR tree-optimization/86083
18606         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
18608 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
18610         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
18612 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
18614         PR target/85755
18615         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
18616         on the correct operand.
18617         (*movdi_internal64): Ditto.
18619 2018-06-11  Martin Liska  <mliska@suse.cz>
18621         PR tree-optimization/86089
18622         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
18624 2018-06-11  Julia Koval  <julia.koval@intel.com>
18626         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
18627         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
18628         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
18630 2018-06-11  Olivier Hainque  <hainque@adacore.com>
18632         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
18633         for Ada with strict dwarf2.
18635 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
18637         PR target/85755
18638         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
18639         addresses.
18641 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
18643         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
18645 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
18647         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
18648         TARGET_ELF.
18650 2018-06-08  Martin Liska  <mliska@suse.cz>
18652         * tree-cfg.h (debug_function): Fix argument type to match
18653         implementation.
18655 2018-06-08  Martin Liska  <mliska@suse.cz>
18657         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
18658         Remove usage of MPX-related (and removed) fields.
18659         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
18661 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
18663         * cfg.c (debug): Use TDF_NONE rather than 0.
18664         * cfghooks.c (debug): Likewise.
18665         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
18666         (struct dump_option_value_info): Convert to...
18667         (struct kv_pair): ...this template type.
18668         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
18669         rather than 0.
18670         (optinfo_verbosity_options): Likewise.
18671         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
18672         OPTGROUP_NONE.
18673         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
18674         than int for "optgroup_flags" param.
18675         (dump_generic_expr_loc): Use dump_flags_t rather than int for
18676         "dump_kind" param.
18677         (dump_dec): Likewise.
18678         (dump_finish): Use TDF_NONE rather than 0.
18679         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
18680         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
18681         than 0.  Update for change to option_ptr.
18682         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
18683         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
18684         0.  Update for changes to optinfo_verbosity_options and
18685         optgroup_options.
18686         (opt_info_switch_p): Convert optgroup_flags from int to
18687         optgroup_flags_t.
18688         (dump_basic_block): Use dump_flags_t rather than int
18689         for "dump_kind" param.
18690         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
18691         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
18692         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
18693         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
18694         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
18695         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
18696         TDF_NONE): Convert from macros to...
18697         (enum dump_flag): ...this new enum.
18698         (dump_flags_t): Update to use enum.
18699         (operator|, operator&, operator~, operator|=, operator&=):
18700         Implement for dump_flags_t.
18701         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
18702         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
18703         Convert from macros to...
18704         (enum optgroup_flag): ...this new enum.
18705         (optgroup_flags_t): New typedef.
18706         (operator|, operator|=): Implement for optgroup_flags_t.
18707         (struct dump_file_info): Convert field "alt_flags" to
18708         dump_flags_t.  Convert field "optgroup_flags" to
18709         optgroup_flags_t.
18710         (dump_basic_block): Use dump_flags_t rather than int for param.
18711         (dump_generic_expr_loc): Likewise.
18712         (dump_dec): Likewise.
18713         (dump_register): Convert param "optgroup_flags" to
18714         optgroup_flags_t.
18715         (opt_info_enable_passes): Likewise.
18716         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
18717         than 0.
18718         * gimple-pretty-print.c (debug): Likewise.
18719         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
18720         (merged_store_group::apply_stores): Likewise.
18721         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
18722         * gimple.c (verify_gimple_pp): Likewise.
18723         * graphite-poly.c (print_pbb_body): Likewise.
18724         * passes.c (pass_manager::register_one_dump_file): Convert
18725         local "optgroup_flags" to optgroup_flags_t.
18726         * print-tree.c (print_node): Use TDF_NONE rather than 0.
18727         (debug): Likewise.
18728         (debug_body): Likewise.
18729         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
18730         to optgroup_flags_t.
18731         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
18732         than 0.
18733         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
18734         (convert_mult_to_fma): Likewise.
18735         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
18736         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
18737         * tree-vect-data-refs.c (dump_lower_bound): Convert param
18738         "dump_kind" to dump_flags_t.
18740 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
18742         * config/rs6000/rs6000.c (min, max): Delete.
18744 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
18746         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
18747         -mabi=spe and -mabi=no-spe.
18749 2018-06-08  Martin Liska  <mliska@suse.cz>
18751         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
18752         where we expect an existing summary.
18754 2018-06-08  Martin Liska  <mliska@suse.cz>
18756         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
18757         * ipa-inline.h (estimate_edge_growth): Likewise.
18759 2018-06-08  Martin Liska  <mliska@suse.cz>
18761         * cgraph.c (function_version_hasher::hash): Use
18762         cgraph_node::get_uid ().
18763         (function_version_hasher::equal):
18764         * cgraph.h (cgraph_node::get_uid): New method.
18765         * ipa-inline.c (update_caller_keys): Use
18766         cgraph_node::get_uid ().
18767         (update_callee_keys): Likewise.
18768         * ipa-utils.c (searchc): Likewise.
18769         (ipa_reduced_postorder): Likewise.
18770         * lto-cgraph.c (input_node): Likewise.
18771         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
18772         * symbol-summary.h (symtab_insertion): Likewise.
18773         (symtab_removal): Likewise.
18774         (symtab_duplication): Likewise.
18775         * tree-pretty-print.c (dump_function_header): Likewise.
18776         * tree-sra.c (convert_callers_for_node): Likewise.
18778 2018-06-08  Martin Liska  <mliska@suse.cz>
18780         * cgraph.c (symbol_table::create_edge): Always assign a new
18781         unique number.
18782         (symbol_table::free_edge): Do not recycle numbers.
18783         * cgraph.h (cgraph_edge::get): New method.
18784         * symbol-summary.h (symtab_removal): Use it.
18785         (symtab_duplication): Likewise.
18786         (call_summary::hashable_uid): Remove.
18788 2018-06-08  Martin Liska  <mliska@suse.cz>
18790         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
18791         (initialize_growth_caches): Remove.
18792         (free_growth_caches): Likewise.
18793         (do_estimate_edge_time): Use edge_growth_cache.
18794         (do_estimate_edge_size): Likewise.
18795         (do_estimate_edge_hints): Likewise.
18796         * ipa-inline.c (reset_edge_caches): Likewise.
18797         (recursive_inlining): Likewise.
18798         (inline_small_functions): Likewise.
18799         * ipa-inline.h (initialize_growth_caches): Remove.
18800         (estimate_edge_size): Likewise.
18801         (estimate_edge_time): Likewise.
18802         (estimate_edge_hints): Likewise.
18803         (reset_edge_growth_cache): Likewise.
18804         * symbol-summary.h (call_summary::remove): New method.
18806 2018-06-08  Martin Liska  <mliska@suse.cz>
18808         * ipa-cp.c (class edge_clone_summary): New summary.
18809         (grow_edge_clone_vectors): Remove.
18810         (ipcp_edge_duplication_hook): Remove.
18811         (class edge_clone_summary_t): New call_summary class.
18812         (ipcp_edge_removal_hook): Remove.
18813         (edge_clone_summary_t::duplicate): New function.
18814         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
18815         (create_specialized_node): Likewise.
18816         (ipcp_driver): Initialize edge_clone_summaries and do not
18817         register hooks.
18819 2018-06-08  Martin Liska  <mliska@suse.cz>
18821         * symbol-summary.h (get): New function.
18822         (call_summary::m_initialize_when_cloning): New class member.
18824 2018-06-08  Martin Liska  <mliska@suse.cz>
18826         * cgraph.c (cgraph_node::remove): Do not recycle uid.
18827         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
18828         (symbol_table::allocate_cgraph_symbol): Do not set uid.
18829         * passes.c (uid_hash_t): Record removed_nodes by their uids.
18830         (remove_cgraph_node_from_order): Use the removed_nodes set.
18831         (do_per_function_toporder): Likwise.
18832         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
18833         instead of summary_uid.
18834         (symtab_removal): Likewise.
18835         (symtab_duplication): Likewise.
18837 2018-06-08  Martin Liska  <mliska@suse.cz>
18839         * ipa-cp.c (ipcp_store_bits_results): Use
18840         ipcp_transformation_sum.
18841         (ipcp_store_vr_results): Likewise.
18842         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
18843         to ...
18844         (ipcp_transformation_initialize): ... this.
18845         (ipa_set_node_agg_value_chain):
18846         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
18847         (write_ipcp_transformation_info): Likewise.
18848         (read_ipcp_transformation_info): Likewise.
18849         (ipcp_update_bits): Likewise.
18850         (ipcp_update_vr): Likewise.
18851         (ipcp_transform_function): Likewise.
18852         * ipa-prop.h: Rename ipcp_transformation_summary to
18853         ipcp_transformation.
18854         (class ipcp_transformation_t): New function summary.
18855         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
18856         (ipa_get_agg_replacements_for_node): Likewise.
18858 2018-06-08  Martin Liska  <mliska@suse.cz>
18860         * ipa-pure-const.c (struct funct_state_d): Do it class instead
18861         of struct.
18862         (class funct_state_summary_t): New function_summary class.
18863         (has_function_state): Remove.
18864         (get_function_state): Likewise.
18865         (set_function_state): Likewise.
18866         (add_new_function): Likewise.
18867         (funct_state_summary_t::insert): New function.
18868         (duplicate_node_data): Remove.
18869         (remove_node_data): Remove.
18870         (funct_state_summary_t::duplicate): New function.
18871         (register_hooks): Create new funct_state_summaries.
18872         (pure_const_generate_summary): Use it.
18873         (pure_const_write_summary): Likewise.
18874         (pure_const_read_summary): Likewise.
18875         (propagate_pure_const): Likewise.
18876         (propagate_nothrow): Likewise.
18877         (dump_malloc_lattice): Likewise.
18878         (propagate_malloc): Likewise.
18879         (execute): Do not register hooks, just remove summary
18880         instead.
18881         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
18882         constructor.
18884 2018-06-08  Martin Liska  <mliska@suse.cz>
18886         * ipa-reference.c (remove_node_data): Remove.
18887         (duplicate_node_data): Likewise.
18888         (class ipa_ref_var_info_summary_t): New class.
18889         (class ipa_ref_opt_summary_t): Likewise.
18890         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
18891         (get_reference_optimization_summary): Use
18892         ipa_ref_opt_sum_summaries.
18893         (set_reference_vars_info): Remove.
18894         (set_reference_optimization_summary): Likewise.
18895         (ipa_init): Create summaries.
18896         (init_function_info): Use function summary.
18897         (ipa_ref_opt_summary_t::duplicate): New function.
18898         (ipa_ref_opt_summary_t::remove): New function.
18899         (get_read_write_all_from_node): Fix GNU coding style.
18900         (propagate): Use function summary.
18901         (write_node_summary_p): Fix GNU coding style.
18902         (stream_out_bitmap): Likewise.
18903         (ipa_reference_read_optimization_summary): Use function summary.
18904         (ipa_reference_c_finalize): Do not release hooks.
18906 2018-06-08  Martin Liska  <mliska@suse.cz>
18908         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
18909         (analyze_function_body): Extract multiple calls of get_create.
18910         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
18911         * ipa-inline.c (recursive_inlining): Use ::get method.
18912         * ipa-inline.h (estimate_edge_growth): Likewise.
18914 2018-06-08  Martin Liska  <mliska@suse.cz>
18916         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
18917         HSA_INVALID.
18918         (hsa_function_summary::hsa_function_summary): Use the new enum
18919         value.
18920         (hsa_gpu_implementation_p): Use hsa_summaries::get.
18921         * hsa-gen.c (hsa_get_host_function): Likewise.
18922         (get_brig_function_name): Likewise.
18923         * ipa-hsa.c (process_hsa_functions): Likewise.
18924         (ipa_hsa_write_summary): Likewise.
18925         * symbol-summary.h (symtab_duplication): Use ::get function/
18926         (get): New function.
18928 2018-06-08  Martin Liska  <mliska@suse.cz>
18930         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
18931         of get.
18932         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
18933         (hsa_register_kernel): Likewise.
18934         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
18935         * hsa-gen.c (hsa_get_host_function): Likewise.
18936         (get_brig_function_name): Likewise.
18937         (generate_hsa): Likewise.
18938         (pass_gen_hsail::execute): Likewise.
18939         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
18940         (devirtualization_time_bonus): Likewise.
18941         (ipcp_propagate_stage): Likewise.
18942         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
18943         (edge_set_predicate): Likewise.
18944         (evaluate_conditions_for_known_args): Likewise.
18945         (evaluate_properties_for_edge): Likewise.
18946         (ipa_fn_summary::reset): Likewise.
18947         (ipa_fn_summary_t::duplicate): Likewise.
18948         (dump_ipa_call_summary): Likewise.
18949         (ipa_dump_fn_summary): Likewise.
18950         (analyze_function_body): Likewise.
18951         (compute_fn_summary): Likewise.
18952         (estimate_edge_devirt_benefit): Likewise.
18953         (estimate_edge_size_and_time): Likewise.
18954         (estimate_calls_size_and_time): Likewise.
18955         (estimate_node_size_and_time): Likewise.
18956         (inline_update_callee_summaries): Likewise.
18957         (remap_edge_change_prob): Likewise.
18958         (remap_edge_summaries): Likewise.
18959         (ipa_merge_fn_summary_after_inlining): Likewise.
18960         (ipa_update_overall_fn_summary): Likewise.
18961         (read_ipa_call_summary): Likewise.
18962         (inline_read_section): Likewise.
18963         (write_ipa_call_summary): Likewise.
18964         (ipa_fn_summary_write): Likewise.
18965         (ipa_free_fn_summary): Likewise.
18966         * ipa-hsa.c (process_hsa_functions): Likewise.
18967         (ipa_hsa_write_summary): Likewise.
18968         (ipa_hsa_read_section): Likewise.
18969         * ipa-icf.c (sem_function::merge): Likewise.
18970         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
18971         (do_estimate_edge_time): Likewise.
18972         (estimate_size_after_inlining): Likewise.
18973         (estimate_growth): Likewise.
18974         (growth_likely_positive): Likewise.
18975         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
18976         (inline_call): Likewise.
18977         * ipa-inline.c (caller_growth_limits): Likewise.
18978         (can_inline_edge_p): Likewise.
18979         (can_inline_edge_by_limits_p): Likewise.
18980         (compute_uninlined_call_time): Likewise.
18981         (compute_inlined_call_time): Likewise.
18982         (want_inline_small_function_p): Likewise.
18983         (edge_badness): Likewise.
18984         (update_caller_keys): Likewise.
18985         (update_callee_keys): Likewise.
18986         (recursive_inlining): Likewise.
18987         (inline_small_functions): Likewise.
18988         (inline_to_all_callers_1): Likewise.
18989         (dump_overall_stats): Likewise.
18990         (early_inline_small_functions): Likewise.
18991         (early_inliner): Likewise.
18992         * ipa-inline.h (estimate_edge_growth): Likewise.
18993         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
18994         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18995         * ipa-prop.h (IPA_NODE_REF): Likewise.
18996         (IPA_EDGE_REF): Likewise.
18997         * ipa-pure-const.c (malloc_candidate_p): Likewise.
18998         (propagate_malloc): Likewise.
18999         * ipa-split.c (execute_split_functions): Likewise.
19000         * symbol-summary.h: Rename get to get_create.
19001         (get): Likewise.
19002         (get_create): Likewise.
19003         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
19005 2018-06-08  Martin Liska  <mliska@suse.cz>
19007         * symbol-summary.h (release): Move definition out of class
19008         declaration.
19009         (symtab_removal): Likewise.
19010         (symtab_duplication): Likewise.
19012 2018-06-08  Martin Liska  <mliska@suse.cz>
19014         * symbol-summary.h (function_summary): Move constructor
19015         implementation out of class declaration.
19016         (release): Likewise.
19017         (symtab_insertion): Likewise.
19018         (symtab_removal): Likewise.
19019         (symtab_duplication): Likewise.
19020         (get): Likewise.
19022 2018-06-08  Martin Liska  <mliska@suse.cz>
19024         * Makefile.in: Remove support for MPX (macros, related functions,
19025         fields in cgraph_node, ...).
19026         * builtin-types.def (BT_BND): Likewise.
19027         (BT_FN_BND_CONST_PTR): Likewise.
19028         (BT_FN_CONST_PTR_BND): Likewise.
19029         (BT_FN_VOID_PTR_BND): Likewise.
19030         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
19031         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
19032         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
19033         (expand_builtin_mempcpy_with_bounds): Likewise.
19034         (expand_builtin_memset_with_bounds): Likewise.
19035         (expand_builtin_memset_args): Likewise.
19036         (std_expand_builtin_va_start): Likewise.
19037         (expand_builtin): Likewise.
19038         (expand_builtin_with_bounds): Likewise.
19039         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
19040         (DEF_LIB_BUILTIN_CHKP): Likewise.
19041         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
19042         (DEF_CHKP_BUILTIN): Likewise.
19043         (BUILT_IN_MEMCPY): Likewise.
19044         (BUILT_IN_MEMMOVE): Likewise.
19045         (BUILT_IN_MEMPCPY): Likewise.
19046         (BUILT_IN_MEMSET): Likewise.
19047         (BUILT_IN_STPCPY): Likewise.
19048         (BUILT_IN_STRCAT): Likewise.
19049         (BUILT_IN_STRCHR): Likewise.
19050         (BUILT_IN_STRCPY): Likewise.
19051         (BUILT_IN_STRLEN): Likewise.
19052         (BUILT_IN_MEMCPY_CHK): Likewise.
19053         (BUILT_IN_MEMMOVE_CHK): Likewise.
19054         (BUILT_IN_MEMPCPY_CHK): Likewise.
19055         (BUILT_IN_MEMSET_CHK): Likewise.
19056         (BUILT_IN_STPCPY_CHK): Likewise.
19057         (BUILT_IN_STRCAT_CHK): Likewise.
19058         (BUILT_IN_STRCPY_CHK): Likewise.
19059         * calls.c (store_bounds): Likewise.
19060         (emit_call_1): Likewise.
19061         (special_function_p): Likewise.
19062         (maybe_warn_nonstring_arg): Likewise.
19063         (initialize_argument_information): Likewise.
19064         (finalize_must_preallocate): Likewise.
19065         (compute_argument_addresses): Likewise.
19066         (expand_call): Likewise.
19067         * cfgexpand.c (expand_call_stmt): Likewise.
19068         (expand_return): Likewise.
19069         (expand_gimple_stmt_1): Likewise.
19070         (pass_expand::execute): Likewise.
19071         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
19072         (cgraph_node::remove): Likewise.
19073         (cgraph_node::dump): Likewise.
19074         (cgraph_node::verify_node): Likewise.
19075         * cgraph.h (chkp_function_instrumented_p): Likewise.
19076         (symtab_node::get_alias_target): Likewise.
19077         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
19078         (cgraph_local_p): Likewise.
19079         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
19080         (cgraph_edge::rebuild_references): Likewise.
19081         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
19082         (walk_polymorphic_call_targets): Likewise.
19083         (cgraph_node::expand_thunk): Likewise.
19084         (symbol_table::output_weakrefs): Likewise.
19085         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
19086         (ix86_handle_option): Likewise.
19087         * config/i386/constraints.md: Likewise.
19088         * config/i386/i386-builtin-types.def (BND): Likewise.
19089         (VOID): Likewise.
19090         (PVOID): Likewise.
19091         (ULONG): Likewise.
19092         * config/i386/i386-builtin.def (BDESC_END): Likewise.
19093         (BDESC_FIRST): Likewise.
19094         (BDESC): Likewise.
19095         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
19096         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
19097         * config/i386/i386.c (enum reg_class): Likewise.
19098         (ix86_target_string): Likewise.
19099         (ix86_option_override_internal): Likewise.
19100         (ix86_conditional_register_usage): Likewise.
19101         (ix86_valid_target_attribute_inner_p): Likewise.
19102         (ix86_set_indirect_branch_type): Likewise.
19103         (ix86_set_current_function): Likewise.
19104         (ix86_function_arg_regno_p): Likewise.
19105         (init_cumulative_args): Likewise.
19106         (ix86_function_arg_advance): Likewise.
19107         (ix86_function_arg): Likewise.
19108         (ix86_pass_by_reference): Likewise.
19109         (ix86_function_value_regno_p): Likewise.
19110         (ix86_function_value_1): Likewise.
19111         (ix86_function_value_bounds): Likewise.
19112         (ix86_return_in_memory): Likewise.
19113         (ix86_setup_incoming_vararg_bounds): Likewise.
19114         (ix86_va_start): Likewise.
19115         (indirect_thunk_need_prefix): Likewise.
19116         (print_reg): Likewise.
19117         (ix86_print_operand): Likewise.
19118         (ix86_expand_call): Likewise.
19119         (ix86_output_function_return): Likewise.
19120         (reg_encoded_number): Likewise.
19121         (BDESC_VERIFYS): Likewise.
19122         (ix86_init_mpx_builtins): Likewise.
19123         (ix86_init_builtins): Likewise.
19124         (ix86_emit_cmove): Likewise.
19125         (ix86_emit_move_max): Likewise.
19126         (ix86_expand_builtin): Likewise.
19127         (ix86_builtin_mpx_function): Likewise.
19128         (ix86_get_arg_address_for_bt): Likewise.
19129         (ix86_load_bounds): Likewise.
19130         (ix86_store_bounds): Likewise.
19131         (ix86_load_returned_bounds): Likewise.
19132         (ix86_store_returned_bounds): Likewise.
19133         (ix86_class_likely_spilled_p): Likewise.
19134         (ix86_hard_regno_mode_ok): Likewise.
19135         (x86_order_regs_for_local_alloc): Likewise.
19136         (ix86_mitigate_rop): Likewise.
19137         (ix86_bnd_prefixed_insn_p): Likewise.
19138         (ix86_mpx_bound_mode): Likewise.
19139         (ix86_make_bounds_constant): Likewise.
19140         (ix86_initialize_bounds): Likewise.
19141         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
19142         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
19143         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
19144         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
19145         (TARGET_CHKP_BOUND_MODE): Likewise.
19146         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
19147         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
19148         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
19149         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
19150         * config/i386/i386.h (TARGET_MPX): Likewise.
19151         (TARGET_MPX_P): Likewise.
19152         (VALID_BND_REG_MODE): Likewise.
19153         (FIRST_BND_REG): Likewise.
19154         (LAST_BND_REG): Likewise.
19155         (enum reg_class): Likewise.
19156         (BND_REG_P): Likewise.
19157         (BND_REGNO_P): Likewise.
19158         (BNDmode): Likewise.
19159         (ADJUST_INSN_LENGTH): Likewise.
19160         * config/i386/i386.md: Likewise.
19161         * config/i386/i386.opt: Likewise.
19162         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
19163         (defined): Likewise.
19164         (LINK_MPX): Likewise.
19165         (MPX_SPEC): Likewise.
19166         (LIBMPX_SPEC): Likewise.
19167         (LIBMPXWRAPPERS_SPEC): Likewise.
19168         (CHKP_SPEC): Likewise.
19169         * config/i386/predicates.md: Likewise.
19170         * dbxout.c (dbxout_type): Likewise.
19171         * doc/extend.texi: Likewise.
19172         * doc/invoke.texi: Likewise.
19173         * doc/md.texi: Likewise.
19174         * doc/tm.texi: Likewise.
19175         * doc/tm.texi.in: Likewise.
19176         * dwarf2out.c (is_base_type): Likewise.
19177         (gen_formal_types_die): Likewise.
19178         (gen_subprogram_die): Likewise.
19179         (gen_type_die_with_usage): Likewise.
19180         (gen_decl_die): Likewise.
19181         (dwarf2out_late_global_decl): Likewise.
19182         * expr.c (expand_assignment): Likewise.
19183         (emit_storent_insn): Likewise.
19184         (store_expr_with_bounds): Likewise.
19185         (store_expr): Likewise.
19186         (expand_expr_real_1): Likewise.
19187         * expr.h (store_expr_with_bounds): Likewise.
19188         * function.c (use_register_for_decl): Likewise.
19189         (struct bounds_parm_data): Likewise.
19190         (assign_parms_augmented_arg_list): Likewise.
19191         (assign_parm_find_entry_rtl): Likewise.
19192         (assign_parm_is_stack_parm): Likewise.
19193         (assign_parm_load_bounds): Likewise.
19194         (assign_bounds): Likewise.
19195         (assign_parms): Likewise.
19196         (expand_function_start): Likewise.
19197         * gcc.c (CHKP_SPEC): Likewise.
19198         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
19199         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
19200         (wrestrict_dom_walker::check_call): Likewise.
19201         * gimple.c (gimple_build_call_from_tree): Likewise.
19202         * gimple.h (enum gf_mask): Likewise.
19203         (gimple_call_with_bounds_p): Likewise.
19204         (gimple_call_set_with_bounds): Likewise.
19205         * gimplify.c (gimplify_init_constructor): Likewise.
19206         * ipa-cp.c (initialize_node_lattices): Likewise.
19207         (propagate_constants_across_call): Likewise.
19208         (find_more_scalar_values_for_callers_subset): Likewise.
19209         * ipa-hsa.c (process_hsa_functions): Likewise.
19210         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
19211         * ipa-icf.c (sem_function::merge): Likewise.
19212         * ipa-inline.c (early_inliner): Likewise.
19213         * ipa-pure-const.c (warn_function_noreturn): Likewise.
19214         (warn_function_cold): Likewise.
19215         (propagate_pure_const): Likewise.
19216         * ipa-ref.h (enum GTY): Likewise.
19217         * ipa-split.c (find_retbnd): Likewise.
19218         (consider_split): Likewise.
19219         (split_function): Likewise.
19220         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
19221         * ipa.c (walk_polymorphic_call_targets): Likewise.
19222         (symbol_table::remove_unreachable_nodes): Likewise.
19223         (process_references): Likewise.
19224         (cgraph_build_static_cdtor_1): Likewise.
19225         * lto-cgraph.c (lto_output_node): Likewise.
19226         (output_refs): Likewise.
19227         (compute_ltrans_boundary): Likewise.
19228         (input_overwrite_node): Likewise.
19229         (input_node): Likewise.
19230         (input_cgraph_1): Likewise.
19231         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
19232         * passes.c (pass_manager::execute_early_local_passes): Likewise.
19233         (class pass_chkp_instrumentation_passes): Likewise.
19234         (make_pass_chkp_instrumentation_passes): Likewise.
19235         * passes.def: Likewise.
19236         * rtl.h (struct GTY): Likewise.
19237         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
19238         * stor-layout.c (layout_type): Likewise.
19239         * symtab.c: Likewise.
19240         * target.def: Likewise.
19241         * targhooks.c (default_chkp_bound_type): Likewise.
19242         (default_chkp_bound_mode): Likewise.
19243         (default_builtin_chkp_function): Likewise.
19244         (default_chkp_function_value_bounds): Likewise.
19245         (default_chkp_make_bounds_constant): Likewise.
19246         (default_chkp_initialize_bounds): Likewise.
19247         * targhooks.h (default_chkp_bound_type): Likewise.
19248         (default_chkp_bound_mode): Likewise.
19249         (default_builtin_chkp_function): Likewise.
19250         (default_chkp_function_value_bounds): Likewise.
19251         (default_chkp_make_bounds_constant): Likewise.
19252         (default_chkp_initialize_bounds): Likewise.
19253         * toplev.c (compile_file): Likewise.
19254         (process_options): Likewise.
19255         * tree-core.h (DEF_BUILTIN): Likewise.
19256         (DEF_BUILTIN_CHKP): Likewise.
19257         * tree-inline.c (declare_return_variable): Likewise.
19258         (remap_gimple_stmt): Likewise.
19259         (copy_bb): Likewise.
19260         (initialize_inlined_parameters): Likewise.
19261         (expand_call_inline): Likewise.
19262         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
19263         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
19264         (make_pass_ipa_chkp_produce_thunks): Likewise.
19265         (make_pass_chkp): Likewise.
19266         (make_pass_chkp_opt): Likewise.
19267         (make_pass_chkp_instrumentation_passes): Likewise.
19268         * tree-pretty-print.c (dump_generic_node): Likewise.
19269         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
19270         * tree-ssa-dce.c (propagate_necessity): Likewise.
19271         (eliminate_unnecessary_stmts): Likewise.
19272         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
19273         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
19274         * tree-ssa-sccvn.h: Likewise.
19275         * tree-ssa-strlen.c (get_string_length): Likewise.
19276         (valid_builtin_call): Likewise.
19277         (adjust_last_stmt): Likewise.
19278         (handle_builtin_strchr): Likewise.
19279         (handle_builtin_strcpy): Likewise.
19280         (handle_builtin_stxncpy): Likewise.
19281         (handle_builtin_memcpy): Likewise.
19282         (handle_builtin_strcat): Likewise.
19283         (strlen_check_and_optimize_stmt): Likewise.
19284         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
19285         * tree-streamer-in.c: Likewise.
19286         * tree-streamer.c (record_common_node): Likewise.
19287         * tree.c (tree_code_size): Likewise.
19288         (wide_int_to_tree_1): Likewise.
19289         (type_contains_placeholder_1): Likewise.
19290         (build_common_tree_nodes): Likewise.
19291         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
19292         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
19293         (POINTER_BOUNDS_P): Likewise.
19294         (BOUNDED_TYPE_P): Likewise.
19295         (BOUNDED_P): Likewise.
19296         (CALL_WITH_BOUNDS_P): Likewise.
19297         (pointer_bounds_type_node): Likewise.
19298         * value-prof.c (gimple_ic): Likewise.
19299         * var-tracking.c (vt_add_function_parameters): Likewise.
19300         * varasm.c (make_decl_rtl): Likewise.
19301         (assemble_start_function): Likewise.
19302         (output_constant): Likewise.
19303         (maybe_assemble_visibility): Likewise.
19304         * varpool.c (ctor_for_folding): Likewise.
19305         * chkp-builtins.def: Remove.
19306         * ipa-chkp.c: Remove.
19307         * ipa-chkp.h: Remove.
19308         * rtl-chkp.c: Remove.
19309         * rtl-chkp.h: Remove.
19310         * tree-chkp-opt.c: Remove.
19311         * tree-chkp.c: Remove.
19312         * tree-chkp.h: Remove.
19314 2018-06-07  Carl Love  <cel@us.ibm.com>
19316         * config/rs6000/vsx.md (vextract_fp_from_shorth,
19317         vextract_fp_from_shortl): Add BE support.
19319 2018-06-07  Paul Koning  <ni1d@arrl.net>
19321         * compare-elim.c (try_merge_compare): Don't merge compare if
19322         address contains a side effect.
19323         (try_eliminate_compare): Likewise.
19325 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
19327         * config.gcc: Support "tremont".
19328         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
19329         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19330         PROCESSOR_TREMONT.
19331         * config/i386/i386.c (m_TREMONT): Define.
19332         (processor_target_table): Add "tremont".
19333         (PTA_TREMONT): Define.
19334         (ix86_lea_outperforms): Add TARGET_TREMONT.
19335         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
19336         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
19337         and M_INTEL_GOLDMONT_PLUS.
19338         (fold_builtin_cpu): Add "tremont".
19339         (ix86_add_stmt_cost): Add TARGET_TREMONT.
19340         (ix86_option_override_internal): Add "tremont".
19341         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
19342         (processor_type): Add PROCESSOR_TREMONT.
19343         * config/i386/x86-tune.def: Add m_TREMONT.
19344         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
19346 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19348         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
19349         symbol defined for msp430i* devices to be lower case.
19351 2018-06-07  Richard Biener  <rguenther@suse.de>
19353         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
19354         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
19355         Properly wrap signed arithmetic if overflow wraps.
19357 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
19359         PR tree-optimization/69615
19360         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
19361         of a cast from a same precision integral SSA_NAME in a bb dominated
19362         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
19363         cast to utype if rhs2 has already a compatible type.
19365 2018-06-07  Richard Biener  <rguenther@suse.de>
19367         PR tree-optimization/85935
19368         * graphite-scop-detection.c (find_params_in_bb): Analyze
19369         condition operands with respect to the correct loop.  Assert
19370         the analysis doesn't fail.
19372 2018-06-04  Carl Love  <cel@us.ibm.com>
19374         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
19375         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
19376         as it is slightly cheaper.
19377         (first_match_or_eos_index_<mode>):
19378         Calculate index using natural element order.
19379         (first_match_index_<mode>):
19380         Calculate index using natural element order.
19381         (first_match_or_eos_index_<mode>):
19382         Calculate index using natural order.
19383         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
19384         for BE and LE modes.
19385         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
19386         P9V_BUILTIN_VCLZLSBB_V16QI.
19387         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
19388         specific.
19390 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19392         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
19393         indentation and line wrap for many prototypes.  Add missing
19394         @smallexample directives around block of prototypes for vec_xl and
19395         vec_xst.
19397 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
19399         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
19400         track if we pass or return IEEE 128-bit floating point.
19401         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
19402         C++ mangling that is compatible with GCC 8.1.
19403         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
19404         (init_cumulative_args): Note if we pass or return IEEE 128-bit
19405         floating point types.
19406         (rs6000_function_arg_advance_1): Likewise.
19407         (rs6000_mangle_type): Optionally generate mangled names that match
19408         what GCC 8.1 generated for IEEE 128-bit floating point types.
19409         (rs6000_globalize_decl_name): If we have an external function that
19410         passes or returns IEEE 128-bit types, generate a weak reference
19411         from the mangled name used in GCC 8.1 to the current mangled
19412         name.
19413         (rs6000_init_builtins): Make __ibm128 use the long double type if
19414         long double is IBM extended double.  Make __float128 use the long
19415         double type if long double is IEEE 128-bit.
19417         PR target/85657
19418         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
19419         macro for __ibm128 built-in functions.
19420         (PACK_IF): Add __ibm128 pack/unpack functions.
19421         (UNPACK_IF): Likewise.
19422         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
19423         enable long double built-in functions if long double is IEEE
19424         128-bit floating point.
19425         (rs6000_invalid_builtin): Update long double built-in function
19426         error message.
19427         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
19428         functions, adjust the built-in function to use the long double
19429         built-in function if __ibm128 and long double are the same type.
19430         * doc/extend.texi (PowerPC builtins): Update documention for
19431         __builtin_{,un}pack_longdouble.  Add documentation for
19432         __builtin_{,un}pack_ibm128.
19434 2018-06-06  Jim Wilson  <jimw@sifive.com>
19436         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
19437         (struct machine_function): New field interrupt_mode.
19438         (riscv_handle_type_attribute): New function.  Add forward declaration.
19439         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
19440         (riscv_expand_epilogue): Check interrupt_mode field.
19441         (riscv_set_current_function): Check interrupt attribute args and
19442         set interrupt_mode field.
19443         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
19444         (riscv_sret, riscv_uret): New.
19445         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
19446         new arguments to interrupt attribute.
19448 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
19450         PR target/63177
19451         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
19452         Don't handle -mcpu=power8 if -mpower9-vector is also used.
19454 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19456         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
19457         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
19458         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19459         several redundant entries.
19461 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
19463         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
19464         type from "rtx" to "rtx_insn *".
19465         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
19466         for local "call_insn", removing cast.
19467         (ix86_expand_call): Likewise, introducing a "call_insn" local.
19469 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
19471         PR tree-optimization/86066
19472         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
19473         for BIT_INSERT_EXPR stores.
19475 2018-06-06  Richard Biener  <rguenther@suse.de>
19477         PR tree-optimization/86062
19478         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
19479         component refs ontop
19480         of to be offsetted base.
19482 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19484         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
19485         to be static and remove check on interrupt attribute name.
19487 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19489         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
19490         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
19492 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
19494         PR target/79924
19495         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
19496         second argument.
19497         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
19498         Remove second argument, change how error is called.
19499         (aarch64_layout_arg): Remove second argument from
19500         aarch64_err_no_fpadvsimd call.
19501         (aarch64_init_cumulative_args): Ditto.
19502         (aarch64_gimplify_va_arg_expr): Ditto.
19503         * config/aarch64/aarch64.md (mov<mode>): Ditto.
19505 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
19507         * config/i386/i386.md (simple_return_indirect_internal): New expander.
19508         (*simple_return_indirect_internal<mode>): Rename from
19509         simple_return_indirect_internal.  Use W mode iterator.
19510         (rstorssp): New expander.
19511         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
19512         (clrssbsy): New expander.
19513         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
19515 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
19517         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
19518         __typeof__.
19519         (cmse_check_pointed_object): Likewise.
19521 2018-06-05  Martin Liska  <mliska@suse.cz>
19523         PR gcov-profile/47618
19524         * doc/invoke.texi: Document how -fprofile-dir format
19525         is extended.
19527 2018-06-05  Richard Biener  <rguenther@suse.de>
19529         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
19530         removal pretend DOM info isn't available so we do not update
19531         it and only remove edges, not dominated blocks.  Actually free
19532         DOM info in case we removed something.  Remove unreachable blocks.
19533         (mfb_keep_latches): Work with either DOM info or marked backedges.
19534         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
19535         first.  Mark backedges if DOM info isn't available.
19536         (Re-)compute DOM info after cleanup_control_flow_pre.
19538 2018-06-05  Richard Biener  <rguenther@suse.de>
19540         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
19541         (locus_discrim_hasher::hash): Adjust.
19542         (locus_discrim_hasher::equal): Likewise.
19543         (next_discriminator_for_locus): Work on line directly.
19544         (same_line_p): Pass in expanded locus1 as well.
19545         (assign_discriminators): Avoid redundant location expansions.
19547 2018-06-05  Richard Biener  <rguenther@suse.de>
19549         PR tree-optimization/86046
19550         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
19551         if required after clearing TREE_ADDRESSABLE.
19553 2018-06-05  Richard Biener  <rguenther@suse.de>
19555         PR tree-optimization/86047
19556         * tree-ssa-loop.c (for_each_index): Glob handling of all
19557         decls and constants and really handle all of them.
19559 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19561         PR target/81497
19562         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
19563         qualifier_void_pointer and qualifier_const_void_pointer.
19564         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
19565         (arm_init_builtins): Handle the above.
19566         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
19567         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
19568         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
19569         void intrinsics.
19571 2018-06-05  Martin Liska  <mliska@suse.cz>
19573         * auto-profile.c (read_autofdo_file): Do not use
19574         gcov_ctr_summary struct.
19575         (afdo_callsite_hot_enough_for_early_inline): Likewise.
19576         * coverage.c (struct counts_entry): Likewise.
19577         (read_counts_file): Read just single summary entry.
19578         (get_coverage_counts): Use gcov_summary struct.
19579         * coverage.h (get_coverage_counts): Likewise.
19580         * gcov-dump.c (dump_working_sets): Likewise.
19581         (tag_summary): Dump just single summary.
19582         * gcov-io.c (gcov_write_summary): Write just histogram
19583         summary.
19584         (gcov_read_summary): Read just single summary.
19585         (compute_working_sets): Use gcov_summary struct.
19586         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
19587         of GCOV_COUNTERS_SUMMABLE.
19588         (GCOV_COUNTERS_SUMMABLE): Remove.
19589         (GCOV_FIRST_VALUE_COUNTER): Replace with
19590         GCOV_COUNTER_V_INTERVAL.
19591         (struct gcov_ctr_summary): Remove.
19592         (struct gcov_summary): Directly use fields of former
19593         gcov_ctr_summary.
19594         (compute_working_sets): Use gcov_summary struct.
19595         * gcov.c (read_count_file): Do not use ctrs fields.
19596         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
19597         struct.
19598         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
19599         struct.
19600         * profile.c: Likewise.
19601         * profile.h: Likewise.
19603 2018-06-05  Martin Liska  <mliska@suse.cz>
19605         PR gcov-profile/84846
19606         * gcov.c (output_lines): Print working directory only
19607         in intermediate format.
19609 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
19611         * config/s390/s390-builtin-types.def: Add void function type.
19612         * config/s390/s390-builtins.def: Use the function type for the
19613         tbeginc builtin.
19615 2018-06-04  Jim Wilson  <jimw@sifive.com>
19617         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
19618         to int.
19619         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
19620         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
19621         handle EH_RETURN_DATA_REGNO registers properly.
19622         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
19623         (riscv_expand_epilogue): Update comment.  Change argument name and
19624         type.  Update code to use new name and type.  Pass new args to
19625         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
19626         EXCEPTION_RETURN.
19627         * config/riscv/riscv.md (NORMAL_RETURN): New.
19628         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
19629         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
19630         (eh_return): Call gen_eh_return_internal and emit barrier.
19631         (eh_return_internal): Call riscv_expand_epilogue.
19633 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
19635         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
19636         bit_insertion field and declare can_be_merged_into method.
19637         (merged_store_group::can_be_merged_into): New method.
19638         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
19639         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
19640         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
19642 2018-06-04  Richard Biener  <rguenther@suse.de>
19644         PR tree-optimization/85955
19645         * builtins.c (fold_builtin_sincos): Convert pointers to
19646         destination to appropriate type before dereferencing.
19648 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
19650         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
19652 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
19654         * expr.c (expand_expr_real_1): Force the operand into memory if
19655         its TYPE_MODE is BLKmode and if there is no integer mode for
19656         the number of bits being extracted.
19658 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
19660         PR target/85832
19661         PR target/86036
19662         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
19663         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
19665 2018-06-04  Richard Biener  <rguenther@suse.de>
19667         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
19668         (cleanup_tree_cfg_noloop): ... single caller.  Do
19669         start_recording_case_labels later.
19671 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
19673         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
19674         to _IMMINTRIN_H_INCLUDED.
19675         * config/i386/pconfigintrin.h: Ditto.
19676         * config/i386/waitpkgintrin.h: Ditto.
19677         * config/i386/immintrin.h: Add includes for sgxintrin.h,
19678         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
19679         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
19680         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
19681         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
19682         waitpkgintrin.h and cldemoteintrin.h.
19684 2018-06-04  Richard Biener  <rguenther@suse.de>
19686         PR tree-optimization/86038
19687         * tracer.c (find_best_successor): Check probability for
19688         being initialized, bail out if not.
19690 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
19692         PR target/86003
19693         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
19694         of bits to ignore when comparing architectures.
19696 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
19698         PR tree-optimization/69615
19699         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
19700         maximum or minimum of the type, try to merge it also as if
19701         range1 is + [-, x - 1] or + [x + 1, -].
19703         PR c++/86025
19704         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
19706 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
19708         PR tree-optimization/86034
19709         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
19710         the unsigned bitfield type in a bit insertion sequence if it does not
19711         have a larger precision than the bitfield size.
19712         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
19714 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
19716         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
19718 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19720         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
19721         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
19722         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
19723         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
19725 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19727         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
19728         Disable -fdelete-null-pointer-checks for ELF toolchain.
19730 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19731             Kito Cheng  <kito.cheng@gmail.com>
19733         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
19734         (nds32le-*-*, nds32be-*-*): Integrate checking process.
19735         (nds32*-*-*): Add glibc and uclibc conditions.
19736         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
19737         (TARGET_EXCEPT_UNWIND_INFO): Define.
19738         * config/nds32/elf.h: New file.
19739         * config/nds32/linux.h: New file.
19740         * config/nds32/nds32-elf.opt: New file.
19741         * config/nds32/nds32-linux.opt: New file.
19742         * config/nds32/nds32-fp-as-gp.c
19743         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
19744         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
19745         TARGET_LINUX_ABI.
19746         (nds32_asm_file_end): Ditto.
19747         (nds32_print_operand): Ditto.
19748         (nds32_insert_attributes): Ditto.
19749         (nds32_init_libfuncs): New function.
19750         (TARGET_HAVE_TLS): Define.
19751         (TARGET_INIT_LIBFUNCS): Define.
19752         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
19753         spec content.
19754         (TARGET_ELF): Apply different mcmodel setting.
19755         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
19756         been migrated into elf.h and linux.h files.
19757         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
19758         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
19759         (mcmodel): The content has been migrated into nds32-elf.opt and
19760         nds32-linux.opt files.
19761         * config/nds32/t-elf: New file.
19762         * config/nds32/t-linux: New file.
19764 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19765             Shiva Chen  <shiva0217@gmail.com>
19767         * config/nds32/constants.md (unspec_volatile_element): Add
19768         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
19769         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
19770         optimization.
19771         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
19772         (make_pass_nds32_fp_as_gp): Declare.
19773         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
19774         optmization pass.
19775         (nds32_asm_function_end_prologue): Remove unused asm output.
19776         (nds32_asm_function_begin_epilogue): Remove unused asm output.
19777         (nds32_asm_file_start): Output necessary fp_as_gp information.
19778         (nds32_option_override): Adjust register usage.
19779         (nds32_expand_prologue): Consider fp_as_gp situation.
19780         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
19781         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
19782         (epilogue): Ditto.
19783         (return): Ditto.
19784         (simple_return): Ditto.
19785         (omit_fp_begin): Output special directive for fp_as_gp.
19786         (omit_fp_end): Output special directive for fp_as_gp.
19787         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
19788         mforbid-fp-as-gp): New options.
19790 2018-06-01  Mark Wielaard  <mark@klomp.org>
19792         * dwarf2out.c (dwarf2out_finish): Remove generation of
19793         DW_AT_loclists_base.
19795 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
19797         * gimple-ssa-store-merging.c: Include gimple-fold.h.
19798         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
19799         (struct merged_store_group): Add bit_insertion field.
19800         (dump_char_array): Use standard hexadecimal format.
19801         (merged_store_group::merged_store_group): Set bit_insertion to false.
19802         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
19803         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
19804         also print the mask in the dump file.
19805         (pass_store_merging::gate): Minor tweak.
19806         (imm_store_chain_info::coalesce_immediate): Fix wrong association
19807         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
19808         stores with INTEGER_CST stores.
19809         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
19810         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
19811         and use it throughout.  Generate bit insertion sequences if need be.
19812         (pass_store_merging::process_store): Remove redundant condition.
19813         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
19815 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
19817         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
19818         the 128-bit floating point types.  Fix function comment.
19820 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19822         * config/aarch64/aarch64-simd.md
19823         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
19824         mnemonics.
19825         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
19826         mnemonics.
19828 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
19830         PR tree-optimization/85989
19831         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
19832         variable.
19833         (backprop::intersect_uses): Check it when deciding whether this
19834         is a backedge reference.
19835         (backprop::process_block): Add each phi to m_visited_phis
19836         after visiting it, then clear it at the end.
19838 2018-06-01  Richard Biener  <rguenther@suse.de>
19840         * tree-vectorizer.h (vect_dr_stmt): New function.
19841         (vect_get_load_cost): Adjust.
19842         (vect_get_store_cost): Likewise.
19843         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
19844         Use vect_dr_stmt instead of DR_SMTT.
19845         (vect_record_base_alignments): Likewise.
19846         (vect_calculate_target_alignment): Likewise.
19847         (vect_compute_data_ref_alignment): Likewise and make static.
19848         (vect_update_misalignment_for_peel): Likewise.
19849         (vect_verify_datarefs_alignment): Likewise.
19850         (vector_alignment_reachable_p): Likewise.
19851         (vect_get_data_access_cost): Likewise.  Pass down
19852         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
19853         (vect_get_peeling_costs_all_drs): Likewise.
19854         (vect_peeling_hash_get_lowest_cost): Likewise.
19855         (vect_enhance_data_refs_alignment): Likewise.
19856         (vect_find_same_alignment_drs): Likewise.
19857         (vect_analyze_data_refs_alignment): Likewise.
19858         (vect_analyze_group_access_1): Likewise.
19859         (vect_analyze_group_access): Likewise.
19860         (vect_analyze_data_ref_access): Likewise.
19861         (vect_analyze_data_ref_accesses): Likewise.
19862         (vect_vfa_segment_size): Likewise.
19863         (vect_small_gap_p): Likewise.
19864         (vectorizable_with_step_bound_p): Likewise.
19865         (vect_prune_runtime_alias_test_list): Likewise.
19866         (vect_analyze_data_refs): Likewise.
19867         (vect_supportable_dr_alignment): Likewise.
19868         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
19869         (vect_gen_prolog_loop_niters): Likewise.
19870         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
19871         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
19872         modify DR_STMT.
19873         (vect_recog_mask_conversion_pattern): Likewise.
19874         (vect_try_gather_scatter_pattern): Likewise.
19875         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
19876         to vect_get_store_cost.
19877         (vect_get_store_cost): Get stmt_info instead of DR.
19878         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
19879         (vect_get_load_cost): Get stmt_info instead of DR.
19881 2018-06-01  Richard Biener  <rguenther@suse.de>
19883         PR middle-end/86017
19884         * gimple-fold.c (var_decl_component_p): Also allow offsetted
19885         vars wrapped in MEM_REFs.
19887 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
19889         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
19890         Fix subreg tests so that we only return a choice between
19891         GENERAL_REGS and FP_REGS if the original classes included both.
19893 2018-06-01  Richard Biener  <rguenther@suse.de>
19895         PR ipa/85960
19896         * tree-ssa-structalias.c (get_function_part_constraint):
19897         Handle NULL fi->decl.
19898         (find_func_aliases_for_call): Properly handle indirect
19899         fi from direct call.
19900         (find_func_clobbers): Likewise.
19901         (ipa_pta_execute): Likewise.
19902         (create_variable_info_for): For functions that are ifunc_resolver
19903         resolve to a varinfo that contains the result of the resolver call.
19904         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
19905         aliases.
19907 2018-05-31  Michael Collison  <michael.collison@arm.com>
19909         * config/aarch64/aarch64.md:
19910         (*fix_to_zero_extenddfdi2): New pattern.
19911         * gcc.target/aarch64/fix_extend1.c: New testcase.
19913 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
19915         PR middle-end/78809
19916         PR middle-end/83026
19917         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
19918         and BUILT_IN_STRNCMP_EQ.
19919         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
19920         BUILT_IN_STRNCMP_EQ.
19921         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
19922         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
19923         (gimple_fold_builtin): Likewise.
19924         * tree-ssa-strlen.c (compute_string_length): New function.
19925         (determine_min_obsize): New function.
19926         (handle_builtin_string_cmp): New function to handle calls to
19927         string compare functions.
19928         (strlen_optimize_stmt): Add handling to builtin string compare
19929         calls.
19930         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
19931         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
19932         * tree.c (build_common_builtin_nodes): Add new defines of
19933         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
19935 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
19937         PR target/85984
19938         * bb-reorder.c (pass_partition_blocks::gate): Return false for
19939         functions with naked attribute.
19941 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
19943         * config/i386/sse.md (avx_vec_concat<mode>):
19944         Substitute concat_tg_mode mode attribute with xtg_mode.
19945         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
19946         (concat_tg_mode): Remove mode attribute.
19948 2018-05-31  Martin Sebor  <msebor@redhat.com>
19950         PR c/82063
19951         * calls.c (alloc_max_size): Correct a logic error/typo.
19952         Treat excessive arguments as infinite.  Warn for invalid arguments.
19953         * doc/invoke.texi (-Walloc-size-larger-than): Update.
19955 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
19957         PR target/85829
19958         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
19959         and movx for Haswell.
19961 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
19962             Cesar Philippidis  <cesar@codesourcery.com>
19964         PR middle-end/85879
19965         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
19966         when emitting error on private/firstprivate reductions.
19967         * omp-low.c (lower_omp_target): Avoid reference-type processing
19968         on pointers for firstprivate clause.
19970 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
19972         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
19973         (st1x2): Likewise.
19974         (st1x3): Likewise.
19975         * config/aarch64/aarch64-simd.md
19976         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
19977         (aarch64_ld1_x3_<mode>): Likewise
19978         (aarch64_st1x2<VALLDIF:mode>): Likewise
19979         (aarch64_st1_x2_<mode>): Likewise
19980         (aarch64_st1x3<VALLDIF:mode>): Likewise
19981         (aarch64_st1_x3_<mode>): Likewise
19982         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
19983         (vld1_s8_x3): Likewise.
19984         (vld1_u16_x3): Likewise.
19985         (vld1_s16_x3): Likewise.
19986         (vld1_u32_x3): Likewise.
19987         (vld1_s32_x3): Likewise.
19988         (vld1_u64_x3): Likewise.
19989         (vld1_s64_x3): Likewise.
19990         (vld1_f16_x3): Likewise.
19991         (vld1_f32_x3): Likewise.
19992         (vld1_f64_x3): Likewise.
19993         (vld1_p8_x3): Likewise.
19994         (vld1_p16_x3): Likewise.
19995         (vld1_p64_x3): Likewise.
19996         (vld1q_u8_x3): Likewise.
19997         (vld1q_s8_x3): Likewise.
19998         (vld1q_u16_x3): Likewise.
19999         (vld1q_s16_x3): Likewise.
20000         (vld1q_u32_x3): Likewise.
20001         (vld1q_s32_x3): Likewise.
20002         (vld1q_u64_x3): Likewise.
20003         (vld1q_s64_x3): Likewise.
20004         (vld1q_f16_x3): Likewise.
20005         (vld1q_f32_x3): Likewise.
20006         (vld1q_f64_x3): Likewise.
20007         (vld1q_p8_x3): Likewise.
20008         (vld1q_p16_x3): Likewise.
20009         (vld1q_p64_x3): Likewise.
20010         (vst1_s64_x2): Likewise.
20011         (vst1_u64_x2): Likewise.
20012         (vst1_f64_x2): Likewise.
20013         (vst1_s8_x2): Likewise.
20014         (vst1_p8_x2): Likewise.
20015         (vst1_s16_x2): Likewise.
20016         (vst1_p16_x2): Likewise.
20017         (vst1_s32_x2): Likewise.
20018         (vst1_u8_x2): Likewise.
20019         (vst1_u16_x2): Likewise.
20020         (vst1_u32_x2): Likewise.
20021         (vst1_f16_x2): Likewise.
20022         (vst1_f32_x2): Likewise.
20023         (vst1_p64_x2): Likewise.
20024         (vst1q_s8_x2): Likewise.
20025         (vst1q_p8_x2): Likewise.
20026         (vst1q_s16_x2): Likewise.
20027         (vst1q_p16_x2): Likewise.
20028         (vst1q_s32_x2): Likewise.
20029         (vst1q_s64_x2): Likewise.
20030         (vst1q_u8_x2): Likewise.
20031         (vst1q_u16_x2): Likewise.
20032         (vst1q_u32_x2): Likewise.
20033         (vst1q_u64_x2): Likewise.
20034         (vst1q_f16_x2): Likewise.
20035         (vst1q_f32_x2): Likewise.
20036         (vst1q_f64_x2): Likewise.
20037         (vst1q_p64_x2): Likewise.
20038         (vst1_s64_x3): Likewise.
20039         (vst1_u64_x3): Likewise.
20040         (vst1_f64_x3): Likewise.
20041         (vst1_s8_x3): Likewise.
20042         (vst1_p8_x3): Likewise.
20043         (vst1_s16_x3): Likewise.
20044         (vst1_p16_x3): Likewise.
20045         (vst1_s32_x3): Likewise.
20046         (vst1_u8_x3): Likewise.
20047         (vst1_u16_x3): Likewise.
20048         (vst1_u32_x3): Likewise.
20049         (vst1_f16_x3): Likewise.
20050         (vst1_f32_x3): Likewise.
20051         (vst1_p64_x3): Likewise.
20052         (vst1q_s8_x3): Likewise.
20053         (vst1q_p8_x3): Likewise.
20054         (vst1q_s16_x3): Likewise.
20055         (vst1q_p16_x3): Likewise.
20056         (vst1q_s32_x3): Likewise.
20057         (vst1q_s64_x3): Likewise.
20058         (vst1q_u8_x3): Likewise.
20059         (vst1q_u16_x3): Likewise.
20060         (vst1q_u32_x3): Likewise.
20061         (vst1q_u64_x3): Likewise.
20062         (vst1q_f16_x3): Likewise.
20063         (vst1q_f32_x3): Likewise.
20064         (vst1q_f64_x3): Likewise.
20065         (vst1q_p64_x3): Likewise.
20067 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20069         * config/msp430/msp430.c (msp430_output_labelref): Prepend
20070         user_label_prefix to name.
20072         * tree-core.h: Update comment about the format of NAME string
20073         passed to handler in attribute_spec.
20075         * config/msp430/msp430.md: Remove erroneous subreg expression from
20076         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
20077         zero_extend{q,h}isi2.
20079 2018-05-30  Borislav Petkov  <bp@suse.de>
20081         * doc/extend.texi: Document some architecture specific
20082         constraints and sort entries.
20084 2018-05-30  Martin Sebor  <msebor@redhat.com>
20086         PR middle-end/85369
20087         * builtins.c (expand_builtin_stpcpy_1): New function.
20088         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
20089         only if the former succeeds.
20091 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
20093         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
20094         in saphira.
20096 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20098         * doc/invoke.texi (-flinker-output): Document
20100 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20102         * passes.c (ipa_write_summaries): Only modify statements if body
20103         is in memory.
20104         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
20105         incrementally linking.
20106         (ipa_passes): Likewise.
20107         * lto-cgraph.c (lto_output_node): When incrementally linking do not
20108         pass down resolution info.
20109         * common.opt (flag_incremental_link): Update info.
20110         * gcc.c (plugin specs): Turn flinker-output=* to
20111         -plugin-opt=-linker-output-known
20112         * toplev.c (compile_file): Also cut compilation when doing incremental
20113         link.
20114         * flag-types. (enum lto_partition_model): Add
20115         LTO_LINKER_OUTPUT_NOLTOREL.
20116         (invoke.texi): Add -flinker-output docs.
20117         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
20118         link same way as WPA; do not stream in dead initializers.
20120         * dwarf2out.c (dwarf2out_die_ref_for_decl,
20121         darf2out_register_external_decl): Support incremental link.
20123 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20125         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
20127 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
20129         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
20130         it down to simple_object_copy_lto_debug_sections.
20131         (run_gcc): Determine incremental LTO link time and configure
20132         lto1 into non-wpa mode, disable renaming of debug sections.
20134 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20136         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20137         descriptions of various incorrectly documented functions.
20139 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
20141         Revert:
20142         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
20143         address check not strict.
20145 2018-05-30  Richard Biener  <rguenther@suse.de>
20147         PR tree-optimization/85964
20148         * tracer.c (better_p): Drop initialized count check, we only
20149         call the function with initialized counts now.
20150         (find_best_successor): Do find a best edge if one
20151         has uninitialized count.
20152         (find_best_predecessor): Likewise.  Do BB frequency check only
20153         if count is initialized.
20155 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
20157         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
20158         (aarch64_ldrstr_offset_compare): New.
20159         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
20160         load/store orderings.
20161         (aarch64_gen_adjusted_ldpstp): Likewise.
20163 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
20165         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
20166         Check for subset of GENERAL_REGS and FP_REGS.
20167         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
20168         r=w alternative.
20170 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
20172         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
20173         and wi::to_poly_offset.  Add the current offset and then check
20174         whether the sum fits, rather than using an unchecked addition of
20175         a checked term.  Check for a shwi rather than a uhwi.
20176         * expr.c (get_bit_range): Use tree_to_poly_uint64.
20177         (store_constructor): Use poly_int_tree_p.
20178         (expand_expr_real_1): Likewise.
20179         * function.c (assign_temp): Likewise.
20180         * fold-const.c (const_binop): Use poly_int_tree_p and
20181         wi::to_poly_offset.
20182         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
20183         division.
20184         * ipa-icf-gimple.c (func_checker::compare_operand): Use
20185         to_poly_offset for MEM offsets.
20186         * ipa-icf.c (sem_variable::equals): Likewise.
20187         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
20188         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20189         wi::to_poly_offset for BIT_FIELD_REF offsets.
20190         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
20191         wi::to_poly_offset.
20192         * var-tracking.c (emit_note_insn_var_location): Use
20193         tree_to_poly_uint64.
20195 2018-05-29  Jim Wilson  <jimw@sifive.com>
20197         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
20199 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
20201         PR target/85950
20202         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
20203         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
20204         sequence.
20205         (sse4_1_round<mode>2): Use nonimmediate_operand
20206         for operand 1 predicate.
20208 2018-05-29  Martin Sebor  <msebor@redhat.com>
20209             Richard Biener  <rguenther@suse.de>
20211         PR testsuite/85888
20212         * calls.c (get_size_range): Call determine_value_range instead
20213         of get_value_range..
20214         * tree-vrp.h (determine_value_range): Declared new function.
20215         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
20217 2018-05-29  Richard Biener  <rguenther@suse.de>
20219         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
20220         sure to use non-pattern stmts for get_earlier_stmt arguments.
20221         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
20222         called on pattern stmts.
20223         (get_later_stmt): Likewise.
20225 2018-05-29  Martin Liska  <mliska@suse.cz>
20227         PR gcov-profile/85759
20228         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
20229         env variables.
20231 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
20233         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
20234         VEC_UNPACK_*_EXPR.
20235         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
20236         VEC_PACK_*_EXPR.
20238         PR target/85918
20239         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
20240         VEC_PACK_FLOAT_EXPR): New tree codes.
20241         * tree-pretty-print.c (op_code_prio): Handle
20242         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
20243         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20244         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
20245         * tree-inline.c (estimate_operator_cost): Likewise.
20246         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
20247         * fold-const.c (const_binop): Likewise.
20248         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
20249         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
20250         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
20251         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
20252         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20253         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
20254         * expr.c (expand_expr_real_2): Likewise.
20255         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
20256         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
20257         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
20258         optabs.
20259         * optabs.c (expand_widen_pattern_expr): For
20260         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
20261         sign from result type rather than operand's type.
20262         (expand_binop_directly): For vec_packu_float_optab and
20263         vec_packs_float_optab allow result type to be different from operand's
20264         type.
20265         * optabs-tree.c (optab_for_tree_code): Handle
20266         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
20267         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
20268         * tree-vect-generic.c (expand_vector_operations_1):  Handle
20269         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
20270         VEC_PACK_FLOAT_EXPR.
20271         * tree-vect-stmts.c (supportable_widening_operation): Handle
20272         FIX_TRUNC_EXPR.
20273         (supportable_narrowing_operation): Handle FLOAT_EXPR.
20274         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
20275         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
20276         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
20277         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
20278         mode attributes.
20279         (vec_pack<floatprefix>_float_<mode>): New expander.
20280         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
20281         attributes.
20282         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
20283         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
20284         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
20285         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
20286         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
20287         Document.
20288         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
20289         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
20290         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
20291         VEC_PACK_FLOAT_EXPR): Document.
20293 2018-05-29  Richard Biener  <rguenther@suse.de>
20295         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
20296         member.
20297         (stmt_vec_info_vec): Make pointer.
20298         (init_stmt_vec_info_vec): Remove.
20299         (free_stmt_vec_info_vec): Likewise.
20300         (set_stmt_vec_info_vec): New function.
20301         (free_stmt_vec_infos): Likewise.
20302         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
20303         (set_vinfo_for_stmt): Likewise.
20304         (get_earlier_stmt): Likewise.
20305         (get_later_stmt): Likewise.
20306         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
20307         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
20308         (vec_info::~vec_info): Free stmt_vec_infos.
20309         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
20310         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
20311         (pass_slp_vectorize::execute): Likewise.
20312         * tree-vect-stmts.c (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         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
20317         the global stmt_vec_info_vec.
20318         * tree-parloops.c (gather_scalar_reductions): Use
20319         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
20320         vector.
20322 2018-05-29  Richard Biener  <rguenther@suse.de>
20324         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
20326 2018-05-29  Martin Liska  <mliska@suse.cz>
20327             David Malcolm  <dmalcolm@redhat.com>
20329         * vec.c (test_reverse): New.
20330         (vec_c_tests): Add new test.
20331         * vec.h (vl_ptr>::reverse): New function.
20333 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
20335         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
20337         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
20338         and later.
20340 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
20342         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
20344 2018-05-28  Richard Biener  <rguenther@suse.de>
20346         PR tree-optimization/85933
20347         * tree-vect-data-refs.c (vect_record_base_alignments): Only
20348         look at stmts marked as vectorizable.
20350 2018-05-28  Richard Biener  <rguenther@suse.de>
20352         PR tree-optimization/85934
20353         * tree-vect-generic.c (expand_vector_operations_1): Hoist
20354         vector boolean check before scalar optimization.
20356 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
20358         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
20359         for armv5te.
20361 2018-05-28  Mark Wielaard  <mark@klomp.org>
20363         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
20364         if it is an expression containing a minus sign.
20366 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
20368         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
20370 2018-05-27  Paul Koning  <ni1d@arrl.net>
20372         * config/pdp11/pdp11.md (truncsihi2): Remove.
20374 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
20375             Chung-Ju Wu  <jasonwucj@gmail.com>
20377         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
20378         implementation.
20379         (unaligned_store_dw): Ditto.
20380         * config/nds32/nds32-memory-manipulation.c
20381         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
20382         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
20383         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
20384         (emit_setmem_word_loop): Rename to ...
20385         (emit_setmem_doubleword_loop): ... this.
20386         (nds32_gen_dup_4_byte_to_word_value): New function.
20387         (nds32_gen_dup_8_byte_to_double_word_value): New function.
20388         (nds32_expand_setmem_loop): Refine implementation.
20389         (nds32_expand_setmem_loop_v3m): Ditto.
20390         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
20391         pattern.
20393 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20395         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
20397 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20399         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
20400         (nds32_init_machine_status): Initialize machine->attr_naked_p and
20401         machine->attr_no_prologue_p.
20402         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
20403         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
20404         (nds32_expand_epilogue): Consider attr_naked_p.
20405         (nds32_expand_epilogue_v3pop): Likewise.
20406         (nds32_can_use_return_insn): Likewise.
20407         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
20408         attr_no_prologue_p fields.
20409         * config/nds32/nds32.opt (mret-in-naked-func): New option.
20411 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
20413         PR target/85918
20414         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
20415         attributes.
20416         * config/i386/sse.md
20417         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
20418         Rename to ...
20419         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
20420         ... this.
20421         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
20422         Rename to ...
20423         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
20424         ... this.
20425         (*<floatsuffix>floatv2div2sf2): Rename to ...
20426         (*float<floatunssuffix>v2div2sf2): ... this.
20427         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
20428         (float<floatunssuffix>v2div2sf2_mask): ... this.
20429         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
20430         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
20431         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
20432         to ...
20433         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
20434         ... this.
20435         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
20436         Rename to ...
20437         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
20438         ... this.
20439         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
20440         Rename to ...
20441         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
20442         ... this.
20443         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
20444         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
20445         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
20446         gen_ufix_truncv8dfv8si2.
20447         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
20448         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
20449         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
20450         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
20451         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
20452         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
20453         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
20454         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
20456 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
20458         PR target/85900
20459         PR target/85345
20460         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
20462 2018-05-25  Jim Wilson  <jimw@sifive.com>
20464         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
20465         * config/riscv/riscv.c (struct machine_function): Add
20466         interrupt_handler_p and attribute_checked_p fields.
20467         (riscv_attribute_table): Add interrupt.
20468         (riscv_interrupt_type_p): New.
20469         (riscv_save_reg_p): Save extra regs for interrupt handler.
20470         (riscv_use_save_libcall): Return false  for interrupt handler.
20471         (riscv_first_stack_step): Add forward declaration.
20472         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
20473         for interrupt handler with large frame.  Use it for saved reg list.
20474         (riscv_expand_prologue): Move flag_stack_usage_info support to
20475         eliminate duplication.
20476         (riscv_expand_epilogue): Generate mret for interrupt handler.
20477         (riscv_epilogue_uses): New.
20478         (riscv_can_use_return_insn): Return false for interrupt handler.
20479         (riscv_function_ok_for_sibcall): Likewise.
20480         (riscv_set_current_function): Add interrupt handler support.
20481         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
20482         * config/riscv/riscv.md (UNSPECV_MRET): New.
20483         (GP_REGNUM): New.
20484         (riscv_frflags, riscv_fsflags): Use tab after opcode.
20485         (riscv_mret): New.
20486         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
20488 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
20490         PR tree-optimization/85712
20491         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
20492         this candidate has already been replaced in-situ by a copy.
20494 2018-05-25  Jason Merrill  <jason@redhat.com>
20496         PR c++/80485 - inline function non-zero address.
20497         * symtab.c (nonzero_address): Check DECL_COMDAT.
20499 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
20501         PR target/83628
20502         * config/alpha/alpha.md (ashlsi3): New insn pattern.
20503         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
20504         extension of SImode operation.  Use const123_operand predicate.
20505         (*saddsi_1): Remove.
20506         (*saddl_se_1): Ditto.
20507         (*ssubsi_1): Ditto.
20508         (*ssubl_se_1): Ditto.
20509         * config/alpha/predicates.md (const123_operand): New predicate.
20510         * config/alpha/constraints.md (P): Use IN_RANGE.
20512 2018-05-25  Richard Biener  <rguenther@suse.de>
20514         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
20515         defaulted to true.
20516         (ref_maybe_used_by_stmt_p): Likewise.
20517         (stmt_may_clobber_ref_p): Likewise.
20518         (stmt_may_clobber_ref_p_1): Likewise.
20519         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
20520         and pass it along.
20521         (ref_maybe_used_by_stmt_p): Likewise.
20522         (stmt_may_clobber_ref_p): Likewise.
20523         (stmt_may_clobber_ref_p_1): Likewise.
20524         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
20525         the alias oracle to disambiguate DRs with stmts DR analysis
20526         couldn't handle.
20527         (vect_analyze_data_refs): Do not give up on not analyzable
20528         DRs for BB vectorization.  Remove code truncating the dataref
20529         vector.
20531 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
20533         PR target/85832
20534         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
20535         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
20536         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
20538 2018-05-25  Richard Biener  <rguenther@suse.de>
20540         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
20541         function, combining stmt data ref gathering and fatal analysis
20542         parts.
20543         (vect_analyze_data_refs): Remove now redudnant code and simplify.
20544         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
20545         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
20546         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
20547         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
20549 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
20551         PR tree-optimization/85720
20552         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
20553         SCC if all partitions are builtins.
20554         (version_loop_by_alias_check): New parameter.  Generate cancelable
20555         runtime alias check if all partitions are builtins.
20556         (distribute_loop): Update call to above function.
20558 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
20560         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
20561         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
20562         (parm_default_def_partition_arg): Ditto.
20563         (set_parm_default_def_partition): Ditto.
20564         (get_parm_default_def_partitions): Ditto and make it static.
20565         (get_undefined_value_partitions): Ditto and make it static.
20566         (remove_ssa_form): Refactor call to init_var_map here.
20567         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
20568         computation for loop region.
20569         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
20570         (register_default_def): Delete.
20571         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
20572         (parm_default_def_partition_arg): Ditto.
20573         (set_parm_default_def_partition): Ditto.
20574         (get_parm_default_def_partitions): Ditto and make it static.
20575         (get_undefined_value_partitions): Ditto and make it static.
20576         (coalesce_with_default, coalesce_with_default): Update comment.
20577         (create_coalesce_list_for_region): New func factored out from
20578         create_outofssa_var_map.
20579         (populate_coalesce_list_for_outofssa): New func factored out from
20580         create_outofssa_var_map and coalesce_ssa_name.
20581         (create_outofssa_var_map): Delete.
20582         (coalesce_ssa_name): Refactor to support live range computation.
20583         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
20584         (get_parm_default_def_partitions): Delete.
20585         (get_undefined_value_partitions): Ditto.
20586         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
20587         computation for loop region.
20588         (new_tree_live_info, loe_visit_block): Ditto.
20589         (live_worklist, set_var_live_on_entry): Ditto.
20590         (calculate_live_on_exit, verify_live_on_entry): Ditto.
20591         * tree-ssa-live.h (struct _var_map): New fields.
20592         (init_var_map): Change decl.
20593         (region_contains_p): New.
20595 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
20597         * tree-ssa-live.h (live_merge_and_clear): Delete.
20599 2018-05-25  Richard Biener  <rguenther@suse.de>
20601         PR c++/85912
20602         * tree-dump.c (dequeue_and_dump): Remove access to removed
20603         operand 2 of a SWITCH_EXPR.
20605 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20607         * doc/sourcebuild.texi (vect_double_cond_arith): Include
20608         multiplication and division.
20609         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
20610         (cond_udiv@var{m}, cond_umod@var{m}): Document.
20611         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
20612         (cond_udiv_optab, cond_umod_optab): New optabs.
20613         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
20614         (IFN_COND_RDIV): New internal functions.
20615         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
20616         TRUNC_MOD_EXPR and RDIV_EXPR.
20617         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
20618         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
20619         New unspecs.
20620         (SVE_INT_BINARY): Include mult.
20621         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
20622         (optab, sve_int_op): Handle mult.
20623         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
20624         UNSPEC_COND_DIV.
20625         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
20626         for SVE_INT_BINARY_SD.
20628 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20630         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
20631         (optab, sve_int_op): Handle div and udiv.
20632         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
20633         for SVE_INT_BINARY_SD.
20634         (*<optab><mode>3): New insn for the same.
20636 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20638         * tree-vect-patterns.c: Include predict.h.
20639         (vect_recog_divmod_pattern): Restrict check for division support
20640         to when optimizing for size.
20642 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20644         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
20645         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
20646         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
20647         (gimple_match_op::set_op): Likewise.
20648         (gimple_resimplify4): Declare.
20649         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
20650         (expr::gen_transform): Likewise.
20651         (decision_tree::gen): Generate a simplification routine for 4 operands.
20652         * gimple-match-head.c (gimple_simplify): Add an overload for
20653         4 operands.  In the top-level function, handle up to 4 call
20654         arguments and call gimple_resimplify4.
20655         (gimple_resimplify4): New function.
20656         (build_call_internal): Pass a fourth operand.
20657         (maybe_push_to_seq): Likewise.
20658         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
20659         Fold VEC_COND_EXPRs of an operation and a default value into
20660         an IFN_COND_* function if possible.
20661         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
20662         New unspecs.
20663         (SVE_COND_FP_BINARY): Include them.
20664         (optab, sve_fp_op): Handle them.
20665         (SVE_INT_BINARY_REV): New code iterator.
20666         (SVE_COND_FP_BINARY_REV): New int iterator.
20667         (commutative): New int attribute.
20668         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
20669         Declare.
20670         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
20671         function.
20672         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
20673         (*cond_<optab><mode>): New patterns for reversed operands.
20675 2018-05-25  Richard Biener  <rguenther@suse.de>
20677         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
20678         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
20679         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
20680         (STMT_VINFO_GROUPED_ACCESS): Adjust.
20681         * tree-vect-data-refs.c (everywhere): Adjust users.
20682         * tree-vect-loop.c (everywhere): Likewise.
20683         * tree-vect-slp.c (everywhere): Likewise.
20684         * tree-vect-stmts.c (everywhere): Likewise.
20685         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
20687 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20689         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
20690         Rename to...
20691         (gcc_cv_as_section_exclude): ... this.
20692         Try Solaris as #exclude syntax.
20693         * configure: Regenerate.
20694         * config.in: Regenerate.
20695         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
20696         SECTION_EXCLUDE.
20697         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
20698         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
20700         * varasm.c (default_elf_asm_named_section): Don't check if
20701         HAVE_GAS_SECTION_EXCLUDE is defined.
20703 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20705         * doc/md.texi: Update the documentation of the cond_* optabs
20706         to mention the new final operand.  Fix GET_MODE_NUNITS call.
20707         Describe the scalar case too.
20708         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
20709         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
20710         instead of 2.
20711         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
20712         (get_conditional_internal_fn): Update comment.
20713         * tree-vect-loop.c (vectorizable_reduction): Pass the original
20714         accumulator value as a final argument to conditional functions.
20715         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
20716         a define_expand and add an "else" operand.  Assert for now that
20717         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
20718         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
20719         (*cond_<optab><mode>): New patterns.
20720         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
20721         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
20722         (UNSPEC_COND_EOR): Delete.
20723         (optab): Remove associated mappings.
20724         (SVE_INT_BINARY): New code iterator.
20725         (sve_int_op): Remove int attribute and add "minus" to the code
20726         attribute.
20727         (SVE_COND_INT_OP): Delete.
20728         (SVE_COND_FP_OP): Rename to...
20729         (SVE_COND_FP_BINARY): ...this.
20731 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
20733         * optabs.c (can_reuse_operands_p): New function.
20734         (maybe_legitimize_operands): Try to reuse the results for
20735         earlier operands.
20737 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
20739         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
20740         Add {q} suffix to insn mnemonic.
20742 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
20744         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
20745         (msp430_warn_func_return): New.
20747 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
20749         * fold-const.c (tree_nonzero_bits): New function.
20750         * fold-const.h (tree_nonzero_bits): Likewise.
20751         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
20752         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
20754 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
20756         PR target/85900
20757         PR target/85345
20758         * varasm.c (assemble_alias): Check ifunc_resolver only on
20759         FUNCTION_DECL.
20761 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
20763         PR target/85903
20764         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
20765         when memory input operand is handled.
20767 2018-05-24  Luis Machado  <luis.machado@linaro.org>
20769         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
20770         global.
20771         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
20773 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
20775         * match.pd: Delay FMA folds until after vectorization.
20777 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
20779         PR target/83009
20780         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
20781         address check not strict.
20783 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
20785         * gimple-match.h (gimple_match_op): New class.
20786         (mprts_hook): Replace parameters with a gimple_match_op *.
20787         (maybe_build_generic_op): Likewise.
20788         (gimple_simplified_result_is_gimple_val): Replace parameters with
20789         a const gimple_match_op *.
20790         (gimple_simplify): Replace code_helper * and tree * parameters with
20791         a gimple_match_op * parameter.
20792         (gimple_resimplify1): Replace code_helper *, tree and tree *
20793         parameters with a gimple_match_op * parameter.
20794         (gimple_resimplify2): Likewise.
20795         (gimple_resimplify3): Likewise.
20796         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
20797         parameters with a gimple_match_op * parameter.
20798         * gimple-match-head.c (gimple_simplify): Change prototypes of
20799         auto-generated functions to take a gimple_match_op * instead of
20800         separate code_helper * and tree * parameters.  Make the same
20801         change in the top-level overload and update calls to the
20802         gimple_resimplify routines.  Update calls to the auto-generated
20803         functions and to maybe_push_res_to_seq in the publicly-facing
20804         operation-specific gimple_simplify overloads.
20805         (gimple_match_op::MAX_NUM_OPS): Define.
20806         (gimple_resimplify1): Replace rcode and ops with a single res_op
20807         parameter.  Update call to gimple_simplify.
20808         (gimple_resimplify2): Likewise.
20809         (gimple_resimplify3): Likewise.
20810         (mprts_hook): Replace parameters with a gimple_match_op *.
20811         (maybe_build_generic_op): Likewise.
20812         (build_call_internal): Replace type, nargs and ops with
20813         a gimple_match_op *.
20814         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
20815         with a single gimple_match_op *.  Update calls to mprts_hook,
20816         build_call_internal and gimple_simplified_result_is_gimple_val.
20817         Factor out code that is common to the tree_code and combined_fn cases.
20818         * genmatch.c (expr::gen_transform): Replace tem_code and
20819         tem_ops with a gimple_match_op called tem_op.  Update calls
20820         to the gimple_resimplify functions and maybe_push_res_to_seq.
20821         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
20822         res_ops.  Update call to the gimple_resimplify functions.
20823         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
20824         (decision_tree::gen): Make the functions take a gimple_match_op *
20825         called res_op instead of separate res_code and res_ops parameters.
20826         Update call accordingly.
20827         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
20828         and ops with a single res_op parameter.  Update calls to
20829         maybe_build_generic_op and maybe_push_res_to_seq.
20830         (fold_stmt_1): Update calls to gimple_simplify and
20831         replace_stmt_with_simplification.
20832         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
20833         and gimple_simplified_result_is_gimple_val.
20834         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
20835         gimple_simplify.
20836         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
20837         with a gimple_match_op *.
20838         (vn_nary_build_or_lookup): Likewise.  Update call to
20839         vn_nary_build_or_lookup_1.
20840         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
20841         gimple_match_op *.  Update calls to the gimple_resimplify routines
20842         and to gimple_simplified_result_is_gimple_val.
20843         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
20844         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
20845         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
20846         (visit_nary_op): Likewise.
20847         (visit_reference_op_load): Likewise.
20849 2018-05-23  Luis Machado  <luis.machado@linaro.org>
20851         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
20852         modifier for printing the step amount.
20854 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20856         PR target/78849
20857         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
20858         types.
20860 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
20862         * doc/sourcebuild.texi (Endianness): New subsubsection.
20864 2018-05-23  Luis Machado  <luis.machado@linaro.org>
20866         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20867         <prefetch_dynamic_strides>: New const bool field.
20868         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20869         prefetch_dynamic_strides.
20870         (exynosm1_prefetch_tune): Likewise.
20871         (thunderxt88_prefetch_tune): Likewise.
20872         (thunderx_prefetch_tune): Likewise.
20873         (thunderx2t99_prefetch_tune): Likewise.
20874         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
20875         false.
20876         (aarch64_override_options_internal): Update to set
20877         PARAM_PREFETCH_DYNAMIC_STRIDES.
20878         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20879         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20880         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20881         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20882         prefetch-dynamic-strides setting.
20884 2018-05-23  Luis Machado  <luis.machado@linaro.org>
20886         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20887         <minimum_stride>: New const int field.
20888         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20889         minimum_stride field defaulting to -1.
20890         (exynosm1_prefetch_tune): Likewise.
20891         (thunderxt88_prefetch_tune): Likewise.
20892         (thunderx_prefetch_tune): Likewise.
20893         (thunderx2t99_prefetch_tune): Likewise.
20894         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
20895         <default_opt_level>: Set to 3.
20896         (aarch64_override_options_internal): Update to set
20897         PARAM_PREFETCH_MINIMUM_STRIDE.
20898         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20899         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20900         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20901         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20902         stride is constant and is below the minimum stride threshold.
20904 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20906         * config/arm/arm-cpus.in (mode26): Delete.
20907         (armv4): Delete mode26 reference.
20908         * config/arm/arm.c (arm_configure_build_target): Delete use of
20909         isa_bit_mode26.
20911 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
20913         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
20914         New insn pattern.
20915         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
20916         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
20917         for non-SSE modes.
20918         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
20919         (floatunsdidf2): Ditto.
20921 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
20923         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
20924         (fixuns_trunc<mode>si2_avx512f): Ditto.
20925         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
20926         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
20927         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
20929 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
20931         PR rtl-optimization/79985
20932         * df-scan.c (df_insn_refs_collect): Remove special case for
20933         global registers and asm statements.
20935 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
20937         * extend.texi (Global Register Variables): Rewrite the bullet list.
20938         Note that the register is available for allocation. Note that access
20939         via inline asm must use constraints. Add note about async-signal
20940         handlers. Remove paragraph about automagic register selection.
20942 2018-05-23  Richard Biener  <rguenther@suse.de>
20944         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
20945         of fixed offset from memset VN.
20947 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
20949         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
20950         first_interp field.
20951         (alloc_cand_and_find_basis): Initialize first_interp field.
20952         (slsr_process_mul): Modify first_interp field.
20953         (slsr_process_add): Likewise.
20954         (slsr_process_cast): Modify first_interp field for each new
20955         interpretation.
20956         (slsr_process_copy): Likewise.
20957         (dump_candidate): Dump first_interp field.
20958         (replace_mult_candidate): Process all interpretations, not just
20959         subsequent ones.
20960         (replace_rhs_if_not_dup): Likewise.
20961         (replace_one_candidate): Likewise.
20963 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
20965         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
20966         Add new boolean.
20967         (aarch64_needs_frame_chain): New function.
20968         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
20970 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
20972         PR target/84882
20973         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
20974         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
20975         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
20976         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
20977         as true for strict-align.
20978         (aarch64_can_inline_p): Perform checks even when callee has no
20979         attributes to check for strict alignment.
20980         * doc/extend.texi (AArch64 Function Attributes): Document
20981         no-strict-align.
20982         * doc/invoke.texi: (AArch64 Options): Likewise.
20984 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
20986         PR tree-optimization/85853
20987         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
20988         the handling of the root of the node to...
20989         (vect_slp_analyze_node_operations_1): ...this new function,
20990         and run the whole thing with the child nodes' def types
20991         set according to their SLP node's def type.
20993 2018-05-23  Richard Biener  <rguenther@suse.de>
20995         PR middle-end/85874
20996         * tree-data-ref.c (create_runtime_alias_checks): Defer
20997         and ignore overflow warnings.
20999 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
21001         PR tree-optimization/85822
21002         * tree-vrp.c (is_masked_range_test): Fix handling of negative
21003         constants.
21005 2018-05-23  Richard Biener  <rguenther@suse.de>
21007         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
21008         memset constants via native_interpret_expr.
21010 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
21012         PR target/85345
21013         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
21014         attribute.
21015         (cgraph_node::create_alias): Likewise.
21016         (cgraph_node::get_availability): Check ifunc_resolver instead
21017         of looking up ifunc attribute.
21018         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
21019         * varasm.c (do_assemble_alias): Likewise.
21020         (assemble_alias): Likewise.
21021         (default_binds_local_p_3): Likewise.
21022         * cgraph.h (cgraph_node): Add ifunc_resolver.
21023         (cgraph_node::only_called_directly_or_aliased_p): Return false
21024         for IFUNC resolver.
21025         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
21026         attribute.
21027         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
21028         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
21029         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
21030         instead of looking up ifunc attribute.
21032 2018-05-22  Luis Machado  <luis.machado@linaro.org>
21034         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
21036 2018-05-22  Martin Sebor  <msebor@redhat.com>
21038         PR middle-end/85359
21039         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
21040         only when expasion succeeds.
21041         (expand_builtin_strcmp): Same.
21042         (expand_builtin_strncmp): Same.
21044 2018-05-22  Martin Sebor  <msebor@redhat.com>
21046         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
21048 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
21049             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21051         * config/aarch64/aarch64-ldpstp.md: Replace uses of
21052         aarch64_mem_pair_operand with memory_operand and delete operand swapping
21053         code.
21054         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
21055         Add check for legitimate_address.
21056         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
21057         (aarch64_swap_ldrstr_operands): New.
21058         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
21059         Define prototype.
21061 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
21062             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21064         * config/aarch64/aarch64.md: New patterns to generate stp
21065         and ldp.
21066         (store_pair_sw, store_pair_dw): New patterns to generate stp for
21067         single words and double words.
21068         (load_pair_sw, load_pair_dw): Likewise.
21069         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
21070         Delete.
21071         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
21072         Delete.
21073         * config/aarch64/aarch64-ldpstp.md: Modify peephole
21074         for different mode ldpstp and add peephole for merged zero stores.
21075         Likewise for loads.
21076         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
21077         Add size check.
21078         (aarch64_gen_store_pair): Rename calls to match new patterns.
21079         (aarch64_gen_load_pair): Rename calls to match new patterns.
21080         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
21081         (load_pair<DREG:mode><DREG2:mode>): ... This.
21082         (store_pair<mode>): Rename to...
21083         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
21084         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
21085         New mode iterators.
21086         (V_INT_EQUIV): Handle SImode.
21087         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
21088         New predicate.
21090 2018-05-22  Martin Sebor  <msebor@redhat.com>
21092         PR c/85623
21093         * calls.c (maybe_warn_nonstring_arg): Use string length to set
21094         or ajust the presumed bound on an operation to avoid unnecessary
21095         warnings.
21097 2018-05-22  Martin Sebor  <msebor@redhat.com>
21099         PR tree-optimization/85826
21100         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
21101         assuming that a DECL necesarily has a constant size.
21103 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
21105         PR middle-end/85862
21106         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
21108 2018-05-22  Richard Biener  <rguenther@suse.de>
21110         PR tree-optimization/85834
21111         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
21112         non-constant and non-zero memset arguments.
21114 2018-05-22  Martin Liska  <mliska@suse.cz>
21116         PR ipa/85607
21117         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
21119 2018-05-22  Richard Biener  <rguenther@suse.de>
21121         PR tree-optimization/85863
21122         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
21123         comparisons when vectype is specified.
21124         (vectorizable_condition): Do not specify vectype for
21125         vect_is_simple_cond when SLP vectorizing.
21127 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
21129         PR target/85657
21130         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
21131         define __ibm128 as long double.
21132         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
21133         as a distinct type when IEEE 128-bit support is enabled.
21134         (init_float128_ieee): Fix up conversions between IFmode and IEEE
21135         128-bit types to use the correct functions.
21136         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
21137         convert between 128-bit floating point types that have different
21138         modes but the same representation, instead of using gen_lowpart to
21139         makean alias.
21140         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
21141         KFmode.
21142         (IFKF_reg): New attributes to give the register constraints for
21143         IFmode and KFmode.
21144         (extend<mode>tf2_internal): New insns to mark an explicit
21145         conversion between 128-bit floating point types that have a
21146         different mode but share the same representation.
21148 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
21150         PR tree-optimization/85814
21151         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
21152         a null return from get_strinfo when unsharing the next
21153         strinfo in the chain.
21155 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21157         PR gcc/84923
21158         * varasm.c (weak_finish): Clean up weak_decls.
21160 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21162         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
21163         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
21164         UNSPEC_UADALP values.
21165         * config/aarch64/iterators.md (ABAL): New int iterator.
21166         (ABDL2): Likewise.
21167         (ADALP): Likewise.
21168         (sur): Add mappings for the above.
21169         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
21170         New define_insn.
21171         (aarch64_<sur>abal<mode>_4): Likewise.
21172         (aarch64_<sur>adalp<mode>_3): Likewise.
21173         (<sur>sadv16qi): New define_expand.
21175 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
21177         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
21178         (*movdf_internal): Ditto.
21179         (*rcpsf2_sse): Ditto.
21180         (*rsqrtsf2_sse): Ditto.
21181         (*sqrt<mode>2_sse): Ditto.
21183 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
21185         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
21186         eor3q<mode>4.
21187         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
21188         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
21189         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
21190         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
21191         vbcaxq_s64): New.
21192         * config/aarch64/arm_neon.h: Likewise.
21193         * config/aarch64/iterators.md (VQ_I): New.
21195 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
21197         * config.gcc: Add arc/t-multilib-linux to tmake_file for
21198         arc*-*-linux*.
21199         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
21200         MULTILIB_DIRNAMES
21202 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
21204         * config/nds32/constraints.md (S): New constraint.
21205         * config/nds32/nds32.md (call_internal): Use constraint S.
21206         (call_value_internal): Likewise.
21207         (sibcall_internal): Likewise.
21208         (sibcall_value_internal): Likewise.
21210 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
21211             Chung-Ju Wu  <jasonwucj@gmail.com>
21213         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
21214         into consideration.
21216 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
21217             Chung-Ju Wu  <jasonwucj@gmail.com>
21219         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
21220         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
21221         (nds32_rtx_costs_impl): Simplify.
21222         (nds32_address_cost_impl): Simplify.
21223         (nds32_init_rtx_costs): New function.
21224         (nds32_rtx_costs_speed_prefer): Likewise.
21225         (nds32_rtx_costs_size_prefer): Likewise.
21226         (nds32_address_cost_speed_prefer): Likewise.
21227         (nds32_address_cost_speed_fwprop): Likewise.
21228         (nds32_address_cost_size_prefer): Likewise.
21229         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
21230         * config/nds32/nds32.c (nds32_option_override): Use
21231         nds32_init_rtx_costs function.
21233 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
21235         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
21236         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
21237         (TARGET_PIPELINE_N8): Likewise.
21238         (TARGET_PIPELINE_N10): Likewise.
21239         (TARGET_PIPELINE_N13): Likewise.
21240         (TARGET_PIPELINE_GRAYWOLF): Likewise.
21242 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
21244         * config/nds32/nds32-fpu.md: Update copyright year.
21246 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21248         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
21250 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21252         * config/nds32/nds32.c
21253         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
21254         * config/nds32/nds32.opt (minline-asm-r15): New option.
21256 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21258         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
21259         MASK_HW_ABS.
21260         * config/nds32/nds32.md (abssi2): New pattern.
21262 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
21264         * config/i386/i386.md (rex64namesuffix): New mode attribute.
21265         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
21266         Merge insn pattern from sse_cvtsi2ss<round_name> and
21267         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
21268         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
21269         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
21270         using SWI48 mode iterator.
21271         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
21272         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
21273         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
21274         pattern from sse_cvttss2si<round_saeonly_name>
21275         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
21276         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
21277         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
21278         using SWI48 mode iterator.
21279         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
21280         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
21281         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
21282         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
21283         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
21284         using SWI48 mode iterator.
21285         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
21286         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
21287         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
21288         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
21289         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
21290         SWI48 mode iterator.
21291         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
21292         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
21293         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
21294         pattern from sse_cvttsd2si<round_saeonly_name>
21295         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
21297 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21299         * config/nds32/nds32-md-auxiliary.c
21300         (nds32_valid_smw_lwm_base_p): Refine.
21301         (nds32_output_smw_single_word): Refine.
21302         (nds32_output_smw_double_word): New.
21303         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
21305 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21307         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
21308         (nds32_output_stack_pop): Refine.
21309         (nds32_expand_unaligned_load): Refine.
21310         (nds32_expand_unaligned_store): Refine.
21312 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
21313             Chung-Ju Wu  <jasonwucj@gmail.com>
21315         * config/nds32/constants.md: Add TP_REGNUM constant.
21316         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
21317         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
21318         UNSPEC_ADD32.
21319         * config/nds32/nds32-doubleword.md: Consider flag_pic.
21320         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
21321         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
21322         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
21323         and PIC code generation.
21324         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
21325         code generation.
21326         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
21327         optimization.
21328         * config/nds32/nds32.md: Support TLS and PIC.
21329         * config/nds32/nds32.c: Support TLS and PIC.
21330         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
21331         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
21332         predicate.
21334 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
21336         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
21337         mode with E_ prefix.
21339 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
21340             Chung-Ju Wu  <jasonwucj@gmail.com>
21342         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
21343         * config/nds32/nds32-md-auxiliary.c
21344         (symbolic_reference_mentioned_p): New.
21345         (nds32_legitimize_ict_address): New.
21346         (nds32_expand_ict_move): New.
21347         (nds32_indirect_call_referenced_p): New.
21348         (nds32_symbol_binds_local_p): Delete.
21349         (nds32_long_call_p): Modify.
21350         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
21351         * config/nds32/nds32-protos.h
21352         (symbolic_reference_mentioned_p): Declare.
21353         (nds32_legitimize_ict_address): Declare.
21354         (nds32_expand_ict_move): Declare.
21355         (nds32_indirect_call_referenced_p): Declare.
21356         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
21357         (nds32_relax_group): Use nds32_ict_const_p as condition.
21358         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
21359         (nds32_asm_file_start): Output ict_model directive in asm code.
21360         (nds32_legitimate_address_p): Consider indirect call.
21361         (nds32_print_operand): Consider indirect call.
21362         (nds32_print_operand_address): Consider indirect call.
21363         (nds32_insert_attributes): Handle "indirect_call" attribute.
21364         (TARGET_LEGITIMATE_ADDRESS_P): Define.
21365         (TARGET_LEGITIMATE_CONSTANT_P): Define.
21366         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
21367         (TARGET_DELEGITIMIZE_ADDRESS): Define.
21368         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
21369         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
21370         (TARGET_ICT_MODEL_SMALL): Define.
21371         (TARGET_ICT_MODEL_LARGE): Define.
21372         * config/nds32/nds32.md (movsi): Consider ict model.
21373         (call, call_value): Consider ict model.
21374         (sibcall, sibcall_value): Consider ict model.
21375         * config/nds32/nds32.opt (mict-model): New option.
21376         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
21377         model.
21379 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
21380             Monk Chiang  <sh.chiang04@gmail.com>
21381             Jim Wilson <jimw@sifive.com>
21383         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
21384         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
21385         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
21386         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
21387         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
21388         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
21389         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
21390         compute save_libcall_adjustment properly.
21391         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
21392         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
21393         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
21394         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
21395         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
21396         (ABI_SPEC): Handle mabi=ilp32e.
21397         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
21398         (RVE): Add RVE mask.
21399         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
21400         <-march>: Add rv32e as an example.
21402 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
21404         PR c++/82899
21405         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
21406         (intra_create_variable_infos): Handle C++ constructors.
21408 2018-05-18  Martin Liska  <mliska@suse.cz>
21410         * passes.def: Remove a redundant pass.
21412 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
21414         PR bootstrap/85838
21415         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
21417 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21419         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
21420         (ARMv4): Update.
21421         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
21422         (ARMv6m): Update.
21423         (armv2, armv2a, armv3, armv3m): Delete architectures.
21424         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
21425         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
21426         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
21427         Delete cpus.
21428         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
21429         (*mulsidi3adddi): Likewise.
21430         (mulsidi3): Likewise.
21431         (*mulsidi3_nov6): Likewise.
21432         (umulsidi3): Likewise.
21433         (umulsidi3_nov6): Likewise.
21434         (umaddsidi4): Likewise.
21435         (*umulsidi3adddi): Likewise.
21436         (smulsi3_highpart): Likewise.
21437         (*smulsi3_highpart_nov6): Likewise.
21438         (umulsi3_highpart): Likewise.
21439         (*umulsi3_highpart_nov6): Likewise.
21440         * config/arm/arm.h (arm_arch3m): Delete.
21441         * config/arm/arm.c (arm_arch3m): Delete.
21442         (arm_option_override_internal): Update armv3-related comment.
21443         (arm_configure_build_target): Delete use of isa_bit_mode32.
21444         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
21445         (arm_rtx_costs_internal): Delete check of arm_arch3m.
21446         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
21447         (mulsa3): Likewise.
21448         (mulusa3): Likewise.
21449         * config/arm/arm-protos.h (arm_arch3m): Delete.
21450         * config/arm/arm-tables.opt: Regenerate.
21451         * config/arm/arm-tune.md: Likewise.
21452         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
21453         deleted architectures.
21455 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21457         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
21458         (armv5t, armv5te): New features.
21459         (ARMv5, ARMv5e): Delete fgroups.
21460         (ARMv5t, ARMv5te): Adjust for above changes.
21461         (ARMv6m): Likewise.
21462         (armv5, armv5e): Delete arches.
21463         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
21464         arm_arch5.
21465         (*call_reg_arm): Likewise.
21466         (*call_value_reg_armv5): Likewise.
21467         (*call_value_reg_arm): Likewise.
21468         (*call_symbol): Likewise.
21469         (*call_value_symbol): Likewise.
21470         (*sibcall_insn): Likewise.
21471         (*sibcall_value_insn): Likewise.
21472         (clzsi2): Likewise.
21473         (prefetch): Likewise.
21474         (define_split and define_peephole2 dependent on arm_arch5):
21475         Likewise.
21476         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
21477         arm_arch5e.
21478         (TARGET_ARM_QBIT): Likewise.
21479         (TARGET_DSP_MULTIPLY): Likewise.
21480         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
21481         (arm_arch5, arm_arch5e): Delete.
21482         (arm_arch5t, arm_arch5te): Declare.
21483         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
21484         (arm_arch5t): Declare.
21485         (arm_option_reconfigure_globals): Update for the above.
21486         (arm_options_perform_arch_sanity_checks): Update comment, replace
21487         use of arm_arch5 with arm_arch5t.
21488         (use_return_insn): Likewise.
21489         (arm_emit_call_insn): Likewise.
21490         (output_return_instruction): Likewise.
21491         (arm_final_prescan_insn): Likewise.
21492         (arm_coproc_builtin_available): Likewise.
21493         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
21494         arm_arch5e with arm_arch5t and arm_arch5te.
21495         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
21496         (arm_arch5t, arm_arch5te): Declare.
21497         * config/arm/arm-tables.opt: Regenerate.
21498         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
21499         * config/arm/t-multilib: Likewise.
21500         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
21501         instead of arm_arch5.
21502         (*call_reg_thumb1): Likewise.
21503         (*call_value_reg_thumb1_v5): Likewise.
21504         (*call_value_reg_thumb1): Likewise.
21505         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
21506         unreachable path.
21507         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
21509 2018-05-18  Martin Liska  <mliska@suse.cz>
21511         PR gcov-profile/84846
21512         * doc/gcov.texi: Document -t option of gcov tool.
21514 2018-05-18  Martin Liska  <mliska@suse.cz>
21516         PR gcov-profile/84846
21517         * gcov.c (print_usage): Add new -t option.
21518         (process_args): Handle the option.
21519         (generate_results): Use stdout as output when requested by
21520         the option.
21522 2018-05-18  Martin Liska  <mliska@suse.cz>
21524         PR gcov-profile/84846
21525         * coverage.c (coverage_init): Write PWD to .gcno file.
21526         * doc/gcov.texi: Document how working directory is printed.
21527         * gcov-dump.c (dump_gcov_file): Print PWD.
21528         * gcov.c (output_intermediate_file): Likewise.
21529         (read_graph_file): Read PWD string.
21530         (output_lines): Print PWD.
21532 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
21534         PR middle-end/85817
21535         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
21536         for retval and return false if all args to phi are zero.
21538 2018-05-18  Richard Biener  <rguenther@suse.de>
21540         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
21541         method.
21542         (evrp_dom_walker::before_dom_children): Call it.
21544 2018-05-18  Richard Biener  <rguenther@suse.de>
21546         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
21547         results when processing array refs with variable index.
21549 2018-05-18  Toon Moene  <toon@moene.org>
21551         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
21552         directly after that of -floop-interchange. Indicate that both
21553         options are enabled by default when specifying -O3.
21555 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21557         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
21558         iterator.  Delete separate integer-mode vec_set<mode> expander.
21559         (aarch64_simd_vec_setv2di): Delete.
21560         (vec_setv2di): Delete.
21561         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
21562         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
21563         the "w, r" alternative.
21565 2018-05-18  Martin Liska  <mliska@suse.cz>
21567         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
21568         * tree-pass.h (make_pass_lower_switch_O0): New function.
21569         * tree-switch-conversion.c (node_has_low_bound): Remove.
21570         (node_has_high_bound): Likewise.
21571         (node_is_bounded): Likewise.
21572         (class pass_lower_switch): Make it a template type and create
21573         two instances.
21574         (pass_lower_switch::execute): Add template argument.
21575         (make_pass_lower_switch): New function.
21576         (make_pass_lower_switch_O0): New function.
21577         (do_jump_if_equal): Remove.
21578         (emit_case_nodes): Simplify to just handle all 3 cases and leave
21579         all the hard work to tree optimization passes.
21581 2018-05-18  Martin Liska  <mliska@suse.cz>
21583         * dbgcnt.c (limit_low): Renamed from limit.
21584         (limit_high): New variable.
21585         (dbg_cnt_is_enabled): Check for upper limit.
21586         (dbg_cnt): Adjust dumping.
21587         (dbg_cnt_set_limit_by_index): Add new argument for high
21588         value.
21589         (dbg_cnt_set_limit_by_name): Likewise.
21590         (dbg_cnt_process_single_pair): Parse new format.
21591         (dbg_cnt_process_opt): Use strtok.
21592         (dbg_cnt_list_all_counters): Remove 'value' and add
21593         'limit_high'.
21594         * doc/invoke.texi: Document changes.
21596 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
21598         * doc/sourcebuild.texi (scalar_all_fma): Document.
21599         * tree.def (FMA_EXPR): Delete.
21600         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
21601         * internal-fn.c (ternary_direct): New macro.
21602         (expand_ternary_optab_fn): Likewise.
21603         (direct_ternary_optab_supported_p): Likewise.
21604         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
21605         * builtins.c (fold_builtin_fma): Delete.
21606         (fold_builtin_3): Don't call it.
21607         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
21608         * expr.c (expand_expr_real_2): Likewise.
21609         * fold-const.c (operand_equal_p): Likewise.
21610         (fold_ternary_loc): Likewise.
21611         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
21612         * gimple.c (DEFTREECODE): Likewise.
21613         * gimplify.c (gimplify_expr): Likewise.
21614         * optabs-tree.c (optab_for_tree_code): Likewise.
21615         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
21616         * tree-eh.c (operation_could_trap_p): Likewise.
21617         (stmt_could_throw_1_p): Likewise.
21618         * tree-inline.c (estimate_operator_cost): Likewise.
21619         * tree-pretty-print.c (dump_generic_node): Likewise.
21620         (op_code_prio): Likewise.
21621         * tree-ssa-loop-im.c (stmt_cost): Likewise.
21622         * tree-ssa-operands.c (get_expr_operands): Likewise.
21623         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
21624         * fold-const-call.h (fold_fma): Delete.
21625         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
21626         CFN_FNMA and CFN_FNMS.
21627         (fold_fma): Delete.
21628         * genmatch.c (combined_fn): New enum.
21629         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
21630         (commutative_op): New function.
21631         (commutate): Use it.  Handle more than 2 operands.
21632         (dt_operand::gen_gimple_expr): Use commutative_op.
21633         (parser::parse_expr): Allow :c to be used with non-binary
21634         operators if the commutative operand is known.
21635         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
21636         CFN_FMS, CFN_FNMA and CFN_FNMS.
21637         (backprop::process_assign_use): Remove FMA_EXPR handling.
21638         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
21639         (gen_hsa_fma): New function.
21640         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
21641         IFN_FNMA and IFN_FNMS.
21642         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
21643         * gimple-fold.h (follow_all_ssa_edges): Declare.
21644         * gimple-fold.c (follow_all_ssa_edges): New function.
21645         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
21646         gimple_build interface and use follow_all_ssa_edges to fold the result.
21647         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
21648         instead of checking for optabs directly.
21649         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
21650         rather than FMA_EXPRs.
21651         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
21652         call to IFN_FMA instead of an FMA_EXPR.
21654 2018-05-17  Jim Wilson  <jimw@sifive.com>
21656         * expr.c (do_tablejump): When converting index to Pmode, if we have a
21657         sign extended promoted subreg, and the range does not have the sign bit
21658         set, then do a sign extend.
21660         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
21661         test, check for sign extended subreg and/or constant operands, and
21662         do a sign extend in that case.
21664 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
21666         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
21667         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
21668         Add untyped.
21669         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
21670         Change logics_shift_reg to logics_shift_imm.
21671         (thunderx2t99_fp_loadpair_basic): Delete.
21672         (thunderx2t99_fp_storepair_basic): Delete.
21673         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
21674         (thunderx2t99_asimd_polynomial): Delete.
21675         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
21676         and neon_fp_mul_d_scalar_q.
21677         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
21678         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
21679         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
21680         (thunderx2t99_asimd_lut): Add missing tbl types.
21681         (thunderx2t99_asimd_ext): Delete.
21682         (thunderx2t99_asimd_load1_1_mult): Delete.
21683         (thunderx2t99_asimd_load1_2_mult): Delete.
21684         (thunderx2t99_asimd_load1_ldp): New.
21685         (thunderx2t99_asimd_load1): New.
21686         (thunderx2t99_asimd_load2): Add missing *load2* types.
21687         (thunderx2t99_asimd_load3): New.
21688         (thunderx2t99_asimd_load4): New.
21689         (thunderx2t99_asimd_store1_1_mult): Delete.
21690         (thunderx2t99_asimd_store1_2_mult): Delete.
21691         (thunderx2t99_asimd_store2_mult): Delete.
21692         (thunderx2t99_asimd_store2_onelane): Delete.
21693         (thunderx2t99_asimd_store_stp): New.
21694         (thunderx2t99_asimd_store1): New.
21695         (thunderx2t99_asimd_store2): New.
21696         (thunderx2t99_asimd_store3): New.
21697         (thunderx2t99_asimd_store4): New.
21699 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
21701         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
21702         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
21704 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
21705             Segher Boessenkool  <segher@kernel.crashing.org>
21707         PR target/85698
21708         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
21709         operand.
21711 2018-05-17  Richard Biener  <rguenther@suse.de>
21713         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
21714         for pruning loop and prune defs feeding only already visited PHIs.
21716 2018-05-17  Richard Biener  <rguenther@suse.de>
21718         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
21720 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
21721             Richard Biener  <rguenther@suse.de>
21723         PR tree-optimization/85793
21724         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
21725         for VMAT_ELEMENTWISE.
21727 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
21729         * internal-fn.h (lookup_internal_fn): Declare
21730         * internal-fn.c (lookup_internal_fn): New function.
21731         * gimple.c (gimple_build_call_from_tree): Handle calls to
21732         internal functions.
21733         * gimple-pretty-print.c (dump_gimple_call): Print "." before
21734         internal function names.
21735         * tree-pretty-print.c (dump_generic_node): Likewise.
21736         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
21738 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
21740         * gimple-fold.h (gimple_build): Make the function forms take
21741         combined_fn rather than built_in_function.
21742         (gimple_simplify): Likewise.
21743         * gimple-match-head.c (gimple_simplify): Likewise.
21744         * gimple-fold.c (gimple_build): Likewise.
21745         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
21746         rather than gimple_build_call_internal.
21747         (get_initial_defs_for_reduction): Likewise.
21748         (vect_create_epilog_for_reduction): Likewise.
21749         (vectorizable_live_operation): Likewise.
21751 2018-05-17  Martin Liska  <mliska@suse.cz>
21753         * gimple-ssa-sprintf.c (format_directive): Do not use
21754         space in between 'G_' and '('.
21756 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
21758         PR target/85323
21759         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
21760         even if the mask is not all ones.
21762         PR target/85323
21763         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
21764         vector.
21765         (ix86_gimple_fold_builtin): Likewise.
21767         PR target/85323
21768         * config/i386/i386.c: Include tree-vector-builder.h.
21769         (ix86_vector_shift_count): New function.
21770         (ix86_fold_builtin): Fold shift builtins by scalar count.
21771         (ix86_gimple_fold_builtin): Likewise.
21773         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
21774         _mm512_setzero): New intrinsics.
21776 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
21777             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21779         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
21780         code generation for cases where splatting a value is not useful.
21781         * simplify-rtx.c (simplify_ternary_operation): Simplify
21782         vec_merge across a vec_duplicate and a paradoxical subreg forming
21783         a vector mode to a vec_concat.
21785 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
21787         * config.gcc: Support "goldmont-plus".
21788         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
21789         "goldmont-plus".
21790         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21791         PROCESSOR_GOLDMONT_PLUS.
21792         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
21793         (processor_target_table): Add "goldmont-plus".
21794         (PTA_GOLDMONT_PLUS): Define.
21795         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
21796         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
21797         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
21798         (fold_builtin_cpu): Add "goldmont-plus".
21799         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
21800         (ix86_option_override_internal): Add "goldmont-plus".
21801         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
21802         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
21803         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
21804         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
21806 2018-05-17  Richard Biener  <rguenther@suse.de>
21808         PR tree-optimization/85757
21809         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
21810         remove defs that only feed that PHI from further processing.
21812 2018-05-16  Jim Wilson  <jimw@sifive.com>
21814         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
21815         asterisk to name.
21816         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
21818 2018-05-16  Mark Wielaard  <mark@klomp.org>
21820         * dwarf2out.c (count_index_strings): New function.
21821         (output_indirect_strings): Call count_index_strings and generate
21822         header for dwarf_version >= 5.
21824 2018-05-16  Mark Wielaard  <mark@klomp.org>
21826         * dwarf2out.c (dwarf_FORM): New function.
21827         (set_indirect_string): Use dwarf_FORM.
21828         (reset_indirect_string): Likewise.
21829         (size_of_die): Likewise.
21830         (value_format): Likewise.
21831         (output_die): Likewise.
21832         (add_skeleton_AT_string): Likewise.
21833         (output_macinfo_op): Likewise.
21834         (index_string): Likewise.
21835         (output_index_string_offset): Likewise.
21836         (output_index_string): Likewise.
21837         (count_index_strings): Likewise.
21839 2018-05-16  Carl Love  <cel@us.ibm.com>
21841         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
21842         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
21844 2018-05-16  Martin Jambor  <mjambor@suse.cz>
21846         * ipa-prop.c (ipa_free_all_edge_args): Remove.
21847         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
21849 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
21851         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
21852         (fnma<mode>4): Likewise.
21853         (fms<mode>4): Likewise.
21854         (fnms<mode>4): Likewise.
21855         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
21856         (aarch64_fnma<mode>4): Likewise.
21857         (aarch64_fms<mode>4): Likewise.
21858         (aarch64_fnms<mode>4): Likewise.
21859         (aarch64_fnmadd<mode>4): Likewise.
21861 2018-05-16  Jason Merrill  <jason@redhat.com>
21863         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
21865 2018-05-16  Richard Biener  <rguenther@suse.de>
21867         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
21868         (dump_stmt_cost): Declare.
21869         (add_stmt_cost): Dump cost we add.
21870         (add_stmt_costs): New function.
21871         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
21872         No longer exported.
21873         (vect_analyze_stmt): Adjust prototype.
21874         (vectorizable_condition): Likewise.
21875         (vectorizable_live_operation): Likewise.
21876         (vectorizable_reduction): Likewise.
21877         (vectorizable_induction): Likewise.
21878         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
21879         cost vector to pass to vectorizable_ and record afterwards.
21880         (vect_model_reduction_cost): Take cost vector argument and adjust.
21881         (vect_model_induction_cost): Likewise.
21882         (vectorizable_reduction): Likewise.
21883         (vectorizable_induction): Likewise.
21884         (vectorizable_live_operation): Likewise.
21885         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
21886         SLP_TREE_NUMBER_OF_VEC_STMTS.
21887         (vect_analyze_slp_cost_1): Remove.
21888         (vect_analyze_slp_cost): Likewise.
21889         (vect_slp_analyze_node_operations): Take visited args and
21890         a target cost vector.  Avoid processing already visited stmt sets.
21891         (vect_slp_analyze_operations): Use a local cost vector to gather
21892         costs and register those of non-discarded instances.
21893         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
21894         (vect_schedule_slp_instance): Remove copying of
21895         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
21896         zero.
21897         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
21898         adding cost.  Record cost entry location.
21899         (vect_prologue_cost_for_slp_op): Function to compute cost of
21900         a constant or invariant generated for SLP vect in the prologue,
21901         split out from vect_analyze_slp_cost_1.
21902         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
21903         (vect_model_promotion_demotion_cost): Likewise.
21904         (vect_model_store_cost): Likewise, make static.
21905         (vect_model_load_cost): Likewise.
21906         (vectorizable_bswap): Add cost vector arg and adjust.
21907         (vectorizable_call): Likewise.
21908         (vectorizable_simd_clone_call): Likewise.
21909         (vectorizable_conversion): Likewise.
21910         (vectorizable_assignment): Likewise.
21911         (vectorizable_shift): Likewise.
21912         (vectorizable_operation): Likewise.
21913         (vectorizable_store): Likewise.
21914         (vectorizable_load): Likewise.
21915         (vectorizable_condition): Likewise.
21916         (vectorizable_comparison): Likewise.
21917         (can_vectorize_live_stmts): Likewise.
21918         (vect_analyze_stmt): Likewise.
21919         (vect_transform_stmt): Adjust calls to vectorizable_*.
21920         * tree-vectorizer.c: Include gimple-pretty-print.h.
21921         (dump_stmt_cost): New function.
21923 2018-05-16  Richard Biener  <rguenther@suse.de>
21925         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
21926         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
21927         * tree-ssa-dse.c: Include tree-ssa-loop.h.
21928         (check_name): New callback.
21929         (dse_classify_store): Track cycles via a visited bitmap of PHI
21930         defs and simplify handling of in-loop and across loop dead stores
21931         and properly fail for loop-variant refs.  Handle byte-tracking with
21932         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
21933         limiting the walk.
21935 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
21937         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
21938         (vect_get_mask_type_for_stmt): Likewise.
21939         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
21940         split out from...
21941         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
21942         to determine the statement's vector type and the vector type that
21943         should be used for calculating nunits.  Deal with cases in which
21944         the type has to be deferred.
21945         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
21946         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
21947         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
21948         (vect_determine_vf_for_stmt): New functions, split out from...
21949         (vect_determine_vectorization_factor): ...here.
21950         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
21951         (vect_get_mask_type_for_stmt): New functions, split out from
21952         vect_determine_vectorization_factor.
21954 2018-05-16  Richard Biener  <rguenther@suse.de>
21956         * tree-cfg.c (verify_gimple_assign_ternary): Properly
21957         verify the [VEC_]COND_EXPR embedded comparison.
21959 2018-05-15  Martin Sebor  <msebor@redhat.com>
21961         PR tree-optimization/85753
21962         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
21963         RECORD_TYPE in addition to ARRAY_TYPE.
21965 2018-05-15  Martin Sebor  <msebor@redhat.com>
21967         PR middle-end/85643
21968         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
21970 2018-05-15  Richard Biener  <rguenther@suse.de>
21972         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
21973         add by_clobber_p one.  Change algorithm to collect all defs
21974         representing uses we need to walk and try reducing them to
21975         a single one before failing.
21976         (dse_dom_walker::dse_optimize_stmt): Adjust.
21978 2018-05-13  Mark Wielaard  <mark@klomp.org>
21980         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
21981         (size_of_loc_descr): Likewise.
21982         (output_loc_operands): Likewise.
21983         (output_loc_operands_raw): Likewise.
21984         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
21985         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
21986         (hash_loc_operands): Likewise.
21987         (compare_loc_operands): Likewise.
21989 2018-05-14  Mark Wielaard  <mark@klomp.org>
21991         * dwarf2out.c (count_index_addrs): New function.
21992         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
21994 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
21996         PR tree-optimization/83648
21997         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
21998         return value as malloc candidate.
22000 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22002         PR ipa/85734
22003         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
22004         param as true in call to suggest_attribute.
22006 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
22008         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
22009         -mreadonly-in-sdata.
22011 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22013         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
22014         New pattern.
22015         (aarch64_crypto_aesd_fused): Likewise.
22017 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
22019         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
22020         (movsi_aarch64): Likewise.
22021         (load_pairsi): Likewise.
22022         (load_pairdi): Likewise.
22023         (store_pairsi): Likewise.
22024         (store_pairdi): Likewise.
22025         (load_pairsf): Likewise.
22026         (load_pairdf): Likewise.
22027         (store_pairsf): Likewise.
22028         (store_pairdf): Likewise.
22029         (zero_extend): Likewise.
22030         (trunc): Swap alternatives.
22031         (fcvt_target): Add '?' to prefer w over r.
22033 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
22035         PR target/85756
22036         * config/i386/i386.md: Disallow non-commutative arithmetics in
22037         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
22038         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
22039         in the peephole2 before it.
22041 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
22043         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
22044         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
22045         (ix86_handle_option): Handle -mcldemote.
22046         * config.gcc: New header.
22047         * config/i386/cldemoteintrin.h: New file.
22048         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
22049         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
22050         -mcldemote.
22051         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22052         OPTION_MASK_ISA_CLDEMOTE.
22053         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
22054         (ix86_valid_target_attribute_inner_p): Ditto.
22055         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
22056         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
22057         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
22058         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
22059         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
22060         (cldemote): New.
22061         * config/i386/i386.opt: Add -mcldemote.
22062         * config/i386/x86intrin.h: New header.
22063         * doc/invoke.texi: Add -mcldemote.
22065 2018-05-14  Richard Biener  <rguenther@suse.de>
22067         * doc/match-and-simplify.texi: Adjust :s documentation.
22069 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
22071         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
22072         intended memcpy size.
22073         (REORDER_45): Likewise.
22075 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
22077         * sort.cc: New file.
22078         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
22079         * vec.c (qsort_chk): Use gcc_qsort.
22080         * Makefile.in (OBJS-libcommon): Add sort.o.
22081         (build/sort.o): New target.  Use it...
22082         (BUILD_RTL): ... here, and...
22083         (build/gencfn-macros): ... here, and...
22084         (build/genmatch): ... here.
22086 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22087             Chung-Ju Wu  <jasonwucj@gmail.com>
22089         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
22090         * config/nds32/nds32-graywolf.md: New file.
22091         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
22092         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
22093         pipeline.
22094         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
22095         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
22096         * config/nds32/nds32.md (pipeline_model): Add graywolf.
22097         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
22098         * config/nds32/pipelines.md: Include n15 settings.
22100 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22101             Chung-Ju Wu  <jasonwucj@gmail.com>
22103         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
22104         * config/nds32/nds32-n13.md: New file.
22105         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
22106         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
22107         pipeline.
22108         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
22109         * config/nds32/nds32.md (pipeline_model): Add n13.
22110         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
22111         * config/nds32/pipelines.md: Include n13 settings.
22113 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
22114             Chung-Ju Wu  <jasonwucj@gmail.com>
22116         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
22117         * config/nds32/nds32-n10.md: New file.
22118         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
22119         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
22120         pipeline.
22121         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
22122         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
22123         * config/nds32/nds32.md (pipeline_model): Add n10.
22124         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
22125         * config/nds32/pipelines.md: Include n10 settings.
22127 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
22128             Kito Cheng  <kito.cheng@gmail.com>
22129             Chung-Ju Wu  <jasonwucj@gmail.com>
22131         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
22132         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22133         Add enum values for DSP extension instructions.
22134         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
22135         New constraints.
22136         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
22137         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
22138         New code iterators.
22139         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
22140         * config/nds32/nds32-dspext.md: New file for DSP implementation.
22141         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
22142         * config/nds32/nds32-intrinsic.md: Likewise.
22143         * config/nds32/nds32_intrinsic.h: Likewise.
22144         * config/nds32/nds32-md-auxiliary.c: Likewise.
22145         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
22146         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
22147         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
22148         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
22149         * config/nds32/nds32-protos.h: New declarations for DSP extension.
22150         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
22151         TYPE_DMAC in switch statement.
22152         * config/nds32/nds32.c: New checking and implementation for DSP
22153         extension instructions.
22154         * config/nds32/nds32.h: Likewise.
22155         * config/nds32/nds32.md: Likewise.
22156         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
22157         * config/nds32/predicates.md: Implement new predicates for DSP
22158         extension.
22160 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
22162         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
22163         Reformat alternatives and attributes so it is easier to identify
22164         which constraints/attributes go with which instruction.
22165         (mov<mode>_hardfloat32, FMOVE64): Likewise.
22166         (mov<mode>_softfloat32, FMOVE64): Likewise.
22167         (mov<mode>_hardfloat64, FMOVE64): Likewise.
22168         (mov<mode>_softfloat64, FMOVE64): Likewise.
22170 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22172         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22173         subsection.
22174         (Basic PowerPC Built-in Functions): The new name of the
22175         subsection previously known as "PowerPC Built-in Functions".
22176         (Basic PowerPC Built-in Functions Available on all Configurations):
22177         New subsubsection.
22178         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
22179         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
22180         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
22181         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
22183 2018-05-11  Martin Jambor  <mjambor@suse.cz>
22185         PR ipa/85655
22186         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
22187         single const.
22189 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
22191         PR target/85733
22192         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
22194 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
22196         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
22197         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
22198         (ix86_handle_option): Handle -mwaitpkg.
22199         * config.gcc: New header.
22200         * config/i386/cpuid.h (bit_WAITPKG): New bit.
22201         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
22202         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
22203         function type.
22204         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22205         OPTION_MASK_ISA_WAITPKG.
22206         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
22207         (ix86_option_override_internal): Add PTA_WAITPKG.
22208         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
22209         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
22210         IX86_BUILTIN_TPAUSE.
22211         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
22212         __builtin_ia32_umwait and __builtin_ia32_tpause.
22213         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
22214         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
22215         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
22216         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
22217         UNSPECV_TPAUSE): New.
22218         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
22219         * config/i386/i386.opt: Add -mwaitpkg.
22220         * config/i386/waitpkgintrin.h: New file.
22221         * config/i386/x86intrin.h: New header.
22222         * doc/invoke.texi: Add -mwaitpkg.
22224 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
22226         PR target/85606
22227         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
22228         equivalent.
22229         (cortex-m0): Use armv6s-m isa.
22230         (cortex-m0plus): Likewise.
22231         (cortex-m1): Likewise.
22232         (cortex-m0.small-multiply): Likewise.
22233         (cortex-m0plus.small-multiply): Likewise.
22234         (cortex-m1.small-multiply): Likewise.
22236 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
22237             Jakub Jelinek  <jakub@redhat.com>
22239         PR tree-optimization/85692
22240         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
22241         source permute as well.
22243 2018-05-11  Martin Liska  <mliska@suse.cz>
22245         PR sanitizer/85556
22246         * doc/extend.texi: Document LLVM style format for no_sanitize
22247         attribute.
22249 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
22251         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
22252         mode_supports_vsx_dform_quad to mode_supports_dq_form.
22253         (mode_supports_vsx_dform_quad): Likewise.
22254         (mode_supports_vmx_dform): Move these functions to be next to the
22255         other mode_supports functions.
22256         (mode_supports_dq_form): Likewise.
22257         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
22258         mode_supports_dq_form.
22259         (reg_offset_addressing_ok_p): Likewise.
22260         (offsettable_ok_by_alignment): Likewise.
22261         (rs6000_legitimate_offset_address_p): Likewise.
22262         (legitimate_lo_sum_address_p): Likewise.
22263         (rs6000_legitimize_address): Likewise.
22264         (rs6000_legitimize_reload_address): Likewise.
22265         (rs6000_secondary_reload_inner): Likewise.
22266         (rs6000_preferred_reload_class): Likewise.
22267         (rs6000_output_move_128bit): Likewise.
22269 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
22271         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
22272         Generate SImode target register for null target.
22273         <case IX86_BUILTIN_XGETBV>: Ditto.
22274         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
22275         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
22277 2018-05-10  Carl Love  <cel@us.ibm.com>
22279         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
22280         dcbtt and dcbtstt if operands[2] is 0.
22282 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
22284         PR target/85693
22285         * config/i386/sse.md (usadv64qi): New expander.
22287 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
22289         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
22290         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
22291         -maltivec=be support.
22292         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
22293         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
22294         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
22295         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
22296         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
22297         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
22298         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
22299         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
22300         altivec_vsumsws): Adjust.
22301         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
22302         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
22303         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
22304         support.
22305         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
22306         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
22307         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
22308         (altivec_lve<VI_char>x): Delete expand.
22309         (*altivec_lve<VI_char>x_internal): Rename to...
22310         (altivec_lve<VI_char>x): ... this.
22311         (altivec_lvxl_<mode>): Delete expand.
22312         (*altivec_lvxl_<mode>_internal): Rename to ...
22313         (altivec_lvxl_<mode>): ... this.
22314         (altivec_stvxl_<mode>): Delete expand.
22315         (*altivec_stvxl_<mode>_internal): Rename to ...
22316         (altivec_stvxl_<mode>): ... this.
22317         (altivec_stve<VI_char>x): Delete expand.
22318         (*altivec_stve<VI_char>x_internal): Rename to ...
22319         (altivec_stve<VI_char>x): ... this.
22320         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
22321         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
22322         reduc_plus_scal_<mode>): Adjust.
22323         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
22324         comment.
22325         (rs6000_cpu_cpp_builtins): Adjust.
22326         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
22327         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
22328         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
22329         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
22330         -maltivec=be support.
22331         (rs6000_split_vec_extract_var): Adjust.
22332         (rs6000_split_v4si_init): Adjust.
22333         (swap_selector_for_mode): Delete.
22334         (altivec_expand_lvx_be, altivec_expand_stvx_be,
22335         altivec_expand_stvex_be): Delete.
22336         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
22337         -maltivec=be support.
22338         (rs6000_gimple_fold_builtin): Ditto.
22339         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
22340         Adjust.
22341         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
22342         (TARGET_DIRECT_MOVE_64BIT): Adjust.
22343         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
22344         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
22345         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
22346         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
22347         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
22348         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
22349         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
22350         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
22351         anonymous split): Adjust.
22352         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
22353         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
22355 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
22357         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
22358         when --with-gxx-include-dir is also specified.
22359         * configure: Regenerate.
22361 2018-05-09  Jim Wilson  <jimw@sifive.com>
22363         PR target/84797
22364         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
22365         * config/riscv/t-withmultilib: New.
22366         * config/riscv/withmultilib.h: New.
22367         * doc/install.texi: Document RISC-V --with-multilib-list support.
22369 2018-05-09  Richard Biener  <rguenther@suse.de>
22371         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
22372         vector.
22373         (vect_bb_vectorization_profitable_p): Adjust.  Compute
22374         actual scalar cost using the cost vector and the add_stmt_cost
22375         machinery.
22377 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22379         PR rtl-optimization/85645
22380         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
22381         in the REG_CFA_REGISTER note for LR, don't leave it empty.
22383 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22385         PR rtl-optimization/85645
22386         * shrink-wrap.c (spread_components): Return a boolean saying if
22387         anything was changed.
22388         (try_shrink_wrapping_separate): Iterate spread_components until
22389         nothing changes anymore.
22391 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22393         PR rtl-optimization/85645
22394         * regrename.c (build_def_use): Also kill the chains that include the
22395         destination of a REG_CFA_REGISTER note.
22397 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
22399         PR rtl-optimization/85645
22400         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
22401         insn that has a REG_CFA_REGISTER note.
22403 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
22405         * cfgexpand.c (expand_clobber): New function.
22406         (expand_gimple_stmt_1): Use it.
22407         * tree-vect-stmts.c (vect_clobber_variable): New function,
22408         split out from...
22409         (vectorizable_simd_clone_call): ...here.
22410         (vectorizable_store): Emit a clobber either side of an
22411         IFN_STORE_LANES sequence.
22412         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
22414 2018-05-09  Tom de Vries  <tom@codesourcery.com>
22416         PR target/85626
22417         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
22418         (define_insn "trap_if_false"): Add exit after trap.
22420 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
22422         PR rtl-optimization/85638
22423         * bb-reorder.c: Include common/common-target.h.
22424         (create_forwarder_block): New function extracted from...
22425         (fix_up_crossing_landing_pad): ...here.  Rename into...
22426         (dw2_fix_up_crossing_landing_pad): ...this.
22427         (sjlj_fix_up_crossing_landing_pad): New function.
22428         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
22429         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
22430         from both partitions and exit the loop after one iteration.
22432 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22434         Revert:
22435         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22436         subsection.
22437         (Basic PowerPC Built-in Functions): The new name of the
22438         subsection previously known as "PowerPC Built-in Functions".
22439         (Basic PowerPC Built-in Functions Available on all Configurations):
22440         New subsubsection.
22441         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
22442         subsubsection.
22443         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
22444         subsubsection.
22445         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
22446         subsubsection.
22447         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
22448         subsubsection.
22450 2018-05-08  Jim Wilson  <jimw@sifive.com>
22452         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
22453         (LD_EMUL_SUFFIX): New.
22454         (LINK_SPEC): Use it.
22456 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22458         * doc/extend.texi (PowerPC Built-in Functions): Rename this
22459         subsection.
22460         (Basic PowerPC Built-in Functions): The new name of the
22461         subsection previously known as "PowerPC Built-in Functions".
22462         (Basic PowerPC Built-in Functions Available on all Configurations):
22463         New subsubsection.
22464         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
22465         subsubsection.
22466         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
22467         subsubsection.
22468         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
22469         subsubsection.
22470         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
22471         subsubsection.
22473 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
22475         PR target/85683
22476         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
22477         after cmpelim optimization.
22479 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
22481         * config.gcc: Support "goldmont".
22482         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
22483         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22484         PROCESSOR_GOLDMONT.
22485         * config/i386/i386.c (m_GOLDMONT): Define.
22486         (processor_target_table): Add "goldmont".
22487         (PTA_GOLDMONT): Define.
22488         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
22489         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
22490         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
22491         (fold_builtin_cpu): Add "goldmont".
22492         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
22493         (ix86_option_override_internal): Add "goldmont".
22494         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
22495         (processor_type): Add PROCESSOR_GOLDMONT.
22496         * config/i386/i386.md: Add CPU "glm".
22497         * config/i386/glm.md: New file.
22498         * config/i386/x86-tune.def: Add m_GOLDMONT.
22499         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
22501 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
22503         PR target/85572
22504         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
22505         E_V4DImode.
22506         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
22507         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
22508         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
22510         PR target/85317
22511         * config/i386/i386.c (ix86_fold_builtin): Handle
22512         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
22514         PR target/85480
22515         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
22516         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
22518 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
22520         PR target/85658
22521         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
22522         (check_arch): Likewise.
22523         (check_fpu): Return the result rather than printing it.
22524         (end arch): Fix operator precedence.
22525         (end cpu): Likewise.
22526         (END): Print the result from check_fpu.
22528 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
22529             Alan Hayward  <alan.hayward@arm.com>
22530             David Sherwood  <david.sherwood@arm.com>
22532         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
22533         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
22534         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
22535         (*fcmuo<mode>_and): New patterns.
22537 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
22539         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
22540         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
22541         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
22542         (cmp_op, sve_imm_con): New code attributes.
22543         (SVE_COND_INT_CMP, imm_con): Delete.
22544         (cmp_op): Remove above unspecs from int attribute.
22545         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
22546         to...
22547         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
22548         comparison-specific unspecs.
22549         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
22550         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
22551         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
22552         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
22553         (*vec_fcm<cmp_op><mode>): Rename to...
22554         (*fcm<cmp_op><mode>): ...this and adjust likewise.
22555         (*vec_fcmuo<mode>): Rename to...
22556         (*fcmuo<mode>): ...this and adjust likewise.
22557         (*pred_fcm<cmp_op><mode>): New pattern.
22558         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
22559         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
22560         functions.
22561         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
22562         and UNORDERED.
22563         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
22564         (aarch64_emit_sve_predicated_cond): New function.
22565         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
22566         (aarch64_emit_unspec_cond_or): Replace with...
22567         (aarch64_emit_sve_or_conds): ...this new function.  Use
22568         aarch64_emit_sve_ptrue_op for the individual comparisons and
22569         aarch64_emit_binop to OR them together.
22570         (aarch64_emit_inverted_unspec_cond): Replace with...
22571         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
22572         aarch64_emit_sve_ptrue_op for the comparison and
22573         aarch64_emit_unop to invert the result.
22574         (aarch64_expand_sve_vec_cmp_float): Update after the above
22575         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
22577 2018-05-07  Nathan Sidwell  <nathan@acm.org>
22579         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
22580         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
22581         (Backwards Compatibility): Likewise.
22583 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22585         PR bootstrap/85681
22586         Revert:
22587         2018-05-07  Luis Machado  <luis.machado@linaro.org>
22589         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22590         <prefetch_dynamic_strides>: New const bool field.
22591         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22592         prefetch_dynamic_strides.
22593         (exynosm1_prefetch_tune): Likewise.
22594         (thunderxt88_prefetch_tune): Likewise.
22595         (thunderx_prefetch_tune): Likewise.
22596         (thunderx2t99_prefetch_tune): Likewise.
22597         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
22598         to false.
22599         (aarch64_override_options_internal): Update to set
22600         PARAM_PREFETCH_DYNAMIC_STRIDES.
22601         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22602         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22603         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22604         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22605         prefetch-dynamic-strides setting.
22607         2018-05-07  Luis Machado  <luis.machado@linaro.org>
22609         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22610         <minimum_stride>: New const int field.
22611         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22612         minimum_stride field.
22613         (exynosm1_prefetch_tune): Likewise.
22614         (thunderxt88_prefetch_tune): Likewise.
22615         (thunderx_prefetch_tune): Likewise.
22616         (thunderx2t99_prefetch_tune): Likewise.
22617         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
22618         (aarch64_override_options_internal): Update to set
22619         PARAM_PREFETCH_MINIMUM_STRIDE.
22620         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22621         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22622         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22623         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22624         stride is constant and is below the minimum stride threshold.
22626 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22628         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
22629         to 512.
22631 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22633         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22634         <prefetch_dynamic_strides>: New const bool field.
22635         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22636         prefetch_dynamic_strides.
22637         (exynosm1_prefetch_tune): Likewise.
22638         (thunderxt88_prefetch_tune): Likewise.
22639         (thunderx_prefetch_tune): Likewise.
22640         (thunderx2t99_prefetch_tune): Likewise.
22641         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
22642         to false.
22643         (aarch64_override_options_internal): Update to set
22644         PARAM_PREFETCH_DYNAMIC_STRIDES.
22645         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22646         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22647         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22648         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22649         prefetch-dynamic-strides setting.
22651 2018-05-07  Luis Machado  <luis.machado@linaro.org>
22653         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22654         <minimum_stride>: New const int field.
22655         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22656         minimum_stride field.
22657         (exynosm1_prefetch_tune): Likewise.
22658         (thunderxt88_prefetch_tune): Likewise.
22659         (thunderx_prefetch_tune): Likewise.
22660         (thunderx2t99_prefetch_tune): Likewise.
22661         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
22662         (aarch64_override_options_internal): Update to set
22663         PARAM_PREFETCH_MINIMUM_STRIDE.
22664         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22665         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22666         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22667         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22668         stride is constant and is below the minimum stride threshold.
22670 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
22672         PR c++/85659
22673         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
22674         the type is addressable.  Don't force op into register if it has
22675         BLKmode.
22677 2018-05-05  Roland McGrath  <mcgrathr@google.com>
22679         PR other/77609
22680         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
22681         any section for which we don't know a specific type it should have,
22682         regardless of name.  Previously this was done only for the exact
22683         names ".init_array", ".fini_array", and ".preinit_array".
22684         (default_elf_asm_named_section): Add comment about
22685         relationship with default_section_type_flags and SECTION_NOTYPE.
22686         (get_section): Don't consider it a type conflict if one side has
22687         SECTION_NOTYPE and the other doesn't, as long as neither has the
22688         SECTION_BSS et al used in the default_section_type_flags logic.
22690 2018-05-05  Tom de Vries  <tom@codesourcery.com>
22692         PR target/85653
22693         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
22694         (workaround_barsyncs): New function.
22695         (nvptx_reorg): Use workaround_barsyncs.
22696         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22697         (define_expand "nvptx_membar_cta"): New define_expand.
22698         (define_insn "*nvptx_membar_cta"): New insn.
22700 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
22702         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
22703         To improve optimization opportunities.
22704         * builtin-types.def: The new needed builtin types for the above.
22706 2018-05-04  Richard Biener  <rguenther@suse.de>
22708         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
22709         * gimple-ssa-store-merging.c
22710         (imm_store_chain_info::output_merged_store): Remove redundant create,
22711         release split_store vector contents on failure.
22712         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
22713         scalar stmt vector on cache hit.
22715 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
22717         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
22718         Xilinx FP support.
22719         * config.gcc (powerpc-xilinx-eabi*): Remove.
22720         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
22721         support.
22722         (fusion_addis_mem_combo_load): Ditto.
22723         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
22724         FP support.
22725         (rs6000_cpu_cpp_builtins): Ditto.
22726         * config/rs6000/rs6000-linux.c
22727         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
22728         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
22729         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
22730         support.
22731         (rs6000_setup_reg_addr_masks): Ditto.
22732         (rs6000_init_hard_regno_mode_ok): Ditto.
22733         (rs6000_option_override_internal): Ditto.
22734         (legitimate_lo_sum_address_p): Ditto.
22735         (rs6000_legitimize_address): Ditto.
22736         (rs6000_legitimize_reload_address): Ditto.
22737         (rs6000_legitimate_address_p): Ditto.
22738         (abi_v4_pass_in_fpr): Ditto.
22739         (setup_incoming_varargs): Ditto.
22740         (rs6000_gimplify_va_arg): Ditto.
22741         (rs6000_split_multireg_move): Ditto.
22742         (rs6000_savres_strategy): Ditto.
22743         (rs6000_emit_prologue_components): Ditto.
22744         (rs6000_emit_epilogue_components): Ditto.
22745         (rs6000_emit_prologue): Ditto.
22746         (rs6000_emit_epilogue): Ditto.
22747         (rs6000_elf_file_end): Ditto.
22748         (rs6000_function_value): Ditto.
22749         (rs6000_libcall_value): Ditto.
22750         * config/rs6000/rs6000.h: Ditto.
22751         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
22752         (TARGET_MINMAX): ... this.  New.
22753         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
22754         * config/rs6000/rs6000.md: Remove Xilinx FP support.
22755         (*movsi_internal1_single): Delete.
22756         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
22757         mfpu=, mxilinx-fpu): Delete.
22758         * config/rs6000/singlefp.h: Delete.
22759         * config/rs6000/sysv4.h: Remove Xilinx FP support.
22760         * config/rs6000/t-rs6000: Ditto.
22761         * config/rs6000/t-xilinx: Delete.
22762         * config/rs6000/titan.md: Adjust for fp_type removal.
22763         * config/rs6000/vsx.md: Remove Xilinx FP support.
22764         (VStype_simple): Delete.
22765         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
22766         * config/rs6000/xfpu.h: Delete.
22767         * config/rs6000/xfpu.md: Delete.
22768         * config/rs6000/xilinx.h: Delete.
22769         * config/rs6000/xilinx.opt: Delete.
22770         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
22771         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
22773 2018-05-04  Tom de Vries  <tom@codesourcery.com>
22775         PR libgomp/85639
22776         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
22777         if ignore == 0.
22779 2018-05-04  Richard Biener  <rguenther@suse.de>
22781         PR middle-end/85627
22782         * tree-complex.c (update_complex_assignment): We are always in SSA form.
22783         (expand_complex_div_wide): Likewise.
22784         (expand_complex_operations_1): Likewise.
22785         (expand_complex_libcall): Preserve EH info of the original stmt.
22786         (tree_lower_complex): Handle removed blocks.
22787         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
22788         on complex multiplication and division libcall builtins.
22790 2018-05-04  Richard Biener  <rguenther@suse.de>
22792         PR middle-end/85574
22793         * fold-const.c (negate_expr_p): Restrict negation of operand
22794         zero of a division to when we know that can happen without
22795         overflow.
22796         (fold_negate_expr_1): Likewise.
22798 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
22800         PR libstdc++/85466
22801         * real.h (real_nextafter): Declare.
22802         * real.c (real_nextafter): New function.
22803         * fold-const-call.c (fold_const_nextafter): New function.
22804         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
22805         CASE_CFN_NEXTTOWARD.
22806         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
22807         even when arg1_mode is different from arg0_mode.
22809 2018-05-03  Nathan Sidwell  <nathan@acm.org>
22811         * doc/extend.texi (Deprecated Features): Remove
22812         -ffriend-injection.
22813         (Backwards Compatibility): Likewise.
22814         * doc/invoke.texi (C++ Language Options): Likewise.
22815         (C++ Dialect Options): Likewise.
22817 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
22819         PR target/85530
22820         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
22821         _mm512_mask_mullox_epi64): New intrinsics.
22823 2018-05-03  Tom de Vries  <tom@codesourcery.com>
22825         PR testsuite/85106
22826         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22827         dump files): Add offload-tree.
22829 2018-05-03  Richard Biener  <rguenther@suse.de>
22831         PR tree-optimization/85615
22832         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
22833         to loops not nested in BBs loop father to avoid creating multi-entry
22834         loops.
22836 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22838         PR tree-optimization/70291
22839         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
22840         arguments.  Change return type to tree.  Emit libcall as a new
22841         statement rather than replacing existing one when inplace_p is true.
22842         (expand_complex_multiplication_components): New function.
22843         (expand_complex_multiplication): Expand floating-point complex
22844         multiplication using the above.
22845         (expand_complex_division): Rename inner_type parameter to type.
22846         Update expand_complex_libcall call-site.
22847         (expand_complex_operations_1): Update expand_complex_multiplication
22848         and expand_complex_division call-sites.
22850 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
22852         PR target/85582
22853         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
22854         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
22855         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
22856         the highest significant bit of the shift count mask is clear.  In
22857         check whether and[sq]i3 is needed verify that all significant bits
22858         of the shift count other than the highest are set.
22860 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22862         PR libgomp/82428
22863         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
22864         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
22865         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
22866         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
22867         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
22868         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
22869         __builtin_goacc_parlevel_size.
22871 2018-05-02  Richard Biener  <rguenther@suse.de>
22873         PR tree-optimization/85597
22874         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
22875         do not use split vect_get_vec_defs call but call vect_get_slp_defs
22876         directly.
22878 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22880         PR testsuite/85106
22881         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22882         dump files): Add ltrans-tree.
22884 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22886         PR testsuite/85106
22887         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22888         dump files): Add wpa-ipa.
22890 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
22892         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
22893         powerpc*-*-linux*paired* target.
22894         * config/rs6000/750cl.h: Delete.
22895         * config/rs6000/paired.h: Delete.
22896         * config/rs6000/paired.md: Delete.
22897         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
22898         float support.
22899         * config/rs6000/rs6000-builtin.def: Remove paired float support.
22900         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
22901         comment.  Remove paired float support.
22902         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
22903         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
22904         VECTOR_PAIRED.
22905         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
22906         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
22907         declarations.
22908         * config/rs6000/rs6000.c: Remove paired float support.
22909         (paired_expand_vector_init, paired_expand_vector_move,
22910         paired_emit_vector_compare, paired_emit_vector_cond_expr,
22911         (paired_expand_lv_builtin, paired_expand_stv_builtin,
22912         paired_expand_builtin, paired_expand_predicate_builtin,
22913         paired_init_builtins): Delete.
22914         * config/rs6000/rs6000.h: Remove paired float support.
22915         * config/rs6000/rs6000.md: Remove paired float support.
22916         (move_from_CR_ov_bit): Delete.
22917         * config/rs6000/rs6000.opt (mpaired): Delete.
22918         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
22919         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
22921 2018-05-02  Richard Biener  <rguenther@suse.de>
22923         PR middle-end/85567
22924         * gimplify.c (gimplify_save_expr): When in SSA form allow
22925         SAVE_EXPRs to compute to SSA vars.
22927 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
22929         PR target/85582
22930         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
22931         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
22932         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
22933         clobber operands[2], instead use a new pseudo.  Formatting fixes.
22935 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
22937         PR tree-optimization/85586
22938         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
22939         exit early for statements in the same group if the accesses are
22940         not strided.
22942 2018-05-02  Tom de Vries  <tom@codesourcery.com>
22944         PR lto/85451
22945         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
22946         error message.
22948 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
22950         PR tree-optimization/85143
22951         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
22953 2018-05-01  Tom de Vries  <tom@codesourcery.com>
22955         PR lto/85451
22956         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
22957         not found" error message.
22959 2018-05-01  Tom de Vries  <tom@codesourcery.com>
22961         PR other/83786
22962         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
22963         * vec.c (test_ordered_remove_if): New function.
22964         (vec_c_tests): Call test_ordered_remove_if.
22965         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
22966         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
22967         * tree-vect-patterns.c (vect_pattern_recog_1): Use
22968         VEC_ORDERED_REMOVE_IF.
22970 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22972         PR tree-optimization/82665
22973         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
22974         pointer subtraction where arguments come from a memchr call.
22976 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
22978         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
22979         --push-state --as-needed and --pop-state instead of --as-needed and
22980         --no-as-needed if ld supports it.
22981         * configure: Regenerated.
22983         PR web/85578
22984         * doc/install.texi2html: Replace _002d with - and _002a with * in
22985         generated html files using sed.
22987 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
22989         PR c++/85523
22990         * gcc-rich-location.c (blank_line_before_p): New function.
22991         (use_new_line): New function.
22992         (gcc_rich_location::add_fixit_insert_formatted): New function.
22993         * gcc-rich-location.h
22994         (gcc_rich_location::add_fixit_insert_formatted): New function.
22996 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
22998         * selftest.c (assert_streq): Rename "expected" and "actual" to
22999         "val1" and "val2".  Extend NULL-handling to cover both inputs
23000         symmetrically, while still requiring both to be non-NULL for a pass.
23001         * selftest.h (assert_streq): Rename "expected" and "actual" to
23002         "val1" and "val2".
23003         (ASSERT_EQ): Likewise.
23004         (ASSERT_EQ_AT): Likewise.
23005         (ASSERT_KNOWN_EQ): Likewise.
23006         (ASSERT_KNOWN_EQ_AT): Likewise.
23007         (ASSERT_NE): Likewise.
23008         (ASSERT_MAYBE_NE): Likewise.
23009         (ASSERT_MAYBE_NE_AT): Likewise.
23010         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
23011         the assertion to pass.
23012         (ASSERT_STREQ_AT): Likewise.
23014 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
23016         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
23017         interaction with -pie.
23019 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23021         * selftest.h: Fix alphabetization of per-source-file selftest
23022         declarations.
23024 2018-04-30  Jason Merrill  <jason@redhat.com>
23026         PR c++/61982 - dead stores to destroyed objects.
23027         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
23028         of clobber.
23030 2018-04-30  Jason Merrill  <jason@redhat.com>
23032         * tree.c (build_clobber): New.
23033         * tree.h: Declare it.
23034         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
23036 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23038         * diagnostic-show-locus.c (layout::layout): Update for
23039         location_get_source_line returning a char_span.
23040         (struct char_span): Move to input.h.
23041         (struct correction): Update for fields in char_span becoming
23042         private.
23043         (struct source_line): Update for location_get_source_line
23044         returning a char_span.
23045         (layout::print_line): Likewise.
23046         * edit-context.c (edited_file::print_content): Likewise.
23047         (edited_file::print_diff_hunk): Likewise.
23048         (edited_file::print_run_of_changed_lines): Likewise.
23049         (edited_file::get_num_lines): Likewise.
23050         (edited_line::edited_line): Likewise.
23051         * final.c (asm_show_source): Likewise.
23052         * input.c (location_get_source_line): Convert return type
23053         from const char * to char_span, losing the final "line_len"
23054         param.
23055         (dump_location_info): Update for the above.
23056         (get_substring_ranges_for_loc): Likewise.  Use a char_span
23057         when handling the literal within the line.
23058         (test_reading_source_line): Update for location_get_source_line
23059         returning a char_span.
23060         * input.h (class char_span): Move here from
23061         diagnostic-show-locus.c, converting from a struct to a class.
23062         Make data members private.
23063         (char_span::operator bool): New.
23064         (char_span::length): New.
23065         (char_span::get_buffer): New.
23066         (char_span::operator[]): New.
23067         (char_span::subspan): Make const.
23068         (char_span::xstrdup): New.
23069         (location_get_source_line): Convert return type from const char *
23070         to char_span, losing the final "line_size" param.
23072 2018-04-30  Jan Hubicka  <jh@suse.cz>
23074         * lto-wrapper.c (ltrans_priorities): New static var.
23075         (cmp_priority): New.
23076         (run_gcc): Read priorities and if doing parallel build order
23077         the Makefile by them.
23079 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
23081         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
23083 2018-04-30  Richard Biener  <rguenther@suse.de>
23085         * tree-cfg.c (verify_address): Remove base argument, add
23086         flag whether to check TREE_ADDRESSABLE and do that.
23087         (verify_expr): Remove.
23088         (verify_types_in_gimple_reference): Add pieces from verify_expr.
23089         (verify_gimple_assign_single): Likewise.
23090         (verify_gimple_switch): Likewise.
23091         (verify_expr_location_1): Dereference tp once.  Add (disabled)
23092         piece from verify_expr.
23093         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
23095 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23097         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
23099 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23101         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
23102         (small_data_pattern): Likewise.
23103         (arc_rewrite_small_data): Likewise.
23104         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
23105         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
23106         (get_symbol_alignment): New function.
23107         (legitimate_small_data_address_p): Likewise.
23108         (legitimate_scaled_address): Update, call
23109         legitimate_small_data_address_p.
23110         (output_sdata): New static variable.
23111         (arc_print_operand): Update how we handle small data operands.
23112         (arc_print_operand_address): Likewise.
23113         (arc_legitimate_address_p): Update, use
23114         legitimate_small_data_address_p.
23115         (arc_rewrite_small_data_p): Remove.
23116         (arc_rewrite_small_data_1): Likewise.
23117         (arc_rewrite_small_data): Likewise.
23118         (small_data_pattern): Likewise.
23119         (compact_sda_memory_operand): Update to use
23120         legitimate_small_data_address_p and get_symbol_alignment.
23121         (prepare_move_operands): Don't rewite sdata pattern.
23122         (prepare_extend_operands): Remove.
23123         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
23124         pattern.
23125         (zero_extendqisi2): Likewise.
23126         (zero_extendhisi2): Likewise.
23127         (extendqihi2): Likewise.
23128         (extendqisi2): Likewise.
23129         (extendhisi2): Likewise.
23130         (addsi3): Likewise.
23131         (subsi3): Likewise.
23132         (andsi3): Likewise.
23133         * config/arc/constraints.md (Usd): Change it to memory constraint.
23135 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
23137         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
23138         as source of std instructions.
23139         * config/arc/arc.md (movsi_insn): Update pattern predicate to
23140         allow 6-bit constants as source for store instructions.
23141         (movdi_insn): Update instruction pattern to allow 6-bit constants
23142         as source for store instructions.
23144 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
23146         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
23148 2018-04-30  Nathan Sidwell  <nathan@acm.org>
23149             Sandra Loosemore <sandra@codesourcery.com>
23151         * dumpfile.c (dump_open): Allow '-' for stdout.
23152         * doc/invoke.texi (Developer Options): Document dump filename
23153         determination early.  Document stdin/stdout selection.
23155 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
23157         Microblaze Target: PIC data text relative
23159         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
23160         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
23161         Add declaration.
23162         * config/microblaze/microblaze.h (microblaze_constant_address_p):
23163         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
23164         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
23165         New addressing mode for data-text relative position indepenedent code.
23166         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
23167         'ADDRESS_SYMBOLIC_TXT_REL'.
23168         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
23169         (microblaze_legitimate_pic_operand): Exclude function calls from
23170         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
23171         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
23172         addresses cases.
23173         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
23174         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
23175         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
23176         for 'address + offset'.
23177         (microblaze_expand_prologue): Add new function prologue call for
23178         'r20' assignation.
23179         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
23180         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
23181         table in case of TARGET_PIC_DATA_TEXT_REL.
23182         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
23183         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
23184         Add new macros 'UNSPEC_TEXT',
23185         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
23186         + exclude function calls from 'UNSPEC_PLT' in case of data text
23187         relative mode.
23188         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
23189         new target hook for generating address diff vector tables in case of
23190         flag_pic.
23191         * doc/tm.texi : Regenerate.
23192         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
23193         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
23194         of addr diff vector generation.
23195         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
23196         target hook definition.
23197         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
23198         Add default function for generate_pic_addr_diff_vec -> flag_pic.
23199         * doc/invoke.texi (Add new pic option): Add new microblaze pic
23200         option for data text relative.
23202 2018-04-30  Richard Biener  <rguenther@suse.de>
23204         * tree-chrec.h (evolution_function_is_constant_p): Remove
23205         redundant check.
23206         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
23208 2018-04-30  Richard Biener  <rguenther@suse.de>
23210         PR bootstrap/85571
23211         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
23213 2018-04-30  Richard Biener  <rguenther@suse.de>
23215         PR tree-optimization/28364
23216         PR tree-optimization/85275
23217         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
23218         copying first exit test.
23220 2018-04-28  Mark Wielaard  <mark@klomp.org>
23222         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
23223         dwarf_version >= 5.
23224         (dwarf_AT): Handle DW_AT_addr_base.
23225         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
23227 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
23229         PR target/84431
23230         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
23231         (*ashl<dwi>3_doubleword_mask_1): Ditto.
23232         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
23233         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
23235 2018-04-28  Richard Biener  <rguenther@suse.de>
23237         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
23238         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
23239         to reflect use.  Only add interesting stmts.
23241 2018-04-27  Martin Jambor  <mjambor@suse.cz>
23243         PR ipa/85549
23244         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
23245         the jump function allows for passing through aggregate values.
23247 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
23249         * input.h (in_system_header_at): Convert from macro to inline
23250         function.
23251         (from_macro_expansion_at): Likewise.
23252         (from_macro_definition_at): Likewise.
23254 2018-04-27  Jeff Law  <law@redhat.com>
23256         * config.gcc: Mark tile* targets as deprecated/obsolete.
23258 2018-04-27  Richard Biener  <rguenther@suse.de>
23260         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
23261         fix for ILP32.
23263 2018-04-27  Richard Biener  <rguenther@suse.de>
23265         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
23267 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
23269         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
23270         with Yd constraint. Set "preferred_for_speed" attribute from
23271         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
23272         with Yd constraint.
23273         (*movdi_internal): Ditto.
23274         (movti_interunit splitters): Remove
23275         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
23276         (movdi_interunit splitters): Ditto.
23277         * config/i386/constraints.md (Ye): Remove.
23278         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
23280 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23282         PR target/85512
23283         * config/aarch64/constraints.md (Usg): Limit to 31.
23284         (Usj): Limit to 63.
23286 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
23288         PR tree-optimization/85529
23289         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
23290         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
23291         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
23292         zero extension or masking of the MSB bit.
23293         (optimize_range_tests): Add FIRST_BB argument, pass it through
23294         to optimize_range_tests_var_bound.
23295         (maybe_optimize_range_tests, reassociate_bb): Adjust
23296         optimize_range_tests callers.
23298 2018-04-26  Richard Biener  <rguenther@suse.de>
23299             Jakub Jelinek  <jakub@redhat.com>
23301         * cgraph.h (symbol_table): Just declare debug method here.
23302         * symtab.c (symbol_table::debug): Define.
23304 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
23306         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
23308 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
23310         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
23311         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
23312         (*movdi_internal): Substitute Yi and Yj constraint with x
23313         and Ym and Yn constraint with y constraint.  Update "isa"
23314         attribute and set "preferred_for_speed" attribute from
23315         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
23316         (*movsi_internal): Ditto.
23317         (*movdf_internal): Ditto.
23318         (*movsf_internal): Ditto.
23319         (*zero_extendsidi2): Ditto.
23320         * config/i386/sse.md (vec_set<mode>_0): Ditto.
23321         (sse2_loadld): Ditto.
23322         (*vec_extract<ssevecmodelower>_0): Ditto.
23323         (*vec_extractv4si_0_zext_sse4): Ditto.
23324         (vec_concatv2di): Ditto.
23325         (*vec_dup<mode>): Ditto.
23326         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
23327         * config/i386/constraints.md (Yi): Remove.
23328         (Yj): Remove.
23329         (Ym): Remove.
23330         (Yn): Remove.
23332 2018-04-26  Nathan Sidwell  <nathan@acm.org>
23334         * dumpfile.c (dump_open): New.
23335         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
23336         (dump_finish): Detect stdio/stderr by value not name.
23338 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
23340         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
23342 2018-04-26  Tom de Vries  <tom@codesourcery.com>
23344         PR target/84952
23345         * config/nvptx/nvptx.c (verify_neutering_jumps)
23346         (verify_neutering_labels): New function
23347         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
23349 2018-04-26  Tom de Vries  <tom@codesourcery.com>
23351         PR target/84025
23352         * config/nvptx/nvptx.c (needs_neutering_p): New function.
23353         (nvptx_single): Use needs_neutering_p to skip over insns that do not
23354         need neutering.
23356 2018-04-26  Richard Biener <rguenther@suse.de>
23357             Tom de Vries  <tom@codesourcery.com>
23359         PR lto/85422
23360         * lto-streamer-out.c (output_function): Fixup loops if required to match
23361         discovery done in the reader.
23363 2018-04-26  Richard Biener  <rguenther@suse.de>
23365         PR tree-optimization/85116
23366         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
23367         have a loop exit from the single latch predecessor.  Remove
23368         case of header with just condition.
23369         (ch_base::copy_headers): Exclude infinite loops from any
23370         processing.
23371         (pass_ch::execute): Record exits.
23373 2018-04-26  Richard Biener  <rguenther@suse.de>
23375         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
23376         prologue cost vector and pass it to vect_get_load_cost.
23377         (vect_get_peeling_costs_all_drs): Likewise.
23378         (vect_peeling_hash_get_lowest_cost): Likewise.
23379         (vect_enhance_data_refs_alignment): Likewise.
23381 2018-04-26  Richard Biener  <rguenther@suse.de>
23383         PR middle-end/85450
23384         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
23385         checking of integer<->pointer conversions.
23386         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
23387         sign-/zero-extending pointer types.
23388         (expand_omp_for_static_chunk): Likewise.
23390 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
23391             Jean Lee  <xiaoyur347@gmail.com>
23393         * config/mips/mips.c (mips_asan_shadow_offset): New function.
23394         (TARGET_ASAN_SHADOW_OFFSET): Define.
23395         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
23396         true for -fsanitize=address.
23398 2018-04-25  Mark Wielaard  <mark@klomp.org>
23400         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
23401         shorter ones.
23403 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
23405         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
23406         than "alu", remove explicit "memory" and "imm_disp" attributes.
23407         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
23409         PR middle-end/85414
23410         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
23411         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
23412         gen_lowpart_no_emit.
23414 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
23416         PR target/85473
23417         * config/i386/i386.c (ix86_expand_builtin): Change memory
23418         operand to XI, extend p0 to Pmode.
23419         * config/i386/i386.md: Change unspec volatile and operand
23420         1 mode to XI, change operand 0 mode to P.
23422 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23424         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
23425         GET_MODE_MASK before any checking.
23426         (nds32_can_use_bset_p): Likewise.
23427         (nds32_can_use_btgl_p): Likewise.
23429 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23431         * config/nds32/nds32-doubleword.md: New define_split pattern for
23432         illegal register number.
23434 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23436         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
23438 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
23440         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
23442 2018-04-25  Richard Biener  <rguenther@suse.de>
23444         * lto-streamer.h (LTO_major_version): Bump to 8.
23446 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
23448         * BASE-VER: Set to 9.0.0.
23450 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
23452         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
23453         in __abskf2 and __powikf2.
23455 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23457         PR target/85512
23458         * config/aarch64/constraints.md (Usg, Usj): New constraints.
23459         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
23460         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
23461         Use the above on operand 2.  Reindent.
23462         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
23464 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
23466         PR target/85485
23467         * common/config/i386/i386-common.c (ix86_handle_option): Don't
23468         handle OPT_mcet.
23469         * config/i386/i386.opt (mcet): Removed.
23470         * doc/install.texi: Remove -mcet documentation.
23471         * doc/invoke.texi: Likewise.
23473 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
23475         PR target/85485
23476         * doc/install.texi: Remove -mcet from bootstrap-cet.
23478 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
23480         PR target/85511
23481         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
23482         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
23483         if TARGET_64BIT.
23485         PR target/85503
23486         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
23487         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
23488         containing a CONST_VECTOR.
23490 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
23492         * doc/install.texi: Update newlib dependency for nvptx.
23494 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
23496         PR target/85508
23497         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
23498         instead of INTVAL when shifting x left.
23500 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
23502         PR tree-optimization/85478
23503         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
23504         vect_grouped_store_supported for single element vectors.
23506 2018-04-24  Richard Biener  <rguenther@suse.de>
23508         PR target/85491
23509         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
23510         load cost increase to the case of non-constant step.
23512 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
23514         PR target/84828
23515         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
23516         destination if any_malformed_asm.
23518 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
23520         PR middle-end/85496
23521         * expr.c (store_field): In the bitfield case, if the value comes from
23522         a function call and is returned in registers by means of a PARALLEL,
23523         do not change the mode of the temporary unless BLKmode and VOIDmode.
23525 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
23527         PR rtl-optimization/85423
23528         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
23529         dependencies to debug insns when the previous insn is non-debug.
23531 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
23533         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
23534         enums into a single definition.
23535         (fls): Fix predicates and printing.
23536         (seti): Likewise.
23538 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
23540         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
23541         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
23542         and short u6 immediate.
23543         (check_if_valid_sleep_operand): Remove.
23544         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
23546 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23548         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
23549         flag_always_save_lp condition.
23550         * config/nds32/nds32.opt (malways-save-lp): New option.
23552 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
23554         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
23555         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
23556         * config/nds32/nds32.h
23557         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
23558         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
23560 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
23562         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
23563         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
23565 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
23566             Chung-Ju Wu  <jasonwucj@gmail.com>
23568         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
23569         Declare.
23570         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
23571         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
23573 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23575         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
23577 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23579         * config/nds32/nds32-protos.h (nds32_data_alignment,
23580         nds32_local_alignment): Declare.
23581         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
23582         nds32_local_alignment): New functions.
23583         (TARGET_CONSTANT_ALIGNMENT): Define.
23584         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
23586 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23588         * config/nds32/nds32.c
23589         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
23590         (TARGET_MODES_TIEABLE_P): Likewise.
23592 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
23594         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
23595         level Ofast and Og.
23597 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
23598             Chung-Ju Wu  <jasonwucj@gmail.com>
23600         * config/nds32/constants.md (unspec_volatile_element): Add enum values
23601         for unaligned access.
23602         * config/nds32/nds32-intrinsic.c: Implementation of expanding
23603         unaligned access.
23604         * config/nds32/nds32-intrinsic.md: Likewise.
23605         * config/nds32/nds32_intrinsic.h: Likewise.
23606         * config/nds32/nds32.h (nds32_builtins): Likewise.
23607         * config/nds32/nds32.opt (munaligned-access): New option.
23608         * config/nds32/nds32.c (nds32_asm_file_start): Display
23609         flag_unaligned_access status.
23611 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
23613         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
23614         -mno-relax is present.
23615         * config/riscv/linux.h (LINK_SPEC): Ditto.
23617 2018-04-20  Martin Sebor  <msebor@redhat.com>
23619         PR c/85365
23620         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
23621         for null pointers.
23622         (gimple_fold_builtin_stxcpy_chk): Same.
23623         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
23625 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
23627         PR target/85456
23628         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
23629         __powikf2 when long double is IEEE 128-bit.
23631 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
23633         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
23634         step to make sure stack always aligned.
23636 2018-04-20  Carl Love  <cel@us.ibm.com>
23638         PR target/83402
23639         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
23640         size check for arg0.
23642 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
23643             Tom de Vries  <tom@codesourcery.com>
23645         PR target/85445
23646         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
23647         Emit insns for calls too.
23648         (nvptx_find_par): Always look for worker-level predecessor insn.
23649         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
23650         calls.
23651         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
23652         (nvptx_process_pars): Propagate frames for calls.
23654 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
23656         PR target/85469
23657         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
23658         Removed.
23659         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
23660         (ix86_handle_option): Don't handle OPT_mibt.
23661         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
23662         __SHSTK__.
23663         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
23664         has_ibt and ibt.
23665         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
23666         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
23667         (ix86_target_macros): Define __CET__ with flag_cf_protection
23668         for -fcf-protection.
23669         * config/i386/i386.c (isa2_opts): Remove -mibt.
23670         * config/i386/i386.h (TARGET_IBT): Removed.
23671         (TARGET_IBT_P): Likewise.
23672         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
23673         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
23674         * config/i386/i386.opt (mcet): Update help message.
23675         (mshstk): Likewise.
23676         (mibt): Removed.
23677         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
23678         -mcet as an alias for -mshstk.
23680 2018-04-20  Richard Biener <rguenther@suse.de>
23682         PR middle-end/85475
23683         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
23684         complexity by forcing a single use of the multiply operand.
23686 2018-04-20  Martin Jambor  <mjambor@suse.cz>
23688         ipa/85449
23689         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
23690         recursion dependency to only apply to non-clones.
23692 2018-04-20  Martin Jambor  <mjambor@suse.cz>
23694         ipa/85447
23695         * ipa-cp.c (create_specialized_node): Check that clones of
23696         self-recursive edges exist during IPA-CP.
23698 2018-04-19  Toon Moene  <toon@moene.org>
23700         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
23701         by -O3.
23703 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
23705         PR tree-optimization/85467
23706         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
23707         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
23708         VECTOR_CST element to type.
23710 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
23712         PR target/85397
23713         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
23714         * config/i386/i386.md (builtin_setjmp_setup): Removed.
23715         (builtin_longjmp): Likewise.
23716         (save_stack_nonlocal): New pattern.
23717         (restore_stack_nonlocal): Likewise.
23719 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
23721         PR target/85404
23722         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
23723         Replace ASM_OUTPUT_LABEL with fprintf.
23725 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
23727         PR target/85417
23728         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
23729         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
23730         * config/i386/i386-c.c (ix86_target_macros_internal): Also
23731         define __IBT__ and __SHSTK__ for -fcf-protection.
23732         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
23733         TARGET_IBT.
23734         (ix86_trampoline_init): Likewise.
23735         (x86_output_mi_thunk): Likewise.
23736         (ix86_notrack_prefixed_insn_p): Likewise.
23737         (ix86_option_override_internal): Don't disallow -fcf-protection.
23738         * config/i386/i386.md (rdssp<mode>): Also enable for
23739         -fcf-protection.
23740         (incssp<mode>): Likewise.
23741         (nop_endbr): Likewise.
23742         * config/i386/i386.opt (mcet): Change help message to built-in
23743         functions only.
23744         (mibt): Likewise.
23745         (mshstk): Likewise.
23746         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
23747         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
23748         enable CET built-in functions.
23750 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
23752         * common/config/i386/i386-common.c
23753         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
23754         OPTION_MASK_ISA_MOVDIRI_UNSET,
23755         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
23756         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
23757         * config.gcc (movdirintrin.h): New header.
23758         * config/i386/cpuid.h (bit_MOVDIRI,
23759         bit_MOVDIR64B): New bits.
23760         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
23761         and -mmvodir64b.
23762         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
23763         (VOID, PVOID, PCVOID)): New function types.
23764         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
23765         __builtin_ia32_directstoreu_u64,
23766         __builtin_ia32_movdir64b): New builtins.
23767         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
23768         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
23769         and -mmovdiri.
23770         (ix86_valid_target_attribute_inner_p): Ditto.
23771         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
23772         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
23773         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
23774         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
23775         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
23776         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
23777         (movdiri<mode>, movdir64b_<mode>): New.
23778         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
23779         * config/i386/immintrin.h: Include movdirintrin.h.
23780         * config/i386/movdirintrin.h: New file.
23781         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
23783 2018-04-19  Richard Biener  <rguenther@suse.de>
23785         PR middle-end/85455
23786         * cfg.c (clear_bb_flags): When loop state says we have
23787         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
23789 2018-04-19  Richard Biener  <rguenther@suse.de>
23791         PR tree-optimization/84737
23792         * tree-vect-data-refs.c (vect_copy_ref_info): New function
23793         copying restrict info.
23794         (vect_setup_realignment): Use it.
23795         * tree-vectorizer.h (vect_copy_ref_info): Declare.
23796         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
23797         the first DR to all generated stores.
23798         (vectorizable_load): Likewise for loads.
23800 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
23802         PR tree-optimization/85446
23803         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
23804         the integral and pointer types to have the same precision.
23806         * doc/install.texi: Document --disable-cet being the default and
23807         --enable-cet=auto.
23809 2018-04-18  Martin Liska  <mliska@suse.cz>
23811         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
23812         style.
23814 2018-04-18  Martin Liska  <mliska@suse.cz>
23816         Revert
23817         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
23819         PR ipa/83983
23820         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23821         arguments if they are comparable.
23823 2018-04-18  Martin Liska  <mliska@suse.cz>
23825         Revert
23826         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23828         PR lto/84805
23829         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23830         incomplete types.
23832 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
23834         PR target/85388
23835         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
23836         ENDBR after calling __morestack.
23838 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
23840         PR jit/85384
23841         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
23842         by using gcc_base_ver to generate a gcc_driver_version, and use
23843         it when generating GCC_DRIVER_NAME.
23844         * configure: Regenerate.
23846 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
23848         PR target/81084
23849         * config.gcc: Obsolete powerpc*-*-*spe*.
23851 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23853         PR debug/84637
23854         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
23855         (stabstr_D): Change type of unum from unsigned int to
23856         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
23857         type.
23859 2018-04-17  Jim Wilson  <jimw@sifive.com>
23861         PR 84856
23862         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
23863         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
23864         Set arg_pointer_offset after using pretend_args_size.
23866 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23868         PR rtl-optimization/85431
23869         * dse.c (record_store): Ignore zero width stores.
23871         PR sanitizer/85230
23872         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
23873         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
23874         __builtin_stack_restore rather than after it.
23875         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
23876         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
23877         argument instead of virtual_dynamic_stack_rtx.
23879 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23881         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
23882         New prototype.
23883         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23884         Add note to error message to explain internal mapping of overloaded
23885         built-in function name to non-overloaded built-in function name.
23886         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
23887         function.
23889 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
23891         PR target/85424
23892         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
23893         where the inputs overlap with the output.
23895 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23897         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
23898         (=v, v) alternative and explicit "memory" attribute.
23899         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
23900         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23901         attributes.
23902         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
23903         "sselog1" type instead of "sselog".
23904         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
23905         "sselog".  Remove explicit "memory" attribute.
23906         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
23907         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23908         attributes.
23909         (vec_extract_hi_v32hi): Merge all alternatives into one, use
23910         "sselog1" type instead of "sselog".  Remove explicit "memory"
23911         attribute.
23912         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
23913         use "sselog1" type instead of "sselog".  Remove explicit "memory"
23914         attribute.
23915         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
23916         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23917         attributes.
23918         (vec_extract_hi_v64qi): Merge all alternatives into one, use
23919         "sselog1" type instead of "sselog".  Remove explicit "memory"
23920         attribute.
23921         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
23922         use "sselog1" type instead of "sselog".  Remove explicit "memory"
23923         attribute.
23925         PR target/85430
23926         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
23928         PR middle-end/85414
23929         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
23930         on a SUBREG.
23932 2018-04-17  Martin Jambor  <mjambor@suse.cz>
23934         PR ipa/85421
23935         * ipa-cp.c (create_specialized_node): Call
23936         expand_all_artificial_thunks if necessary.
23938 2018-04-17  Martin Liska  <mliska@suse.cz>
23940         PR lto/85405
23941         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
23942         in message, remote space in between '_G' and '('.
23944 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
23946         PR target/85281
23947         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
23948         avx512f_vmcmp<mode>3<round_saeonly_name>,
23949         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
23950         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
23951         avx512f_rndscale<mode><round_saeonly_name>,
23952         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
23953         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
23954         Use %<iptr>2 instead of %2 for -masm=intel.
23955         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
23956         avx512f_vcvttss2usi<round_saeonly_name>,
23957         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
23958         -masm=intel.
23959         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
23960         avx512f_vcvttsd2usi<round_saeonly_name>,
23961         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
23962         Use %q1 instead of %1 for -masm=intel.
23963         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
23964         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
23965         of %3 for -masm=intel.
23966         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
23967         -masm=intel.
23968         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
23969         -masm=intel.
23970         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
23971         -masm=intel.
23972         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
23973         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
23974         %g1.
23975         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
23976         -masm=intel.
23977         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
23978         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
23979         %g1 and one with %0 and %1.
23980         (avx512er_vmrcp28<mode><round_saeonly_name>,
23981         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
23982         %1 for -masm=intel.
23983         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
23984         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
23985         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
23986         of %0 and %{%4%} for -masm=intel.
23987         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
23988         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
23989         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
23990         order of %0 and %{%5%}%{z%} for -masm=intel.
23992 2018-04-17  Jan Hubicka  <jh@suse.cz>
23994         PR lto/85405
23995         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
23997 2018-04-17  Martin Liska  <mliska@suse.cz>
23999         PR ipa/85329
24000         * multiple_target.c (create_dispatcher_calls): Set apostrophes
24001         for target_clone error message.  Make default implementation
24002         clone to be a local declaration.
24003         (separate_attrs): Add new argument and check for an empty
24004         string.
24005         (expand_target_clones): Handle it.
24006         (ipa_target_clone): Make redirection just for target_clones
24007         functions.
24009 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
24010             Tom de Vries  <tom@codesourcery.com>
24012         PR middle-end/84955
24013         * omp-expand.c (expand_oacc_for): Add dummy false branch for
24014         tiled basic blocks without omp continue statements.
24016 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
24018         PR target/83660
24019         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
24020         vec_extract expression as having side effects to make sure it gets
24021         a cleanup point.
24023 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
24025         PR target/85403
24026         * config/i386/i386.c (get_builtin_code_for_version): Check
24027         error_mark_node.
24029 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
24031         PR target/84331
24032         * config.gcc: Support "skylake".
24033         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
24034         PROCESSOR_SKYLAKE.
24035         * config/i386/i386.c (m_SKYLAKE): Define.
24036         (processor_target_table): Add "skylake".
24037         (ix86_option_override_internal): Add "skylake".
24038         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
24039         PROCESSOR_CANNONLAKE.
24040         (get_builtin_code_for_version): Fix priority for
24041         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
24042         PROCESSOR_SKYLAKE-AVX512.
24043         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
24044         (processor_type): Add PROCESSOR_SKYLAKE.
24046 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
24047             Jason Merrill  <jason@redhat.com>
24049         PR c++/85112
24050         * convert.c (convert_to_integer_1): Use direct recursion for
24051         enumeral types and types with a precision less than the number
24052         of bits in their mode.
24054 2018-04-16  Julia Koval  <julia.koval@intel.com>
24056         PR target/84413
24057         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
24058         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
24060 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
24062         PR target/85293
24063         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
24064         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
24065         and -mno-direct-move.
24067 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
24069         PR target/83402
24070         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
24071         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
24072         Ensure negative shifts result in {0}.
24074 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
24076         PR rtl-optimization/79916
24077         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
24078         regs (if any) to define how to gnerate SD moves when LRA is in
24079         progress.
24081 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
24083         PR rtl-optimization/85393
24084         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
24085         * except.c (expand_dw2_landing_pad_for_region): Make static.
24086         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
24087         a label and unconditional jump to old_bb, rather than
24088         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
24089         basic block.
24091         PR rtl-optimization/85376
24092         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
24093         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
24094         instead of a specific value.
24096 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
24097             Bin Cheng  <bin.cheng@arm.com>
24099         PR tree-optimization/82965
24100         PR tree-optimization/83991
24101         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
24102         by_profile_only parameter.
24103         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
24104         information if the loop was predicted to iterate too many times.
24105         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
24107 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
24109         PR lto/71991
24110         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
24111         always inline.
24113 2018-04-13  Martin Liska  <mliska@suse.cz>
24114             Jakub Jelinek  <jakub@redhat.com>
24116         PR middle-end/81657
24117         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
24118         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
24119         * builtins.c (expand_builtin_memory_copy_args): Use
24120         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
24121         handle dest_addr == pc_rtx.
24123 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
24125         PR target/85291
24126         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
24127         asked to not generate direct moves.
24128         (fix_trunc<mode>si2_stfiwx): Similar.
24129         (fix_trunc<mode>si2_internal): Similar.
24131 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24133         PR debug/83157
24134         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
24135         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
24136         lookup if dest in some wider mode is known to be const0_rtx and
24137         if so, record permanent equivalence for it to be ZERO_EXTEND of
24138         the narrower mode destination.
24140 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
24142         * lto-streamer-out.c (output_function): Revert 259346.
24143         * omp-expand.c (expand_oacc_for): Likewise.
24145 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
24147         PR rtl-optimization/85354
24148         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
24149         * sel-sched.c (sel_global_init): ... here.
24151 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
24153         PR target/85238
24154         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
24155         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
24156         mode for PE-COFF targets.
24157         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
24158         (i386_pe_asm_lto_end): Likewise.
24159         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
24160         (TARGET_ASM_LTO_END): Likewise.
24161         * config/i386/winnt.c (saved_debug_info_level): New static variable.
24162         (i386_pe_asm_lto_start): New function.
24163         (i386_pe_asm_lto_end): Likewise.
24165 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
24166             Richard Biener  <rguenther@suse.de>
24168         PR middle-end/84955
24169         * lto-streamer-out.c (output_function): Fix CFG loop state before
24170         streaming out.
24171         * omp-expand.c (expand_oacc_for): Handle calls to internal
24172         functions like regular functions.
24174 2018-04-12  Richard Biener  <rguenther@suse.de>
24176         PR lto/85371
24177         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
24178         for the early LTO debug to properly generate references to it
24179         during DIE emission.  Do not re-use that for the skeleton for
24180         split-dwarf.
24181         (dwarf2out_early_finish): Likewise.
24183 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24185         PR target/85328
24186         * config/i386/sse.md
24187         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
24188         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
24189         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
24190         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
24191         and output is a reg, avoid creating invalid lowpart subreg, but
24192         instead split into a 512-bit move.  Don't split if not AVX512VL,
24193         input is xmm16+ reg and output is a mem.
24194         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
24195         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
24196         xmm16+ reg and output is a mem.
24198 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24200         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
24201         also for flag_dwarf2_cfi_asm.
24203 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
24205         PR rtl-optimization/85342
24206         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
24207         a bool scalar var inside of the loop instead.  Don't try to update
24208         recog_data.operand after failed apply_change_group.
24210 2018-04-12  Tom de Vries  <tom@codesourcery.com>
24212         PR target/85296
24213         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
24214         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
24215         array with flexible array member as array without given dimension.
24216         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
24217         argument for undefined param to true.
24219 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
24221         PR target/85321
24222         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
24223         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
24224         from PowerPC section.
24225         * config/rs6000/sysv4.opt (mcall-): Improve help text.
24226         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
24227         help text that is too long.
24228         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
24229         help text that is too long.
24230         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
24231         help text that is too long.
24233 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
24235         * config/alpha/alpha.md (stack_probe_internal): Rename
24236         from "probe_stack".  Update all callers.
24238 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24240         PR rtl-optimization/84566
24241         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
24242         sched_macro_fuse_insns.
24244 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24246         PR target/84301
24247         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
24248         (compute_block_dependences): ... from here.
24250 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24252         PR tree-optimization/85331
24253         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
24254         from int to HOST_WIDE_INT.
24256 2018-04-11  Martin Jambor  <mjambor@suse.cz>
24258         PR ipa/84149
24259         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
24260         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
24261         not the same as the source val.
24262         (cgraph_edge_brings_value_p): New parameter.
24263         (gather_edges_for_value): Pass destination value to
24264         cgraph_edge_brings_value_p.
24265         (perhaps_add_new_callers): Likewise.
24266         (get_info_about_necessary_edges): Likewise and exclude values brought
24267         only by self-recursive edges.
24268         (create_specialized_node): Redirect only clones of self-calling edges.
24269         (+self_recursive_pass_through_p): New function.
24270         (find_more_scalar_values_for_callers_subset): Use it.
24271         (find_aggregate_values_for_callers_subset): Likewise.
24272         (known_aggs_to_agg_replacement_list): Removed.
24273         (decide_whether_version_node): Re-calculate known constants for all
24274         remaining context clones.
24276 2018-04-11  Richard Biener  <rguenther@suse.de>
24278         PR lto/85339
24279         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
24280         from early DWARF output.
24281         (dwarf2out_early_finish): Output line info unconditionally into
24282         early DWARF and add reference to it.
24284 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24286         PR target/85281
24287         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
24288         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
24289         other than V2DFmode using iptr mode attribute.
24290         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
24292 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
24294         PR rtl-optimization/84659
24295         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
24297 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
24299         PR debug/85302
24300         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
24301         SIZEP is NULL.
24302         (output_loc_list): Pass address of a dummy size variable even in the
24303         locview handling loop.
24304         (index_location_lists): Add comment on why skip_loc_list_entry can't
24305         call size_of_locs.
24307 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24309         PR target/85261
24310         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
24311         into register.
24313 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
24315         PR target/85321
24316         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
24317         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
24318         and -mstring-compare-inline-limit.
24320 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24322         PR target/85287
24323         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
24324         for stack clash protection in a register whenever we need it to be in
24325         a register.
24327 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24329         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
24330         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
24332 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
24334         PR target/85321
24335         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
24336         the help text.
24337         (mlong-double-): Ditto.
24338         * config/rs6000/sysv4.opt (msdata=): Ditto.
24339         (mtls-size=): Ditto.
24341 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24343         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
24344         erroneous entries for
24345         "vector int vec_ldl (int, long int *)", and
24346         "vector unsigned int vec_ldl (int, unsigned long int *)".
24347         Add comments and entries for
24348         "vector bool char vec_ldl (int, bool char *)",
24349         "vector bool short vec_ldl (int, bool short *)",
24350         "vector bool int vec_ldl (int, bool int *)",
24351         "vector bool long long vec_ldl (int, bool long long *)",
24352         "vector pixel vec_ldl (int, pixel *)",
24353         "vector long long vec_ldl (int, long long *)",
24354         "vector unsigned long long vec_ldl (int, unsigned long long *)".
24355         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
24356         type tree bool_long_long_type_node and correct definition of
24357         bool_V2DI_type_node to make reference to this new type tree.
24358         (rs6000_mangle_type): Replace erroneous reference to
24359         bool_long_type_node with bool_long_long_type_node.
24360         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
24361         comments to emphasize sign distinctions for char and int types and
24362         replace RS6000_BTI_bool_long constant with
24363         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
24364         use of RS6000_BTI_pixel.
24365         (bool_long_type_node): Remove this macro definition.
24366         (bool_long_long_type_node): New macro definition
24368 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
24370         PR rtl-optimization/85300
24371         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
24372         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
24373         simplify_unary_operation fails.
24375 2018-04-10  Martin Liska  <mliska@suse.cz>
24377         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
24378         cgraph_edge and ipa_ref.
24380 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
24382         PR target/85177
24383         PR target/85255
24384         * config/i386/sse.md
24385         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
24386         computation of the VEC_MERGE selector from mask.
24387         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
24388         Fix decoding of the VEC_MERGE selector into mask.
24390 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
24392         PR tree-optimization/85286
24393         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
24395 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
24397         * final.c (final_1): Set insn_last_address as well as
24398         insn_current_address.
24400 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24402         PR target/85173
24403         * explow.c (emit_stack_probe): Call validize_mem on memory location
24404         before passing it to gen_probe_stack.  Create address operand and
24405         legitimize it for the probe_stack_address case.
24407 2018-04-09  Jan Hubicka  <jh@suse.cz>
24409         PR lto/85078
24410         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
24411         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
24412         * tree.c (free_lang_data_in_type): Fix handling of binfos;
24413         walk basetypes.
24414         (free_lang_data): Rebuild type inheritance graph.
24416 2018-04-09  Martin Sebor  <msebor@redhat.com>
24418         * invoke.texi (-finline-small-functions): Mention other optimization
24419         options.
24420         (-findirect-inlining, -fpartial-inlining): Same.
24421         (-finline-functions-called-once): Same.
24422         (-freorder-blocks-and-partition): Same.
24424 2018-04-09  Jan Hubicka  <jh@suse.cz>
24426         PR rtl/84058
24427         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
24428         jumps; choose last target that matches the criteria (i.e.
24429         no partition changes for non-crossing jumps).
24430         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
24431         support for redirecting crossing jumps to non-crossing.
24433 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
24435         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
24436         also for naked functions.
24438 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
24440         * config/arc/arc.md (add_shift): New pattern.
24441         (add_shift2): Likewise.
24442         (sub_shift): Likewise.
24443         (sub_shift_cmp0_noout): Likewise.
24444         (compare_si_ashiftsi): Likewise.
24445         (xbfu_cmp0_noout): New combine pattern.
24446         (xbfu_cmp0"): Likewise.
24447         (movsi_set_cc_insn): Place the predicable variant first.
24448         (commutative_binary_cmp0_noout): Remove clobber.
24449         (commutative_binary_cmp0): New pattern.
24450         (noncommutative_binary_cmp0): Likewise.
24451         (noncommutative_binary_cmp0_noout): Likewise.
24452         (noncommutative_binary_comparison_result_used): Removed.
24453         (rsub_cmp0): New pattern.
24454         (rsub_cmp0_noout): Likewise.
24455         (extzvsi): Changed, keep only meaningful variants.
24456         (SQH, SEZ): New iterators.
24457         (SQH_postfix): New mode attribute.
24458         (SEZ_prefix): New code attribute.
24459         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
24460         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
24461         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
24462         of numerical value.
24463         (noncommutative_operator): Check the availability of barrel
24464         shifter option.
24466 2018-04-09  Richard Biener  <rguenther@suse.de>
24468         PR tree-optimization/85284
24469         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
24470         Only use the niter constraining form of simple_iv when the exit
24471         is always executed.
24473 2018-04-09  Tom de Vries  <tom@codesourcery.com>
24475         PR target/84041
24476         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
24477         (define_expand "*memory_barrier"): New define_expand.
24478         (define_insn "memory_barrier"): New insn.
24480 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24482         PR rtl-optimization/80463
24483         PR rtl-optimization/83972
24484         PR rtl-optimization/83480
24486         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
24487         correct producer for the insn.
24488         (tidy_control_flow): Fixup seqnos in case of debug insns.
24490 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24492         PR rtl-optimization/83913
24494         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
24495         different sched-times when merging exprs.
24497 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24499         PR rtl-optimization/83962
24501         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
24502         tidy_fallthru_edge and tidy_control_flow.
24504 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
24506         PR rtl-optimization/83530
24508         * sel-sched.c (force_next_insn): New global variable.
24509         (remove_insn_for_debug): When force_next_insn is true, also leave only
24510         next insn in the ready list.
24511         (sel_sched_region): When the region wasn't scheduled, make another pass
24512         over it with force_next_insn set to 1.
24514 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
24516         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
24517         into tm_file.
24518         * config/nds32/constants.md (unspec_volatile_element): Add enum values
24519         for interrupt control.
24520         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
24521         functions for interrupt control.
24522         * config/nds32/nds32-intrinsic.md: Likewise.
24523         * config/nds32/nds32_intrinsic.h: Likewise.
24524         * config/nds32/nds32.h (nds32_builtins): Likewise.
24526 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
24528         * config/nds32/nds32.c (nds32_init_machine_status,
24529         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
24530         strict_aligned_p field.
24531         (nds32_expand_to_rtl_hook): New function.
24532         (TARGET_EXPAND_TO_RTL_HOOK): Define.
24533         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
24535 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24536             Chung-Ju Wu  <jasonwucj@gmail.com>
24538         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
24539         * config/nds32/nds32-n7.md: New file.
24540         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
24541         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
24542         pipeline.
24543         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
24544         * config/nds32/nds32.md (pipeline_model): Add n7.
24545         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
24546         * config/nds32/pipelines.md: Include n7 settings.
24548 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24549             Chung-Ju Wu  <jasonwucj@gmail.com>
24551         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
24552         * config/nds32/nds32-e8.md: New file.
24553         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
24554         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
24555         pipeline.
24556         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
24557         * config/nds32/nds32.md (pipeline_model): Add e8.
24558         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
24559         * config/nds32/pipelines.md: Include e8 settings.
24561 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24562             Chung-Ju Wu  <jasonwucj@gmail.com>
24564         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
24565         * config/nds32/nds32-n8.md: New file.
24566         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
24567         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
24568         pipeline.
24569         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
24570         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
24571         * config/nds32/nds32.md (pipeline_model): Add n8.
24572         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
24573         * config/nds32/pipelines.md: Include n8 settings.
24575 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
24576             Chung-Ju Wu  <jasonwucj@gmail.com>
24578         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
24579         * config/nds32/nds32-n9-2r1w.md: New file.
24580         * config/nds32/nds32-n9-3r2w.md: New file.
24581         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
24582         nds32_register_ports): New or modify for cpu n9.
24583         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
24584         pipeline.
24585         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
24586         * config/nds32/nds32-utils.c: New file.
24587         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
24588         TARGET_MUL_SLOW): Define.
24589         * config/nds32/nds32.md (pipeline_model): New attribute.
24590         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
24591         New options that support cpu n9.
24592         * config/nds32/pipelines.md: Include n9 settings.
24593         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
24595 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
24597         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
24598         information if necessary.
24599         (output_cond_branch_compare_zero): Likewise.
24600         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
24601         (nds32_target_alignment): Refine for alignment.
24602         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
24603         (FUNCTION_BOUNDARY): Modify.
24604         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
24605         align case.
24606         * config/nds32/nds32.opt (malways-align, malign-functions): New.
24608 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
24610         * config/nds32/constants.md (unspec_volatile_element): Add values for
24611         TLB operation and data prefetch.
24612         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
24613         functions for TLB operation and data prefetch.
24614         * config/nds32/nds32-intrinsic.md: Likewise.
24615         * config/nds32/nds32_intrinsic.h: Likewise.
24616         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
24617         (nds32_print_operand): Likewise.
24618         * config/nds32/nds32.h (nds32_builtins): Likewise.
24620 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
24621         Andrew Pinski <pinsika@gcc.gnu.org>
24623         PR middle-end/82976
24624         * match.pd: Use constant_boolean_node of correct type instead of
24625         boolean_true_node or boolean_false_node for simplifying
24626         pointer comparisons to zero.
24628 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
24630         PR tree-optimization/80021
24631         * tree.c (verify_type_variant): Make error call in verify_variant_match
24632         translatable and remove final full stop.
24634 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24636         * config/nds32/constants.md (unspec_volatile_element): Add
24637         UNSPEC_VOLATILE_EH_RETURN.
24638         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
24639         nds32_output_stack_pop): Support dwarf exception handling process.
24640         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
24641         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
24642         exception handling process.
24643         (nds32_compute_stack_frame): Likewise.
24644         (nds32_return_addr_rtx): Likewise.
24645         (nds32_initial_elimination_offset): Likewise.
24646         (nds32_expand_prologue): Likewise.
24647         (nds32_expand_epilogue): Likewise.
24648         (nds32_dynamic_chain_address): New function.
24649         * config/nds32/nds32.h (machine_function): Add fields for dwarf
24650         exception handling.
24651         (DYNAMIC_CHAIN_ADDRESS): Define.
24652         (EH_RETURN_DATA_REGNO): Define.
24653         (EH_RETURN_STACKADJ_RTX): Define.
24654         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
24655         patterns for dwarf exception handling.
24657 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24659         * config/nds32/nds32.h: Clean up obsolete macros.
24661 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24663         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
24664         Add enum values for particular instructions.
24665         * config/nds32/nds32-intrinsic.c: Implementation of expanding
24666         particular intrinsic functions.
24667         * config/nds32/nds32-intrinsic.md: Likewise.
24668         * config/nds32/nds32_intrinsic.h: Likewise.
24669         * config/nds32/nds32.h (nds32_builtins): Likewise.
24670         * config/nds32/nds32.md (type): Add pbsad and pbsada.
24671         (btst, ave): New patterns for particular instructions.
24673 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24675         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
24676         Add enum values for atomic load/store and memory sync.
24677         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
24678         and memory sync.
24679         * config/nds32/nds32-intrinsic.md: Likewise.
24680         * config/nds32/nds32_intrinsic.h: Likewise.
24681         * config/nds32/nds32.h (nds32_builtins): Likewise.
24683 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
24685         PR tree-optimization/85257
24686         * fold-const.c (native_encode_vector): If not all elts could fit
24687         and off is -1, return 0 rather than offset.
24688         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
24689         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
24690         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
24691         adjust buffer in native_interpret_expr call.
24693 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24695         * config/nds32/constants.md (unspec_volatile_element): Add cache
24696         control enum values.
24697         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
24698         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
24699         * config/nds32/nds32.c (nds32_cctl_names): New.
24700         (nds32_print_operand): Handle cache control register names.
24701         * config/nds32/nds32.h (nds32_builtins): New enum values.
24702         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
24703         macros.
24704         * config/nds32/nds32.md (type): Add mmu.
24705         * config/nds32/pipelines.md (simple_insn): Add mmu.
24707 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24709         * config/nds32/nds32.md (type): Remove call.
24710         * config/nds32/pipelines.md (simple_insn): Likewise.
24712 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24714         * config/nds32/constants.md (unspec_volatile_element): Add
24715         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
24716         UNSPEC_VOLATILE_FMFCFG.
24717         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
24718         description for fmfcfg and fmfcsr.
24719         (bdesc_1arg): Add fmtcsr.
24720         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
24721         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
24722         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
24723         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
24724         unspec_fmfcfg): New patterns.
24725         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
24726         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
24727         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
24728         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
24729         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
24730         __nds32__fmfcfg): Define.
24732 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
24734         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
24735         intrinsic register names.
24736         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
24737         intrinsic register enum values and macros.
24739 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
24741         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
24742         for load/store addressing form.
24743         (nds32_print_operand_address): Likewise.
24745 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
24747         PR target/85196
24748         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
24749         based on LABEL_REF.  Remove useless assertion.
24750         (pic_address_needs_scratch): Fix formatting.
24751         (sparc_legitimize_pic_address): Minor tweaks.
24752         (sparc_delegitimize_address): Adjust assertion accordingly.
24753         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
24754         into symbolic_operand.
24755         (movsi_high_pic_label_ref): Likewise.
24756         (movsi_lo_sum_pic_label_ref): Likewise.
24757         (movdi_pic_label_ref): Likewise.
24758         (movdi_high_pic_label_ref): Likewise.
24759         (movdi_lo_sum_pic_label_ref): Likewise.
24761 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
24763         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
24764         custom LIB_SPEC setup.
24766 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
24767             Kito Cheng  <kito.cheng@gmail.com>
24769         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
24770         * config/riscv/freebsd.h: New.
24772 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
24774         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
24775         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
24776         file.
24778 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
24779             Kito Cheng  <kito.cheng@gmail.com>
24781         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
24782         nds32_output_call, nds32_symbol_binds_local_p): New functions.
24783         * config/nds32/nds32-protos.h (nds32_output_call,
24784         nds32_output_return): Declare.
24785         * config/nds32/nds32.md: Refine all the call and return patterns.
24787 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
24789         PR debug/85252
24790         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
24791         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
24793         PR rtl-optimization/84872
24794         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
24795         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
24796         EDGE_CROSSING edge.
24798 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
24800         * expr.c (copy_blkmode_to_reg): Revert 254862.
24801         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
24803 2018-04-06  Richard Biener  <rguenther@suse.de>
24805         PR middle-end/85244
24806         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
24807         after seeing a component reference with an adjacent field.  Treat
24808         refs to arrays at struct end of external decls similar to
24809         refs to unconstrained commons.
24811 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
24813         PR sanitizer/85213
24814         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
24815         look through SAVE_EXPRs with non-side-effects argument.  Adjust
24816         recursive calls.
24817         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
24818         save_p here.
24820 2018-04-06  Richard Biener  <rguenther@suse.de>
24822         PR middle-end/85180
24823         * alias.c (find_base_term): New wrapper around find_base_term
24824         unwinding CSELIB_VAL_PTR changes.
24825         (find_base_term): Do not restore CSELIB_VAL_PTR during the
24826         recursion.
24828 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24830         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
24831         instructions.
24832         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
24833         constant definitions.
24834         ("nop"): lr 0,0 -> nopr r0
24835         ("nop_lr0", "nop_lr1"): New insn definitions.
24837 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
24839         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
24840         NDS32_V3PUSH_AVAILABLE_P macro.
24842 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
24843             Chung-Ju Wu  <jasonwucj@gmail.com>
24845         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
24846         (nds32*-*-*): Add float and fpu_config into supported_defaults.
24847         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
24848         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
24849         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
24850         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
24851         * config/nds32/constraints.md: New constraints and checking for hard
24852         float configuration.
24853         * config/nds32/iterators.md: New mode iterator and attribute for hard
24854         float configuration.
24855         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
24856         patterns.
24857         * config/nds32/nds32-fpu.md: New file.
24858         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
24859         deal with hard float code generation.
24860         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
24861         ARCH_V3S.
24862         (abi_type, float_reg_number): New enum type.
24863         * config/nds32/nds32-predicates.c: New predicates for hard float.
24864         * config/nds32/nds32-protos.h: Declare functions for hard float.
24865         * config/nds32/nds32.c: Implementation for hard float configuration.
24866         * config/nds32/nds32.h: Definitions for hard float configuration.
24867         * config/nds32/nds32.md: Include hard float machine description and
24868         modify patterns for hard float configuration.
24869         * config/nds32/nds32.opt: New options for hard float configuration.
24870         * config/nds32/predicates.md: New predicates for hard float
24871         configuration.
24873 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
24875         * common/config/nds32/nds32-common.c
24876         (nds32_option_optimization_table): Enable -mreleax-hint by default.
24878 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
24880         PR middle-end/85195
24881         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
24882         CONSTRUCTOR_ELT (ctor, ...)->value.
24884 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
24886         PR target/85193
24887         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
24889 2018-04-05  Tom de Vries  <tom@codesourcery.com>
24891         PR target/85204
24892         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
24893         cond jump.
24895 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
24896             Kito Cheng  <kito.cheng@gmail.com>
24898         * config/nds32/constraints.md (U33): Fine-tune checking condition.
24899         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
24900         * config/nds32/nds32.h (nds32_16bit_address_type): Add
24901         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
24903 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
24904             Kito Cheng  <kito.cheng@gmail.com>
24906         * config/nds32/constraints.md (Ufe): New memory constraint.
24907         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
24908         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
24909         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
24910         operands.
24911         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
24912         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
24914 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24916         * config/nds32/nds32.md: Use optimize_size in the condition for
24917         alu-shift instructions.
24919 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24921         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
24923 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24925         * config/nds32/nds32.md (negsi2): Refine pattern.
24927 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
24928             Chung-Ju Wu  <jasonwucj@gmail.com>
24930         * config/nds32/iterators.md (shift_rotate): New code iterator.
24931         (shift): New code attribute.
24932         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
24933         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
24934         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
24935         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
24936         bit-wise operations.
24937         (andsi3, *andsi3): Ditto.
24938         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
24939         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
24940         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
24941         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
24942         nds32_ior_operand, nds32_xor_operand): New predicates.
24944 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24946         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
24947         (addsi3, subsi3): ... this.
24949 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24951         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
24953 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24955         * config/nds32/nds32.md: Adjust indention.
24957 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
24959         * config/nds32/nds32.md (feature): New attribute.
24961 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
24963         * config/nds32/nds32.md (subtype): New attribute.
24965 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24967         PR target/85203
24968         * config/arm/arm-builtins.c (arm_expand_builtin): Change
24969         expansion to perform a bitwise AND of the argument followed by a
24970         boolean negation of the result.
24972 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
24974         PR rtl-optimization/84878
24975         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
24976         the basic block.  Assert the use reference is not artificial and that
24977         it has an associated insn.
24979 2018-04-04  Michael Matz  <matz@suse.de>
24981         * builtins.c (compute_objsize): Pass correct operand
24982         to array_at_struct_end_p.
24984 2018-04-04  Richard Biener  <rguenther@suse.de>
24986         PR lto/85176
24987         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
24988         from contexts for DINFO_LEVEL_TERSE and below.
24990 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
24992         * config/nds32/nds32-doubleword.md (move_<mode>): Require
24993         resiter_operand condition.
24994         * config/nds32/nds32.md (*move<mode>): Ditto.
24996 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
24997             Monk Chiang  <sh.chiang04@gmail.com>
24999         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
25001 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25003         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
25005 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25006             Kito Cheng  <kito.cheng@gmail.com>
25008         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
25009         nds32_cond_code_str, output_cond_branch,
25010         output_cond_branch_compare_zero, nds32_expand_cbranch,
25011         nds32_expand_cstore, nds32_expand_movcc,
25012         nds32_output_cbranchsi4_equality_zero,
25013         nds32_output_cbranchsi4_equality_reg,
25014         nds32_output_cbranchsi4_equality_reg_or_const_int,
25015         nds32_output_cbranchsi4_greater_less_zero: New functions.
25016         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
25017         nds32_expand_cstore, nds32_expand_movcc,
25018         nds32_output_cbranchsi4_equality_zero,
25019         nds32_output_cbranchsi4_equality_reg,
25020         nds32_output_cbranchsi4_equality_reg_or_const_int,
25021         nds32_output_cbranchsi4_greater_less_zero): Declare.
25022         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
25023         nds32_rimm11s_operand): New predicates.
25024         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
25025         * config/nds32/nds32.md: Rewrite all the branch and conditional move
25026         patterns.
25028 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
25030         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
25031         * config/nds32/nds32.md: Ditto.
25032         * config/nds32/pipelines.md: Ditto.
25034 2018-04-04  Richard Biener  <rguenther@suse.de>
25036         PR tree-optimization/85168
25037         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
25038         propagating abnormals.
25040 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25042         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
25044 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
25045             Kito Cheng  <kito.cheng@gmail.com>
25047         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
25048         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
25049         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
25050         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
25051         * config/nds32/nds32.md (sibcall_internal): New.
25052         (sibcall_register): Remove.
25053         (sibcall_immediate): Remove.
25054         (sibcall_value_internal): New.
25055         (sibcall_value_register): Remove.
25056         (sibcall_value_immediate): Remove.
25057         * config/nds32/predicates.md (nds32_general_register_operand): New.
25058         (nds32_call_address_operand): New.
25060 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
25062         PR rtl-optimization/85167
25063         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
25064         bb_defs if *split_p, instead preinitialize it to NULL.
25066         PR tree-optimization/85156
25067         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
25068         evaluating the argument multiple times.
25070 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
25072         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
25073         than vector.
25074         (_mm_cvtpd_ps): Likewise.
25075         (_mm_cvttpd_epi32): Likewise.
25076         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
25077         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
25078         vector, pixel, and bool following altivec.h include.
25080 2018-04-03  Martin Sebor  <msebor@redhat.com>
25082         * doc/extend.texi (Common Function Attributes): Clarify.
25083         (const attribute): Likewise.
25084         (pure attribute): Likewise.
25086 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
25088         PR target/85169
25089         * config/i386/i386.c (ix86_expand_vector_set): Use
25090         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
25092 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
25094         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
25095         instructions when changing rounding bits to preserve precision bits
25096         in the x87 control word.
25098 2018-04-03  Martin Liska  <mliska@suse.cz>
25100         PR tree-optimization/82491
25101         * rtl.h (strip_offset_and_add): Replace += suboffset with
25102         poly_uint64 () + suboffset.
25104 2018-03-29  Martin Liska  <mliska@suse.cz>
25105             Martin Jambor  <mjambor@suse.cz>
25107         PR ipa/84947
25108         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
25109         param_type is not an integral or pointer type.
25111 2018-04-03  Richard Biener  <rguenther@suse.de>
25113         * sese.h (recompute_all_dominators): Remove.
25115 2018-04-02  Martin Sebor  <msebor@redhat.com>
25117         * doc/invoke.texi (-Wrestrict): Fix typos.
25119 2018-04-02  Jim Wilson  <jimw@sifive.com>
25121         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
25122         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
25123         (<optab>di3, <optab>si3_extend): Likewise.
25124         (<optab>si3_mask, <optab>si3_mask_1): New.
25125         (<optab>di3_mask, <optab>di3_mask_1): New.
25126         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
25127         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
25128         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
25130 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
25132         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
25133         example.
25135 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
25137         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
25138         (nds32_canonicalize_comparison): New function.
25140 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25141             Kito Cheng  <kito.cheng@gmail.com>
25142             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
25144         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
25145         * config/nds32/constants.md (unspec_volatile_element): Add
25146         UNSPEC_VOLATILE_RELAX_GROUP.
25147         * config/nds32/nds32-relax-opt.c: New file.
25148         * config/nds32/nds32-predicates.c
25149         (nds32_symbol_load_store_p): New function.
25150         * config/nds32/nds32-protos.h
25151         (nds32_symbol_load_store_p): Declare function.
25152         (make_pass_nds32_relax_opt): Declare new rtl pass function.
25153         * config/nds32/nds32.c
25154         (nds32_register_pass): New function to register pass.
25155         (nds32_register_passes): New function to register passes.
25156         * config/nds32/nds32.md (relax_group): New pattern.
25157         * config/nds32/nds32.opt (mrelax-hint): New option.
25158         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
25160 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
25162         * config/nds32/t-nds32: Modify files dependency.
25164 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25166         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
25167         (PROFILE_HOOK): Define its implementation.
25169 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
25171         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
25172         type and 32-bit size.
25174 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
25176         PR middle-end/85090
25177         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
25178         (V_128_256): New mode iterator.
25179         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
25180         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
25181         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
25182         of V.
25183         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
25184         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
25186 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
25188         PR target/83315
25189         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
25190         NaN inputs correctly.
25192 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
25194         PR target/80546
25195         * config/rs6000/vsx.md (??r): New mode attribute.
25196         (*vsx_mov<mode>_64bit): Use it.
25197         (*vsx_mov<mode>_32bit): Likewise.
25199 2018-03-30  Martin Sebor  <msebor@redhat.com>
25201         PR tree-optimization/84818
25202         * builtins.c (check_access): Use warning_n.
25204 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25206         PR target/83822
25207         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
25208         condition.
25209         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
25210         condition.
25212 2018-03-30  Julia Koval  <julia.koval@intel.com>
25214         PR target/84413
25215         * x86-tune.def (movx, partial_reg_dependency): Enable for
25216         m_SKYLAKE_AVX512.
25218 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
25220         PR inline-asm/84985
25221         * lra-constraints.c (process_alt_operands): Move setting
25222         this_alternative_matches below.
25224 2018-03-29  Martin Liska  <mliska@suse.cz>
25226         PR lto/84995.
25227         * doc/invoke.texi: Document how LTO works with debug info.
25228         Describe auto-load support of binutils.  Mention 'x86-64'
25229         as valid option value of -march option.
25231 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
25233         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
25235         PR c/85094
25236         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
25237         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
25238         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
25239         checking.
25241 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
25243         PR target/84912
25244         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
25245         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
25246         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
25247         for RS6000_BTM_POWERPC64.
25248         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
25249         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
25250         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
25251         definition.
25252         (DIVDE): Use it.
25253         (DIVDEU): Likewise.
25255 2018-03-28  Carl Love  <cel@us.ibm.com>
25257         Revert
25258         2017-09-27  Carl Love  <cel@us.ibm.com>
25260         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
25261         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
25262         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
25263         fctiw instruction.
25265 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25267         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
25268         instead of __vector bool.
25269         (_mm_max_pu8): Likewise.
25270         (_mm_min_pi16): Likewise.
25272 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
25274         PR target/84912
25275         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
25276         (DIVWEUO): Likewise.
25277         (DIVDEO): Likewise.
25278         (DIVDEUO): Likewise.
25279         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
25280         DIVWEUO and DIVDEUO.
25281         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
25282         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
25283         (div_extend): Likewise.
25284         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
25285         builtin function.
25286         (__builtin_divweuo): Likewise.
25287         (__builtin_divdeo): Likewise.
25288         (__builtin_divdeuo): Likewise.
25290 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
25292         PR target/85095
25293         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
25294         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
25296         PR tree-optimization/82004
25297         * gimple-match-head.c (optimize_pow_to_exp): New function.
25298         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
25299         Don't fold to exp if optimize_pow_to_exp is false.
25301 2018-03-28  Martin Liska  <mliska@suse.cz>
25303         PR other/84819
25304         * calls.c (initialize_argument_information): Fix trailing space.
25305         * common.opt: Fix typo and provide better explanation for
25306         -fsanitize-coverage option.
25307         * config/i386/i386.opt: Fix typo.
25309 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
25310             Martin Liska  <mliska@suse.cz>
25312         PR sanitizer/85081
25313         * gimplify.c (asan_poison_variable): Don't do the check for
25314         gimplify_omp_ctxp here.
25315         (gimplify_decl_expr): Do it here.
25316         (gimplify_target_expr): Likewise.
25318 2018-03-28  Martin Liska  <mliska@suse.cz>
25320         PR target/84988
25321         * config/i386/i386.c (ix86_function_arg_advance): Do not call
25322         chkp_type_bounds_count if MPX is not enabled.
25324 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
25326         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
25328 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
25330         PR target/84914
25331         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
25332         function to create the function decl for complex long double
25333         multiply and divide for -mabi=ieeelongdouble.
25334         (init_float128_ieee): Call it.
25336 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
25338         PR target/85044
25339         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
25340         -fcf-protection=branch -mibt.
25341         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
25343 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25345         PR target/81863
25346         * config/arm/arm.c (arm_valid_symbolic_address): Handle
25347         arm_word_relocations.
25349 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
25351         PR target/85056
25352         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
25353         extern array declarations.
25355 2018-03-27  Richard Biener  <rguenther@suse.de>
25357         PR middle-end/84067
25358         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
25359         explicit single_use checks.
25361 2018-03-27  Richard Biener  <rguenther@suse.de>
25363         PR tree-optimization/85082
25364         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
25365         Valueize the VUSE.
25367 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25369         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
25370         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
25371         Turn on fasynchronous-unwind-tables and funwind-tables.
25373 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
25375         PR target/85073
25376         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
25377         (*bmi_blsr_<mode>_ccz): Ditto.
25379 2018-03-26  Tom de Vries  <tom@codesourcery.com>
25381         PR tree-optimization/85063
25382         * omp-general.c (offloading_function_p): New function.  Factor out
25383         of ...
25384         * omp-offload.c (pass_omp_target_link::gate): ... here.
25385         * omp-general.h (offloading_function_p): Declare.
25386         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
25387         with attribute omp declare target for offloading functions.
25389 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
25391         PR tree-optimization/84005
25392         * tree-data-ref.h (get_base_for_alignment): Declare.
25393         * tree-data-ref.c (get_base_for_alignment_1): New function.
25394         (get_base_for_alignment): Likewise.
25395         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
25396         get_base_for_alignment to find a suitable base object, instead
25397         of always using drb->base_address.
25399 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
25401         PR inline-asm/85022
25402         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
25403         known size by default.
25405 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
25407         PR inline-asm/85030
25408         * lra-constraints.c (process_alt_operands): Don't match BLKmode
25409         and non BLKmode operands.
25411 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25413         PR target/85026
25414         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
25415         Clean up attributes.
25417 2018-03-23  Richard Biener  <rguenther@suse.de>
25419         PR debug/85020
25420         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
25421         we are going to emit early debug for LTO.
25423 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
25425         PR inline-asm/85034
25426         * function.c (match_asm_constraints_1): Don't optimize if input
25427         doesn't satisfy general_operand predicate for output's mode.
25429         PR inline-asm/85022
25430         * alias.c (write_dependence_p): Don't require for x_canonicalized
25431         non-VOIDmode if x has VOIDmode.
25433         PR sanitizer/85029
25434         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
25435         just don't try to optimize it rather than assert it never happens.
25437 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25439         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
25440         macro expansions for definition of ST_INTERNAL_<mode> and
25441         LD_INTERNAL_<mode> builtins.
25442         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
25443         Remove prototype.
25444         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
25445         function.
25446         (altivec_expand_st_builtin): Likewise.
25447         (altivec_expand_builtin): Remove calls to deleted functions.
25448         (rs6000_address_for_altivec): Delete this function.
25449         * config/rs6000/vector.md: Remove expands for
25450         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
25452 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
25454         PR target/84826
25455         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
25456         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
25457         re-computing once computed.
25458         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
25459         (arm_init_machine_status): Initialize
25460         machine->static_chain_stack_bytes.
25462 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25464         PR target/84760
25465         * doc/extend.texi: Add four new prototypes for vec_ld.
25466         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
25467         definitions for more logical presentation.
25468         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
25469         entries for V1TI variants of __builtin_altivec_ld builtin.
25470         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
25471         handling of V1TI variant of LVX icode pattern.
25472         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
25473         (rs6000_gimple_fold_builtin): Likewise.
25474         (altivec_init_builtins): Add code to define
25475         __builtin_altivec_lvx_v1ti function.
25477 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
25479         PR inline-asm/84941
25480         * function.c (match_asm_constraints_1): Don't do the optimization
25481         if input isn't a REG, SUBREG, MEM or constant.
25483 2018-03-22  Tom de Vries  <tom@codesourcery.com>
25485         PR tree-optimization/84956
25486         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
25487         bb_has_abnormal_pred.
25489 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
25491         PR sanitizer/85018
25492         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
25493         DECL_INITIAL (decl) to decl at the end.
25494         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
25495         adjust the comment.
25497 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
25499         * doc/extend.texi (__builtin_tgmath): Document when complex
25500         integer types are treated as _Complex _Float64.
25502 2018-03-21  Tom de Vries  <tom@codesourcery.com>
25504         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
25506 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
25508         PR tree-optimization/84960
25509         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
25510         if it is ENTRY block, move them into single succ of ENTRY in that case.
25512 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
25514         PR tree-optimization/84811
25515         * poly-int.h (poly_span_traits): Remove the T3 parameter and
25516         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
25517         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
25518         (known_subrange_p): Update accordingly.  Cast each value involved
25519         in the size comparison, rather than casting the result of the
25520         subtraction.
25522 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
25524         PR tree-optimization/84982
25525         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
25526         by flipping the least significant bit rather than all bits from
25527         bitpos to bitpos + bitsize - 1.
25529 2018-03-21  Nathan Sidwell  <nathan@acm.org>
25531         * doc/extend.texi (Deprecated Features): Remove mention of
25532         long-deleted deprecations.
25534 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25536         PR jit/84288
25537         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
25538         * configure: Regenerate.
25540 2018-03-21  Tom de Vries  <tom@codesourcery.com>
25542         PR tree-optimization/83126
25543         * tree-parloops.c (num_phis): New function.
25544         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
25546 2018-03-21  Nathan Sidwell  <nathan@acm.org>
25548         * doc/extend.texi (Deprecated Features): Update deprecated flags,
25549         mention anon-struct/union members and trailing attributes.
25551 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
25553         PR tree-optimization/84969
25554         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
25555         builtin memset partitions if they set different rhs values.
25557 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
25559         PR rtl-optimization/84989
25560         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
25561         VEC_DUPLICATE with scalar result mode.
25563 2018-03-21  Martin Liska  <mliska@suse.cz>
25565         PR ipa/84963
25566         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
25567         not intended return statement.
25569 2018-03-21  Martin Liska  <mliska@suse.cz>
25571         PR target/84988
25572         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
25573         (chkp_find_bound_slots_1): Limit number of iterations.
25575 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
25577         PR target/84838
25578         * Minor grammar fixes for x86 options.
25580 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
25582         PR debug/84875
25583         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
25584         holding REG_CFA_RESTORE notes, instead turn them into a USE.
25586 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
25588         PR target/83789
25589         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
25590         (altivec_lvx_<mode>_1op): Likewise.
25591         (altivec_stvx_<mode>_2op): Likewise.
25592         (altivec_stvx_<mode>_1op): Likewise.
25593         (altivec_lvx_<VM2:mode>): New define_expand.
25594         (altivec_stvx_<VM2:mode>): Likewise.
25595         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
25596         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
25597         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
25598         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
25599         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
25600         (rs6000_gen_lvx): Likewise.
25601         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
25602         (altivec_expand_stv_builtin): Likewise.
25603         (altivec_expand_builtin): Likewise.
25604         * config/rs6000/vector.md: Likewise.
25606 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25608         PR target/82518
25609         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
25610         BYTES_BIG_ENDIAN.
25612 2018-03-20  Richard Biener  <rguenther@suse.de>
25614         PR target/84986
25615         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
25616         sign-conversions as zero, fall back to standard scalar_stmt
25617         cost for the rest.
25619 2018-03-20  Martin Liska  <mliska@suse.cz>
25621         PR ipa/84825
25622         * predict.c (rebuild_frequencies): Handle case when we have
25623         PROFILE_ABSENT, but flag_guess_branch_prob is false.
25625 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
25627         PR target/84990
25628         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
25629         flag_section_anchors.
25630         * varasm.c (use_blocks_for_decl_p): Remove hack for
25631         dw2_force_const_mem.
25633         PR target/84845
25634         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
25635         to ...
25636         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
25637         be created, use lowpart_subreg of operands[0] rather than operands[0]
25638         itself.
25639         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
25640         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
25641         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
25642         and n constraint instead of aarch64_shift_imm_di and Usd.
25643         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
25644         (*aarch64_<optab>_reg_minus<mode>3): ... this.
25646 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
25648         PR target/82989
25649         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
25650         to favor GPR over NEON registers.
25651         (<shift>di3_neon): Likewise.
25653 2018-03-20  Tom de Vries  <tom@codesourcery.com>
25655         PR target/84952
25656         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
25657         (nvptx_process_pars): Emit bar.sync asap and alap.
25659 2018-03-20  Tom de Vries  <tom@codesourcery.com>
25661         PR target/84954
25662         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
25663         seen_label if seen_label is already set.
25665 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
25667         PR target/84945
25668         * config/i386/i386.c (fold_builtin_cpu): For features above 31
25669         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
25670         Use 1U instead of 1.  Formatting fixes.
25672         PR c/84953
25673         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
25674         instead of TREE_TYPE (s1) for the return value.
25676 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
25678         PR tree-optimization/84946
25679         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
25680         bitsize + bitsize in poly_uint64 rather than poly_int64.
25682         PR sanitizer/78651
25683         * dwarf2asm.c: Include fold-const.c.
25684         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
25685         of decl rather than decl itself.
25687         PR rtl-optimization/84643
25688         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
25690 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
25692         PR sanitizer/78651
25693         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
25694         calling assemble_variable.
25696 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
25698         PR target/81647
25699         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
25700         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
25702 2018-03-19  Jim Wilson  <jimw@sifive.com>
25704         PR bootstrap/84856
25705         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
25706         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
25707         (riscv_first_stack_step): Likewise.
25708         (riscv_option_override): Use STACK_BOUNDARY instead of
25709         MIN_STACK_BOUNDARY.
25710         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
25711         MIN_STACK_BOUNDARY.
25712         (BIGGEST_ALIGNMENT): Set to 128.
25713         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
25714         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
25715         STACK_BOUNDARY.
25717 2018-03-19  Richard Biener  <rguenther@suse.de>
25719         PR tree-optimization/84933
25720         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
25721         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
25723 2018-03-19  Richard Biener  <rguenther@suse.de>
25725         PR tree-optimization/84859
25726         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
25727         (cond_if_else_store_replacement): Perform sinking operation on
25728         single-store BBs regardless of MAX_STORES_TO_SINK setting.
25729         Generalize what a BB with a single eligible store is.
25731 2018-03-19  Richard Biener  <rguenther@suse.de>
25733         PR tree-optimization/84929
25734         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
25735         chrec_is_positive against non-chrec arg.
25737 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
25739         PR target/84711
25740         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
25742 2018-03-18  Martin Liska  <mliska@suse.cz>
25744         PR rtl-optimization/84635
25745         * regrename.c (build_def_use): Use matches_mode only when
25746         matches >= 0.
25748 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
25750         PR tree-optimization/84913
25751         * tree-vect-loop.c (vectorizable_reduction): Don't try to
25752         vectorize chains of COND_EXPRs.
25754 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
25756         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
25758 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
25760         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
25762 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
25764         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
25766 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
25767             Kito Cheng  <kito.cheng@gmail.com>
25769         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
25770         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
25771         (nds32_adjust_reg_alloc_order): New function.
25772         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
25774 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
25776         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
25777         nds32_print_operand, nds32_print_operand_address): Use
25778         HOST_WIDE_INT_PRINT_DEC instead.
25780 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
25782         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
25784 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
25786         PR target/84902
25787         * config/i386/i386.c (initial_ix86_tune_features,
25788         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
25789         unsigned long long.
25790         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
25791         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
25792         rather than 1u << ix86_tune.  Formatting fix.
25793         (ix86_option_override_internal): Change ix86_arch_mask from
25794         unsigned int to unsigned HOST_WIDE_INT, initialize to
25795         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
25796         (ix86_function_specific_restore): Likewise.
25798 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
25800         PR target/84899
25801         * postreload.c (reload_combine_recognize_pattern): Perform
25802         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
25803         truncate_int_for_mode the result for the destination's mode.
25805         PR c/84909
25806         * hsa-gen.c (mem_type_for_type): Fix comment typo.
25807         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
25808         Likewise.
25809         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
25810         Likewise.
25812 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
25814         PR target/84876
25815         * lra-assigns.c (lra_split_hard_reg_for): Don't use
25816         regno_allocno_class_array and sorted_pseudos.
25817         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
25818         insns where regno is used.
25820 2018-03-16  Martin Liska  <mliska@suse.cz>
25822         PR ipa/84833
25823         * multiple_target.c (create_dispatcher_calls): Redirect
25824         reference in the symbol table.
25826 2018-03-16  Martin Liska  <mliska@suse.cz>
25828         PR ipa/84722
25829         * multiple_target.c (create_dispatcher_calls): Redirect also
25830         an alias.
25832 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
25834         PR c++/79937
25835         PR c++/82410
25836         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
25837         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
25838         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
25840 2018-03-16  Julia Koval  <julia.koval@intel.com>
25842         * doc/invoke.texi (Skylake Server): Add CLWB.
25843         Cannonlake): Remove CLWB.
25845 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
25847         PR tree-optimization/84841
25848         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
25849         1 << 3.
25850         (FLOAT_ONE_CONST_TYPE): Define.
25851         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
25852         (sort_by_operand_rank): Put entries with higher constant_type last
25853         rather than first to match comments.
25855 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
25857         * config/nios2/nios2.md (movsi_internal): Fix thinko in
25858         split predicate.
25860 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
25862         PR c++/79085
25863         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
25864         check and use address of target always.
25866 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
25868         PR target/84574
25869         * config/i386/i386.c (indirect_thunk_needed): Update comments.
25870         (indirect_thunk_bnd_needed): Likewise.
25871         (indirect_thunks_used): Likewise.
25872         (indirect_thunks_bnd_used): Likewise.
25873         (indirect_return_needed): New.
25874         (indirect_return_bnd_needed): Likewise.
25875         (output_indirect_thunk_function): Add a bool argument for
25876         function return.
25877         (output_indirect_thunk_function): Don't generate alias for
25878         function return thunk.
25879         (ix86_code_end): Call output_indirect_thunk_function to generate
25880         function return thunks.
25881         (ix86_output_function_return): Set indirect_return_bnd_needed
25882         and indirect_return_needed instead of indirect_thunk_bnd_needed
25883         and indirect_thunk_needed.
25885 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
25887         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
25888         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
25889         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
25891 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
25892             Paul Hua <paul.hua.gm@gmail.com>
25894         PR c/84852
25895         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
25897 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
25899         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
25900         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
25901         resp. SFmode cases.
25903 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
25905         PR target/84711
25906         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
25907         instead of GET_MODE_SIZE when comparing Units.
25909 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
25911         PR target/68256
25912         * varasm.c (hash_section): Return an unchangeble hash value
25913         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
25914         Return !aarch64_can_use_per_function_literal_pools_p ().
25916 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
25918         PR target/84860
25919         * optabs.c (emit_conditional_move): Pass address of cmode's copy
25920         rather than address of cmode as last argument to prepare_cmp_insn.
25922 2018-03-15  Julia Koval  <julia.koval@intel.com>
25924         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
25925         F_AVX512VNNI, F_AVX512BITALG): New.
25927 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
25929         PR target/83451
25930         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
25931         insn for floating-point loads and stores.
25933 2018-03-14  Carl Love  <cel@us.ibm.com>
25935         * config/rs6000/rs6000-c.c: Add macro definitions for
25936         ALTIVEC_BUILTIN_VEC_PERMXOR.
25937         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
25938         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
25939         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
25940         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
25941         UNSPEC_VPERMXOR.
25942         * config/doc/extend.texi: Add prototypes for vec_permxor.
25944 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
25946         PR c/84852
25947         * diagnostic-show-locus.c (class layout_point): Convert m_line
25948         from int to linenum_type.
25949         (line_span::comparator): Use linenum "compare" function when
25950         comparing line numbers.
25951         (test_line_span): New function.
25952         (layout_range::contains_point): Convert param "row" from int to
25953         linenum_type.
25954         (layout_range::intersects_line_p): Likewise.
25955         (layout::will_show_line_p): Likewise.
25956         (layout::print_source_line): Likewise.
25957         (layout::should_print_annotation_line_p): Likewise.
25958         (layout::print_annotation_line): Likewise.
25959         (layout::print_leading_fixits): Likewise.
25960         (layout::annotation_line_showed_range_p): Likewise.
25961         (struct line_corrections): Likewise for field m_row.
25962         (line_corrections::line_corrections): Likewise for param "row".
25963         (layout::print_trailing_fixits): Likewise.
25964         (layout::get_state_at_point): Likewise.
25965         (layout::get_x_bound_for_row): Likewise.
25966         (layout::print_line): Likewise.
25967         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
25968         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
25969         * input.c (selftest::test_linenum_comparisons): New function.
25970         (selftest::input_c_tests): Call it.
25971         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
25972         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
25973         * selftest.h (ASSERT_GT): New macro.
25974         (ASSERT_GT_AT): New macro.
25975         (ASSERT_LT): New macro.
25976         (ASSERT_LT_AT): New macro.
25978 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
25980         PR rtl-optimization/84780
25981         * combine.c (distribute_links): Don't make a link based on pc_rtx.
25983 2018-03-14  Martin Liska  <mliska@suse.cz>
25985         * tree.c (record_node_allocation_statistics): Use
25986         get_stats_node_kind.
25987         (get_stats_node_kind): New function extracted from
25988         record_node_allocation_statistics.
25989         (free_node): Use get_stats_node_kind.
25991 2018-03-14  Richard Biener  <rguenther@suse.de>
25993         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
25994         that the value-set of ANTIC_IN doesn't grow.
25996         Revert
25997         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
25998         member.
25999         (BB_VISITED_WITH_VISITED_SUCCS): New define.
26000         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
26002 2018-03-14  Julia Koval  <julia.koval@intel.com>
26004         * config.gcc (icelake-client, icelake-server): New.
26005         (icelake): Remove.
26006         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
26007         (initial_ix86_arch_features): Ditto.
26008         (PTA_SKYLAKE): Add SGX.
26009         (PTA_ICELAKE): Remove.
26010         (PTA_ICELAKE_CLIENT): New.
26011         (PTA_ICELAKE_SERVER): New.
26012         (ix86_option_override_internal): Split up icelake on icelake client and
26013         icelake server.
26014         (get_builtin_code_for_version): Ditto.
26015         (fold_builtin_cpu): Ditto.
26016         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
26017         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
26018         * config/i386/i386.h (processor_type): Ditto.
26019         * doc/invoke.texi: Ditto.
26021 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
26023         PR sanitizer/83392
26024         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
26025         INTEGER_CST offset, add it together with bitpos / 8 and
26026         sign extend based on POINTER_SIZE.
26028         PR target/84844
26029         Revert
26030         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
26032         PR target/78090
26033         * config/i386/constraints.md (Yc): New register constraint.
26034         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
26035         Use Yc constraint for alternative 2 of operand 0.  Remove
26036         preferred_for_speed attribute.
26038 2018-03-14  Richard Biener  <rguenther@suse.de>
26040         PR tree-optimization/84830
26041         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
26042         with the old one to avoid oscillations.
26044 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
26046         PR target/83712
26047         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
26048         pseudos.
26049         (assign_by_spills): Return a flag of reload assignment failure.
26050         Do not process the reload assignment failures.  Do not spill other
26051         reload pseudos if they has the same reg class.  Update n if
26052         necessary.
26053         (lra_assign): Add a return arg.  Set up from the result of
26054         assign_by_spills call.
26055         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26056         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26057         usage_insns if it is not NULL.
26058         (spill_hard_reg_in_range): New function.
26059         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26060         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26061         function prototypes.
26062         (lra_assign): Change prototype.
26063         * lra.c (lra): Add code to deal with fails by splitting hard reg
26064         live ranges.
26066 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
26068         * config/riscv/riscv.opt (mrelax): New option.
26069         * config/riscv/riscv.c (riscv_file_start): Emit ".option
26070         "norelax" when riscv_mrelax is disabled.
26071         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
26073 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26075         PR target/84743
26076         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
26077         reassociation for int modes.
26079 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26081         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
26082         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
26083         for big-endian.
26084         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
26085         * config/aarch64/aarch64-sve.md
26086         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
26087         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
26088         (*extend<mode><Vwide>2): Rename to...
26089         (aarch64_sve_extend<mode><Vwide>2): ...this.
26090         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
26091         renaming the old pattern to...
26092         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
26093         unsigned packs.
26094         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
26095         define_expand, renaming the old pattern to...
26096         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
26097         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
26098         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
26099         account when deciding which SVE instruction the optab should use.
26100         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
26102 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26104         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
26105         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
26106         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
26107         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
26108         (tlsdesc_small_<mode>): Turn a define_expand and use
26109         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
26110         (tlsdesc_small_advsimd_<mode>): ...this.
26111         (tlsdesc_small_sve_<mode>): New pattern.
26113 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
26115         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
26116         (UNSPEC_UMUL_HIGHPART): New constants.
26117         (MUL_HIGHPART): New int iteraor.
26118         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
26119         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
26120         define_expand.
26121         (*<su>mul<mode>3_highpart): New define_insn.
26123 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
26125         PR lto/84805
26126         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
26127         incomplete types.
26129 2018-03-13  Martin Liska  <mliska@suse.cz>
26131         PR ipa/84658.
26132         * (sem_item_optimizer::sem_item_optimizer): Initialize new
26133         vector.
26134         (sem_item_optimizer::~sem_item_optimizer): Release it.
26135         (sem_item_optimizer::merge_classes): Register variable aliases.
26136         (sem_item_optimizer::fixup_pt_set): New function.
26137         (sem_item_optimizer::fixup_points_to_sets): Likewise.
26138         * ipa-icf.h: Declare new variables and functions.
26140 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
26142         PR middle-end/84834
26143         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
26144         integer_pow2p@2 and test integer_pow2p in condition.
26145         (A < 0 ? C : 0): Similarly for @1.
26147         PR middle-end/84831
26148         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
26149         characters starting at p contain '\0' character, don't look beyond
26150         that.
26152         PR target/84827
26153         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
26154         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
26156         PR target/84828
26157         * reg-stack.c (change_stack): Change update_end var from int to
26158         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
26159         also call set_block_for_insn on the newly added insns and rescan.
26161         PR target/84786
26162         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
26163         on the last operand.
26165         PR c++/84704
26166         * tree.c (stabilize_reference_1): Return save_expr (e) for
26167         STATEMENT_LIST even if it doesn't have side-effects.
26169 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
26171         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
26173 2018-03-12  Renlin Li  <renlin.li@arm.com>
26175         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
26176         aarch64_output_scalar_simd_mov_immediate.
26178 2018-03-12  Martin Sebor  <msebor@redhat.com>
26180         PR tree-optimization/83456
26181         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
26182         for perfectly overlapping calls to memcpy.
26183         (gimple_fold_builtin_memory_chk): Same.
26184         (gimple_fold_builtin_strcpy): Handle no-warning.
26185         (gimple_fold_builtin_stxcpy_chk): Same.
26186         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
26188 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
26190         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
26191         parameter.  Use it for SFmode.
26192         (rs6000_function_arg_advance_1): Adjust.
26193         (rs6000_function_arg): Adjust.
26194         (rs6000_gimplify_va_arg): Pass false for that new parameter.
26196 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
26198         PR rtl-optimization/84169
26199         PR rtl-optimization/84780
26200         * combine.c (can_combine_p): Check for a 2-insn combination whether
26201         the destination register is used between the two insns, too.
26203 2018-03-12  Richard Biener  <rguenther@suse.de>
26205         PR tree-optimization/84803
26206         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
26207         for refs DR analysis didn't process.
26209 2018-03-12  Richard Biener  <rguenther@suse.de>
26211         PR tree-optimization/84777
26212         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
26213         force-vectorize loops ignore whether we are optimizing for size.
26215 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26217         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
26218         (TARGET_MD_ASM_ADJUST): Define.
26220 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
26221             Kito Cheng  <kito.cheng@gmail.com>
26222             Chung-Ju Wu  <jasonwucj@gmail.com>
26224         * config/nds32/nds32.c (nds32_compute_stack_frame,
26225         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
26226         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
26227         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
26228         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
26229         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
26230         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
26231         * config/nds32/nds32.md (prologue, epilogue): Use macro
26232         NDS32_V3PUSH_AVAILABLE_P to do checking.
26234 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
26236         PR debug/58150
26237         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
26238         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
26239         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
26240         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
26241         addition of most attributes on !orig_type_die or the attribute not
26242         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
26244 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26245             Chung-Ju Wu  <jasonwucj@gmail.com>
26247         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
26248         __NDS32_VH__ macro.
26249         * config/nds32/nds32.opt (mvh): New option.
26251 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26252             Chung-Ju Wu  <jasonwucj@gmail.com>
26254         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
26255         function.
26256         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
26257         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
26258         definition.
26260 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
26261             Chung-Ju Wu  <jasonwucj@gmail.com>
26263         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
26264         function.
26265         * config/nds32/nds32-multiple.md (strlensi): New pattern.
26266         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
26268 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
26269             Kito Cheng  <kito.cheng@gmail.com>
26270             Chung-Ju Wu  <jasonwucj@gmail.com>
26272         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
26273         UNSPEC_FFMISM and UNSPEC_FLMISM.
26274         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
26275         for ffb, ffmism and flmism.
26276         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
26277         (unspec_ffmism): Ditto.
26278         (unspec_flmism): Ditto.
26279         (nds32_expand_builtin_impl): Check if string extension is available.
26280         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
26281         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
26283 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
26285         Reverting patch:
26286         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
26288         PR target/83712
26289         * lra-assigns.c (assign_by_spills): Return a flag of reload
26290         assignment failure.  Do not process the reload assignment
26291         failures.  Do not spill other reload pseudos if they has the same
26292         reg class.
26293         (lra_assign): Add a return arg.  Set up from the result of
26294         assign_by_spills call.
26295         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26296         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26297         usage_insns if it is not NULL.
26298         (spill_hard_reg_in_range): New function.
26299         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26300         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26301         function prototypes.
26302         (lra_assign): Change prototype.
26303         * lra.c (lra): Add code to deal with fails by splitting hard reg
26304         live ranges.
26306 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
26308         PR target/84807
26309         * config/i386/i386.opt: Replace Enforcment with Enforcement.
26311 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
26313         PR debug/84620
26314         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
26315         (dw_val_node): Add val_symbolic_view.
26316         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
26317         (symview_upper_bound): New.
26318         (new_line_info_table): Initialize symviews_since_reset.
26319         (dwarf2out_source_line): Count symviews_since_reset and set
26320         symview_upper_bound.
26321         (dw_val_equal_p): Handle symview.
26322         (add_AT_symview): New.
26323         (print_dw_val): Handle symview.
26324         (attr_checksum, attr_checksum_ordered): Likewise.
26325         (same_dw_val_p, size_of_die): Likewise.
26326         (value_format, output_die): Likewise.
26327         (add_high_low_attributes): Use add_AT_symview for entry_view.
26328         (dwarf2out_finish): Reset symview_upper_bound, clear
26329         zero_view_p.
26331 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
26333         PR target/83969
26334         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
26335         Add strict argument and use it.
26336         (rs6000_split_multireg_move): Update for new strict argument.
26337         (mem_operand_gpr): Disallow all non-offsettable addresses.
26338         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
26340 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
26342         PR target/84772
26343         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
26344         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
26345         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
26347         PR c++/84767
26348         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
26349         decl, use remap_type if we want to use the type.
26351 2018-03-09  Martin Sebor  <msebor@redhat.com>
26353         PR tree-optimization/84526
26354         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
26355         Remove dead code.
26356         (builtin_access::generic_overlap): Be prepared to handle non-array
26357         base objects.
26359 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
26361         PR rtl-optimization/84682
26362         * lra-constraints.c (process_address_1): Check is_address flag
26363         for address constraints.
26364         (process_alt_operands): Likewise.
26365         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
26366         preprocess_constraints.
26367         * recog.h (preprocess_constraints): Add oploc parameter.
26368         Adjust callers.
26369         * recog.c (preprocess_constraints): Test address_operand for
26370         CT_ADDRESS constraints.
26372 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
26374         PR target/83712
26375         * lra-assigns.c (assign_by_spills): Return a flag of reload
26376         assignment failure.  Do not process the reload assignment
26377         failures.  Do not spill other reload pseudos if they has the same
26378         reg class.
26379         (lra_assign): Add a return arg.  Set up from the result of
26380         assign_by_spills call.
26381         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
26382         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
26383         usage_insns if it is not NULL.
26384         (spill_hard_reg_in_range): New function.
26385         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
26386         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
26387         function prototypes.
26388         (lra_assign): Change prototype.
26389         * lra.c (lra): Add code to deal with fails by splitting hard reg
26390         live ranges.
26392 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26394         PR target/83193
26395         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
26396         Accept complain bool parameter.  Only emit errors if it is true.
26397         (arm_parse_cpu_option_name): Likewise.
26398         (arm_target_thumb_only): Adjust callers of the above.
26399         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
26400         prototype to take a default true bool parameter.
26401         (arm_parse_arch_option_name): Likewise.
26403 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
26404             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
26406         PR jit/64089
26407         PR jit/84288
26408         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
26409         * configure: Regenerate.
26410         * configure.ac ("linker --version-script option"): New.
26411         ("linker soname option"): New.
26413 2018-03-09  Richard Biener  <rguenther@suse.de>
26415         PR tree-optimization/84775
26416         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
26417         immediate uses of predicate stmts and mark them modified.
26419         Revert
26420         PR tree-optimization/84178
26421         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
26422         to caller.
26423         (version_loop_for_if_conversion): Delay update_ssa call.
26424         (tree_if_conversion): Delay update_ssa until after predicate
26425         insertion.
26427 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
26429         PR target/84763
26430         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
26431         when the function accesses prior frames.
26433 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
26435         PR debug/84456
26436         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
26437         gen_llsym, otherwise call maybe_gen_llsym.
26439         PR inline-asm/84742
26440         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
26441         has ',' character inside of it.
26443 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26445         PR target/84748
26446         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
26447         as clobbering CC_REGNUM.
26449 2018-03-08  Richard Biener  <rguenther@suse.de>
26451         PR middle-end/84552
26452         * tree-scalar-evolution.c: Include tree-into-ssa.h.
26453         (follow_copies_to_constant): Do not follow SSA names registered
26454         for update.
26456 2018-03-08  Richard Biener  <rguenther@suse.de>
26458         PR tree-optimization/84178
26459         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
26460         to caller.
26461         (version_loop_for_if_conversion): Delay update_ssa call.
26462         (tree_if_conversion): Delay update_ssa until after predicate
26463         insertion.
26465 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
26467         PR tree-optimization/84178
26468         * tree-if-conv.c (release_bb_predicate): Remove the
26469         the assertion that the stmts have NULL use_ops.
26470         Discard the statements, asserting that they haven't
26471         yet been added to a BB.
26473 2018-03-08  Richard Biener  <rguenther@suse.de>
26475         PR tree-optimization/84746
26476         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
26477         (phi_translate): Pass in destination ANTIC_OUT set.
26478         (phi_translate_1): Likewise.  For a simplified result lookup
26479         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
26480         (phi_translate_set): Adjust.
26481         (do_pre_regular_insertion): Likewise.
26482         (do_pre_partial_partial_insertion): Likewise.
26484 2018-03-08  Martin Liska  <mliska@suse.cz>
26486         PR gcov-profile/84735
26487         * doc/gcov.texi: Document usage of profile files.
26488         * gcov-io.h: Document changes in the format.
26490 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
26492         PR debug/84404
26493         PR debug/84408
26494         * dwarf2out.c (struct dw_line_info_table): Update comments for
26495         view == -1.
26496         (FORCE_RESET_NEXT_VIEW): New.
26497         (FORCE_RESETTING_VIEW_P): New.
26498         (RESETTING_VIEW_P): Check for -1 too.
26499         (ZERO_VIEW_P): Likewise.
26500         (new_line_info_table): Force-reset next view.
26501         (dwarf2out_begin_function): Likewise.
26502         (dwarf2out_source_line): Simplify zero_view_p initialization.
26503         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
26504         view directly.  Omit view when omitting .loc at line 0.
26506 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
26508         PR tree-optimization/84740
26509         * tree-switch-conversion.c (process_switch): Call build_constructors
26510         only if info.phi_count is non-zero.
26512         PR tree-optimization/84739
26513         * tree-tailcall.c (find_tail_calls): Check call arguments against
26514         DECL_ARGUMENTS (current_function_decl) rather than
26515         DECL_ARGUMENTS (func) when checking for tail recursion.
26517 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
26519         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
26520         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
26521         Volker Reichelt's entry and add entries for people that perform
26522         GCC fuzzy testing and report numerous bugs.
26524 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
26526         PR target/82411
26527         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
26528         readonly data in sdata, if that is disabled.
26529         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
26530         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
26531         -mreadonly-in-sdata option.
26533 2018-03-07  Martin Sebor  <msebor@redhat.com>
26535         PR tree-optimization/84468
26536         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
26537         basic block when looking for nul assignment.
26539 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
26541         PR target/84277
26542         * except.h (output_function_exception_table): Adjust prototype.
26543         * except.c (output_function_exception_table): Remove FNNAME parameter
26544         and add SECTION parameter.  Ouput one part of the table at a time.
26545         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
26546         the first part of the exception table and emit unwind directives.
26547         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
26548         (i386_pe_seh_cold_init): Likewise.
26549         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
26550         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
26551         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
26552         (ix86_output_call_insn): Emit a nop in one more case for SEH.
26553         * config/i386/winnt.c: Include except.h.
26554         (struct seh_frame_state): Add reg_offset, after_prologue and
26555         in_cold_section fields.
26556         (i386_pe_seh_end_prologue): Set seh->after_prologue.
26557         (i386_pe_seh_cold_init): New function.
26558         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
26559         to seh->in_cold_section.
26560         (seh_emit_push): Record the offset of the push.
26561         (seh_emit_save): Record the offet of the save.
26562         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
26563         Test seh->after_prologue to disregard the epilogue.
26564         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
26565         (i386_pe_end_cold_function): New function.
26567 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
26569         PR fortran/84565
26570         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
26571         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
26573         PR c++/84704
26574         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
26575         on tmp_var.
26576         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
26577         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
26579         PR middle-end/84723
26580         * multiple_target.c: Include tree-inline.h and intl.h.
26581         (expand_target_clones): Diagnose and fail if node->definition and
26582         !tree_versionable_function_p (node->decl).
26584 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
26586         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
26587         sprint_ul.
26588         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
26589         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
26590         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
26592 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
26594         PR target/84710
26595         * combine.c (try_combine): Use reg_or_subregno instead of handling
26596         just paradoxical SUBREGs and REGs.
26598 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
26600         * config/arc/arc.c (arc_finalize_pic): Remove function.
26601         (arc_must_save_register): We use single base PIC register, remove
26602         checks to save/restore the PIC register.
26603         (arc_expand_prologue): Likewise.
26604         * config/arc/arc-protos.h (arc_set_default_type_attributes):
26605         Remove.
26606         (arc_verify_short): Likewise.
26607         (arc_attr_type): Likewise.
26608         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
26609         (walk_stores): Likewise.
26610         (arc_address_cost): Make it static.
26611         (arc_verify_short): Likewise.
26612         (branch_dest): Likewise.
26613         (arc_attr_type): Likewise.
26614         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
26615         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
26616         (arc_final_prescan_insn): Remove inserting the nops due to
26617         hardware hazards.  It is done in reorg step.
26618         (insn_length_variant_t): Remove.
26619         (insn_length_parameters_t): Likewise.
26620         (arc_insn_length_parameters): Likewise.
26621         (arc_get_insn_variants): Likewise.
26622         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
26624 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
26626         PR inline-asm/84683
26627         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
26628         assertion failure.
26630         PR tree-optimization/84687
26631         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
26632         on new_node->decl.
26633         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
26635 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26637         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
26638         Rename to ppc_speculation_barrier.
26639         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
26640         __builtin_ppc_speculation_barrier.
26642 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
26644         PR target/84700
26645         * combine.c (combine_simplify_rtx): Don't try to simplify if
26646         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
26647         are equal to x.
26649 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
26651         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
26652         to 32 bytes when compiling for POWER9.
26654 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
26656         PR target/84564
26657         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
26658         regparm >= 3 with no arg reg available also for calls with
26659         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
26661         PR target/84524
26662         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
26663         orig,vex.
26664         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
26666 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
26668         PR target/84264
26669         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
26671 2018-03-05  Richard Biener  <rguenther@suse.de>
26673         PR tree-optimization/84486
26674         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
26675         When inserting a __builtin_assume_aligned call set the LHS
26676         SSA name alignment info accordingly.
26678 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
26680         PR tree-optimization/84114
26681         * config/aarch64/aarch64.c (aarch64_reassociation_width)
26682         Avoid reassociation of FLOAT_MODE addition.
26684 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
26686         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
26687         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
26688         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
26689         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
26690         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
26691         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
26692         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
26693         and -mwbnoinvd.
26694         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
26695         __builtin_ia32_wbinvd): New builtins.
26696         (SPECIAL_ARGS2): New.
26697         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
26698         (SPECIAL_ARGS2): New.
26699         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
26700         (ix86_valid_target_attribute_inner_p): Ditto.
26701         (ix86_init_mmx_sse_builtins): Add special_args2.
26702         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
26703         TARGET_WBNOINVD_P): New.
26704         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
26705         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
26706         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
26707         * config/i386/immintrin.h (_wbinvd): New intrinsic.
26708         * config/i386/pconfigintrin.h: New file.
26709         * config/i386/wbnoinvdintrin.h: Ditto.
26710         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
26711         wbnoinvdintrin.h.
26712         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
26714 2018-03-05  Richard Biener  <rguenther@suse.de>
26716         PR tree-optimization/84670
26717         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
26718         member.
26719         (BB_VISITED_WITH_VISITED_SUCCS): New define.
26720         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
26721         (compute_antic_aux): Only assert the number of values in ANTIC_IN
26722         doesn't grow if all successors (recursively) were visited at least
26723         once.
26725 2018-03-05  Richard Biener  <rguenther@suse.de>
26727         PR tree-optimization/84650
26728         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
26729         if executed in the loop pipeline.
26731 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
26733         * doc/configfiles.texi (Configuration Files): Move info about
26734         conditionalizing $target-protos.h to...
26735         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
26736         differs from $target-protos.h.
26738 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
26739             Chung-Ju Wu  <jasonwucj@gmail.com>
26741         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
26742         * config/nds32/nds32-multiple.md (setmemsi): Define.
26743         * config/nds32/nds32-memory-manipulation.c
26744         (nds32_gen_dup_4_byte_to_word_value): New.
26745         (emit_setmem_word_loop): New.
26746         (emit_setmem_byte_loop): New.
26747         (nds32_expand_setmem_loop): New.
26748         (nds32_expand_setmem_loop_v3m): New.
26749         (nds32_expand_setmem_unroll): New.
26750         (nds32_expand_setmem): New.
26752 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
26753             Chung-Ju Wu  <jasonwucj@gmail.com>
26755         * config/nds32/nds32-memory-manipulation.c
26756         (nds32_emit_load_store): New.
26757         (nds32_emit_post_inc_load_store): New.
26758         (nds32_emit_mem_move): New.
26759         (nds32_emit_mem_move_block): New.
26760         (nds32_expand_movmemsi_loop_unknown_size): New.
26761         (nds32_expand_movmemsi_loop_known_size): New.
26762         (nds32_expand_movmemsi_loop): New.
26763         (nds32_expand_movmemsi_unroll): New.
26764         (nds32_expand_movmemqi): Rename ...
26765         (nds32_expand_movmemsi): ... to this.
26766         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
26767         (movmemsi): ... to this.
26768         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
26769         (nds32_expand_movmemsi): ... to this.
26771 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
26772             Monk Chiang  <sh.chiang04@gmail.com>
26773             Chung-Ju Wu  <jasonwucj@gmail.com>
26775         * config/nds32/nds32-protos.h
26776         (nds32_expand_load_multiple): New arguments.
26777         (nds32_expand_store_multiple): Ditto.
26778         (nds32_valid_multiple_load_store): Rename ...
26779         (nds32_valid_multiple_load_store_p): ... to this.
26780         * config/nds32/nds32-memory-manipulation.c
26781         (nds32_expand_load_multiple): Refine implementation.
26782         (nds32_expand_store_multiple): Ditto.
26783         * config/nds32/nds32-multiple.md
26784         (load_multiple): Update nds32_expand_load_multiple interface.
26785         (store_multiple): Update nds32_expand_store_multiple interface.
26786         * config/nds32/nds32-predicates.c
26787         (nds32_valid_multiple_load_store): Rename ...
26788         (nds32_valid_multiple_load_store_p): ... to this and refine
26789         implementation.
26790         * config/nds32/predicates.md
26791         (nds32_load_multiple_and_update_address_operation): New predicate.
26792         (nds32_store_multiple_and_update_address_operation): New predicate.
26794 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
26795             Chung-Ju Wu  <jasonwucj@gmail.com>
26797         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
26798         (combo): New attribute.
26799         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
26801 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
26803         * config/nds32/nds32.opt: Change -mcmodel= default value.
26805 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
26806             Monk Chiang  <sh.chiang04@gmail.com>
26807             Chung-Ju Wu  <jasonwucj@gmail.com>
26809         * config/nds32/constants.md (unspec_element): New enum.
26810         * config/nds32/constraints.md (Umw): New constraint.
26811         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
26812         * config/nds32/nds32-intrinsic.md: Likewise.
26813         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
26814         (nds32_valid_smw_lwm_base_p): New.
26815         (nds32_output_smw_single_word): New.
26816         (nds32_output_lmw_single_word): New.
26817         (nds32_expand_unaligned_load): New.
26818         (nds32_expand_unaligned_store): New.
26819         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
26820         (nds32_output_smw_single_word): Declare.
26821         (nds32_output_lmw_single_word): Declare.
26822         (nds32_expand_unaligned_load): Declare.
26823         (nds32_expand_unaligned_store): Declare.
26824         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
26825         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
26826         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
26827         NDS32_BUILTIN_UASTORE_DW.
26828         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
26829         predicate.
26831 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
26832             Kito Cheng  <kito.cheng@gmail.com>
26833             Chung-Ju Wu  <jasonwucj@gmail.com>
26835         * config/nds32/nds32-intrinsic.c
26836         (nds32_expand_builtin_null_ftype_reg): Delete.
26837         (nds32_expand_builtin_reg_ftype_imm): Ditto.
26838         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
26839         (nds32_read_argument): New.
26840         (nds32_legitimize_target): Ditto.
26841         (nds32_legitimize_argument): Ditto.
26842         (nds32_check_constant_argument): Ditto.
26843         (nds32_expand_unop_builtin): Ditto.
26844         (nds32_expand_unopimm_builtin): Ditto.
26845         (nds32_expand_binop_builtin): Ditto.
26846         (nds32_builtin_decl_impl): Ditto.
26847         (builtin_description): Ditto.
26848         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
26849         (nds32_init_builtins_impl): Ditto.
26850         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
26851         (nds32_builtin_decl): New.
26852         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
26853         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
26855 2018-03-02  Jeff Law  <law@redhat.com>
26857         * reorg.c (stop_search_p): Handle DEBUG_INSN.
26858         (redundant_insn, fill_simple_delay_slots): Likewise.
26859         (fill_slots_from_thread): Likewise.
26860         * resource.c (mark_referenced_resources): Likewise.
26861         (mark_set_resources, find_dead_or_set_registers): Likewise.
26863 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26865         * substring-locations.h (format_warning_va): Formatting fix for
26866         ATTRIBUTE_GCC_DIAG.
26867         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
26868         argument.
26869         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
26870         * substring-locations.c: Include intl.h.
26871         (format_warning_va): Turned into small wrapper around
26872         format_warning_n_va, renamed to ...
26873         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
26874         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
26875         use ngettext.
26876         (format_warning_at_substring_n): New function.
26877         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
26878         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
26879         format_warning_at_substring with just a shorter name instead of
26880         const function pointer.
26881         (fmtwarn_n): New function.
26882         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
26883         appropriate, get rid of all the fmtstr temporaries, move conditionals
26884         with G_() wrapped string literals directly into fmtwarn arguments,
26885         cast dir.len to (int), formatting fixes.
26887 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
26889         * doc/invoke.texi: Remove "Cilk Plus" references.
26891 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26892             Richard Biener  <rguenther@suse.de>
26894         PR ipa/84628
26895         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
26896         for error or warning attributes if CALL_FROM_THUNK_P is set.
26897         Formatting fixes.
26899 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26901         PR target/56540
26902         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
26903         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
26905         PR target/56540
26906         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
26907         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
26909         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
26910         instead of -1U in last predictors element's probability member.
26912 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
26914         PR ipa/83983
26915         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
26916         arguments if they are comparable.
26918 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
26920         PR tree-optimization/84634
26921         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
26922         masks and masked_loop_p with a single loop_masks, making sure it's
26923         null for bb vectorization.
26925 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
26927         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
26928         (vect_analyze_data_ref_access): Use loop->safe_len rather than
26929         loop->force_vectorize to check whether there is no alias.
26931 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
26933         PR target/84614
26934         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
26935         prototypes.
26936         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
26937         comments.
26938         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
26939         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
26940         instead of a loop around prev_real_insn.
26941         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
26942         prev_real_insn.
26944         PR inline-asm/84625
26945         * config/i386/i386.c (ix86_print_operand): Use conditional
26946         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
26947         zero vector.
26949 2018-03-02  Richard Biener  <rguenther@suse.de>
26951         PR tree-optimization/84427
26952         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
26953         (bitmap_set_subtract_values): Rewrite to handle multiple
26954         exprs per value.
26955         (clean): Likewise.
26956         (prune_clobbered_mems): Likewise.
26957         (phi_translate): Take edge instead of pred/phiblock.
26958         (phi_translate_1): Likewise.
26959         (phi_translate_set): Likewise.  Insert all translated
26960         exprs for a value into the set, keeping possibly multiple
26961         expressions per value.
26962         (compute_antic_aux): Adjust for phi_translate changes.
26963         When intersecting union the expressions and prune those
26964         not in the final value set, keeping possibly multiple
26965         expressions per value.  Do not use value-insertion
26966         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
26967         all expressions.  Add verification that the value-sets
26968         only shrink during iteration.
26969         (compute_partial_antic_aux): Adjust for the phi_translate changes.
26970         (do_pre_regular_insertion): Likewise.
26971         (do_pre_partial_partial_insertion): Likewise.
26973 2018-03-02  Richard Biener  <rguenther@suse.de>
26975         PR target/82005
26976         * config/darwin.c (saved_debug_info_level): New static global.
26977         (darwin_asm_lto_start): Disable debug info generation for LTO out.
26978         (darwin_asm_lto_end): Restore debug info generation settings.
26980 2018-03-01  Martin Liska  <mliska@suse.cz>
26982         PR sanitizer/82484
26983         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
26984         volatile arguments.
26986 2018-03-01  Richard Biener  <rguenther@suse.de>
26988         PR debug/84645
26989         * dwarf2out.c (gen_variable_die): Properly handle late VLA
26990         type annotation with LTO when debug was disabled at compile-time.
26992 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
26994         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
26995         XINT with INTVAL.
26996         (mips_final_postscan_insn): Likewise.
26998 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
27000         PR rtl-optimization/84528
27001         * alias.c (init_alias_target): Add commentary.
27002         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
27003         a unique base value if the frame pointer is not eliminated
27004         to the stack pointer.
27006 2018-03-01  Tom de Vries  <tom@codesourcery.com>
27008         PR rtl-optimization/83327
27009         * lra-int.h (hard_regs_spilled_into): Declare.
27010         * lra.c (hard_regs_spilled_into): Define.
27011         (init_reg_info): Init hard_regs_spilled_into.
27012         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
27013         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
27014         (process_bb_lives): Handle hard_regs_spilled_into.
27015         (lra_create_live_ranges_1): Before doing liveness propagation, clear
27016         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
27018 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
27020         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
27021         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
27022         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
27023         * config/rs6000/aix72.h: New file.
27025 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
27027         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
27028         instead of warning_at with conditional singular and plural messages
27029         where possible.
27031         PR target/52991
27032         * stor-layout.c (update_alignment_for_field): For
27033         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
27034         && !DECL_PACKED (field), do the alignment update, just use
27035         only desired_align instead of MAX (type_align, desired_align)
27036         as the alignment.
27037         (place_field): Don't do known_align < desired_align handling
27038         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
27039         is non-NULL, instead do it after rli->prev_field handling and
27040         only if not within a bitfield word.  For DECL_PACKED (field)
27041         use type_align of BITS_PER_UNIT.
27043 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
27045         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
27046         superfluous parentheses and trailing spaces.
27048 2018-02-28  Richard Biener  <rguenther@suse.de>
27050         PR tree-optimization/84584
27051         * graphite-scop-detection.c (scop_detection::add_scop): Discard
27052         SCoPs with fake exit edge.
27054 2018-02-28  Martin Liska  <mliska@suse.cz>
27056         PR testsuite/84597
27057         * timevar.c (timer::print): Fix format to properly print 100%
27058         values.
27060 2018-02-28  Richard Biener  <rguenther@suse.de>
27062         PR middle-end/84607
27063         * genmatch.c (capture_info::walk_match): Do not mark
27064         captured expressions without operands as expr_p given
27065         they act more like predicates and should be subject to
27066         "lost tail" side-effect preserving.
27068 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
27070         PR rtl-optimization/81611
27071         * auto-inc-dec.c (attempt_change): Move dead note from
27072         mem_insn if it's the next use of regno
27073         (find_address): Take address use of reg holding
27074         non-incremented value.  Add parm to limit search to the named
27075         reg only.
27076         (merge_in_block): Attempt to use a mem insn that is the next
27077         use of the original regno.
27079 2018-02-27  Martin Sebor  <msebor@redhat.com>
27081         PR c++/83871
27082         * doc/invoke.texi (-Wmissing-attributes): New option.
27083         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
27085 2018-02-27  Martin Sebor  <msebor@redhat.com>
27087         PR translation/84207
27088         * diagnostic-core.h (warning_n, error_n, inform_n): Change
27089         n argument to unsigned HOST_WIDE_INT.
27090         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
27091         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
27092         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
27093         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
27095 2018-02-27  Richard Biener  <rguenther@suse.de>
27097         PR tree-optimization/84512
27098         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
27099         Do not use the estimate returned from record_stmt_cost for
27100         the scalar iteration cost but sum properly using add_stmt_cost.
27102 2018-02-27  Richard Biener  <rguenther@suse.de>
27104         PR tree-optimization/84466
27105         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
27106         Adjust last change to less strictly validate use operands.
27108 2018-02-27  Martin Liska  <mliska@suse.cz>
27110         PR gcov-profile/84548
27111         * gcov.c (process_file): Allow partial overlap and consider it
27112         also as group functions.
27113         (output_lines): Properly calculate range of lines for a group.
27115 2018-02-27  Martin Liska  <mliska@suse.cz>
27117         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
27118         'ggc' suffixes.  Change first column width.
27119         (timer::print): Fix formatting of the column.
27121 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
27123         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
27124         preserve inline entry blocks for the sake of debug inline
27125         entry point markers alone.
27126         (remove_unused_locals): Suggest in comments a better place to
27127         force the preservation of inline entry blocks that are
27128         otherwise unused, but do not preserve them.
27130 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27132         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
27134 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27136         PR target/84039
27137         * config/i386/constraints.md (Bs): Replace
27138         ix86_indirect_branch_register with
27139         TARGET_INDIRECT_BRANCH_REGISTER.
27140         (Bw): Likewise.
27141         * config/i386/i386.md (indirect_jump): Likewise.
27142         (tablejump): Likewise.
27143         (*sibcall_memory): Likewise.
27144         (*sibcall_value_memory): Likewise.
27145         Peepholes of indirect call and jump via memory: Likewise.
27146         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
27147         (*sibcall_value_GOT_32): Likewise.
27148         * config/i386/predicates.md (indirect_branch_operand): Likewise.
27149         (GOT_memory_operand): Likewise.
27150         (call_insn_operand): Likewise.
27151         (sibcall_insn_operand): Likewise.
27152         (GOT32_symbol_operand): Likewise.
27153         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
27155 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
27157         PR rtl-optimization/83496
27158         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
27159         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
27160         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
27161         redundant insn, if any.
27162         (relax_delay_slots): Likewise.
27163         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
27165 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
27167         PR tree-optimization/83965
27168         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
27169         that grouped statements are part of a reduction chain.  Return
27170         true if the statement is not marked as a reduction itself but
27171         is part of a group.
27172         (vect_recog_dot_prod_pattern): Don't check whether the statement
27173         is part of a group here.
27174         (vect_recog_sad_pattern): Likewise.
27175         (vect_recog_widen_sum_pattern): Likewise.
27177 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
27179         PR debug/84545
27180         * final.c (rest_of_clean_state): Also look for calls inside sequences.
27182 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
27184         PR target/84530
27185         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
27186         the bool argument.
27187         (ix86_output_indirect_function_return): New prototype.
27188         (ix86_split_simple_return_pop_internal): Likewise.
27189         * config/i386/i386.c (indirect_return_via_cx): New.
27190         (indirect_return_via_cx_bnd): Likewise.
27191         (indirect_thunk_name): Handle return va CX_REG.
27192         (output_indirect_thunk_function): Create alias for
27193         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
27194         (ix86_output_indirect_jmp): Remove the bool argument.
27195         (ix86_output_indirect_function_return): New function.
27196         (ix86_split_simple_return_pop_internal): Likewise.
27197         * config/i386/i386.md (*indirect_jump): Don't pass false
27198         to ix86_output_indirect_jmp.
27199         (*tablejump_1): Likewise.
27200         (simple_return_pop_internal): Change it to define_insn_and_split.
27201         Call ix86_split_simple_return_pop_internal to split it for
27202         -mfunction-return=.
27203         (simple_return_indirect_internal): Call
27204         ix86_output_indirect_function_return instead of
27205         ix86_output_indirect_jmp.
27207 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
27209         PR bootstrap/84405
27210         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
27211         memset and value initialization afterwards.
27213 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
27215         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
27217 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27219         PR target/84521
27220         * common/config/aarch64/aarch64-common.c
27221         (aarch_option_optimization_table[]): Switch
27222         off fomit-frame-pointer
27224 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
27225             Chung-Ju Wu  <jasonwucj@gmail.com>
27227         * config/nds32/nds32-multiple.md (load_multiple): Disallow
27228         volatile memory.
27229         (store_multiple): Ditto.
27231 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
27233         * config.gcc: Add --with-cpu support for nds32 target.
27234         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
27235         * config/nds32/nds32.opt: Add -mcpu= option.
27237 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
27239         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
27240         isel=yes): Warn for these deprecated options.
27242 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
27244         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
27245         ISA_2_5_MASKS_EMBEDDED.
27247 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
27249         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
27250         p->max as pointers rather than using iterative_hash_expr.
27252 2018-02-23  Carl Love  <cel@us.ibm.com>
27254         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
27255         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
27256         BU_P8V_OVERLOAD_2.
27257         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
27258         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
27259         P8V_BUILTIN_VEC_VUNSIGNED2.
27261 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
27263         PR target/81572
27264         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
27265         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
27266         LRA_UNKNOWN_ALT.
27267         * lra-constraints.c (curr_insn_transform): Set up
27268         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
27269         LRA_UNKNOWN_ALT.
27270         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
27271         * lra-eliminations.c (spill_pseudos): Ditto.
27272         (process_insn_for_elimination): Ditto.
27273         * lra-lives.c (reg_early_clobber_p): Use the new macros.
27274         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
27275         LRA_NON_CLOBBERED_ALT.
27277 2018-02-22  Martin Sebor  <msebor@redhat.com>
27279         PR tree-optimization/84480
27280         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
27281         to maybe_diag_stxncpy_trunc.  Call it.
27282         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
27283         from gimple_fold_builtin_strcpy.  Print inlining stack.
27284         (handle_builtin_stxncpy): Print inlining stack.
27285         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
27287 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
27289         PR target/84176
27290         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
27291         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
27292         and -fcheck-pointer-bounds are used together.
27293         (indirect_thunk_prefix): New enum.
27294         (indirect_thunk_need_prefix): New function.
27295         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
27296         "_nt" instead of "_bnd" for NOTRACK prefix.
27297         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
27298         (output_indirect_thunk_function): Likewise.
27299         (): Likewise.
27300         (ix86_code_end): Update output_indirect_thunk_function calls.
27301         (ix86_output_indirect_branch_via_reg): Replace
27302         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
27303         (ix86_output_indirect_branch_via_push): Likewise.
27304         (ix86_output_function_return): Likewise.
27305         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
27306         incompatible with -fcf-protection=branch and
27307         -fcheck-pointer-bounds.
27309 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
27311         PR target/83335
27312         * config/aarch64/aarch64.c (aarch64_print_address_internal):
27313         Change gcc_assert call to output_operand_lossage.
27315 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
27317         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
27319 2018-02-22  DJ Delorie  <dj@redhat.com>
27320             Sebastian Perta  <sebastian.perta@renesas.com>
27321             Oleg Endo  <olegendo@gcc.gnu.org>
27323         * config/rx/rx.c (rx_rtx_costs): New function.
27324         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
27326 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
27328         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
27330 2018-02-22  Martin Liska  <mliska@suse.cz>
27332         PR driver/83193
27333         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
27334         Add "native" as a possible value.
27336 2018-02-22  Martin Liska  <mliska@suse.cz>
27338         PR driver/83193
27339         * config/i386/i386.c (ix86_option_override_internal):
27340         Add "native" as a possible value for -march and -mtune.
27342 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
27344         PR target/84502
27345         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
27346         to all type variants.
27348         PR tree-optimization/84503
27349         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
27350         width as info->bitpos + info->bitsize - start.
27351         (merged_store_group::merge_overlapping): Simplify width computation.
27352         (check_no_overlap): New function.
27353         (imm_store_chain_info::try_coalesce_bswap): Compute expected
27354         start + width and last_order of the group, fail if check_no_overlap
27355         fails.
27356         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
27357         to group if check_no_overlap fails.
27359 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
27361         * config/rs6000/altivec.md: Delete contraint arguments to
27362         define_expand, define_split, and define_peephole2, and in
27363         define_insn_and_split if always unused.
27364         * config/rs6000/darwin.md: Ditto.
27365         * config/rs6000/dfp.md: Ditto.
27366         * config/rs6000/rs6000.md: Ditto.
27367         * config/rs6000/sync.md: Ditto.
27368         * config/rs6000/vector.md: Ditto.
27369         * config/rs6000/vsx.md: Ditto.
27371 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
27373         * config/rs6000/altivec.md: Write output control strings as braced
27374         blocks instead of double-quoted strings.
27375         * config/rs6000/darwin.md: Ditto.
27376         * config/rs6000/rs6000.md: Ditto.
27377         * config/rs6000/vector.md: Ditto.
27378         * config/rs6000/vsx.md: Ditto.
27380 2018-02-21  Jason Merrill  <jason@redhat.com>
27382         PR c++/84314 - ICE with templates and fastcall attribute.
27383         * attribs.c (build_type_attribute_qual_variant): Remove assert.
27385 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
27387         * ipa-cp.c (determine_versionability): Fix comment typos.
27389 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
27391         PR c/84229
27392         * ipa-cp.c (determine_versionability): Do not version functions caling
27393         va_arg_pack.
27395 2018-02-21  Martin Liska  <mliska@suse.cz>
27397         PR driver/83193
27398         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
27399         Add "native" as a possible value.
27400         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
27401         the macro when native cpu detection is available.
27403 2018-02-21  Martin Liska  <mliska@suse.cz>
27405         PR driver/83193
27406         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
27407         Add "native" as a possible value.
27408         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
27409         when native cpu detection is available.
27411 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
27412             Martin Sebor  <msebor@redhat.com>
27414         PR tree-optimization/84478
27415         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
27416         false.
27417         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
27418         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
27419         support which is conservatively correct, for 2 only stay conservative
27420         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
27421         argument to the 2 argument get_range_strlen, adjust 6 arg
27422         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
27423         false.
27424         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
27425         (gimple_fold_builtin_strlen): Pass true as last argument to
27426         get_range_strlen.
27428 2018-02-20  Martin Sebor  <msebor@redhat.com>
27430         PR middle-end/84095
27431         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
27432         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
27433         (builtin_memref::builtin_memref): Factor out parts into
27434         set_base_and_offset and call it.
27436 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
27438         PR middle-end/84406
27439         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
27440         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
27441         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
27442         search at the associated MODE_INT.
27444 2018-02-20  Jeff Law  <law@redhat.com>
27446         PR middle-end/82123
27447         PR tree-optimization/81592
27448         PR middle-end/79257
27449         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
27450         for range data rather than using global data.
27451         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
27452         range data rather than using global data.
27453         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
27454         pass it to children as needed.
27455         (struct directive::fmtresult): Similarly.
27456         (struct directive::set_width): Similarly.
27457         (struct directive::set_precision): Similarly.
27458         (format_integer, format_directive, parse_directive): Similarly.
27459         (format_none): Accept unnamed vr_values parameter.
27460         (format_percent, format_floating, format_character): Similarly.
27461         (format_string, format_plain): Similarly.
27462         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
27463         the EVRP range analyzer for range data rather than using global data.
27464         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
27465         gimple-ssa-evrp-analyze.h
27466         (class sprintf_dom_walker): Add after_dom_children member function.
27467         Add evrp_range_analyzer member.
27468         (sprintf_dom_walker::before_dom_children): Call into the EVRP
27469         range analyzer as needed.
27470         (sprintf_dom_walker::after_dom_children): New member function.
27471         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
27472         if not optimizing.
27473         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
27474         (evrp_range_analyzer::pop_to_marker): Likewise.
27476 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
27478         PR tree-optimization/84419
27479         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
27480         with the required type if its current type is compatible but
27481         different.
27483 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
27485         PR middle-end/82004
27486         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
27487         after vectorization.
27489 2018-02-20  Martin Liska  <mliska@suse.cz>
27491         PR driver/83193
27492         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
27493         possible values if we don't have a hint.
27495 2018-02-20  Martin Liska  <mliska@suse.cz>
27497         PR c/84310
27498         PR target/79747
27499         * final.c (shorten_branches): Build align_tab array with one
27500         more element.
27501         * opts.c (finish_options): Add alignment option limit check.
27502         (MAX_CODE_ALIGN): Likewise.
27503         (MAX_CODE_ALIGN_VALUE): Likewise.
27504         * doc/invoke.texi: Document maximum allowed option value for
27505         all -falign-* options.
27507 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
27509         PR target/84146
27510         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
27511         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
27512         * var-tracking.c (emit_note_insn_var_location): Remove all references
27513         to NOTE_INSN_CALL_ARG_LOCATION.
27514         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
27515         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
27516         Use copy_rtx_if_shared.
27517         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
27518         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
27519         (dwarf2out_var_location): Remove handling of
27520         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
27521         on call_insn.
27522         * final.c (final_scan_insn): Remove all references to
27523         NOTE_INSN_CALL_ARG_LOCATION.
27524         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
27525         before dumping final insns.
27526         * except.c (emit_note_eh_region_end): Remove all references to
27527         NOTE_INSN_CALL_ARG_LOCATION.
27528         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
27529         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
27530         * config/arc/arc.c (hwloop_optimize): Likewise.
27531         * config/arm/arm.c (create_fix_barrier): Likewise.
27532         * config/s390/s390.c (s390_chunkify_start): Likewise.
27533         * config/sh/sh.c (find_barrier): Likewise.
27534         * config/i386/i386.c (rest_of_insert_endbranch,
27535         ix86_seh_fixup_eh_fallthru): Likewise.
27536         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
27537         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
27538         * config/frv/frv.c (frv_function_prologue): Likewise.
27539         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
27540         reg note.
27541         (note_outside_basic_block_p): Remove all references to
27542         NOTE_INSN_CALL_ARG_LOCATION.
27543         * gengtype.c (adjust_field_rtx_def): Likewise.
27544         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
27545         Likewise.
27546         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
27547         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
27549         PR c++/84444
27550         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
27551         is ADDR_EXPR.
27553         PR tree-optimization/84452
27554         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
27555         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
27556         is NULL.
27558 2018-02-19  Martin Liska  <mliska@suse.cz>
27560         PR sanitizer/82183
27561         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
27563 2018-02-19  Martin Liska  <mliska@suse.cz>
27564             Richard Sandiford  <richard.sandiford@linaro.org>
27566         PR tree-optimization/82491
27567         * gimple-fold.c (get_base_constructor): Make earlier bail out
27568         to prevent ubsan.
27570 2018-02-19  Carl Love  <cel@us.ibm.com>
27572         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
27573         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
27574         BU_P8V_OVERLOAD_1.
27575         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
27576         P8V_BUILTIN_VEC_NEG.
27578 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
27580         * config/rl78/rl78.md (movdf): New define expand.
27582 2018-02-19  Martin Liska  <mliska@suse.cz>
27584         PR other/80589
27585         * doc/invoke.texi: Fix typo.
27586         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
27588 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
27590         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
27591         handle rs6000_single_float and rs6000_double_float specially for
27592         e500 family CPUs.
27594 2018-02-16  Jeff Law  <law@redhat.com>
27596         * config/rx/rx.c (add_pop_cfi_notes): New function.;
27597         (pop_regs): Use it.
27599 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
27601         PR ipa/84425
27602         * ipa-inline.c (inline_small_functions): Fix a typo.
27604 2018-02-16  Nathan Sidwell  <nathan@acm.org>
27606         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
27608 2018-02-16  Carl Love  <cel@us.ibm.com>
27610         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
27611         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
27612         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
27613         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
27614         expansion to P8V_BUILTIN_VEC_FLOAT2.
27616 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
27618         PR rtl-optimization/70023
27619         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
27620         src_regno into account.
27622 2018-02-16  Carl Love  <cel@us.ibm.com>
27624         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
27625         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
27626         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
27627         * config/rs6000/rs6000.c: Remove case statements for
27628         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
27629         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
27630         and P9V_BUILTIN_VEC_VINSERT4B.
27631         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
27632         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
27633         * config/rs6000/vsx.md:
27634         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
27635         vec_insert4b.
27637 2018-02-16  Carl Love  <cel@us.ibm.com>
27639         * config/rs6000/altivec.h: Add builtin names vec_extract4b
27640         vec_insert4b.
27641         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
27642         definitions.
27643         * config/rs6000/rs6000-c.c: Add the definitions for
27644         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
27645         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
27646         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
27647         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
27648         definition for insert4b and define insn *insert3b_internal.
27649         * doc/extend.texi: Add documentation for vec_extract4b.
27651 2018-02-16  Nathan Sidwell  <nathan@acm.org>
27653         * doc/extend.texi (Backwards Compatibility): Mention friend
27654         injection.  Note for-scope is deprecated.
27655         * doc/invoke.texi (-ffriend-injection): Deprecate.
27657 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
27659         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
27660         that moved to I2, also allow destinations that are a paradoxical
27661         subreg (instead of a normal reg).
27663 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
27665         PR target/83831
27666         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
27667         to QImode.
27669 2018-02-16  Richard Biener  <rguenther@suse.de>
27671         PR tree-optimization/84037
27672         PR tree-optimization/84016
27673         PR target/82862
27674         * config/i386/i386.c (ix86_builtin_vectorization_cost):
27675         Adjust vec_construct for the fact we need additional higher latency
27676         128bit inserts for AVX256 and AVX512 vector builds.
27677         (ix86_add_stmt_cost): Scale vector construction cost for
27678         elementwise loads.
27680 2018-02-16  Richard Biener  <rguenther@suse.de>
27682         PR tree-optimization/84417
27683         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
27684         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
27685         (non_rewritable_lvalue_p): Likewise, use poly-ints.
27687 2018-02-16  Martin Liska  <mliska@suse.cz>
27689         PR sanitizer/84307
27690         * internal-fn.def (ASAN_CHECK): Set proper flags.
27691         (ASAN_MARK): Likewise.
27693 2018-02-16  Julia Koval  <julia.koval@intel.com>
27695         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
27696         from PTA_CANNONLAKE.
27698 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
27700         PR target/84272
27701         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
27702         Use ++iter rather than iter++ for std::list iterators.
27703         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
27704         defer deleting them until all nodes in the forest are processed.  Do
27705         free even leaf nodes.  Change to_process into auto_vec.
27707         PR bootstrap/84405
27708         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
27709         * vec.h (vec_default_construct): Use memset instead of placement new
27710         if BROKEN_VALUE_INITIALIZATION is defined.
27711         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
27712         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
27713         is defined.
27715         PR rtl-optimization/83723
27716         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
27717         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
27718         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
27719         recursive calls.
27720         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
27721         callers.
27722         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
27724 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
27726         PR rtl-optimization/81443
27727         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
27728         from inner REGs to paradoxical SUBREGs.
27730 2018-02-16  Richard Biener  <rguenther@suse.de>
27732         PR tree-optimization/84399
27733         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
27734         For operands we can analyze at their definition make sure we can
27735         analyze them at each use as well.
27737 2018-02-16  Richard Biener  <rguenther@suse.de>
27739         PR tree-optimization/84190
27740         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
27741         volatile accesses if the decl isn't volatile.
27743 2018-02-15  Jason Merrill  <jason@redhat.com>
27745         PR c++/84314 - ICE with templates and fastcall attribute.
27746         * attribs.c (build_type_attribute_qual_variant): Don't clobber
27747         TYPE_CANONICAL on an existing type.
27749 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
27751         PR tree-optimization/84383
27752         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
27753         dstoff nor call operand_equal_p if dstbase is NULL.
27755         PR tree-optimization/84334
27756         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
27757         also a CONSTANT_CLASS_P, punt.
27759 2018-02-14  Jim Wilson  <jimw@sifive.com>
27761         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
27762         first SMALL_OPERAND check.  New local min_second_step.  Move assert
27763         to where locals are set.  Add TARGET_RVC support.
27764         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
27766 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
27768         * doc/invoke.texi: Correct -Wformat-overflow code sample.
27770 2018-02-14  Martin Sebor  <msebor@redhat.com>
27772         PR tree-optimization/83698
27773         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
27774         arrays constrain the offset range to their bounds.
27775         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
27776         (builtin_access::overlap): Avoid setting the size of overlap if it's
27777         already been set.
27778         (maybe_diag_overlap): Also consider arrays when deciding what values
27779         of offsets to include in diagnostics.
27781 2018-02-14  Martin Sebor  <msebor@redhat.com>
27783         PR c/84108
27784         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
27785         that correspond to the kind of a declaration.
27787 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
27789         PR target/83984
27790         * config/pa/pa.md: Load address of PIC label using the linkage table
27791         if the label is nonlocal.
27793 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27795         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
27796         warning message if user requests -maltivec=be.
27797         * doc/invoke.texi: Document deprecation of -maltivec=be.
27799 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
27801         PR target/84220
27802         * config/rs6000/rs6000-c.c: Update definitions for
27803         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
27804         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
27806 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
27808         PR target/84239
27809         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
27810         add _get_ssp intrinsics. Remove argument from
27811         __builtin_ia32_rdssp[d|q].
27812         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
27813         * config/i386/i386-builtin.def: Remove argument from
27814         __builtin_ia32_rdssp[d|q].
27815         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
27816         ix86_expand_special_args_builtin for _rdssp[d|q].
27817         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
27818         Clear register before usage.
27819         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
27820         Add documentation for new _get_ssp and _inc_ssp intrinsics.
27822 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
27824         PR tree-optimization/84357
27825         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
27826         operand 1 of an ARRAY_REF too.
27828 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
27830         PR target/83831
27831         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
27832         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
27833         declarations.
27834         (set_of_reg): New struct.
27835         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
27836         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
27837         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
27838         functions.
27839         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
27840         Split into bitclr, bitset, bitinvert patterns if appropriate.
27841         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
27842         use rx_fuse_in_memory_bitop.
27843         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
27844         to named insn, correct maximum insn length.
27846 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
27848         PR target/79242
27849         * machmode.def: Define a complex mode for PARTIAL_INT.
27850         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
27851         MODE_PARTIAL_INT.
27852         * doc/rtl.texi: Document CSPImode.
27853         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
27854         handling.
27855         (msp430_hard_regno_nregs_with_padding): Likewise.
27857 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
27859         PR target/84279
27860         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
27862 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
27864         PR rtl-optimization/84169
27865         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
27866         we generated a parallel as new i3 and we split that to new i2 and i3
27867         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
27868         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
27869         those to i2, not i1.  Partially rewrite this scan code.
27871 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
27873         PR c/82210
27874         * stor-layout.c (place_field): For variable length fields, adjust
27875         offset_align afterwards not just based on the field's alignment,
27876         but also on the size.
27878         PR middle-end/84309
27879         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
27880         of exps and logs in the use_exp2 case.
27882 2018-02-13  Jeff Law  <law@redhat.com>
27884         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
27885         entry for "vector".
27887         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
27888         ARGS as unused.
27890 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
27892         PR debug/84342
27893         PR debug/84319
27894         * common.opt (gas-loc-support, gas-locview-support): New.
27895         (ginline-points, ginternal-reset-location-views): New.
27896         * doc/invoke.texi: Document them.  Use @itemx where intended.
27897         (gvariable-location-views): Adjust.
27898         * target.def (reset_location_view): New.
27899         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
27900         (TARGET_RESET_LOCATION_VIEW): New.
27901         * doc/tm.texi: Rebuilt.
27902         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
27903         (dwarf2out_default_as_locview_support): New.
27904         (output_asm_line_debug_info): Use option variables.
27905         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
27906         (output_loc_list): Likewise.
27907         (add_high_low_attributes): Check option variables.
27908         Don't output entry view attribute in strict mode.
27909         (gen_inlined_subroutine_die): Check option variables.
27910         (dwarf2out_inline_entry): Likewise.
27911         (init_sections_and_labels): Likewise.
27912         (dwarf2out_early_finish): Likewise.
27913         (maybe_reset_location_view): New, from...
27914         (dwarf2out_var_location): ... here.  Call it.
27915         * debug.h (dwarf2out_default_as_loc_support): Declare.
27916         (dwarf2out_default_as_locview_support): Declare.
27917         * hooks.c (hook_int_rtx_insn_0): New.
27918         * hooks.h (hook_int_rtx_insn_0): Declare.
27919         * toplev.c (process_options): Take -gas-loc-support and
27920         -gas-locview-support from dwarf2out.  Enable
27921         -gvariable-location-views by default only with locview
27922         assembler support.  Enable -ginternal-reset-location-views by
27923         default only if the target defines the corresponding hook.
27924         Enable -ginline-points by default if location views are
27925         enabled; force it disabled if statement frontiers are
27926         disabled.
27927         * tree-inline.c (expand_call_inline): Check option variables.
27928         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
27930 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
27932         PR tree-optimization/84321
27933         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
27934         handling.  Also check whether the anti-range contains any values
27935         that satisfy the mask; switch to a VR_RANGE if not.
27937 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
27939         PR sanitizer/84340
27940         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
27942 2018-02-13  Martin Jambor  <mjambor@suse.cz>
27944         PR c++/83990
27945         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
27946         of call statements, also set location of a load to a temporary.
27948 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
27950         * config/rl78/rl78.c (add_vector_labels): New function.
27951         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
27952         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
27953         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
27954         which checks that no arguments are passed.
27955         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
27956         * doc/extend.texi: Documentation for the new attribute.
27958 2018-02-13  Andreas Schwab  <schwab@suse.de>
27960         * config/riscv/linux.h (CPP_SPEC): Define.
27962 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
27964         PR target/84335
27965         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
27966         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
27967         OPTION_MASK_ISA_AES as first argument to def_builtin_const
27968         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
27969         instead of OPTION_MASK_ISA_PCLMUL as first argument to
27970         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
27971         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
27972         temporarily for AES and PCLMUL builtins.
27974         PR tree-optimization/84339
27975         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
27976         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
27977         Formatting fixes.
27979         PR middle-end/84309
27980         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
27981         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
27982         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
27983         inline function.
27984         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
27985         inline function.
27986         * omp-simd-clone.h: New file.
27987         * omp-simd-clone.c: Include omp-simd-clone.h.
27988         (expand_simd_clones): No longer static.
27989         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
27990         cgraph.h and omp-simd-clone.h.
27991         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
27992         (vect_recog_widen_shift_pattern): Formatting fix.
27993         (vect_pattern_recog_1): Don't check optab for calls.
27995         PR target/84336
27996         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
27997         operands[2] into a REG before using gen_lowpart on it.
27999 2018-02-12  Jeff Law  <law@redhat.com>
28001         PR target/83760
28002         * config/sh/sh.c (find_barrier): Consider a sibling call
28003         a barrier as well.
28005         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
28006         successfully back substituting a reg.
28008 2018-02-12  Richard Biener  <rguenther@suse.de>
28010         PR tree-optimization/84037
28011         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
28012         parameter, move visited init to caller.
28013         (vect_slp_analyze_operations): Separate cost from validity
28014         check, initialize visited once for all instances.
28015         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
28016         for all instances.
28017         * tree-vect-stmts.c (vect_model_simple_cost): Make early
28018         out an assert.
28019         (vect_model_promotion_demotion_cost): Likewise.
28020         (vectorizable_bswap): Guard cost modeling with !slp_node
28021         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
28022         SLP stmts.
28023         (vectorizable_call): Likewise.
28024         (vectorizable_conversion): Likewise.
28025         (vectorizable_assignment): Likewise.
28026         (vectorizable_shift): Likewise.
28027         (vectorizable_operation): Likewise.
28028         (vectorizable_store): Likewise.
28029         (vectorizable_load): Likewise.
28030         (vectorizable_condition): Likewise.
28031         (vectorizable_comparison): Likewise.
28033 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
28035         PR sanitizer/84307
28036         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
28037         (ASAN_MARK): Fix fnspec to account for return value, change pointer
28038         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
28040 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
28042         PR middle-end/83665
28043         * params.def (inline-min-speedup): Increase from 8 to 15.
28044         (max-inline-insns-auto): Decrease from 40 to 30.
28045         * ipa-split.c (consider_split): Add some buffer for function to
28046         be considered inlining candidate.
28047         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
28048         default values.
28050 2018-02-12  Richard Biener  <rguenther@suse.de>
28052         PR tree-optimization/84037
28053         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
28054         matched stmts if we cannot swap the non-matched ones.
28056 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
28058         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
28059         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
28060         _mm_maskz_scalef_round_ss): New intrinsics.
28061         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
28062         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
28063         __builtin_ia32_scalefss_round): Remove.
28064         (__builtin_ia32_scalefsd_mask_round,
28065         __builtin_ia32_scalefss_mask_round): New intrinsics.
28066         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
28067         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
28068         ((match_operand:VF_128 2 "<round_nimm_predicate>"
28069         "<round_constraint>")): Changed to ...
28070         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
28071         "<round_scalar_constraint>")): ... this.
28072         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
28073         %0, %1, %2<round_op3>}"): Changed to ...
28074         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
28075         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
28076         %2<round_scalar_mask_op3>}"): ... this.
28077         * config/i386/subst.md (round_scalar_nimm_predicate): New.
28079 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
28081         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
28082         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
28083         (_mm_maskz_sqrt_round_ss): New intrinsics.
28084         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
28085         (__builtin_ia32_sqrtsd_mask_round)
28086         (__builtin_ia32_sqrtss_mask_round): New builtins.
28087         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
28088         (__builtin_ia32_sqrtss_round): Remove.
28089         (__builtin_ia32_sqrtsd_mask_round)
28090         (__builtin_ia32_sqrtss_mask_round): New builtins.
28091         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
28092         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
28093         ((match_operand:VF_128 1 "vector_operand"
28094         "xBm,<round_constraint>")): Changed to ...
28095         ((match_operand:VF_128 1 "vector_operand"
28096         "xBm,<round_scalar_constraint>")): ... this.
28097         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
28098         %0, %2, %<iptr>1<round_op3>}): Changed to ...
28099         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
28100         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
28101         %<iptr>1<round_scalar_mask_op3>}): ... this.
28102         ((set_attr "prefix" "<round_prefix>")): Changed to ...
28103         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
28105 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
28107         PR target/84266
28108         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
28109         Cast vec_cmpeq result to correct type.
28110         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
28111         Cast vec_cmpgt result to correct type.
28113 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
28115         * final.c (final_scan_insn_1): Renamed from...
28116         (final_scan_insn): ... this.  New wrapper, to recover
28117         seen from the outermost call in recursive ones.
28118         * config/sparc/sparc.c (output_return): Drop seen from call.
28119         (output_sibcall): Likewise.
28120         * config/visium/visium.c (output_branch): Likewise.
28122 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
28124         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
28125         function label.
28127 2018-02-10  Alan Modra  <amodra@gmail.com>
28129         PR target/84300
28130         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
28131         Specify LR as an input.
28133 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
28135         PR sanitizer/83987
28136         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
28137         remove_member_access_dummy_vars): New functions.
28138         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
28139         lower_omp_1, execute_lower_omp): Use them.
28141         PR rtl-optimization/84308
28142         * shrink-wrap.c (spread_components): Release todo vector.
28144 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
28146         PR rtl-optimization/57193
28147         * ira-color.c (struct allocno_color_data): Add member
28148         conflict_allocno_hard_prefs.
28149         (update_conflict_allocno_hard_prefs): New.
28150         (bucket_allocno_compare_func): Add a preference based on
28151         conflict_allocno_hard_prefs.
28152         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
28153         (color_allocnos): Remove a dead code.  Initiate
28154         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
28156 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
28158         PR target/84226
28159         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
28160         constraint from =wa to wa.  Avoid a subreg on the output operand,
28161         instead use a pseudo and subreg it in a move.
28162         (p9_xxbrd_<mode>): Changed to ...
28163         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
28164         (p9_xxbrd_v2df): New expander.
28165         (p9_xxbrw_<mode>): Changed to ...
28166         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
28167         (p9_xxbrw_v4sf): New expander.
28169 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
28171         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
28173 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
28175         PR target/83926
28176         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
28177         multiply in 32-bit mode.
28178         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
28179         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
28180         mode.
28182 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
28184         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
28185         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
28186         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
28187         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
28189 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
28191         PR lto/84213
28192         * dwarf2out.c (is_trivial_indirect_ref): New function.
28193         (dwarf2out_late_global_decl): Do not generate a location
28194         attribute for variables that have a non-trivial DECL_VALUE_EXPR
28195         and that are not defined in the current unit.
28197 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
28199         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
28200         instead of a libcall for UNORDERED.
28202 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
28204         PR target/82641
28205         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
28206         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
28208 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28210         PR target/PR84295
28211         * config/s390/s390.c (s390_set_current_function): Invoke
28212         s390_indirect_branch_settings also if fndecl didn't change.
28214 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
28216         * config/rs6000/rs6000.md (blockage): Set length to zero.
28218 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
28220         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
28222 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
28224         PR sanitizer/84285
28225         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
28226         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
28227         -static-lib*san.
28229         PR debug/84252
28230         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
28231         PARALLEL incoming that failed vt_get_decl_and_offset check.
28233         PR middle-end/84237
28234         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
28235         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
28236         TREE_READONLY bit.
28237         (get_variable_section): For decls in named .bss* sections pass true as
28238         second argument to bss_initializer_p.
28240 2018-02-09  Marek Polacek  <polacek@redhat.com>
28241             Jakub Jelinek  <jakub@redhat.com>
28243         PR c++/83659
28244         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
28245         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
28246         Sync some changes from cxx_fold_indirect_ref.
28248 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
28250         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
28251         markers.
28252         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
28253         (BLOCK_INLINE_ENTRY_LABEL): New.
28254         (dwarf2out_var_location): Disregard inline entry markers.
28255         (inline_entry_data): New struct.
28256         (inline_entry_data_hasher): New hashtable type.
28257         (inline_entry_data_hasher::hash): New.
28258         (inline_entry_data_hasher::equal): New.
28259         (inline_entry_data_table): New variable.
28260         (add_high_low_attributes): Add DW_AT_entry_pc and
28261         DW_AT_GNU_entry_view attributes if a pending entry is found
28262         in inline_entry_data_table.  Add old entry_pc attribute only
28263         if debug nonbinding markers are disabled.
28264         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
28265         markers are enabled.
28266         (block_within_block_p, dwarf2out_inline_entry): New.
28267         (dwarf2out_finish): Check that no entries remained in
28268         inline_entry_data_table.
28269         * final.c (reemit_insn_block_notes): Handle inline entry notes.
28270         (final_scan_insn, notice_source_line): Likewise.
28271         (rest_of_clean_state): Skip inline entry markers.
28272         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
28273         markers.
28274         * gimple.c (gimple_build_debug_inline_entry): New.
28275         * gimple.h (enum gimple_debug_subcode): Add
28276         GIMPLE_DEBUG_INLINE_ENTRY.
28277         (gimple_build_debug_inline_entry): Declare.
28278         (gimple_debug_inline_entry_p): New.
28279         (gimple_debug_nonbind_marker_p): Adjust.
28280         * insn-notes.def (INLINE_ENTRY): New.
28281         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
28282         inline entry marker notes.
28283         (print_insn): Likewise.
28284         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
28285         (INSN_DEBUG_MARKER_KIND): Likewise.
28286         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
28287         * tree-inline.c (expand_call_inline): Build and insert
28288         debug_inline_entry stmt.
28289         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
28290         inline entry blocks early, if nonbind markers are enabled.
28291         (dump_scope_block): Dump fragment info.
28292         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
28293         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
28294         (gimple_build_debug_inline_entry): New.
28295         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
28296         Enable/disable inline entry points too.
28297         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
28298         (DEBUG_INSN): Describe inline entry markers.
28300         * common.opt (gvariable-location-views): New.
28301         (gvariable-location-views=incompat5): New.
28302         * config.in: Rebuilt.
28303         * configure: Rebuilt.
28304         * configure.ac: Test assembler for view support.
28305         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
28306         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
28307         * dwarf2out.c (var_loc_view): New typedef.
28308         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
28309         (dwarf2out_locviews_in_attribute): New.
28310         (dwarf2out_locviews_in_loclist): New.
28311         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
28312         (enum dw_line_info_opcode): Add LI_adv_address.
28313         (struct dw_line_info_table): Add view.
28314         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
28315         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
28316         (zero_view_p): New variable.
28317         (ZERO_VIEW_P): New macro.
28318         (output_asm_line_debug_info): New.
28319         (struct var_loc_node): Add view.
28320         (add_AT_view_list, AT_loc_list): New.
28321         (add_var_loc_to_decl): Add view param.  Test it against last.
28322         (new_loc_list): Add view params.  Record them.
28323         (AT_loc_list_ptr): Handle loc and view lists.
28324         (view_list_to_loc_list_val_node): New.
28325         (print_dw_val): Handle dw_val_class_view_list.
28326         (size_of_die): Likewise.
28327         (value_format): Likewise.
28328         (loc_list_has_views): New.
28329         (gen_llsym): Set vl_symbol too.
28330         (maybe_gen_llsym, skip_loc_list_entry): New.
28331         (dwarf2out_maybe_output_loclist_view_pair): New.
28332         (output_loc_list): Output view list or entries too.
28333         (output_view_list_offset): New.
28334         (output_die): Handle dw_val_class_view_list.
28335         (output_dwarf_version): New.
28336         (output_compilation_unit_header): Use it.
28337         (output_skeleton_debug_sections): Likewise.
28338         (output_rnglists, output_line_info): Likewise.
28339         (output_pubnames, output_aranges): Update version comments.
28340         (output_one_line_info_table): Output view numbers in asm comments.
28341         (dw_loc_list): Determine current endview, pass it to new_loc_list.
28342         Call maybe_gen_llsym.
28343         (loc_list_from_tree_1): Adjust.
28344         (add_AT_location_description): Create view list attribute if
28345         needed, check it's absent otherwise.
28346         (convert_cfa_to_fb_loc_list): Adjust.
28347         (maybe_emit_file): Call output_asm_line_debug_info for test.
28348         (dwarf2out_var_location): Reset views as needed.  Precompute
28349         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
28350         attribute.  Set view.
28351         (new_line_info_table): Reset next view.
28352         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
28353         (dwarf2out_source_line): Likewise.  Output view resets and labels to
28354         the assembler, or select appropriate line info opcodes.
28355         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
28356         (optimize_string_length): Catch it.  Adjust.
28357         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
28358         dw_val_class_view_list, and remove it if no longer needed.
28359         (hash_loc_list): Hash view numbers.
28360         (loc_list_hasher::equal): Compare them.
28361         (optimize_location_lists): Check whether a view list symbol is
28362         needed, and whether the locview attribute is present, and
28363         whether they match.  Remove the locview attribute if no longer
28364         needed.
28365         (index_location_lists): Call skip_loc_list_entry for test.
28366         (dwarf2out_finish): Call output_asm_line_debug_info for test.
28367         Use output_dwarf_version.
28368         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
28369         (struct dw_val_node): Add val_view_list.
28370         * final.c (SEEN_NEXT_VIEW): New.
28371         (set_next_view_needed): New.
28372         (clear_next_view_needed): New.
28373         (maybe_output_next_view): New.
28374         (final_start_function): Rename to...
28375         (final_start_function_1): ... this.  Take pointer to FIRST,
28376         add SEEN parameter.  Emit param bindings in the initial view.
28377         (final_start_function): Reintroduce SEEN-less interface.
28378         (final): Rename to...
28379         (final_1): ... this.  Take SEEN parameter.  Output final pending
28380         next view at the end.
28381         (final): Reintroduce seen-less interface.
28382         (final_scan_insn): Output pending next view before switching
28383         sections or ending a block.  Mark the next view as needed when
28384         outputting variable locations.  Notify debug backend of section
28385         changes, and of location view changes.
28386         (rest_of_handle_final): Adjust.
28387         * toplev.c (process_options): Autodetect value for debug variable
28388         location views option.  Warn on incompat5 without -gdwarf-5.
28389         * doc/invoke.texi (gvariable-location-views): New.
28390         (gvariable-location-views=incompat5): New.
28391         (gno-variable-location-views): New.
28393 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
28395         PR tree-optimization/84136
28396         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
28397         that the result of find_edge is non-NULL.
28399 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
28401         PR target/83008
28402         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
28403         storing integer register in SImode.  Fix cost of 256 and 512
28404         byte aligned SSE register store.
28406 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
28408         * config/i386/i386.c (ix86_multiplication_cost): Fix
28409         multiplication cost for TARGET_AVX512DQ.
28411 2018-02-08  Marek Polacek  <polacek@redhat.com>
28413         PR tree-optimization/84238
28414         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
28415         get_range_strlen.
28417 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
28419         PR tree-optimization/84265
28420         * tree-vect-stmts.c (vectorizable_store): Don't treat
28421         VMAT_CONTIGUOUS accesses as grouped.
28422         (vectorizable_load): Likewise.
28424 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
28426         PR tree-optimization/81635
28427         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
28428         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
28429         (test_round_for_mask): New functions.
28430         (wide_int_cc_tests): Call test_round_for_mask.
28431         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
28432         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
28433         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
28434         range returned by get_range_info.
28436 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
28438         PR ipa/81360
28439         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
28440         * symtab.c: Include builtins.h
28441         (symtab_node::output_to_lto_symbol_table_p): Move here
28442         from lto-streamer-out.c:output_symbol_p.
28443         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
28444         (output_symbol_p): Move all logic to symtab.c
28445         (produce_symtab): Update.
28447 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28449         * config/s390/s390-opts.h (enum indirect_branch): Define.
28450         * config/s390/s390-protos.h (s390_return_addr_from_memory)
28451         (s390_indirect_branch_via_thunk)
28452         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
28453         (enum s390_indirect_branch_type): Define.
28454         * config/s390/s390.c (struct s390_frame_layout, struct
28455         machine_function): Remove.
28456         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
28457         (indirect_branch_table_label_no, indirect_branch_table_name):
28458         Define variables.
28459         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
28460         (enum s390_indirect_branch_option): Define.
28461         (s390_return_addr_from_memory): New function.
28462         (s390_handle_string_attribute): New function.
28463         (s390_attribute_table): Add new attribute handler.
28464         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
28465         (s390_indirect_branch_via_thunk): New function.
28466         (s390_indirect_branch_via_inline_thunk): New function.
28467         (s390_function_ok_for_sibcall): When jumping via thunk disallow
28468         sibling call optimization for non z10 compiles.
28469         (s390_emit_call): Force indirect branch target to be a single
28470         register.  Add r1 clobber for non-z10 compiles.
28471         (s390_emit_epilogue): Emit return jump via return_use expander.
28472         (s390_reorg): Handle JUMP_INSNs as execute targets.
28473         (s390_option_override_internal): Perform validity checks for the
28474         new command line options.
28475         (s390_indirect_branch_attrvalue): New function.
28476         (s390_indirect_branch_settings): New function.
28477         (s390_set_current_function): Invoke s390_indirect_branch_settings.
28478         (s390_output_indirect_thunk_function):  New function.
28479         (s390_code_end): Implement target hook.
28480         (s390_case_values_threshold): Implement target hook.
28481         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
28482         macros.
28483         * config/s390/s390.h (struct s390_frame_layout)
28484         (struct machine_function): Move here from s390.c.
28485         (TARGET_INDIRECT_BRANCH_NOBP_RET)
28486         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
28487         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
28488         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
28489         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
28490         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
28491         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
28492         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
28493         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
28494         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
28495         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
28496         (mnemonic attribute): Add values which aren't recognized
28497         automatically.
28498         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
28499         pattern for branch conversion.  Fix mnemonic attribute.
28500         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
28501         indirect branch via thunk if requested.
28502         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
28503         ("*indirect_jump"): Disable for branch conversion using out of
28504         line thunks.
28505         ("indirect_jump_via_thunk<mode>_z10")
28506         ("indirect_jump_via_thunk<mode>")
28507         ("indirect_jump_via_inlinethunk<mode>_z10")
28508         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
28509         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
28510         ("casesi_jump_via_inlinethunk<mode>_z10")
28511         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
28512         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
28513         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
28514         ("*indirect2_jump"): Disable for branch conversion.
28515         ("casesi_jump"): Turn into expander and expand patterns for branch
28516         conversion.
28517         ("return_use"): New expander.
28518         ("*return"): Emit return via thunk and rename it to ...
28519         ("*return<mode>"): ... this one.
28520         * config/s390/s390.opt: Add new options and and enum for the
28521         option values.
28523 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
28525         * lra-constraints.c (match_reload): Unconditionally use
28526         gen_lowpart_SUBREG, rather than selecting between that
28527         and equivalent gen_rtx_SUBREG code.
28529 2018-02-08  Richard Biener  <rguenther@suse.de>
28531         PR tree-optimization/84233
28532         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
28533         changed flag instead of boguously re-using phi_inserted.
28535 2018-02-08  Martin Jambor  <mjambor@suse.cz>
28537         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
28538         static local variables.
28540 2018-02-08  Richard Biener  <rguenther@suse.de>
28542         PR tree-optimization/84278
28543         * tree-vect-stmts.c (vectorizable_store): When looking for
28544         smaller vector types to perform grouped strided loads/stores
28545         make sure the mode is supported by the target.
28546         (vectorizable_load): Likewise.
28548 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
28550         * config/aarch64/aarch64.c (aarch64_components_for_bb):
28551         Increase LDP/STP opportunities by adding adjacent callee-saves.
28553 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
28555         PR rtl-optimization/84068
28556         PR rtl-optimization/83459
28557         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
28559 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
28561         PR tree-optimization/84224
28562         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
28563         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
28564         non-zero arguments.
28566 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
28568         PR target/84113
28569         * config/rs6000/altivec.md (*restore_world): Remove LR use.
28570         * config/rs6000/predicates.md (restore_world_operation): Adjust op
28571         count, remove one USE.
28573 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
28575         * doc/install.texi (Configuration): Document the
28576         --with-long-double-format={ibm,ieee} PowerPC configuration
28577         options.
28579         PR target/84154
28580         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
28581         Convert from define_expand to be define_insn_and_split.  Rework
28582         float/double/_Float128 conversions to QI/HI/SImode to work with
28583         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
28584         conversions to QI/HImode types did a store and then a load to
28585         truncate the value.  For conversions to VSX registers, don't split
28586         the insn, instead emit the code directly.  Use the code iterator
28587         any_fix to combine signed and unsigned conversions.
28588         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
28589         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
28590         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
28591         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
28592         (fix_<mode>di2_hw): Likewise.
28593         (fixuns_<mode>di2_hw): Likewise.
28594         (fix_<mode>si2_hw): Likewise.
28595         (fixuns_<mode>si2_hw): Likewise.
28596         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
28597         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
28598         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
28599         fix<uns>_trunc<SFDF:mode>si2_p8.
28600         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
28601         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
28602         (fix<uns>_<mode>_mem): Likewise.
28603         (fctiw<u>z_<mode>_mem): Likewise.
28604         (fix<uns>_<mode>_mem): Likewise.
28605         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
28606         the register allocator from doing a direct move to the GPRs to do
28607         a store, and instead use the ISA 3.0 store byte/half-word from
28608         vector register instruction.  For IEEE 128-bit floating point,
28609         also optimize stores of 32-bit ints.
28610         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
28612 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
28614         * genextract.c (push_pathstr_operand): New function to support
28615         [a-zA-Z].
28616         (walk_rtx): Call push_pathstr_operand.
28617         (print_path): Support [a-zA-Z].
28619 2018-02-07  Richard Biener  <rguenther@suse.de>
28621         PR tree-optimization/84037
28622         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
28623         (cse_and_gimplify_to_preheader): Declare.
28624         (vect_get_place_in_interleaving_chain): Likewise.
28625         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28626         ivexpr_map.
28627         (_loop_vec_info::~_loop_vec_info): Delete it.
28628         (cse_and_gimplify_to_preheader): New function.
28629         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
28630         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
28631         (vectorizable_load): Likewise.  For grouped stores always base
28632         the IV on the first element.
28633         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
28634         condition before gimplifying.
28636 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
28638         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
28639         *DIV_EXPR and *MOD_EXPR.
28641 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
28643         PR target/84248
28644         * config/i386/i386.c (ix86_option_override_internal): Mask out
28645         the CF_SET bit when checking -fcf-protection.
28647 2018-02-07  Tom de Vries  <tom@codesourcery.com>
28649         PR libgomp/84217
28650         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
28651         enough.
28653 2018-02-07  Richard Biener  <rguenther@suse.de>
28655         PR tree-optimization/84204
28656         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
28657         this place.
28659         PR tree-optimization/84205
28660         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
28661         special-case isl_ast_op_zdiv_r.
28663         PR tree-optimization/84223
28664         * graphite-scop-detection.c (gather_bbs::before_dom_children):
28665         Only add conditions from within the region.
28666         (gather_bbs::after_dom_children): Adjust.
28668 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
28670         PR target/84209
28671         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
28672         * config/avr/avr.md: Only post-reload split REG-REG moves if
28673         either register is GENERAL_REG_P.
28675 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
28677         PR tree-optimization/84235
28678         * tree-ssa-scopedtables.c
28679         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
28680         if the subtraction is performed in floating point type where NaNs are
28681         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
28682         build 1.  Formatting fix.
28684 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
28686         PR target/84146
28687         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
28688         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
28689         and skip it regardless of bb boundaries.  Use CALL_P macro,
28690         don't test INSN_P (insn) together with CALL_P or JUMP_P check
28691         unnecessarily, formatting fix.
28693 2018-02-06  Michael Collison  <michael.collison@arm.com>
28695         * config/arm/thumb2.md:
28696         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
28697         (*thumb_mov_notscc): Ditto.
28699 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
28701         PR target/84154
28702         * config/rs6000/rs6000.md (su code attribute): Use "u" for
28703         unsigned_fix, not "s".
28705 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28707         * configure.ac (gcc_fn_eh_frame_ro): New function.
28708         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
28709         correct .eh_frame permissions.
28710         * configure: Regenerate.
28712 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
28714         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
28715         irrelevant options.
28717 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28719         * config/rs6000/rs6000.c (rs6000_option_override_internal):
28720         Display warning message for -mno-speculate-indirect-jumps.
28722 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
28724         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
28725         Undocumented.
28726         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
28728 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
28730         PR tree-optimization/84225
28731         * tree-eh.c (find_trapping_overflow): Only call
28732         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
28734 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
28736         PR target/84145
28737         * config/i386/i386.c: Reimplement the check of possible options
28738         -mibt/-mshstk conbination. Change error messages.
28739         * doc/invoke.texi: Fix a typo: remove extra '='.
28741 2018-02-06  Marek Polacek  <polacek@redhat.com>
28743         PR tree-optimization/84228
28744         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
28746 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
28748         PR target/82641
28749         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
28750         emitted arch directives.
28751         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
28752         __ARM_FEATURE_COPROC before changing architectures.
28754 2018-02-06  Richard Biener  <rguenther@suse.de>
28756         * config/i386/i386.c (print_reg): Fix typo.
28757         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
28759 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
28761         * configure: Regenerate.
28763 2018-02-05  Martin Sebor  <msebor@redhat.com>
28765         PR tree-optimization/83369
28766         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
28767         inlining context.
28769 2018-02-05  Martin Liska  <mliska@suse.cz>
28771         * doc/invoke.texi: Cherry-pick upstream r323995.
28773 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
28775         * ira.c (ira_init_register_move_cost): Adjust comment.
28777 2018-02-05  Martin Liska  <mliska@suse.cz>
28779         PR gcov-profile/84137
28780         * doc/gcov.texi: Fix typo in documentation.
28782 2018-02-05  Martin Liska  <mliska@suse.cz>
28784         PR gcov-profile/83879
28785         * doc/gcov.texi: Document necessity of --dynamic-list-data when
28786         using dlopen functionality.
28788 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
28790         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
28791         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
28792         _mm_maskz_range_ss, _mm_mask_range_round_ss,
28793         _mm_maskz_range_round_ss): New intrinsics.
28794         (__builtin_ia32_rangesd128_round)
28795         (__builtin_ia32_rangess128_round): Remove.
28796         (__builtin_ia32_rangesd128_mask_round,
28797         __builtin_ia32_rangess128_mask_round): New builtins.
28798         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
28799         __builtin_ia32_rangess128_round): Remove.
28800         (__builtin_ia32_rangesd128_mask_round,
28801         __builtin_ia32_rangess128_mask_round): New builtins.
28802         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
28803         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
28804         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
28805         "<round_saeonly_constraint>")): Changed to ...
28806         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
28807         "<round_saeonly_scalar_constraint>")): ... this.
28808         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
28809         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
28810         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
28811         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
28812         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
28814 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
28816         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
28817         options.
28818         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
28819         Remove all values except native, 8540 and 8548.
28821 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
28823         * config/i386/i386.c (ix86_output_function_return): Pass
28824         INVALID_REGNUM, instead of -1, as invalid register number to
28825         indirect_thunk_name and output_indirect_thunk.
28827 2018-02-02  Julia Koval  <julia.koval@intel.com>
28829         * config.gcc: Add -march=icelake.
28830         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
28831         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
28832         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
28833         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
28834         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
28835         (processor_target_table): Add icelake.
28836         (ix86_option_override_internal): Handle new PTAs.
28837         (get_builtin_code_for_version): Handle icelake.
28838         (M_INTEL_COREI7_ICELAKE): New.
28839         (fold_builtin_cpu): Handle icelake.
28840         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
28841         * doc/invoke.texi: Add -march=icelake.
28843 2018-02-02  Julia Koval  <julia.koval@intel.com>
28845         * config/i386/i386.c (ix86_option_override_internal): Change flags type
28846         to wide_int_bitmask.
28847         * wide-int-bitmask.h: New.
28849 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
28851         PR target/84066
28852         * config/i386/i386.md: Replace Pmode with word_mode in
28853         builtin_setjmp_setup and builtin_longjmp to support x32.
28855 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
28857         PR target/56010
28858         PR target/83743
28859         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
28860         #include "opts.h".
28861         (rs6000_supported_cpu_names): New static variable.
28862         (linux_cpu_translation_table): Likewise.
28863         (elf_platform) <cpu>: Define new static variable and use it.
28864         Translate kernel AT_PLATFORM name to canonical name if needed.
28865         Error if platform name is unknown.
28867 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
28869         PR target/84089
28870         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
28872 2018-02-01  Jeff Law  <law@redhat.com>
28874         PR target/84128
28875         * config/i386/i386.c (release_scratch_register_on_entry): Add new
28876         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
28877         the scratch if RELEASE_VIA_POP is false.
28878         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
28879         If we have to save a temporary register, decrement SIZE appropriately.
28880         Pass new arguments to release_scratch_register_on_entry.
28881         (ix86_adjust_stack_and_probe): Likewise.
28882         (ix86_emit_probe_stack_range): Pass new arguments to
28883         release_scratch_register_on_entry.
28885 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
28887         PR rtl-optimization/84157
28888         * combine.c (change_zero_ext): Use REG_P predicate in
28889         front of HARD_REGISTER_P predicate.
28891 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
28893         * config/avr/avr.c (avr_option_override): Move disabling of
28894         -fdelete-null-pointer-checks to...
28895         * common/config/avr/avr-common.c (avr_option_optimization_table):
28896         ...here.
28898 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28900         PR tree-optimization/81635
28901         * tree-data-ref.c (split_constant_offset_1): For types that
28902         wrap on overflow, try to use range info to prove that wrapping
28903         cannot occur.
28905 2018-02-01  Renlin Li  <renlin.li@arm.com>
28907         PR target/83370
28908         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
28909         TAILCALL_ADDR_REGS.
28910         (aarch64_register_move_cost): Likewise.
28911         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
28912         TAILCALL_ADDR_REGS.
28913         (REG_CLASS_NAMES): Likewise.
28914         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
28915         TAILCALL_ADDR_REGS. Remove IP registers.
28916         * config/aarch64/aarch64.md (Ucs): Update register constraint.
28918 2018-02-01  Richard Biener  <rguenther@suse.de>
28920         * domwalk.h (dom_walker::dom_walker): Add additional constructor
28921         for specifying RPO order and allow NULL for that.
28922         * domwalk.c (dom_walker::dom_walker): Likewise.
28923         (dom_walker::walk): Handle NULL RPO order.
28924         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
28925         in RPO order.
28926         (rewrite_update_dom_walker): Likewise.
28927         (mark_def_dom_walker): Likewise.
28929 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28931         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
28932         (aarch64_maybe_expand_sve_subreg_move): Declare.
28933         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
28934         * config/aarch64/predicates.md (aarch64_any_register_operand): New
28935         predicate.
28936         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
28937         that are semantically a reverse operation.
28938         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
28939         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
28940         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
28941         functions.
28942         (aarch64_can_change_mode_class): For big-endian, forbid changes
28943         between two SVE modes if they have different element sizes.
28945 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28947         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
28948         the TImode handling for big-endian targets.
28950 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28952         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
28953         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
28954         not just bytes.
28955         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
28956         Remove BSWAP handing for big-endian targets and use the form of
28957         LD1RQ appropariate for the mode.
28959 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28961         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
28962         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
28963         duplicated element.
28965 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
28967         PR tearget/83845
28968         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
28969         check for operands that need to go through aarch64_sve_reload_be.
28971 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
28973         PR tree-optimization/81661
28974         PR tree-optimization/84117
28975         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
28976         * tree-eh.c: Include gimplify.h.
28977         (find_trapping_overflow, replace_trapping_overflow,
28978         rewrite_to_non_trapping_overflow): New functions.
28979         * tree-vect-loop.c: Include tree-eh.h.
28980         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
28981         * tree-data-ref.c: Include tree-eh.h.
28982         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
28984 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
28986         PR rtl-optimization/84123
28987         * combine.c (change_zero_ext): Check if hard register satisfies
28988         can_change_dest_mode before calling gen_lowpart_SUBREG.
28990 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
28992         PR target/82444
28993         * ira.c (ira_init_register_move_cost): Remove assert.
28995 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
28997         PR rtl-optimization/84071
28998         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
28999         * doc/tm.texi: Regenerate.
29001 2018-01-31  Richard Biener  <rguenther@suse.de>
29003         PR tree-optimization/84132
29004         * tree-data-ref.c (analyze_miv_subscript): Properly
29005         check whether evolution_function_is_affine_multivariate_p
29006         before calling gcd_of_steps_may_divide_p.
29008 2018-01-31  Julia Koval  <julia.koval@intel.com>
29010         PR target/83618
29011         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
29012         * config/i386/i386.md (rdpid_rex64) New.
29013         (rdpid): Make 32bit only.
29015 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
29017         PR lto/84105
29018         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
29019         an IDENTIFIER_NODE for FUNCTION_TYPE's.
29021 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29023         Revert
29024         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29026         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29028 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
29030         PR rtl-optimization/84071
29031         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
29032         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
29034 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
29036         * config/arc/arc.c (arc_handle_aux_attribute): New function.
29037         (arc_attribute_table): Add 'aux' attribute.
29038         (arc_in_small_data_p): Consider aux like variables.
29039         (arc_is_aux_reg_p): New function.
29040         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
29041         (arc_get_aux_arg): New function.
29042         (prepare_move_operands): Handle aux-register access.
29043         (arc_handle_aux_attribute): New function.
29044         * doc/extend.texi (ARC Variable attributes): Add subsection.
29046 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
29048         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
29049         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
29050         (arc_attribute_table): Add 'uncached' attribute.
29051         (arc_print_operand): Print '.di' flag for uncached memory
29052         accesses.
29053         (arc_in_small_data_p): Do not consider for small data the uncached
29054         types.
29055         (arc_is_uncached_mem_p): New function.
29056         * config/arc/predicates.md (compact_store_memory_operand): Check
29057         for uncached memory accesses.
29058         (nonvol_nonimm_operand): Likewise.
29059         * doc/extend.texi (ARC Type Attribute): New subsection.
29061 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
29063         PR c/84100
29064         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
29065         falign-loops=): Add Optimization flag.
29067 2018-01-30  Jeff Law  <law@redhat.com>
29069         PR target/84064
29070         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
29071         INT_REGISTERS_SAVED.  Check it prior to calling
29072         get_scratch_register_on_entry.
29073         (ix86_adjust_stack_and_probe): Similarly.
29074         (ix86_emit_probe_stack_range): Similarly.
29075         (ix86_expand_prologue): Corresponding changes.
29077 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29079         PR target/40411
29080         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
29081         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
29083 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
29085         PR target/84112
29086         * lra-constraints.c (curr_insn_transform): Process AND in the
29087         address.
29089 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
29091         PR rtl-optimization/83986
29092         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
29093         dependence against last_pending_memory_flush in addition to
29094         pending_jump_insns.
29096 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
29098         PR tree-optimization/81611
29099         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
29100         copies.
29102 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29104         PR target/83758
29105         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
29106         a reg rtx.
29108 2018-01-30  Richard Biener  <rguenther@suse.de>
29109             Jakub Jelinek  <jakub@redhat.com>
29111         PR tree-optimization/84111
29112         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
29113         inner loops added during recursion, as they don't have up-to-date
29114         SSA form.
29116 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
29118         PR ipa/81360
29119         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
29120         (can_inline_edge_by_limits_p): ... here.
29121         (can_early_inline_edge_p, check_callers,
29122         update_caller_keys, update_callee_keys, recursive_inlining,
29123         add_new_edges_to_heap, speculation_useful_p,
29124         inline_small_functions,
29125         inline_small_functions, flatten_function,
29126         inline_to_all_callers_1): Update.
29128 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
29130         * profile-count.c (profile_count::combine_with_ipa_count): Handle
29131         zeros correctly.
29133 2018-01-30  Richard Biener  <rguenther@suse.de>
29135         PR tree-optimization/83008
29136         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
29137         invariant and constant vector uses in stmts when they need
29138         more than one stmt.
29140 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29142         PR bootstrap/84017
29143         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
29144         * configure: Regenerate.
29146 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29148         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
29149         pattern.
29150         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
29151         Use gen_rtx_REG rather than gen_lowpart.
29153 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29155         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
29156         rather than 0 when creating partial subregs.
29158 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
29160         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
29161         of usage.
29163 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
29165         PR target/81550
29166         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
29167         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
29168         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
29169         flags.  This restores the settings used before the 2017-07-24.
29170         Turning off pre increment/decrement/modify allows IVOPTS to
29171         optimize DF/SF loops where the index is an int.
29173 2018-01-29  Richard Biener  <rguenther@suse.de>
29174             Kelvin Nilsen  <kelvin@gcc.gnu.org>
29176         PR bootstrap/80867
29177         * tree-vect-stmts.c (vectorizable_call): Don't call
29178         targetm.vectorize_builtin_md_vectorized_function if callee is
29179         NULL.
29181 2018-01-22  Carl Love  <cel@us.ibm.com>
29183         * doc/extend.tex: Fix typo in second arg in
29184         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
29186 2018-01-29  Richard Biener  <rguenther@suse.de>
29188         PR tree-optimization/84086
29189         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
29190         (flush_ssaname_freelist): When SSA names were released reset
29191         the SCEV hash table.
29193 2018-01-29  Richard Biener  <rguenther@suse.de>
29195         PR tree-optimization/84057
29196         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
29197         removed paths when removing edges.
29199 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
29201         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
29202         -mfunction-return=@var{choice}.
29204 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29206         PR diagnostic/84034
29207         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
29208         Handle CR like TAB.
29209         (layout::print_source_line): Likewise.
29210         (test_get_line_width_without_trailing_whitespace): Add test cases.
29212 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
29214         PR middle-end/84040
29215         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
29216         debug insns.
29218 2018-01-26  Jim Wilson  <jimw@sifive.com>
29220         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
29222         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
29223         specified.
29225 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29227         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
29228         and CMP + SUB-immediate -> SUBS.
29230 2018-01-26  Martin Sebor  <msebor@redhat.com>
29232         PR tree-optimization/83896
29233         * tree-ssa-strlen.c (get_string_len): Rename...
29234         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
29235         Avoid assuming length is constant.
29236         (handle_char_store): Use HOST_WIDE_INT for string length.
29238 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
29240         PR target/81763
29241         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
29242         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
29244 2018-01-26  Richard Biener  <rguenther@suse.de>
29246         PR rtl-optimization/84003
29247         * dse.c (record_store): Only record redundant stores when
29248         the earlier store aliases at least all accesses the later one does.
29250 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
29252         PR rtl-optimization/83985
29253         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
29254         REG_CFA_RESTORE insns.
29255         (delete_unmarked_insns): Don't ignore separate shrink wrapping
29256         REG_CFA_RESTORE insns here.
29258         PR c/83989
29259         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
29260         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
29262 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29264         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
29265         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
29266         (arc_init): Likewise.
29267         (arc_override_options): Likewise.
29268         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
29269         value.
29270         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
29271         support.
29272         * config/arc/arc.h (TARGET_DBNZ): Define.
29273         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
29274         properly set the tune attribute.
29275         (dbnz): Use TARGET_DBNZ guard.
29276         * config/arc/arc.opt (mtune): Add core3 option.
29278 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29280         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
29281         recognize new pic like addresses.
29282         (arc_delegitimize_address): Clean up.
29284 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29286         * config/arc/arc-arches.def: Option mrf16 valid for all
29287         architectures.
29288         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
29289         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
29290         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
29291         * config/arc/arc-tables.opt: Regenerate.
29292         * config/arc/arc.c (arc_conditional_register_usage): Handle
29293         reduced register file case.
29294         (arc_file_start): Set must have build attributes.
29295         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
29296         mrf16 option value.
29297         * config/arc/arc.opt (mrf16): Add new option.
29298         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
29299         * config/arc/genmultilib.awk: Handle new mrf16 option.
29300         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
29301         * config/arc/t-multilib: Regenerate.
29302         * doc/invoke.texi (ARC Options): Document mrf16 option.
29304 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29306         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
29307         * config/arc/arc.c (arc_handle_secure_attribute): New function.
29308         (arc_attribute_table): Add 'secure_call' attribute.
29309         (arc_print_operand): Print secure call operand.
29310         (arc_function_ok_for_sibcall): Don't optimize tail calls when
29311         secure.
29312         (arc_is_secure_call_p): New function.  * config/arc/arc.md
29313         (call_i): Add support for sjli instruction.
29314         (call_value_i): Likewise.
29315         * config/arc/constraints.md (Csc): New constraint.
29317 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
29318             John Eric Martin  <John.Martin@emmicro-us.com>
29320         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
29321         * config/arc/arc.c (_arc_jli_section): New struct.
29322         (arc_jli_section): New type.
29323         (rc_jli_sections): New static variable.
29324         (arc_handle_jli_attribute): New function.
29325         (arc_attribute_table): Add jli_always and jli_fixed attribute.
29326         (arc_file_end): New function.
29327         (TARGET_ASM_FILE_END): Define.
29328         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
29329         (arc_add_jli_section): New function.
29330         (jli_call_scan): Likewise.
29331         (arc_reorg): Call jli_call_scan.
29332         (arc_output_addsi): Remove 'S' from printing asm operand.
29333         (arc_is_jli_call_p): New function.
29334         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
29335         operand.
29336         (movhi_insn): Likewise.
29337         (movsi_insn): Likewise.
29338         (movsi_set_cc_insn): Likewise.
29339         (loadqi_update): Likewise.
29340         (load_zeroextendqisi_update): Likewise.
29341         (load_signextendqisi_update): Likewise.
29342         (loadhi_update): Likewise.
29343         (load_zeroextendhisi_update): Likewise.
29344         (load_signextendhisi_update): Likewise.
29345         (loadsi_update): Likewise.
29346         (loadsf_update): Likewise.
29347         (movsicc_insn): Likewise.
29348         (bset_insn): Likewise.
29349         (bxor_insn): Likewise.
29350         (bclr_insn): Likewise.
29351         (bmsk_insn): Likewise.
29352         (bicsi3_insn): Likewise.
29353         (cmpsi_cc_c_insn): Likewise.
29354         (movsi_ne): Likewise.
29355         (movsi_cond_exec): Likewise.
29356         (clrsbsi2): Likewise.
29357         (norm_f): Likewise.
29358         (normw): Likewise.
29359         (swap): Likewise.
29360         (divaw): Likewise.
29361         (flag): Likewise.
29362         (sr): Likewise.
29363         (kflag): Likewise.
29364         (ffs): Likewise.
29365         (ffs_f): Likewise.
29366         (fls): Likewise.
29367         (call_i): Remove 'S' asm letter, add jli instruction.
29368         (call_value_i): Likewise.
29369         * config/arc/arc.op (mjli-always): New option.
29370         * config/arc/constraints.md (Cji): New constraint.
29371         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
29372         operand.
29373         (subsf3_fpx): Likewise.
29374         (mulsf3_fpx): Likewise.
29375         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
29376         asm operand.
29377         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
29378         function attrbutes.
29379         * doc/invoke.texi (ARC): Document mjli-always option.
29381 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
29383         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
29384         avoid addition with 0 and use incw and decw where possible.
29386 2018-01-26  Richard Biener  <rguenther@suse.de>
29388         PR tree-optimization/81082
29389         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
29390         association if it requires casting to unsigned.
29391         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
29392         from fold_plusminus_mult_expr to catch important cases late when
29393         range info is available.
29395 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29397         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
29398         * configure.ac (hidden_linkonce): New test.
29399         * configure: Regenerate.
29400         * config.in: Regenerate.
29402 2018-01-26  Julia Koval  <julia.koval@intel.com>
29404         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
29405         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
29406         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
29407         _mm_mask_bitshuffle_epi64_mask): Fix type.
29408         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
29409         USI_FTYPE_V4DI_V4DI_USI): Remove.
29410         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
29411         __builtin_ia32_vpshufbitqmb256_mask,
29412         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
29413         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
29414         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
29416 2018-01-26  Alan Modra  <amodra@gmail.com>
29418         PR target/84033
29419         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
29420         UNSPEC_VBPERMQ.  Sort other unspecs.
29422 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
29424         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
29426 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
29428         PR middle-end/83055
29429         * predict.c (drop_profile): Do not push/pop cfun; update also
29430         node->count.
29431         (handle_missing_profiles): Fix logic looking for zero profiles.
29433 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
29435         PR middle-end/83977
29436         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
29437         on functions with #pragma omp declare simd or functions with simd
29438         attribute.
29439         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
29440         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
29441         Remove trailing \n from warning_at calls.
29443 2018-01-25  Tom de Vries  <tom@codesourcery.com>
29445         PR target/84028
29446         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
29447         for neutered workers.
29449 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
29451         PR target/68467
29452         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
29453         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
29455 2018-01-24  Jeff Law  <law@redhat.com>
29457         PR target/83994
29458         * i386.c (get_probe_interval): Move to earlier point.
29459         (ix86_compute_frame_layout): If -fstack-clash-protection and
29460         the frame is larger than the probe interval, then use pushes
29461         to save registers rather than reg->mem moves.
29462         (ix86_expand_prologue): Remove conditional for int_registers_saved
29463         assertion.
29465 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
29467         PR target/84014
29468         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
29469         min/max for never referenced object.
29471 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
29473         PR middle-end/83977
29474         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
29475         here.
29476         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
29477         attributes from DECL_ATTRIBUTES (decl) without affecting
29478         DECL_ATTRIBUTES (current_function_decl).
29479         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
29480         functions with non-NULL DECL_ABSTRACT_ORIGIN.
29482 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
29484         PR tree-optimization/83979
29485         * fold-const.c (fold_comparison): Use constant_boolean_node
29486         instead of boolean_{true,false}_node.
29488 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
29490         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
29491         with zero counts.
29493 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29495         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
29496         Simplify the clause that sets the length attribute.
29497         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
29498         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
29499         clause that sets the length attribute.
29500         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
29502 2018-01-24  Tom de Vries  <tom@codesourcery.com>
29504         PR target/83589
29505         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
29506         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
29507         Add strict parameter.
29508         (prevent_branch_around_nothing): Insert dummy insn between branch to
29509         label and label with no ptx insn inbetween.
29510         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
29512 2018-01-24  Tom de Vries  <tom@codesourcery.com>
29514         PR target/81352
29515         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
29516         for neutered threads in warp.
29517         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
29519 2018-01-24  Richard Biener  <rguenther@suse.de>
29521         PR tree-optimization/83176
29522         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
29523         operands.
29525 2018-01-24  Richard Biener  <rguenther@suse.de>
29527         PR tree-optimization/82819
29528         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
29529         code generating pluses that are no-ops in the target precision.
29531 2018-01-24  Richard Biener  <rguenther@suse.de>
29533         PR middle-end/84000
29534         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
29536 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
29538         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
29539         to merge probabilities.
29540         * predict.c (probably_never_executed): Also mark as cold functions
29541         with global 0 profile and guessed local profile.
29542         * profile-count.c (profile_probability::combine_with_count): New
29543         member function.
29544         * profile-count.h (profile_probability::operator*,
29545         profile_probability::operator*=, profile_probability::operator/,
29546         profile_probability::operator/=): Reduce precision to adjusted
29547         and set value to guessed on contradictory divisions.
29548         (profile_probability::combine_with_freq): Remove.
29549         (profile_probability::combine_wiht_count): Declare.
29550         (profile_count::force_nonzero):: Set to adjusted.
29551         (profile_count::probability_in):: Set quality to adjusted.
29552         * tree-ssa-tail-merge.c (replace_block_by): Use
29553         combine_with_count.
29555 2018-01-23  Andrew Waterman  <andrew@sifive.com>
29556             Jim Wilson  <jimw@sifive.com>
29558         * config/riscv/riscv.c (riscv_stack_boundary): New.
29559         (riscv_option_override): Set riscv_stack_boundary.  Handle
29560         riscv_preferred_stack_boundary_arg.
29561         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
29562         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
29563         (STACK_BOUNDARY): Set to riscv_stack_boundary.
29564         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
29565         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
29566         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
29568 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
29570         PR target/83905
29571         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
29572         of struct ix86_frame.
29573         (ix86_expand_epilogue): Likewise.  Add a local variable for
29574         the reg_save_offset field in struct ix86_frame.
29576 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
29578         PR tree-optimization/82604
29579         * tree-loop-distribution.c (enum partition_kind): New enum item
29580         PKIND_PARTIAL_MEMSET.
29581         (partition_builtin_p): Support above new enum item.
29582         (generate_code_for_partition): Ditto.
29583         (compute_access_range): Differentiate cases that equality can be
29584         proven at all loops, the innermost loops or no loops.
29585         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
29586         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
29587         (finalize_partitions, distribute_loop): Don't fuse partition of
29588         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
29589         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
29590         parloop is enabled.
29592 2018-01-23  Martin Liska  <mliska@suse.cz>
29594         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
29595         order to ignore the predictor.
29596         (PRED_POLYMORPHIC_CALL): Likewise.
29597         (PRED_RECURSIVE_CALL): Likewise.
29599 2018-01-23  Martin Liska  <mliska@suse.cz>
29601         * tree-profile.c (tree_profiling): Print function header to
29602         aware reader which function we are working on.
29603         * value-prof.c (gimple_find_values_to_profile): Do not print
29604         not interesting value histograms.
29606 2018-01-23  Martin Liska  <mliska@suse.cz>
29608         * profile-count.h (enum profile_quality): Add
29609         profile_uninitialized as the first value. Do not number values
29610         as they are zero based.
29611         (profile_count::verify): Update sanity check.
29612         (profile_probability::verify): Likewise.
29614 2018-01-23  Nathan Sidwell  <nathan@acm.org>
29616         * doc/invoke.texi (ffor-scope): Deprecate.
29618 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
29620         PR tree-optimization/83510
29621         * domwalk.c (set_all_edges_as_executable): New function.
29622         (dom_walker::dom_walker): Convert bool param
29623         "skip_unreachable_blocks" to enum reachability.  Move setup of
29624         edge flags to set_all_edges_as_executable and only do it when
29625         reachability is REACHABLE_BLOCKS.
29626         * domwalk.h (enum dom_walker::reachability): New enum.
29627         (dom_walker::dom_walker): Convert bool param
29628         "skip_unreachable_blocks" to enum reachability.
29629         (set_all_edges_as_executable): New decl.
29630         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
29631         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
29632         "reachability".
29633         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
29634         but converting true to REACHABLE_BLOCKS.
29635         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
29636         * tree-vrp.c
29637         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
29638         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
29639         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
29640         REACHABLE_BLOCKS.
29641         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
29642         if check_all_array_refs will be called.
29644 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
29646         * tree.c (selftest::test_location_wrappers): Add more test
29647         coverage.
29649 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
29651         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
29652         (selftest::test_bit_in_range): Likewise.
29654 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
29656         PR testsuite/83888
29657         * doc/sourcebuild.texi (vect_float): Say that the selector
29658         only describes the situation when -funsafe-math-optimizations is on.
29659         (vect_float_strict): Document.
29661 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
29663         PR tree-optimization/83965
29664         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
29665         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
29666         instead of checking only for a reduction.
29667         (vect_recog_widen_sum_pattern): Likewise.
29669 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
29671         * predict.c (probably_never_executed): Only use precise profile info.
29672         (compute_function_frequency): Skip after inlining hack since we now
29673         have quality checking.
29675 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
29677         * profile-count.h (profile_probability::very_unlikely,
29678         profile_probability::unlikely, profile_probability::even): Set
29679         precision to guessed.
29681 2018-01-23  Richard Biener  <rguenther@suse.de>
29683         PR tree-optimization/83963
29684         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
29685         Properly terminate dominator walk when crossing the exit edge not
29686         when visiting its source block.
29688 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
29690         PR c++/83918
29691         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
29692         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
29694 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
29696         PR tree-optimization/83957
29697         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
29698         semicolon after for body surrounded by braces.
29700         PR tree-optimization/83081
29701         * profile-count.h (profile_probability::split): New method.
29702         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
29703         Use profile_probability::split.
29704         (do_compare_rtx_and_jump): Fix adjustment of probabilities
29705         when splitting a single conditional jump into 2.
29707 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
29709         PR tree-optimization/69452
29710         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
29711         decl.
29713 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29715         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
29716         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
29717         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
29719 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29721         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
29722         declaration.
29723         * config/rl78/rl78.md (movdi): New define_expand.
29724         * config/rl78/rl78.c (rl78_split_movdi): New function.
29726 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
29728         PR target/83862
29729         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
29730         no longer used.
29731         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
29732         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
29733         128-bit to produce an UNSPEC move to get the double word with the
29734         signbit and then a shift directly to do signbit.
29735         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
29736         implementation with a new version that just does either a direct
29737         move or a regular move.  Move memory interface to separate insns.
29738         Move insns so they are next to the expander.
29739         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
29740         with signbit move.  Split big and little endian case.
29741         (signbit<mode>2_dm_mem_le): Likewise.
29742         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
29743         (signbit<mode>2_dm2): Likewise.
29745 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29747         * config/rl78/rl78.md (anddi3): New define_expand.
29749 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29751         * config/rl78/rl78.md (umindi3): New define_expand.
29753 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29755         * config/rl78/rl78.md (smindi3): New define_expand.
29757 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29759         * config/rl78/rl78.md (smaxdi3): New define_expand.
29761 2018-01-22  Carl Love  <cel@us.ibm.com>
29763         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
29764         LVX_V1TI): Add macro expansion.
29765         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
29766         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
29767         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
29768         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
29769         Change check to determine if the instruction is a byte reversing
29770         entry.  Fix typo in comment.
29771         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
29772         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
29773         Add def_builtin calls for new builtins.
29774         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
29775         Add define_insn expansion.
29777 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29779         * config/rl78/rl78.md (umaxdi3): New define_expand.
29781 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
29783         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
29784         for non-QImode registers.
29786 2018-01-22  Richard Biener  <rguenther@suse.de>
29788         PR tree-optimization/83963
29789         * graphite-scop-detection.c (scop_detection::get_sese): Delay
29790         including the loop exit block.
29791         (scop_detection::merge_sese): Likewise.
29792         (scop_detection::add_scop): Do it here instead.
29794 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29796         * doc/sourcebuild.texi (arm_softfloat): Document.
29798 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
29800         PR gcc/77734
29801         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
29802         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
29803         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
29805 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29806             David Edelsohn  <dje.gcc@gmail.com>
29808         PR target/83946
29809         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
29810         Change "crset eq" to "crset 2".
29811         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
29812         (*call_indirect_aix<mode>_nospec): Likewise.
29813         (*call_value_indirect_aix<mode>_nospec): Likewise.
29814         (*call_indirect_elfv2<mode>_nospec): Likewise.
29815         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
29816         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
29817         change assembly output from . to $.
29818         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
29819         (indirect_jump<mode>_nospec): Change assembly output from . to $.
29820         (*tablejump<mode>_internal1_nospec): Likewise.
29822 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
29824         PR target/80870
29825         * config/sh/sh_optimize_sett_clrt.cc:
29826         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
29828 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
29830         PR tree-optimization/83940
29831         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
29832         offset_dt to vect_constant_def rather than vect_unknown_def_type.
29833         (vect_check_load_store_mask): Add a mask_dt_out parameter and
29834         use it to pass back the definition type.
29835         (vect_check_store_rhs): Likewise rhs_dt_out.
29836         (vect_build_gather_load_calls): Add a mask_dt argument and use
29837         it instead of a call to vect_is_simple_use.
29838         (vectorizable_store): Update calls to vect_check_load_store_mask
29839         and vect_check_store_rhs.  Use the dt returned by the latter instead
29840         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
29841         instead of calls to vect_is_simple_use.  Pass the scalar rather
29842         than the vector operand to vect_is_simple_use when handling
29843         second and subsequent copies of an rhs value.
29844         (vectorizable_load): Update calls to vect_check_load_store_mask
29845         and vect_build_gather_load_calls.  Use the cached mask_dt and
29846         gs_info.offset_dt instead of calls to vect_is_simple_use.
29848 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
29850         PR middle-end/83945
29851         * tree-emutls.c: Include gimplify.h.
29852         (lower_emutls_2): New function.
29853         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
29854         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
29855         it before further processing.
29857         PR target/83930
29858         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
29859         UINTVAL (trueop1) instead of INTVAL (op1).
29861 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
29863         PR debug/81570
29864         PR debug/83728
29865         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
29866         INCOMING_FRAME_SP_OFFSET if not defined.
29867         (scan_trace): Add ENTRY argument.  If true and
29868         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
29869         emit a note to adjust the CFA offset.
29870         (create_cfi_notes): Adjust scan_trace callers.
29871         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
29872         INCOMING_FRAME_SP_OFFSET in the CIE.
29873         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
29874         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
29875         Likewise.
29876         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
29877         * doc/tm.texi: Regenerated.
29879 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29881         PR rtl-optimization/83147
29882         * lra-constraints.c (remove_inheritance_pseudos): Use
29883         lra_substitute_pseudo_within_insn.
29885 2018-01-19  Tom de Vries  <tom@codesourcery.com>
29886             Cesar Philippidis  <cesar@codesourcery.com>
29888         PR target/83920
29889         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
29891 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
29893         PR target/83790
29894         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
29895         spaces for function labels.
29897 2018-01-19  Martin Liska  <mliska@suse.cz>
29899         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
29900         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
29901         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
29902         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
29903         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
29904         (PRED_CONST_RETURN): Change from 69 to 65.
29905         (PRED_NULL_RETURN): Change from 91 to 71.
29906         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
29907         (PRED_LOOP_GUARD): Change from 66 to 73.
29909 2018-01-19  Martin Liska  <mliska@suse.cz>
29911         * predict.c (predict_insn_def): Add new assert.
29912         (struct branch_predictor): Change type to signed integer.
29913         (test_prediction_value_range): Amend test to cover
29914         PROB_UNINITIALIZED.
29915         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
29916         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
29917         (PRED_LOOP_ITERATIONS_MAX): Likewise.
29918         (PRED_LOOP_IV_COMPARE): Likewise.
29919         * predict.h (PROB_UNINITIALIZED): Define new constant.
29921 2018-01-19  Martin Liska  <mliska@suse.cz>
29923         * predict.c (dump_prediction): Add new format for
29924         analyze_brprob.py script which is enabled with -details
29925         suboption.
29926         * profile-count.h (precise_p): New function.
29928 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
29930         PR tree-optimization/83922
29931         * tree-vect-loop.c (vect_verify_full_masking): Return false if
29932         there are no statements that need masking.
29933         (vect_active_double_reduction_p): New function.
29934         (vect_analyze_loop_operations): Use it when handling phis that
29935         are not in the loop header.
29937 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
29939         PR tree-optimization/83914
29940         * tree-vect-loop.c (vectorizable_induction): Don't convert
29941         init_expr or apply the peeling adjustment for inductions
29942         that are nested within the vectorized loop.
29944 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29946         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
29947         instead of NEG.
29949 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
29951         PR sanitizer/81715
29952         PR testsuite/83882
29953         * function.h (gimplify_parameters): Add gimple_seq * argument.
29954         * function.c: Include gimple.h and options.h.
29955         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
29956         for the added local temporaries if needed.
29957         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
29958         if there are any parameter cleanups, wrap whole body into a
29959         try/finally with the cleanups.
29961 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
29963         PR target/82964
29964         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
29965         Use GET_MODE_CLASS for scalar floating point.
29967 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
29969         PR ipa/82256
29970         patch by PaX Team
29971         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
29972         Fix call of call_cgraph_insertion_hooks.
29974 2018-01-18  Martin Sebor  <msebor@redhat.com>
29976         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
29978 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
29980         PR ipa/83619
29981         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
29982         frequencies.
29984 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
29986         PR other/70268
29987         * common.opt: (-ffile-prefix-map): New option.
29988         * opts.c (common_handle_option): Defer it.
29989         * opts-global.c (handle_common_deferred_options): Handle it.
29990         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
29991         * file-prefix-map.h: New file.
29992         (remap_debug_filename, add_debug_prefix_map): ...here.
29993         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
29994         * final.c (debug_prefix_map, add_debug_prefix_map
29995         remap_debug_filename): Move to...
29996         * file-prefix-map.c: New file.
29997         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
29998         generalize, get rid of alloca(), use strrchr() instead of strchr().
29999         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
30000         Implement in terms of add_prefix_map().
30001         (remap_macro_filename, remap_debug_filename): Implement in term of
30002         remap_filename().
30003         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
30004         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
30005         * dbxout.c: Include file-prefix-map.h.
30006         * varasm.c: Likewise.
30007         * vmsdbgout.c: Likewise.
30008         * xcoffout.c: Likewise.
30009         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
30010         * doc/cppopts.texi (-fmacro-prefix-map): Document.
30011         * doc/invoke.texi (-ffile-prefix-map): Document.
30012         (-fdebug-prefix-map): Update description.
30014 2018-01-18  Martin Liska  <mliska@suse.cz>
30016         * config/i386/i386.c (indirect_thunk_name): Document that also
30017         lfence is emitted.
30018         (output_indirect_thunk): Document why both instructions
30019         (pause and lfence) are generated.
30021 2018-01-18  Richard Biener  <rguenther@suse.de>
30023         PR tree-optimization/83887
30024         * graphite-scop-detection.c
30025         (scop_detection::get_nearest_dom_with_single_entry): Remove.
30026         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
30027         (scop_detection::merge_sese): Re-implement with a flood-fill
30028         algorithm that properly finds a SESE region if it exists.
30030 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
30032         PR c/61240
30033         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
30034         pointer_diff optimizations use view_convert instead of convert.
30036 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30038         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30039         Generate different code for -mno-speculate-indirect-jumps.
30040         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30041         (*call_indirect_aix<mode>): Disable for
30042         -mno-speculate-indirect-jumps.
30043         (*call_indirect_aix<mode>_nospec): New define_insn.
30044         (*call_value_indirect_aix<mode>): Disable for
30045         -mno-speculate-indirect-jumps.
30046         (*call_value_indirect_aix<mode>_nospec): New define_insn.
30047         (*sibcall_nonlocal_sysv<mode>): Generate different code for
30048         -mno-speculate-indirect-jumps.
30049         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30051 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
30053         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
30054         long double type, set the flags for noting the default long double
30055         type, even if we don't pass or return a long double type.
30057 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
30059         PR ipa/83051
30060         * ipa-inline.c (flatten_function): Do not overwrite final inlining
30061         failure.
30063 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
30065         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
30066         support for merge[hl].
30067         (fold_mergehl_helper): New helper function.
30068         (tree-vector-builder.h): New #include for tree_vector_builder usage.
30069         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
30070         (altivec_vmrglw_direct): Add xxmrglw insn.
30072 2018-01-17  Andrew Waterman  <andrew@sifive.com>
30074         * config/riscv/riscv.c (riscv_conditional_register_usage): If
30075         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
30077 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
30079         PR lto/83121
30080         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
30081         call the lto_location_cache before reading the
30082         DECL_SOURCE_LOCATION of the types.
30084 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
30085             Richard Sandiford  <richard.sandiford@linaro.org>
30087         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
30088         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
30089         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
30090         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
30091         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
30092         Add declaration.
30093         * config/aarch64/constraints.md (aarch64_movti_operand):
30094         Limit immediates.
30095         * config/aarch64/predicates.md (Uti): Add new constraint.
30097 2018-01-17  Carl Love  <cel@us.ibm.com>
30099         * config/rs6000/vsx.md (define_expand xl_len_r,
30100         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
30101         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
30102         lxvll.
30103         (define_expand, define_insn): Move the shift left from  the
30104         define_insn to the define_expand for lxvl and stxvl instructions.
30105         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
30106         and XL_LEN_R definitions to PURE.
30108 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
30110         * config/i386/i386.c (indirect_thunk_name): Declare regno
30111         as unsigned int.  Compare regno with INVALID_REGNUM.
30112         (output_indirect_thunk): Ditto.
30113         (output_indirect_thunk_function): Ditto.
30114         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
30115         in the call to output_indirect_thunk_function.
30117 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
30119         PR middle-end/83884
30120         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
30121         rather than the size of inner_type to determine the stack slot size
30122         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
30124 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
30126         PR target/83546
30127         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
30128         to PTA_SILVERMONT.
30130 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
30132         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
30133         endian Linux systems to optionally enable multilibs for selecting
30134         the long double type if the user configured an explicit type.
30135         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
30136         have no long double multilibs if not defined.
30137         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
30138         warn if the user used -mabi={ieee,ibm}longdouble and we built
30139         multilibs for long double.
30140         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
30141         appropriate multilib option.
30142         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
30143         multilib options.
30144         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
30145         for building long double multilibs.
30146         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
30148 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
30150         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
30151         copies.
30153         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
30154         64 bits.
30155         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
30156         128 bits.
30158         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
30159         variables.
30161         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
30162         return value.
30164 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
30166         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
30167         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
30169 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
30171         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
30172         different rtl trees depending on TARGET_64BIT.
30173         (rs6000_gen_lvx): Likewise.
30175 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
30177         * config/visium/visium.md (nop): Tweak comment.
30178         (hazard_nop): Likewise.
30180 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30182         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
30183         -mspeculate-indirect-jumps.
30184         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
30185         for -mno-speculate-indirect-jumps.
30186         (*call_indirect_elfv2<mode>_nospec): New define_insn.
30187         (*call_value_indirect_elfv2<mode>): Disable for
30188         -mno-speculate-indirect-jumps.
30189         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
30190         (indirect_jump): Emit different RTL for
30191         -mno-speculate-indirect-jumps.
30192         (*indirect_jump<mode>): Disable for
30193         -mno-speculate-indirect-jumps.
30194         (*indirect_jump<mode>_nospec): New define_insn.
30195         (tablejump): Emit different RTL for
30196         -mno-speculate-indirect-jumps.
30197         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
30198         (tablejumpsi_nospec): New define_expand.
30199         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
30200         (tablejumpdi_nospec): New define_expand.
30201         (*tablejump<mode>_internal1): Disable for
30202         -mno-speculate-indirect-jumps.
30203         (*tablejump<mode>_internal1_nospec): New define_insn.
30204         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
30205         option.
30207 2018-01-16  Artyom Skrobov tyomitch@gmail.com
30209         * caller-save.c (insert_save): Drop unnecessary parameter.  All
30210         callers updated.
30212 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30213             Richard Biener  <rguenth@suse.de>
30215         PR libgomp/83590
30216         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
30217         return early, inline manually is_gimple_sizepos.  Make sure if we
30218         call gimplify_expr we don't end up with a gimple constant.
30219         * tree.c (variably_modified_type_p): Don't return true for
30220         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
30221         * gimplify.h (is_gimple_sizepos): Remove.
30223 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30225         PR tree-optimization/83857
30226         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
30227         vectorizable_live_operation for pure SLP statements.
30228         (vectorizable_live_operation): Handle PHIs.
30230 2018-01-16  Richard Biener  <rguenther@suse.de>
30232         PR tree-optimization/83867
30233         * tree-vect-stmts.c (vect_transform_stmt): Precompute
30234         nested_in_vect_loop_p since the scalar stmt may get invalidated.
30236 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30238         PR c/83844
30239         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
30240         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
30241         If off is not INTEGER_CST, issue a may not be aligned warning
30242         rather than isn't aligned.  Use isn%'t rather than isn't.
30243         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
30244         into MULT_EXPR.
30245         <case MULT_EXPR>: Improve the case when bottom and one of the
30246         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
30247         operand, in that case check if the other operand is multiple of
30248         bottom divided by the INTEGER_CST operand.
30250 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30252         PR target/83858
30253         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
30254         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
30255         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
30256         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
30257         * config/pa/pa.c (pa_function_arg_advance): Likewise.
30258         (pa_function_arg, pa_arg_partial_bytes): Likewise.
30259         (pa_function_arg_size): New function.
30261 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30263         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
30264         in a separate statement.
30266 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
30268         PR tree-optimization/83847
30269         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
30270         group gathers and scatters.
30272 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
30274         PR rtl-optimization/86620
30275         * params.def (max-sched-ready-insns): Bump minimum value to 1.
30277         PR rtl-optimization/83213
30278         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
30279         to last if both are JUMP_INSNs.
30281         PR tree-optimization/83843
30282         * gimple-ssa-store-merging.c
30283         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
30284         store_immediate_info for bswap/nop orig_stores.
30286 2018-01-15  Andrew Waterman  <andrew@sifive.com>
30288         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
30289         !TARGET_MUL.
30290         <UDIV>: Increase cost if !TARGET_DIV.
30292 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
30294         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
30295         (define_attr "cr_logical_3op"): New.
30296         (cceq_ior_compare): Adjust.
30297         (cceq_ior_compare_complement): Adjust.
30298         (*cceq_rev_compare): Adjust.
30299         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
30300         (is_cracked_insn): Adjust.
30301         (insn_must_be_first_in_group): Adjust.
30302         * config/rs6000/40x.md: Adjust.
30303         * config/rs6000/440.md: Adjust.
30304         * config/rs6000/476.md: Adjust.
30305         * config/rs6000/601.md: Adjust.
30306         * config/rs6000/603.md: Adjust.
30307         * config/rs6000/6xx.md: Adjust.
30308         * config/rs6000/7450.md: Adjust.
30309         * config/rs6000/7xx.md: Adjust.
30310         * config/rs6000/8540.md: Adjust.
30311         * config/rs6000/cell.md: Adjust.
30312         * config/rs6000/e300c2c3.md: Adjust.
30313         * config/rs6000/e500mc.md: Adjust.
30314         * config/rs6000/e500mc64.md: Adjust.
30315         * config/rs6000/e5500.md: Adjust.
30316         * config/rs6000/e6500.md: Adjust.
30317         * config/rs6000/mpc.md: Adjust.
30318         * config/rs6000/power4.md: Adjust.
30319         * config/rs6000/power5.md: Adjust.
30320         * config/rs6000/power6.md: Adjust.
30321         * config/rs6000/power7.md: Adjust.
30322         * config/rs6000/power8.md: Adjust.
30323         * config/rs6000/power9.md: Adjust.
30324         * config/rs6000/rs64.md: Adjust.
30325         * config/rs6000/titan.md: Adjust.
30327 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30329         * config/i386/predicates.md (indirect_branch_operand): Rewrite
30330         ix86_indirect_branch_register logic.
30332 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30334         * config/i386/constraints.md (Bs): Update
30335         ix86_indirect_branch_register check.  Don't check
30336         ix86_indirect_branch_register with GOT_memory_operand.
30337         (Bw): Likewise.
30338         * config/i386/predicates.md (GOT_memory_operand): Don't check
30339         ix86_indirect_branch_register here.
30340         (GOT32_symbol_operand): Likewise.
30342 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30344         * config/i386/predicates.md (constant_call_address_operand):
30345         Rewrite ix86_indirect_branch_register logic.
30346         (sibcall_insn_operand): Likewise.
30348 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30350         * config/i386/constraints.md (Bs): Replace
30351         ix86_indirect_branch_thunk_register with
30352         ix86_indirect_branch_register.
30353         (Bw): Likewise.
30354         * config/i386/i386.md (indirect_jump): Likewise.
30355         (tablejump): Likewise.
30356         (*sibcall_memory): Likewise.
30357         (*sibcall_value_memory): Likewise.
30358         Peepholes of indirect call and jump via memory: Likewise.
30359         * config/i386/i386.opt: Likewise.
30360         * config/i386/predicates.md (indirect_branch_operand): Likewise.
30361         (GOT_memory_operand): Likewise.
30362         (call_insn_operand): Likewise.
30363         (sibcall_insn_operand): Likewise.
30364         (GOT32_symbol_operand): Likewise.
30366 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
30368         PR middle-end/83837
30369         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
30370         type rather than type addr's type points to.
30371         (expand_omp_atomic_mutex): Likewise.
30372         (expand_omp_atomic): Likewise.
30374 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
30376         PR target/83839
30377         * config/i386/i386.c (output_indirect_thunk_function): Use
30378         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
30379         for  __x86_return_thunk.
30381 2018-01-15  Richard Biener  <rguenther@suse.de>
30383         PR middle-end/83850
30384         * expmed.c (extract_bit_field_1): Fix typo.
30386 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30388         PR target/83687
30389         * config/arm/iterators.md (VF): New mode iterator.
30390         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
30391         Remove integer-related logic from pattern.
30392         (neon_vabd<mode>_3): Likewise.
30394 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
30396         PR middle-end/82694
30397         * common.opt (fstrict-overflow): No longer an alias.
30398         (fwrapv-pointer): New option.
30399         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
30400         also for pointer types based on flag_wrapv_pointer.
30401         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
30402         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
30403         opts->x_flag_wrapv got set.
30404         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
30405         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
30406         POINTER_TYPE_OVERFLOW_UNDEFINED.
30407         * match.pd: Likewise in address comparison pattern.
30408         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
30410 2018-01-15  Richard Biener  <rguenther@suse.de>
30412         PR lto/83804
30413         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
30414         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
30415         Reset type names to their identifier if their TYPE_DECL doesn't
30416         have linkage (and thus is used for ODR and devirt).
30417         (save_debug_info_for_decl): Remove.
30418         (save_debug_info_for_type): Likewise.
30419         (add_tree_to_fld_list): Adjust.
30420         * tree-pretty-print.c (dump_generic_node): Make dumping of
30421         type names more robust.
30423 2018-01-15  Richard Biener  <rguenther@suse.de>
30425         * BASE-VER: Bump to 8.0.1.
30427 2018-01-14  Martin Sebor  <msebor@redhat.com>
30429         PR other/83508
30430         * builtins.c (check_access): Avoid warning when the no-warning bit
30431         is set.
30433 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
30435         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
30436         * ira-color (allocno_hard_regs_compare): Likewise.
30438 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
30440         PR target/83013
30441         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
30442         Use .pushsection/.popsection.
30444 2018-01-14  Martin Sebor  <msebor@redhat.com>
30446         PR c++/81327
30447         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
30449 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
30451         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
30452         entry from extra_headers.
30453         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
30454         extra_headers, make the list bitwise identical to the i?86-*-* one.
30456 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30458         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
30459         -mcmodel=large with -mindirect-branch=thunk,
30460         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
30461         -mfunction-return=thunk-extern.
30462         * doc/invoke.texi: Document -mcmodel=large is incompatible with
30463         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
30464         -mfunction-return=thunk and -mfunction-return=thunk-extern.
30466 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30468         * config/i386/i386.c (print_reg): Print the name of the full
30469         integer register without '%'.
30470         (ix86_print_operand): Handle 'V'.
30471         * doc/extend.texi: Document 'V' modifier.
30473 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30475         * config/i386/constraints.md (Bs): Disallow memory operand for
30476         -mindirect-branch-register.
30477         (Bw): Likewise.
30478         * config/i386/predicates.md (indirect_branch_operand): Likewise.
30479         (GOT_memory_operand): Likewise.
30480         (call_insn_operand): Likewise.
30481         (sibcall_insn_operand): Likewise.
30482         (GOT32_symbol_operand): Likewise.
30483         * config/i386/i386.md (indirect_jump): Call convert_memory_address
30484         for -mindirect-branch-register.
30485         (tablejump): Likewise.
30486         (*sibcall_memory): Likewise.
30487         (*sibcall_value_memory): Likewise.
30488         Disallow peepholes of indirect call and jump via memory for
30489         -mindirect-branch-register.
30490         (*call_pop): Replace m with Bw.
30491         (*call_value_pop): Likewise.
30492         (*sibcall_pop_memory): Replace m with Bs.
30493         * config/i386/i386.opt (mindirect-branch-register): New option.
30494         * doc/invoke.texi: Document -mindirect-branch-register option.
30496 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30498         * config/i386/i386-protos.h (ix86_output_function_return): New.
30499         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
30500         set function_return_type.
30501         (indirect_thunk_name): Add ret_p to indicate thunk for function
30502         return.
30503         (output_indirect_thunk_function): Pass false to
30504         indirect_thunk_name.
30505         (ix86_output_indirect_branch_via_reg): Likewise.
30506         (ix86_output_indirect_branch_via_push): Likewise.
30507         (output_indirect_thunk_function): Create alias for function
30508         return thunk if regno < 0.
30509         (ix86_output_function_return): New function.
30510         (ix86_handle_fndecl_attribute): Handle function_return.
30511         (ix86_attribute_table): Add function_return.
30512         * config/i386/i386.h (machine_function): Add
30513         function_return_type.
30514         * config/i386/i386.md (simple_return_internal): Use
30515         ix86_output_function_return.
30516         (simple_return_internal_long): Likewise.
30517         * config/i386/i386.opt (mfunction-return=): New option.
30518         (indirect_branch): Mention -mfunction-return=.
30519         * doc/extend.texi: Document function_return function attribute.
30520         * doc/invoke.texi: Document -mfunction-return= option.
30522 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
30524         * config/i386/i386-opts.h (indirect_branch): New.
30525         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
30526         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
30527         with local indirect jump when converting indirect call and jump.
30528         (ix86_set_indirect_branch_type): New.
30529         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
30530         (indirectlabelno): New.
30531         (indirect_thunk_needed): Likewise.
30532         (indirect_thunk_bnd_needed): Likewise.
30533         (indirect_thunks_used): Likewise.
30534         (indirect_thunks_bnd_used): Likewise.
30535         (INDIRECT_LABEL): Likewise.
30536         (indirect_thunk_name): Likewise.
30537         (output_indirect_thunk): Likewise.
30538         (output_indirect_thunk_function): Likewise.
30539         (ix86_output_indirect_branch_via_reg): Likewise.
30540         (ix86_output_indirect_branch_via_push): Likewise.
30541         (ix86_output_indirect_branch): Likewise.
30542         (ix86_output_indirect_jmp): Likewise.
30543         (ix86_code_end): Call output_indirect_thunk_function if needed.
30544         (ix86_output_call_insn): Call ix86_output_indirect_branch if
30545         needed.
30546         (ix86_handle_fndecl_attribute): Handle indirect_branch.
30547         (ix86_attribute_table): Add indirect_branch.
30548         * config/i386/i386.h (machine_function): Add indirect_branch_type
30549         and has_local_indirect_jump.
30550         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
30551         to true.
30552         (tablejump): Likewise.
30553         (*indirect_jump): Use ix86_output_indirect_jmp.
30554         (*tablejump_1): Likewise.
30555         (simple_return_indirect_internal): Likewise.
30556         * config/i386/i386.opt (mindirect-branch=): New option.
30557         (indirect_branch): New.
30558         (keep): Likewise.
30559         (thunk): Likewise.
30560         (thunk-inline): Likewise.
30561         (thunk-extern): Likewise.
30562         * doc/extend.texi: Document indirect_branch function attribute.
30563         * doc/invoke.texi: Document -mindirect-branch= option.
30565 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
30567         PR ipa/83051
30568         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
30570 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
30572         * ipa-inline.c (want_inline_small_function_p): Return false if
30573         inlining has already failed with CIF_FINAL_ERROR.
30574         (update_caller_keys): Call want_inline_small_function_p before
30575         can_inline_edge_p.
30576         (update_callee_keys): Likewise.
30578 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
30580         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
30581         New function.
30582         (rs6000_quadword_masked_address_p): Likewise.
30583         (quad_aligned_load_p): Likewise.
30584         (quad_aligned_store_p): Likewise.
30585         (const_load_sequence_p): Add comment to describe the outer-most loop.
30586         (mimic_memory_attributes_and_flags): New function.
30587         (rs6000_gen_stvx): Likewise.
30588         (replace_swapped_aligned_store): Likewise.
30589         (rs6000_gen_lvx): Likewise.
30590         (replace_swapped_aligned_load): Likewise.
30591         (replace_swapped_load_constant): Capitalize argument name in
30592         comment describing this function.
30593         (rs6000_analyze_swaps): Add a third pass to search for vector loads
30594         and stores that access quad-word aligned addresses and replace
30595         with stvx or lvx instructions when appropriate.
30596         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
30597         New function prototype.
30598         (rs6000_quadword_masked_address_p): Likewise.
30599         (rs6000_gen_lvx): Likewise.
30600         (rs6000_gen_stvx): Likewise.
30601         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
30602         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
30603         when memory address is aligned.
30604         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
30605         this split to select lvx instruction when memory address is aligned.
30606         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
30607         instruction when memory address is aligned.
30608         (*vsx_le_perm_load_v16qi): Likewise.
30609         (four unnamed splitters): Modify to select the stvx instruction
30610         when memory is aligned.
30612 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
30614         * predict.c (determine_unlikely_bbs): Handle correctly BBs
30615         which appears in the queue multiple times.
30617 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30618             Alan Hayward  <alan.hayward@arm.com>
30619             David Sherwood  <david.sherwood@arm.com>
30621         * tree-vectorizer.h (vec_lower_bound): New structure.
30622         (_loop_vec_info): Add check_nonzero and lower_bounds.
30623         (LOOP_VINFO_CHECK_NONZERO): New macro.
30624         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
30625         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
30626         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
30627         fields.  Make seg_len the distance travelled, not including the
30628         access size.
30629         (dr_direction_indicator): Declare.
30630         (dr_zero_step_indicator): Likewise.
30631         (dr_known_forward_stride_p): Likewise.
30632         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
30633         tree-ssanames.h.
30634         (runtime_alias_check_p): Allow runtime alias checks with
30635         variable strides.
30636         (operator ==): Compare access_size and align.
30637         (prune_runtime_alias_test_list): Rework for new distinction between
30638         the access_size and seg_len.
30639         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
30640         segment lengths.
30641         (get_segment_min_max): New function.
30642         (create_intersect_range_checks): Use it.
30643         (dr_step_indicator): New function.
30644         (dr_direction_indicator): Likewise.
30645         (dr_zero_step_indicator): Likewise.
30646         (dr_known_forward_stride_p): Likewise.
30647         * tree-loop-distribution.c (data_ref_segment_size): Return
30648         DR_STEP * (niters - 1).
30649         (compute_alias_check_pairs): Update call to the dr_with_seg_len
30650         constructor.
30651         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
30652         (vect_preserves_scalar_order_p): New function, split out from...
30653         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
30654         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
30655         (vect_vfa_access_size): New function.
30656         (vect_vfa_align): Likewise.
30657         (vect_compile_time_alias): Take access_size_a and access_b arguments.
30658         (dump_lower_bound): New function.
30659         (vect_check_lower_bound): Likewise.
30660         (vect_small_gap_p): Likewise.
30661         (vectorizable_with_step_bound_p): Likewise.
30662         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
30663         depencies if the vectorization factor is 1.  Convert the checks
30664         for nonzero steps into checks on the bounds of DR_STEP.  Try using
30665         a bunds check for variable steps if the minimum required step is
30666         relatively small. Update calls to the dr_with_seg_len
30667         constructor and to vect_compile_time_alias.
30668         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
30669         function.
30670         (vect_loop_versioning): Call it.
30671         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
30672         when retrying.
30673         (vect_estimate_min_profitable_iters): Account for any bounds checks.
30675 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30676             Alan Hayward  <alan.hayward@arm.com>
30677             David Sherwood  <david.sherwood@arm.com>
30679         * doc/sourcebuild.texi (vect_scatter_store): Document.
30680         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
30681         optabs.
30682         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
30683         Document.
30684         * genopinit.c (main): Add supports_vec_scatter_store and
30685         supports_vec_scatter_store_cached to target_optabs.
30686         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
30687         IFN_MASK_SCATTER_STORE.
30688         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
30689         functions.
30690         * internal-fn.h (internal_store_fn_p): Declare.
30691         (internal_fn_stored_value_index): Likewise.
30692         * internal-fn.c (scatter_store_direct): New macro.
30693         (expand_scatter_store_optab_fn): New function.
30694         (direct_scatter_store_optab_supported_p): New macro.
30695         (internal_store_fn_p): New function.
30696         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
30697         IFN_MASK_SCATTER_STORE.
30698         (internal_fn_mask_index): Likewise.
30699         (internal_fn_stored_value_index): New function.
30700         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
30701         for scatter stores.
30702         * optabs-query.h (supports_vec_scatter_store_p): Declare.
30703         * optabs-query.c (supports_vec_scatter_store_p): New function.
30704         * tree-vectorizer.h (vect_get_store_rhs): Declare.
30705         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
30706         true for scatter stores.
30707         (vect_gather_scatter_fn_p): Handle scatter stores too.
30708         (vect_check_gather_scatter): Consider using scatter stores if
30709         supports_vec_scatter_store_p.
30710         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
30711         scatter stores too.
30712         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
30713         internal_fn_stored_value_index.
30714         (check_load_store_masking): Handle scatter stores too.
30715         (vect_get_store_rhs): Make public.
30716         (vectorizable_call): Use internal_store_fn_p.
30717         (vectorizable_store): Handle scatter store internal functions.
30718         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
30719         when deciding whether the end of the group has been reached.
30720         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
30721         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
30722         (mask_scatter_store<mode>): New insns.
30724 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30725             Alan Hayward  <alan.hayward@arm.com>
30726             David Sherwood  <david.sherwood@arm.com>
30728         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
30729         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
30730         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
30731         function.
30732         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
30733         Use vect_truncate_gather_scatter_offset if we can't treat the
30734         operation as a normal gather load or scatter store.
30735         (get_group_load_store_type): Take the gather_scatter_info
30736         as argument.  Try using a gather load or scatter store for
30737         single-element groups.
30738         (get_load_store_type): Update calls to get_group_load_store_type
30739         and vect_use_strided_gather_scatters_p.
30741 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30742             Alan Hayward  <alan.hayward@arm.com>
30743             David Sherwood  <david.sherwood@arm.com>
30745         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
30746         optional tree argument.
30747         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
30748         null target hooks.
30749         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
30750         but continue to use the current value as a fallback.
30751         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
30752         to compare the updates.
30753         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
30754         (get_load_store_type): Use it when handling a strided access.
30755         (vect_get_strided_load_store_ops): New function.
30756         (vect_get_data_ptr_increment): Likewise.
30757         (vectorizable_load): Handle strided gather loads.  Always pass
30758         a step to vect_create_data_ref_ptr and bump_vector_ptr.
30760 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30761             Alan Hayward  <alan.hayward@arm.com>
30762             David Sherwood  <david.sherwood@arm.com>
30764         * doc/md.texi (gather_load@var{m}): Document.
30765         (mask_gather_load@var{m}): Likewise.
30766         * genopinit.c (main): Add supports_vec_gather_load and
30767         supports_vec_gather_load_cached to target_optabs.
30768         * optabs-tree.c (init_tree_optimization_optabs): Use
30769         ggc_cleared_alloc to allocate target_optabs.
30770         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
30771         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
30772         functions.
30773         * internal-fn.h (internal_load_fn_p): Declare.
30774         (internal_gather_scatter_fn_p): Likewise.
30775         (internal_fn_mask_index): Likewise.
30776         (internal_gather_scatter_fn_supported_p): Likewise.
30777         * internal-fn.c (gather_load_direct): New macro.
30778         (expand_gather_load_optab_fn): New function.
30779         (direct_gather_load_optab_supported_p): New macro.
30780         (direct_internal_fn_optab): New function.
30781         (internal_load_fn_p): Likewise.
30782         (internal_gather_scatter_fn_p): Likewise.
30783         (internal_fn_mask_index): Likewise.
30784         (internal_gather_scatter_fn_supported_p): Likewise.
30785         * optabs-query.c (supports_at_least_one_mode_p): New function.
30786         (supports_vec_gather_load_p): Likewise.
30787         * optabs-query.h (supports_vec_gather_load_p): Declare.
30788         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
30789         and memory_type field.
30790         (NUM_PATTERNS): Bump to 15.
30791         * tree-vect-data-refs.c: Include internal-fn.h.
30792         (vect_gather_scatter_fn_p): New function.
30793         (vect_describe_gather_scatter_call): Likewise.
30794         (vect_check_gather_scatter): Try using internal functions for
30795         gather loads.  Recognize existing calls to a gather load function.
30796         (vect_analyze_data_refs): Consider using gather loads if
30797         supports_vec_gather_load_p.
30798         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
30799         (vect_get_gather_scatter_offset_type): Likewise.
30800         (vect_convert_mask_for_vectype): Likewise.
30801         (vect_add_conversion_to_patterm): Likewise.
30802         (vect_try_gather_scatter_pattern): Likewise.
30803         (vect_recog_gather_scatter_pattern): New pattern recognizer.
30804         (vect_vect_recog_func_ptrs): Add it.
30805         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
30806         internal_fn_mask_index and internal_gather_scatter_fn_p.
30807         (check_load_store_masking): Take the gather_scatter_info as an
30808         argument and handle gather loads.
30809         (vect_get_gather_scatter_ops): New function.
30810         (vectorizable_call): Check internal_load_fn_p.
30811         (vectorizable_load): Likewise.  Handle gather load internal
30812         functions.
30813         (vectorizable_store): Update call to check_load_store_masking.
30814         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
30815         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
30816         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
30817         (aarch64_gather_scale_operand_d): New predicates.
30818         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
30819         (mask_gather_load<mode>): New insns.
30821 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30822             Alan Hayward  <alan.hayward@arm.com>
30823             David Sherwood  <david.sherwood@arm.com>
30825         * optabs.def (fold_left_plus_optab): New optab.
30826         * doc/md.texi (fold_left_plus_@var{m}): Document.
30827         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
30828         * internal-fn.c (fold_left_direct): Define.
30829         (expand_fold_left_optab_fn): Likewise.
30830         (direct_fold_left_optab_supported_p): Likewise.
30831         * fold-const-call.c (fold_const_fold_left): New function.
30832         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
30833         * tree-parloops.c (valid_reduction_p): New function.
30834         (gather_scalar_reductions): Use it.
30835         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
30836         (vect_finish_replace_stmt): Declare.
30837         * tree-vect-loop.c (fold_left_reduction_fn): New function.
30838         (needs_fold_left_reduction_p): New function, split out from...
30839         (vect_is_simple_reduction): ...here.  Accept reductions that
30840         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
30841         (vect_force_simple_reduction): Also store the reduction type in
30842         the assignment's STMT_VINFO_REDUC_TYPE.
30843         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
30844         (merge_with_identity): New function.
30845         (vect_expand_fold_left): Likewise.
30846         (vectorize_fold_left_reduction): Likewise.
30847         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
30848         scalar phi in place for it.  Check for target support and reject
30849         cases that would reassociate the operation.  Defer the transform
30850         phase to vectorize_fold_left_reduction.
30851         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
30852         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
30853         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
30855 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30857         * tree-if-conv.c (predicate_mem_writes): Remove redundant
30858         call to ifc_temp_var.
30860 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30861             Alan Hayward  <alan.hayward@arm.com>
30862             David Sherwood  <david.sherwood@arm.com>
30864         * target.def (legitimize_address_displacement): Take the original
30865         offset as a poly_int.
30866         * targhooks.h (default_legitimize_address_displacement): Update
30867         accordingly.
30868         * targhooks.c (default_legitimize_address_displacement): Likewise.
30869         * doc/tm.texi: Regenerate.
30870         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
30871         as an argument, moving assert of ad->disp == ad->disp_term to...
30872         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
30873         Try calling targetm.legitimize_address_displacement before expanding
30874         the address rather than afterwards, and adjust for the new interface.
30875         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
30876         Match the new hook interface.  Handle SVE addresses.
30877         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
30878         new hook interface.
30880 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30882         * Makefile.in (OBJS): Add early-remat.o.
30883         * target.def (select_early_remat_modes): New hook.
30884         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
30885         * doc/tm.texi: Regenerate.
30886         * targhooks.h (default_select_early_remat_modes): Declare.
30887         * targhooks.c (default_select_early_remat_modes): New function.
30888         * timevar.def (TV_EARLY_REMAT): New timevar.
30889         * passes.def (pass_early_remat): New pass.
30890         * tree-pass.h (make_pass_early_remat): Declare.
30891         * early-remat.c: New file.
30892         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
30893         function.
30894         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
30896 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30897             Alan Hayward  <alan.hayward@arm.com>
30898             David Sherwood  <david.sherwood@arm.com>
30900         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
30901         vfm1 with a bound_epilog parameter.
30902         (vect_do_peeling): Update calls accordingly, and move the prologue
30903         call earlier in the function.  Treat the base bound_epilog as 0 for
30904         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
30905         this base when peeling for gaps.
30906         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
30907         with fully-masked loops.
30908         (vect_estimate_min_profitable_iters): Handle the single peeled
30909         iteration in that case.
30911 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30912             Alan Hayward  <alan.hayward@arm.com>
30913             David Sherwood  <david.sherwood@arm.com>
30915         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
30916         single-element interleaving even if the size is not a power of 2.
30917         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
30918         accesses for single-element interleaving if the group size is
30919         not a power of 2.
30921 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30922             Alan Hayward  <alan.hayward@arm.com>
30923             David Sherwood  <david.sherwood@arm.com>
30925         * doc/md.texi (fold_extract_last_@var{m}): Document.
30926         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
30927         * optabs.def (fold_extract_last_optab): New optab.
30928         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
30929         * internal-fn.c (fold_extract_direct): New macro.
30930         (expand_fold_extract_optab_fn): Likewise.
30931         (direct_fold_extract_optab_supported_p): Likewise.
30932         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
30933         * tree-vect-loop.c (vect_model_reduction_cost): Handle
30934         EXTRACT_LAST_REDUCTION.
30935         (get_initial_def_for_reduction): Do not create an initial vector
30936         for EXTRACT_LAST_REDUCTION reductions.
30937         (vectorizable_reduction): Leave the scalar phi in place for
30938         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
30939         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
30940         epilogue code for EXTRACT_LAST_REDUCTION and defer the
30941         transform phase to vectorizable_condition.
30942         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
30943         split out from...
30944         (vect_finish_stmt_generation): ...here.
30945         (vect_finish_replace_stmt): New function.
30946         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
30947         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
30948         pattern.
30949         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
30951 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30952             Alan Hayward  <alan.hayward@arm.com>
30953             David Sherwood  <david.sherwood@arm.com>
30955         * doc/md.texi (extract_last_@var{m}): Document.
30956         * optabs.def (extract_last_optab): New optab.
30957         * internal-fn.def (EXTRACT_LAST): New internal function.
30958         * internal-fn.c (cond_unary_direct): New macro.
30959         (expand_cond_unary_optab_fn): Likewise.
30960         (direct_cond_unary_optab_supported_p): Likewise.
30961         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
30962         loops using EXTRACT_LAST.
30963         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
30964         (extract_last_<mode>): ...this optab.
30965         (vec_extract<mode><Vel>): Update accordingly.
30967 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30968             Alan Hayward  <alan.hayward@arm.com>
30969             David Sherwood  <david.sherwood@arm.com>
30971         * target.def (empty_mask_is_expensive): New hook.
30972         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
30973         * doc/tm.texi: Regenerate.
30974         * targhooks.h (default_empty_mask_is_expensive): Declare.
30975         * targhooks.c (default_empty_mask_is_expensive): New function.
30976         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
30977         if the target says that empty masks are expensive.
30978         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
30979         New function.
30980         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
30982 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30983             Alan Hayward  <alan.hayward@arm.com>
30984             David Sherwood  <david.sherwood@arm.com>
30986         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
30987         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
30988         (vect_use_loop_mask_for_alignment_p): New function.
30989         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
30990         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
30991         niters_skip argument.  Make sure that the first niters_skip elements
30992         of the first iteration are inactive.
30993         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
30994         Update call to vect_set_loop_masks_directly.
30995         (get_misalign_in_elems): New function, split out from...
30996         (vect_gen_prolog_loop_niters): ...here.
30997         (vect_update_init_of_dr): Take a code argument that specifies whether
30998         the adjustment should be added or subtracted.
30999         (vect_update_init_of_drs): Likewise.
31000         (vect_prepare_for_masked_peels): New function.
31001         (vect_do_peeling): Skip prologue peeling if we're using a mask
31002         instead.  Update call to vect_update_inits_of_drs.
31003         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
31004         mask_skip_niters.
31005         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
31006         alignment.  Do not include the number of peeled iterations in
31007         the minimum threshold in that case.
31008         (vectorizable_induction): Adjust the start value down by
31009         LOOP_VINFO_MASK_SKIP_NITERS iterations.
31010         (vect_transform_loop): Call vect_prepare_for_masked_peels.
31011         Take the number of skipped iterations into account when calculating
31012         the loop bounds.
31013         * tree-vect-stmts.c (vect_gen_while_not): New function.
31015 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31016             Alan Hayward  <alan.hayward@arm.com>
31017             David Sherwood  <david.sherwood@arm.com>
31019         * doc/sourcebuild.texi (vect_fully_masked): Document.
31020         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
31021         default value to 0.
31022         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
31023         split out from...
31024         (vect_analyze_loop_2): ...here. Don't check the vectorization
31025         factor against the number of loop iterations if the loop is
31026         fully-masked.
31028 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31029             Alan Hayward  <alan.hayward@arm.com>
31030             David Sherwood  <david.sherwood@arm.com>
31032         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
31033         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
31034         (dump_groups): Update accordingly.
31035         (iv_use::mem_type): New member variable.
31036         (address_p): New function.
31037         (record_use): Add a mem_type argument and initialize the new
31038         mem_type field.
31039         (record_group_use): Add a mem_type argument.  Use address_p.
31040         Remove obsolete null checks of base_object.  Update call to record_use.
31041         (find_interesting_uses_op): Update call to record_group_use.
31042         (find_interesting_uses_cond): Likewise.
31043         (find_interesting_uses_address): Likewise.
31044         (get_mem_type_for_internal_fn): New function.
31045         (find_address_like_use): Likewise.
31046         (find_interesting_uses_stmt): Try find_address_like_use before
31047         calling find_interesting_uses_op.
31048         (addr_offset_valid_p): Use the iv mem_type field as the type
31049         of the addressed memory.
31050         (add_autoinc_candidates): Likewise.
31051         (get_address_cost): Likewise.
31052         (split_small_address_groups_p): Use address_p.
31053         (split_address_groups): Likewise.
31054         (add_iv_candidate_for_use): Likewise.
31055         (autoinc_possible_for_pair): Likewise.
31056         (rewrite_groups): Likewise.
31057         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
31058         (determine_group_iv_cost): Update after split of USE_ADDRESS.
31059         (get_alias_ptr_type_for_ptr_address): New function.
31060         (rewrite_use_address): Rewrite address uses in calls that were
31061         identified by find_address_like_use.
31063 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31064             Alan Hayward  <alan.hayward@arm.com>
31065             David Sherwood  <david.sherwood@arm.com>
31067         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
31068         TARGET_MEM_REFs.
31069         * gimple-expr.h (is_gimple_addressable: Likewise.
31070         * gimple-expr.c (is_gimple_address): Likewise.
31071         * internal-fn.c (expand_call_mem_ref): New function.
31072         (expand_mask_load_optab_fn): Use it.
31073         (expand_mask_store_optab_fn): Likewise.
31075 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31076             Alan Hayward  <alan.hayward@arm.com>
31077             David Sherwood  <david.sherwood@arm.com>
31079         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
31080         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
31081         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
31082         (cond_umax@var{mode}): Document.
31083         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
31084         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
31085         (cond_umin_optab, cond_umax_optab): New optabs.
31086         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
31087         (COND_IOR, COND_XOR): New internal functions.
31088         * internal-fn.h (get_conditional_internal_fn): Declare.
31089         * internal-fn.c (cond_binary_direct): New macro.
31090         (expand_cond_binary_optab_fn): Likewise.
31091         (direct_cond_binary_optab_supported_p): Likewise.
31092         (get_conditional_internal_fn): New function.
31093         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
31094         Cope with reduction statements that are vectorized as calls rather
31095         than assignments.
31096         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
31097         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
31098         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
31099         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
31100         (UNSPEC_COND_EOR): New unspecs.
31101         (optab): Add mappings for them.
31102         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
31103         (sve_int_op, sve_fp_op): New int attributes.
31105 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31106             Alan Hayward  <alan.hayward@arm.com>
31107             David Sherwood  <david.sherwood@arm.com>
31109         * optabs.def (while_ult_optab): New optab.
31110         * doc/md.texi (while_ult@var{m}@var{n}): Document.
31111         * internal-fn.def (WHILE_ULT): New internal function.
31112         * internal-fn.h (direct_internal_fn_supported_p): New override
31113         that takes two types as argument.
31114         * internal-fn.c (while_direct): New macro.
31115         (expand_while_optab_fn): New function.
31116         (convert_optab_supported_p): Likewise.
31117         (direct_while_optab_supported_p): New macro.
31118         * wide-int.h (wi::udiv_ceil): New function.
31119         * tree-vectorizer.h (rgroup_masks): New structure.
31120         (vec_loop_masks): New typedef.
31121         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
31122         and fully_masked_p.
31123         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
31124         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
31125         (vect_max_vf): New function.
31126         (slpeel_make_loop_iterate_ntimes): Delete.
31127         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
31128         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
31129         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
31130         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
31131         internal-fn.h, stor-layout.h and optabs-query.h.
31132         (vect_set_loop_mask): New function.
31133         (add_preheader_seq): Likewise.
31134         (add_header_seq): Likewise.
31135         (interleave_supported_p): Likewise.
31136         (vect_maybe_permute_loop_masks): Likewise.
31137         (vect_set_loop_masks_directly): Likewise.
31138         (vect_set_loop_condition_masked): Likewise.
31139         (vect_set_loop_condition_unmasked): New function, split out from
31140         slpeel_make_loop_iterate_ntimes.
31141         (slpeel_make_loop_iterate_ntimes): Rename to..
31142         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
31143         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
31144         (vect_do_peeling): Update call accordingly.
31145         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
31146         loops.
31147         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
31148         mask_compare_type, can_fully_mask_p and fully_masked_p.
31149         (release_vec_loop_masks): New function.
31150         (_loop_vec_info): Use it to free the loop masks.
31151         (can_produce_all_loop_masks_p): New function.
31152         (vect_get_max_nscalars_per_iter): Likewise.
31153         (vect_verify_full_masking): Likewise.
31154         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
31155         retries, and free the mask rgroups before retrying.  Check loop-wide
31156         reasons for disallowing fully-masked loops.  Make the final decision
31157         about whether use a fully-masked loop or not.
31158         (vect_estimate_min_profitable_iters): Do not assume that peeling
31159         for the number of iterations will be needed for fully-masked loops.
31160         (vectorizable_reduction): Disable fully-masked loops.
31161         (vectorizable_live_operation): Likewise.
31162         (vect_halve_mask_nunits): New function.
31163         (vect_double_mask_nunits): Likewise.
31164         (vect_record_loop_mask): Likewise.
31165         (vect_get_loop_mask): Likewise.
31166         (vect_transform_loop): Handle the case in which the final loop
31167         iteration might handle a partial vector.  Call vect_set_loop_condition
31168         instead of slpeel_make_loop_iterate_ntimes.
31169         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
31170         (check_load_store_masking): New function.
31171         (prepare_load_store_mask): Likewise.
31172         (vectorizable_store): Handle fully-masked loops.
31173         (vectorizable_load): Likewise.
31174         (supportable_widening_operation): Use vect_halve_mask_nunits for
31175         booleans.
31176         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
31177         (vect_gen_while): New function.
31178         * config/aarch64/aarch64.md (umax<mode>3): New expander.
31179         (aarch64_uqdec<mode>): New insn.
31181 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31182             Alan Hayward  <alan.hayward@arm.com>
31183             David Sherwood  <david.sherwood@arm.com>
31185         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
31186         (reduc_xor_scal_optab): New optabs.
31187         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
31188         (reduc_xor_scal_@var{m}): Document.
31189         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
31190         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
31191         internal functions.
31192         * fold-const-call.c (fold_const_call): Handle them.
31193         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
31194         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
31195         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
31196         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
31197         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
31198         (UNSPEC_XORV): New unspecs.
31199         (optab): Add entries for them.
31200         (BITWISEV): New int iterator.
31201         (bit_reduc_op): New int attributes.
31203 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31204             Alan Hayward  <alan.hayward@arm.com>
31205             David Sherwood  <david.sherwood@arm.com>
31207         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
31208         * internal-fn.def (VEC_SHL_INSERT): New internal function.
31209         * optabs.def (vec_shl_insert_optab): New optab.
31210         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
31211         (duplicate_and_interleave): Likewise.
31212         * tree-vect-loop.c: Include internal-fn.h.
31213         (neutral_op_for_slp_reduction): New function, split out from
31214         get_initial_defs_for_reduction.
31215         (get_initial_def_for_reduction): Handle option 2 for variable-length
31216         vectors by loading the neutral value into a vector and then shifting
31217         the initial value into element 0.
31218         (get_initial_defs_for_reduction): Replace the code argument with
31219         the neutral value calculated by neutral_op_for_slp_reduction.
31220         Use gimple_build_vector for constant-length vectors.
31221         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
31222         but the first group_size elements have a neutral value.
31223         Use duplicate_and_interleave otherwise.
31224         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
31225         Update call to get_initial_defs_for_reduction.  Handle SLP
31226         reductions for variable-length vectors by creating one vector
31227         result for each scalar result, with the elements associated
31228         with other scalar results stubbed out with the neutral value.
31229         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
31230         Require IFN_VEC_SHL_INSERT for double reductions on
31231         variable-length vectors, or SLP reductions that have
31232         a neutral value.  Require can_duplicate_and_interleave_p
31233         support for variable-length unchained SLP reductions if there
31234         is no neutral value, such as for MIN/MAX reductions.  Also require
31235         the number of vector elements to be a multiple of the number of
31236         SLP statements when doing variable-length unchained SLP reductions.
31237         Update call to vect_create_epilog_for_reduction.
31238         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
31239         and remove initial values.
31240         (duplicate_and_interleave): Make public.
31241         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
31242         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
31244 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31245             Alan Hayward  <alan.hayward@arm.com>
31246             David Sherwood  <david.sherwood@arm.com>
31248         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
31249         (can_duplicate_and_interleave_p): New function.
31250         (vect_get_and_check_slp_defs): Take the vector of statements
31251         rather than just the current one.  Remove excess parentheses.
31252         Restriction rejectinon of vect_constant_def and vect_external_def
31253         for variable-length vectors to boolean types, or types for which
31254         can_duplicate_and_interleave_p is false.
31255         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
31256         (duplicate_and_interleave): New function.
31257         (vect_get_constant_vectors): Use gimple_build_vector for
31258         constant-length vectors and suitable variable-length constant
31259         vectors.  Use duplicate_and_interleave for other variable-length
31260         vectors.  Don't defer the update when inserting new statements.
31262 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31263             Alan Hayward  <alan.hayward@arm.com>
31264             David Sherwood  <david.sherwood@arm.com>
31266         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
31267         min_profitable_iters doesn't go negative.
31269 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31270             Alan Hayward  <alan.hayward@arm.com>
31271             David Sherwood  <david.sherwood@arm.com>
31273         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
31274         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
31275         * optabs.def (vec_mask_load_lanes_optab): New optab.
31276         (vec_mask_store_lanes_optab): Likewise.
31277         * internal-fn.def (MASK_LOAD_LANES): New internal function.
31278         (MASK_STORE_LANES): Likewise.
31279         * internal-fn.c (mask_load_lanes_direct): New macro.
31280         (mask_store_lanes_direct): Likewise.
31281         (expand_mask_load_optab_fn): Handle masked operations.
31282         (expand_mask_load_lanes_optab_fn): New macro.
31283         (expand_mask_store_optab_fn): Handle masked operations.
31284         (expand_mask_store_lanes_optab_fn): New macro.
31285         (direct_mask_load_lanes_optab_supported_p): Likewise.
31286         (direct_mask_store_lanes_optab_supported_p): Likewise.
31287         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
31288         parameter.
31289         (vect_load_lanes_supported): Likewise.
31290         * tree-vect-data-refs.c (strip_conversion): New function.
31291         (can_group_stmts_p): Likewise.
31292         (vect_analyze_data_ref_accesses): Use it instead of checking
31293         for a pair of assignments.
31294         (vect_store_lanes_supported): Take a masked_p parameter.
31295         (vect_load_lanes_supported): Likewise.
31296         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
31297         vect_store_lanes_supported and vect_load_lanes_supported.
31298         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
31299         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
31300         parameter.  Don't allow gaps for masked accesses.
31301         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
31302         and vect_load_lanes_supported.
31303         (get_load_store_type): Take a masked_p parameter and update
31304         call to get_group_load_store_type.
31305         (vectorizable_store): Update call to get_load_store_type.
31306         Handle IFN_MASK_STORE_LANES.
31307         (vectorizable_load): Update call to get_load_store_type.
31308         Handle IFN_MASK_LOAD_LANES.
31310 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31311             Alan Hayward  <alan.hayward@arm.com>
31312             David Sherwood  <david.sherwood@arm.com>
31314         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
31315         modes for SVE.
31316         * config/aarch64/aarch64-protos.h
31317         (aarch64_sve_struct_memory_operand_p): Declare.
31318         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
31319         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
31320         (VPRED, vpred): Handle SVE structure modes.
31321         * config/aarch64/constraints.md (Utx): New constraint.
31322         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
31323         (aarch64_sve_struct_nonimmediate_operand): New predicates.
31324         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
31325         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
31326         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
31327         structure modes.  Split into pieces after RA.
31328         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
31329         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
31330         New patterns.
31331         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
31332         SVE structure modes.
31333         (aarch64_classify_address): Likewise.
31334         (sizetochar): Move earlier in file.
31335         (aarch64_print_operand): Handle SVE register lists.
31336         (aarch64_array_mode): New function.
31337         (aarch64_sve_struct_memory_operand_p): Likewise.
31338         (TARGET_ARRAY_MODE): Redefine.
31340 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31341             Alan Hayward  <alan.hayward@arm.com>
31342             David Sherwood  <david.sherwood@arm.com>
31344         * target.def (array_mode): New target hook.
31345         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
31346         * doc/tm.texi: Regenerate.
31347         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
31348         * hooks.c (hook_optmode_mode_uhwi_none): New function.
31349         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
31350         targetm.array_mode.
31351         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
31352         type sizes.
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         * fold-const.c (fold_binary_loc): Check the argument types
31359         rather than the result type when testing for a vector operation.
31361 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31363         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
31364         * doc/tm.texi: Regenerate.
31366 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31367             Alan Hayward  <alan.hayward@arm.com>
31368             David Sherwood  <david.sherwood@arm.com>
31370         * doc/invoke.texi (-msve-vector-bits=): Document new option.
31371         (sve): Document new AArch64 extension.
31372         * doc/md.texi (w): Extend the description of the AArch64
31373         constraint to include SVE vectors.
31374         (Upl, Upa): Document new AArch64 predicate constraints.
31375         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
31376         enum.
31377         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
31378         (msve-vector-bits=): New option.
31379         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
31380         SVE when these are disabled.
31381         (sve): New extension.
31382         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
31383         modes.  Adjust their number of units based on aarch64_sve_vg.
31384         (MAX_BITSIZE_MODE_ANY_MODE): Define.
31385         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
31386         aarch64_addr_query_type.
31387         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
31388         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
31389         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
31390         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
31391         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
31392         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
31393         (aarch64_simd_imm_zero_p): Delete.
31394         (aarch64_check_zero_based_sve_index_immediate): Declare.
31395         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
31396         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
31397         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
31398         (aarch64_sve_float_mul_immediate_p): Likewise.
31399         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
31400         rather than an rtx.
31401         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
31402         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
31403         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
31404         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
31405         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
31406         (aarch64_regmode_natural_size): Likewise.
31407         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
31408         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
31409         left one place.
31410         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
31411         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
31412         for VG and the SVE predicate registers.
31413         (V_ALIASES): Add a "z"-prefixed alias.
31414         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
31415         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
31416         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
31417         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
31418         (REG_CLASS_NAMES): Add entries for them.
31419         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
31420         and the predicate registers.
31421         (aarch64_sve_vg): Declare.
31422         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
31423         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
31424         (REGMODE_NATURAL_SIZE): Define.
31425         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
31426         SVE macros.
31427         * config/aarch64/aarch64.c: Include cfgrtl.h.
31428         (simd_immediate_info): Add a constructor for series vectors,
31429         and an associated step field.
31430         (aarch64_sve_vg): New variable.
31431         (aarch64_dbx_register_number): Handle VG and the predicate registers.
31432         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
31433         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
31434         (VEC_ANY_DATA, VEC_STRUCT): New constants.
31435         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
31436         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
31437         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
31438         (aarch64_get_mask_mode): New functions.
31439         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
31440         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
31441         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
31442         predicate modes and predicate registers.  Explicitly restrict
31443         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
31444         to store a vector mode if it is recognized by
31445         aarch64_classify_vector_mode.
31446         (aarch64_regmode_natural_size): New function.
31447         (aarch64_hard_regno_caller_save_mode): Return the original mode
31448         for predicates.
31449         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
31450         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
31451         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
31452         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
31453         functions.
31454         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
31455         does not overlap dest if the function is frame-related.  Handle
31456         SVE constants.
31457         (aarch64_split_add_offset): New function.
31458         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
31459         them aarch64_add_offset.
31460         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
31461         and update call to aarch64_sub_sp.
31462         (aarch64_add_cfa_expression): New function.
31463         (aarch64_expand_prologue): Pass extra temporary registers to the
31464         functions above.  Handle the case in which we need to emit new
31465         DW_CFA_expressions for registers that were originally saved
31466         relative to the stack pointer, but now have to be expressed
31467         relative to the frame pointer.
31468         (aarch64_output_mi_thunk): Pass extra temporary registers to the
31469         functions above.
31470         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
31471         IP0 and IP1 values for SVE frames.
31472         (aarch64_expand_vec_series): New function.
31473         (aarch64_expand_sve_widened_duplicate): Likewise.
31474         (aarch64_expand_sve_const_vector): Likewise.
31475         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
31476         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
31477         into the register, rather than emitting a SET directly.
31478         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
31479         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
31480         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
31481         (offset_9bit_signed_scaled_p): New functions.
31482         (aarch64_replicate_bitmask_imm): New function.
31483         (aarch64_bitmask_imm): Use it.
31484         (aarch64_cannot_force_const_mem): Reject expressions involving
31485         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
31486         (aarch64_classify_index): Handle SVE indices, by requiring
31487         a plain register index with a scale that matches the element size.
31488         (aarch64_classify_address): Handle SVE addresses.  Assert that
31489         the mode of the address is VOIDmode or an integer mode.
31490         Update call to aarch64_classify_symbol.
31491         (aarch64_classify_symbolic_expression): Update call to
31492         aarch64_classify_symbol.
31493         (aarch64_const_vec_all_in_range_p): New function.
31494         (aarch64_print_vector_float_operand): Likewise.
31495         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
31496         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
31497         and the FP immediates 1.0 and 0.5.
31498         (aarch64_print_address_internal): Handle SVE addresses.
31499         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
31500         (aarch64_regno_regclass): Handle predicate registers.
31501         (aarch64_secondary_reload): Handle big-endian reloads of SVE
31502         data modes.
31503         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
31504         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
31505         (aarch64_convert_sve_vector_bits): New function.
31506         (aarch64_override_options): Use it to handle -msve-vector-bits=.
31507         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
31508         rather than an rtx.
31509         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
31510         Handle SVE vector and predicate modes.  Accept VL-based constants
31511         that need only one temporary register, and VL offsets that require
31512         no temporary registers.
31513         (aarch64_conditional_register_usage): Mark the predicate registers
31514         as fixed if SVE isn't available.
31515         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
31516         Return true for SVE vector and predicate modes.
31517         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
31518         rather than an unsigned int.  Handle SVE modes.
31519         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
31520         SVE modes.
31521         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
31522         if SVE is enabled.
31523         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
31524         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
31525         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
31526         (aarch64_sve_float_mul_immediate_p): New functions.
31527         (aarch64_sve_valid_immediate): New function.
31528         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
31529         Explicitly reject structure modes.  Check for INDEX constants.
31530         Handle PTRUE and PFALSE constants.
31531         (aarch64_check_zero_based_sve_index_immediate): New function.
31532         (aarch64_simd_imm_zero_p): Delete.
31533         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
31534         vector modes.  Accept constants in the range of CNT[BHWD].
31535         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
31536         ask for an Advanced SIMD mode.
31537         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
31538         (aarch64_simd_vector_alignment): Handle SVE predicates.
31539         (aarch64_vectorize_preferred_vector_alignment): New function.
31540         (aarch64_simd_vector_alignment_reachable): Use it instead of
31541         the vector size.
31542         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
31543         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
31544         functions.
31545         (MAX_VECT_LEN): Delete.
31546         (expand_vec_perm_d): Add a vec_flags field.
31547         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
31548         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
31549         (aarch64_evpc_ext): Don't apply a big-endian lane correction
31550         for SVE modes.
31551         (aarch64_evpc_rev): Rename to...
31552         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
31553         (aarch64_evpc_rev_global): New function.
31554         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
31555         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
31556         MAX_VECT_LEN.
31557         (aarch64_evpc_sve_tbl): New function.
31558         (aarch64_expand_vec_perm_const_1): Update after rename of
31559         aarch64_evpc_rev.  Handle SVE permutes too, trying
31560         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
31561         than aarch64_evpc_tbl.
31562         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
31563         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
31564         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
31565         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
31566         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
31567         (aarch64_expand_sve_vcond): New functions.
31568         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
31569         of aarch64_vector_mode_p.
31570         (aarch64_dwarf_poly_indeterminate_value): New function.
31571         (aarch64_compute_pressure_classes): Likewise.
31572         (aarch64_can_change_mode_class): Likewise.
31573         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
31574         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
31575         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
31576         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
31577         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
31578         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
31579         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
31580         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
31581         constraints.
31582         (Dn, Dl, Dr): Accept const as well as const_vector.
31583         (Dz): Likewise.  Compare against CONST0_RTX.
31584         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
31585         of "vector" where appropriate.
31586         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
31587         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
31588         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
31589         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
31590         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
31591         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
31592         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
31593         (v_int_equiv): Extend to SVE modes.
31594         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
31595         mode attributes.
31596         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
31597         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
31598         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
31599         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
31600         (SVE_COND_FP_CMP): New int iterators.
31601         (perm_hilo): Handle the new unpack unspecs.
31602         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
31603         attributes.
31604         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
31605         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
31606         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
31607         (aarch64_equality_operator, aarch64_constant_vector_operand)
31608         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
31609         (aarch64_sve_nonimmediate_operand): Likewise.
31610         (aarch64_sve_general_operand): Likewise.
31611         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
31612         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
31613         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
31614         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
31615         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
31616         (aarch64_sve_float_arith_immediate): Likewise.
31617         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
31618         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
31619         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
31620         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
31621         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
31622         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
31623         (aarch64_sve_float_arith_operand): Likewise.
31624         (aarch64_sve_float_arith_with_sub_operand): Likewise.
31625         (aarch64_sve_float_mul_operand): Likewise.
31626         (aarch64_sve_vec_perm_operand): Likewise.
31627         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
31628         (aarch64_mov_operand): Accept const_poly_int and const_vector.
31629         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
31630         as well as const_vector.
31631         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
31632         in file.  Use CONST0_RTX and CONSTM1_RTX.
31633         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
31634         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
31635         Use aarch64_simd_imm_zero.
31636         * config/aarch64/aarch64-sve.md: New file.
31637         * config/aarch64/aarch64.md: Include it.
31638         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
31639         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
31640         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
31641         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
31642         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
31643         (sve): New attribute.
31644         (enabled): Disable instructions with the sve attribute unless
31645         TARGET_SVE.
31646         (movqi, movhi): Pass CONST_POLY_INT operaneds through
31647         aarch64_expand_mov_immediate.
31648         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
31649         CNT[BHSD] immediates.
31650         (movti): Split CONST_POLY_INT moves into two halves.
31651         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
31652         Split additions that need a temporary here if the destination
31653         is the stack pointer.
31654         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
31655         (*add<mode>3_poly_1): New instruction.
31656         (set_clobber_cc): New expander.
31658 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31660         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
31661         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
31662         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
31663         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
31664         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
31665         Change innermode from fixed_mode_size to machine_mode.
31666         (simplify_subreg): Update call accordingly.  Handle a constant-sized
31667         subreg of a variable-length CONST_VECTOR.
31669 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
31670             Alan Hayward  <alan.hayward@arm.com>
31671             David Sherwood  <david.sherwood@arm.com>
31673         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
31674         (add_offset_to_base): New function, split out from...
31675         (create_mem_ref): ...here.  When handling a scale other than 1,
31676         check first whether the address is valid without the offset.
31677         Add it into the base if so, leaving the index and scale as-is.
31679 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
31681         PR c++/83778
31682         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
31683         fold_for_warn before checking if arg2 is INTEGER_CST.
31685 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
31687         * config/rs6000/predicates.md (load_multiple_operation): Delete.
31688         (store_multiple_operation): Delete.
31689         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
31690         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
31691         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
31692         guarded by TARGET_STRING.
31693         (rs6000_output_load_multiple): Delete.
31694         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
31695         OPTION_MASK_STRING / TARGET_STRING handling.
31696         (print_operand) <'N', 'O'>: Add comment that these are unused now.
31697         (const rs6000_opt_masks) <"string">: Change mask to 0.
31698         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
31699         (MASK_STRING): Delete.
31700         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
31701         parts.  Simplify.
31702         (load_multiple): Delete.
31703         (*ldmsi8): Delete.
31704         (*ldmsi7): Delete.
31705         (*ldmsi6): Delete.
31706         (*ldmsi5): Delete.
31707         (*ldmsi4): Delete.
31708         (*ldmsi3): Delete.
31709         (store_multiple): Delete.
31710         (*stmsi8): Delete.
31711         (*stmsi7): Delete.
31712         (*stmsi6): Delete.
31713         (*stmsi5): Delete.
31714         (*stmsi4): Delete.
31715         (*stmsi3): Delete.
31716         (movmemsi_8reg): Delete.
31717         (corresponding unnamed define_insn): Delete.
31718         (movmemsi_6reg): Delete.
31719         (corresponding unnamed define_insn): Delete.
31720         (movmemsi_4reg): Delete.
31721         (corresponding unnamed define_insn): Delete.
31722         (movmemsi_2reg): Delete.
31723         (corresponding unnamed define_insn): Delete.
31724         (movmemsi_1reg): Delete.
31725         (corresponding unnamed define_insn): Delete.
31726         * config/rs6000/rs6000.opt (mno-string): New.
31727         (mstring): Replace by deprecation warning stub.
31728         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
31730 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
31732         * regrename.c (regrename_do_replace): If replacing the same
31733         reg multiple times, try to reuse last created gen_raw_REG.
31735         PR debug/81155
31736         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
31737         main to workaround a bug in GDB.
31739 2018-01-12  Tom de Vries  <tom@codesourcery.com>
31741         PR target/83737
31742         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
31744 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
31746         PR rtl-optimization/80481
31747         * ira-color.c (get_cap_member): New function.
31748         (allocnos_conflict_by_live_ranges_p): Use it.
31749         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
31750         (setup_slot_coalesced_allocno_live_ranges): Ditto.
31752 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
31754         PR target/83628
31755         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
31756         (*saddl_se_1): Ditto.
31757         (*ssubsi_1): Ditto.
31758         (*ssubl_se_1): Ditto.
31760 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
31762         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
31763         rather than wi::to_widest for DR_INITs.
31764         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
31765         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
31766         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
31767         INTEGER_CSTs.
31768         (vect_analyze_group_access_1): Note that here.
31770 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
31772         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
31773         polynomial type sizes.
31775 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
31777         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
31778         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
31779         (gimple_add_tmp_var): Likewise.
31781 2018-01-12  Martin Liska  <mliska@suse.cz>
31783         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
31784         (gimple_alloc_sizes): Likewise.
31785         (dump_gimple_statistics): Use PRIu64 in printf format.
31786         * gimple.h: Change uint64_t to int.
31788 2018-01-12  Martin Liska  <mliska@suse.cz>
31790         * tree-core.h: Use uint64_t instead of int.
31791         * tree.c (tree_node_counts): Likewise.
31792         (tree_node_sizes): Likewise.
31793         (dump_tree_statistics): Use PRIu64 in printf format.
31795 2018-01-12  Martin Liska  <mliska@suse.cz>
31797         * Makefile.in: As qsort_chk is implemented in vec.c, add
31798         vec.o to linkage of gencfn-macros.
31799         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
31800         passing the info to record_node_allocation_statistics.
31801         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
31802         and pass the info.
31803         * ggc-common.c (struct ggc_usage): Add operator== and use
31804         it in operator< and compare function.
31805         * mem-stats.h (struct mem_usage): Likewise.
31806         * vec.c (struct vec_usage): Remove operator< and compare
31807         function. Can be simply inherited.
31809 2018-01-12  Martin Jambor  <mjambor@suse.cz>
31811         PR target/81616
31812         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
31813         * tree-ssa-math-opts.c: Include domwalk.h.
31814         (convert_mult_to_fma_1): New function.
31815         (fma_transformation_info): New type.
31816         (fma_deferring_state): Likewise.
31817         (cancel_fma_deferring): New function.
31818         (result_of_phi): Likewise.
31819         (last_fma_candidate_feeds_initial_phi): Likewise.
31820         (convert_mult_to_fma): Added deferring logic, split actual
31821         transformation to convert_mult_to_fma_1.
31822         (math_opts_dom_walker): New type.
31823         (math_opts_dom_walker::after_dom_children): New method, body moved
31824         here from pass_optimize_widening_mul::execute, added deferring logic
31825         bits.
31826         (pass_optimize_widening_mul::execute): Moved most of code to
31827         math_opts_dom_walker::after_dom_children.
31828         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
31829         * config/i386/i386.c (ix86_option_override_internal): Added
31830         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
31832 2018-01-12  Richard Biener  <rguenther@suse.de>
31834         PR debug/83157
31835         * dwarf2out.c (gen_variable_die): Do not reset old_die for
31836         inline instance vars.
31838 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
31840         PR target/81819
31841         * config/rx/rx.c (rx_is_restricted_memory_address):
31842         Handle SUBREG case.
31844 2018-01-12  Richard Biener  <rguenther@suse.de>
31846         PR tree-optimization/80846
31847         * target.def (split_reduction): New target hook.
31848         * targhooks.c (default_split_reduction): New function.
31849         * targhooks.h (default_split_reduction): Declare.
31850         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
31851         target requests first reduce vectors by combining low and high
31852         parts.
31853         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
31854         (get_vectype_for_scalar_type_and_size): Export.
31855         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
31856         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
31857         * doc/tm.texi: Regenerate.
31858         * config/i386/i386.c (ix86_split_reduction): Implement
31859         TARGET_VECTORIZE_SPLIT_REDUCTION.
31861 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
31863         PR target/83368
31864         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
31865         in PIC mode except for TARGET_VXWORKS_RTP.
31866         * config/sparc/sparc.c: Include cfgrtl.h.
31867         (TARGET_INIT_PIC_REG): Define.
31868         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
31869         (sparc_pic_register_p): New predicate.
31870         (sparc_legitimate_address_p): Use it.
31871         (sparc_legitimize_pic_address): Likewise.
31872         (sparc_delegitimize_address): Likewise.
31873         (sparc_mode_dependent_address_p): Likewise.
31874         (gen_load_pcrel_sym): Remove 4th parameter.
31875         (load_got_register): Adjust call to above.  Remove obsolete stuff.
31876         (sparc_expand_prologue): Do not call load_got_register here.
31877         (sparc_flat_expand_prologue): Likewise.
31878         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
31879         (sparc_use_pseudo_pic_reg): New function.
31880         (sparc_init_pic_reg): Likewise.
31881         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
31882         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
31884 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
31886         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
31887         Add item for branch_cost.
31889 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
31891         PR rtl-optimization/83565
31892         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
31893         not extend the result to a larger mode for rotate operations.
31894         (num_sign_bit_copies1): Likewise.
31896 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
31898         PR target/40411
31899         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
31900         -symbolic.
31901         Use values-Xc.o for -pedantic.
31902         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
31904 2018-01-12  Martin Liska  <mliska@suse.cz>
31906         PR ipa/83054
31907         * ipa-devirt.c (final_warning_record::grow_type_warnings):
31908         New function.
31909         (possible_polymorphic_call_targets): Use it.
31910         (ipa_devirt): Likewise.
31912 2018-01-12  Martin Liska  <mliska@suse.cz>
31914         * profile-count.h (enum profile_quality): Use 0 as invalid
31915         enum value of profile_quality.
31917 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
31919         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
31920         -mext-string options.
31922 2018-01-12  Richard Biener  <rguenther@suse.de>
31924         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
31925         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
31926         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
31927         Likewise.
31928         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
31930 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31932         * configure.ac (--with-long-double-format): Add support for the
31933         configuration option to change the default long double format on
31934         PowerPC systems.
31935         * config.gcc (powerpc*-linux*-*): Likewise.
31936         * configure: Regenerate.
31937         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
31938         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
31939         used without modification.
31941 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31943         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
31944         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
31945         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
31946         MISC_BUILTIN_SPEC_BARRIER.
31947         (rs6000_init_builtins): Likewise.
31948         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
31949         enum value.
31950         (speculation_barrier): New define_insn.
31951         * doc/extend.texi: Document __builtin_speculation_barrier.
31953 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
31955         PR target/83203
31956         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
31957         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
31958         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
31959         iterators.
31960         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
31961         integral modes instead of "ss" and "sd".
31962         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
31963         vectors with 32-bit and 64-bit elements.
31964         (vecdupssescalarmodesuffix): New mode attribute.
31965         (vec_dup<mode>): Use it.
31967 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
31969         PR target/83330
31970         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
31971         frame if argument is passed on stack.
31973 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
31975         PR target/82682
31976         * ree.c (combine_reaching_defs): Optimize also
31977         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
31978         reg2=any_extend(exp); reg1=reg2;, formatting fix.
31980 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
31982         PR middle-end/83189
31983         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
31985 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
31987         PR middle-end/83718
31988         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
31989         after they are computed.
31991 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
31993         PR tree-optimization/83695
31994         * gimple-loop-linterchange.cc
31995         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
31996         reset cached scev information after interchange.
31997         (pass_linterchange::execute): Remove call to scev_reset_htab.
31999 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32001         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
32002         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
32003         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
32004         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
32005         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
32006         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
32007         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
32008         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
32009         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
32010         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
32011         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
32012         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
32013         (V_lane_reg): Likewise.
32014         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
32015         New define_expand.
32016         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
32017         (vfmal_lane_low<mode>_intrinsic,
32018         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
32019         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
32020         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
32021         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
32022         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
32023         vfmsl_lane_high<mode>_intrinsic): New define_insns.
32025 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32027         * config/arm/arm-cpus.in (fp16fml): New feature.
32028         (ALL_SIMD): Add fp16fml.
32029         (armv8.2-a): Add fp16fml as an option.
32030         (armv8.3-a): Likewise.
32031         (armv8.4-a): Add fp16fml as part of fp16.
32032         * config/arm/arm.h (TARGET_FP16FML): Define.
32033         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
32034         when appropriate.
32035         * config/arm/arm-modes.def (V2HF): Define.
32036         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
32037         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
32038         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
32039         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
32040         vfmsl_low, vfmsl_high): New set of builtins.
32041         * config/arm/iterators.md (PLUSMINUS): New code iterator.
32042         (vfml_op): New code attribute.
32043         (VFMLHALVES): New int iterator.
32044         (VFML, VFMLSEL): New mode attributes.
32045         (V_reg): Define mapping for V2HF.
32046         (V_hi, V_lo): New mode attributes.
32047         (VF_constraint): Likewise.
32048         (vfml_half, vfml_half_selector): New int attributes.
32049         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
32050         define_expand.
32051         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
32052         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
32053         New define_insn.
32054         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
32055         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
32056         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
32057         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
32058         documentation.
32059         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
32060         Document new effective target and option set.
32062 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32064         * config/arm/arm-cpus.in (armv8_4): New feature.
32065         (ARMv8_4a): New fgroup.
32066         (armv8.4-a): New arch.
32067         * config/arm/arm-tables.opt: Regenerate.
32068         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
32069         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
32070         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
32071         Add matching rules for -march=armv8.4-a and extensions.
32072         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
32074 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
32076         PR target/81821
32077         * config/rx/rx.md (BW): New mode attribute.
32078         (sync_lock_test_and_setsi): Add mode suffix to insn output.
32080 2018-01-11  Richard Biener  <rguenther@suse.de>
32082         PR tree-optimization/83435
32083         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
32084         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
32085         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
32087 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32088             Alan Hayward  <alan.hayward@arm.com>
32089             David Sherwood  <david.sherwood@arm.com>
32091         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
32092         field.
32093         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
32094         (aarch64_print_address_internal): Use it to check for a zero offset.
32096 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32097             Alan Hayward  <alan.hayward@arm.com>
32098             David Sherwood  <david.sherwood@arm.com>
32100         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
32101         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
32102         Return a poly_int64 rather than a HOST_WIDE_INT.
32103         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
32104         rather than a HOST_WIDE_INT.
32105         * config/aarch64/aarch64.h (aarch64_frame): Protect with
32106         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
32107         hard_fp_offset, frame_size, initial_adjust, callee_offset and
32108         final_offset from HOST_WIDE_INT to poly_int64.
32109         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
32110         to_constant when getting the number of units in an Advanced SIMD
32111         mode.
32112         (aarch64_builtin_vectorized_function): Check for a constant number
32113         of units.
32114         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
32115         GET_MODE_SIZE.
32116         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
32117         attribute instead of GET_MODE_NUNITS.
32118         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
32119         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
32120         GET_MODE_SIZE for fixed-size registers.
32121         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
32122         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
32123         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
32124         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
32125         (aarch64_print_operand, aarch64_print_address_internal)
32126         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
32127         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
32128         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
32129         Handle polynomial GET_MODE_SIZE.
32130         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
32131         wider than SImode without modification.
32132         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
32133         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
32134         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
32135         passing and returning SVE modes.
32136         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
32137         rather than GEN_INT.
32138         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
32139         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
32140         (aarch64_allocate_and_probe_stack_space): Likewise.
32141         (aarch64_layout_frame): Cope with polynomial offsets.
32142         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
32143         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
32144         polynomial offsets.
32145         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
32146         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
32147         poly_int64 rather than a HOST_WIDE_INT.
32148         (aarch64_get_separate_components, aarch64_process_components)
32149         (aarch64_expand_prologue, aarch64_expand_epilogue)
32150         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
32151         (aarch64_anchor_offset): New function, split out from...
32152         (aarch64_legitimize_address): ...here.
32153         (aarch64_builtin_vectorization_cost): Handle polynomial
32154         TYPE_VECTOR_SUBPARTS.
32155         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
32156         GET_MODE_NUNITS.
32157         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
32158         number of elements from the PARALLEL rather than the mode.
32159         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
32160         rather than GET_MODE_BITSIZE.
32161         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
32162         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
32163         (aarch64_expand_vec_perm_const_1): Handle polynomial
32164         d->perm.length () and d->perm elements.
32165         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
32166         Apply to_constant to d->perm elements.
32167         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
32168         polynomial CONST_VECTOR_NUNITS.
32169         (aarch64_move_pointer): Take amount as a poly_int64 rather
32170         than an int.
32171         (aarch64_progress_pointer): Avoid temporary variable.
32172         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
32173         the mode attribute instead of GET_MODE.
32175 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32176             Alan Hayward  <alan.hayward@arm.com>
32177             David Sherwood  <david.sherwood@arm.com>
32179         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
32180         x exists before using it.
32181         (aarch64_add_constant_internal): Rename to...
32182         (aarch64_add_offset_1): ...this.  Replace regnum with separate
32183         src and dest rtxes.  Handle the case in which they're different,
32184         including when the offset is zero.  Replace scratchreg with an rtx.
32185         Use 2 additions if there is no spare register into which we can
32186         move a 16-bit constant.
32187         (aarch64_add_constant): Delete.
32188         (aarch64_add_offset): Replace reg with separate src and dest
32189         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
32190         Use aarch64_add_offset_1.
32191         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
32192         an rtx rather than an int.  Take the delta as a poly_int64
32193         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
32194         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
32195         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
32196         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
32197         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
32198         and aarch64_add_sp.
32199         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
32200         aarch64_add_constant.
32202 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32204         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
32205         Use scalar_float_mode.
32207 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32209         * config/aarch64/aarch64-simd.md
32210         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
32211         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
32212         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
32213         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
32214         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
32215         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
32216         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
32217         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
32218         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
32219         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
32221 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32223         PR target/83514
32224         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
32225         targ_options->x_arm_arch_string is non NULL.
32227 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
32229         * config/aarch64/aarch64.h
32230         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
32232 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
32234         PR target/82096
32235         * expmed.c (emit_store_flag_force): Swap if const op0
32236         and change VOIDmode to mode of op0.
32238 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
32240         PR rtl-optimization/83761
32241         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
32242         than bytes to mode_for_size.
32244 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32246         PR middle-end/83189
32247         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
32248         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
32249         profile.
32251 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32253         PR middle-end/83575
32254         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
32255         when in layout mode.
32256         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
32257         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
32258         partition fixup.
32260 2018-01-10  Michael Collison  <michael.collison@arm.com>
32262         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
32263         * config/aarch64/aarch64-option-extension.def: Add
32264         AARCH64_OPT_EXTENSION of 'fp16fml'.
32265         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32266         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
32267         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
32268         * config/aarch64/constraints.md (Ui7): New constraint.
32269         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
32270         (VFMLA_SEL_W): Ditto.
32271         (f16quad): Ditto.
32272         (f16mac1): Ditto.
32273         (VFMLA16_LOW): New int iterator.
32274         (VFMLA16_HIGH): Ditto.
32275         (UNSPEC_FMLAL): New unspec.
32276         (UNSPEC_FMLSL): Ditto.
32277         (UNSPEC_FMLAL2): Ditto.
32278         (UNSPEC_FMLSL2): Ditto.
32279         (f16mac): New code attribute.
32280         * config/aarch64/aarch64-simd-builtins.def
32281         (aarch64_fmlal_lowv2sf): Ditto.
32282         (aarch64_fmlsl_lowv2sf): Ditto.
32283         (aarch64_fmlalq_lowv4sf): Ditto.
32284         (aarch64_fmlslq_lowv4sf): Ditto.
32285         (aarch64_fmlal_highv2sf): Ditto.
32286         (aarch64_fmlsl_highv2sf): Ditto.
32287         (aarch64_fmlalq_highv4sf): Ditto.
32288         (aarch64_fmlslq_highv4sf): Ditto.
32289         (aarch64_fmlal_lane_lowv2sf): Ditto.
32290         (aarch64_fmlsl_lane_lowv2sf): Ditto.
32291         (aarch64_fmlal_laneq_lowv2sf): Ditto.
32292         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
32293         (aarch64_fmlalq_lane_lowv4sf): Ditto.
32294         (aarch64_fmlsl_lane_lowv4sf): Ditto.
32295         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
32296         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
32297         (aarch64_fmlal_lane_highv2sf): Ditto.
32298         (aarch64_fmlsl_lane_highv2sf): Ditto.
32299         (aarch64_fmlal_laneq_highv2sf): Ditto.
32300         (aarch64_fmlsl_laneq_highv2sf): Ditto.
32301         (aarch64_fmlalq_lane_highv4sf): Ditto.
32302         (aarch64_fmlsl_lane_highv4sf): Ditto.
32303         (aarch64_fmlalq_laneq_highv4sf): Ditto.
32304         (aarch64_fmlsl_laneq_highv4sf): Ditto.
32305         * config/aarch64/aarch64-simd.md:
32306         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
32307         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
32308         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
32309         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
32310         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
32311         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
32312         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
32313         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
32314         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
32315         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
32316         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
32317         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
32318         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
32319         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
32320         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
32321         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
32322         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
32323         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
32324         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
32325         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
32326         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
32327         (vfmlsl_low_u32): Ditto.
32328         (vfmlalq_low_u32): Ditto.
32329         (vfmlslq_low_u32): Ditto.
32330         (vfmlal_high_u32): Ditto.
32331         (vfmlsl_high_u32): Ditto.
32332         (vfmlalq_high_u32): Ditto.
32333         (vfmlslq_high_u32): Ditto.
32334         (vfmlal_lane_low_u32): Ditto.
32335         (vfmlsl_lane_low_u32): Ditto.
32336         (vfmlal_laneq_low_u32): Ditto.
32337         (vfmlsl_laneq_low_u32): Ditto.
32338         (vfmlalq_lane_low_u32): Ditto.
32339         (vfmlslq_lane_low_u32): Ditto.
32340         (vfmlalq_laneq_low_u32): Ditto.
32341         (vfmlslq_laneq_low_u32): Ditto.
32342         (vfmlal_lane_high_u32): Ditto.
32343         (vfmlsl_lane_high_u32): Ditto.
32344         (vfmlal_laneq_high_u32): Ditto.
32345         (vfmlsl_laneq_high_u32): Ditto.
32346         (vfmlalq_lane_high_u32): Ditto.
32347         (vfmlslq_lane_high_u32): Ditto.
32348         (vfmlalq_laneq_high_u32): Ditto.
32349         (vfmlslq_laneq_high_u32): Ditto.
32350         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
32351         (AARCH64_FL_FOR_ARCH8_4): New.
32352         (AARCH64_ISA_F16FML): New ISA flag.
32353         (TARGET_F16FML): New feature flag for fp16fml.
32354         (doc/invoke.texi): Document new fp16fml option.
32356 2018-01-10  Michael Collison  <michael.collison@arm.com>
32358         * config/aarch64/aarch64-builtins.c:
32359         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
32360         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32361         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
32362         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
32363         (AARCH64_ISA_SHA3): New ISA flag.
32364         (TARGET_SHA3): New feature flag for sha3.
32365         * config/aarch64/iterators.md (sha512_op): New int attribute.
32366         (CRYPTO_SHA512): New int iterator.
32367         (UNSPEC_SHA512H): New unspec.
32368         (UNSPEC_SHA512H2): Ditto.
32369         (UNSPEC_SHA512SU0): Ditto.
32370         (UNSPEC_SHA512SU1): Ditto.
32371         * config/aarch64/aarch64-simd-builtins.def
32372         (aarch64_crypto_sha512hqv2di): New builtin.
32373         (aarch64_crypto_sha512h2qv2di): Ditto.
32374         (aarch64_crypto_sha512su0qv2di): Ditto.
32375         (aarch64_crypto_sha512su1qv2di): Ditto.
32376         (aarch64_eor3qv8hi): Ditto.
32377         (aarch64_rax1qv2di): Ditto.
32378         (aarch64_xarqv2di): Ditto.
32379         (aarch64_bcaxqv8hi): Ditto.
32380         * config/aarch64/aarch64-simd.md:
32381         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
32382         (aarch64_crypto_sha512su0qv2di): Ditto.
32383         (aarch64_crypto_sha512su1qv2di): Ditto.
32384         (aarch64_eor3qv8hi): Ditto.
32385         (aarch64_rax1qv2di): Ditto.
32386         (aarch64_xarqv2di): Ditto.
32387         (aarch64_bcaxqv8hi): Ditto.
32388         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
32389         (vsha512h2q_u64): Ditto.
32390         (vsha512su0q_u64): Ditto.
32391         (vsha512su1q_u64): Ditto.
32392         (veor3q_u16): Ditto.
32393         (vrax1q_u64): Ditto.
32394         (vxarq_u64): Ditto.
32395         (vbcaxq_u16): Ditto.
32396         * config/arm/types.md (crypto_sha512): New type attribute.
32397         (crypto_sha3): Ditto.
32398         (doc/invoke.texi): Document new sha3 option.
32400 2018-01-10  Michael Collison  <michael.collison@arm.com>
32402         * config/aarch64/aarch64-builtins.c:
32403         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
32404         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32405         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
32406         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
32407         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
32408         (AARCH64_ISA_SM4): New ISA flag.
32409         (TARGET_SM4): New feature flag for sm4.
32410         * config/aarch64/aarch64-simd-builtins.def
32411         (aarch64_sm3ss1qv4si): Ditto.
32412         (aarch64_sm3tt1aq4si): Ditto.
32413         (aarch64_sm3tt1bq4si): Ditto.
32414         (aarch64_sm3tt2aq4si): Ditto.
32415         (aarch64_sm3tt2bq4si): Ditto.
32416         (aarch64_sm3partw1qv4si): Ditto.
32417         (aarch64_sm3partw2qv4si): Ditto.
32418         (aarch64_sm4eqv4si): Ditto.
32419         (aarch64_sm4ekeyqv4si): Ditto.
32420         * config/aarch64/aarch64-simd.md:
32421         (aarch64_sm3ss1qv4si): Ditto.
32422         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
32423         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
32424         (aarch64_sm4eqv4si): Ditto.
32425         (aarch64_sm4ekeyqv4si): Ditto.
32426         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
32427         (sm3part_op): Ditto.
32428         (CRYPTO_SM3TT): Ditto.
32429         (CRYPTO_SM3PART): Ditto.
32430         (UNSPEC_SM3SS1): New unspec.
32431         (UNSPEC_SM3TT1A): Ditto.
32432         (UNSPEC_SM3TT1B): Ditto.
32433         (UNSPEC_SM3TT2A): Ditto.
32434         (UNSPEC_SM3TT2B): Ditto.
32435         (UNSPEC_SM3PARTW1): Ditto.
32436         (UNSPEC_SM3PARTW2): Ditto.
32437         (UNSPEC_SM4E): Ditto.
32438         (UNSPEC_SM4EKEY): Ditto.
32439         * config/aarch64/constraints.md (Ui2): New constraint.
32440         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
32441         * config/arm/types.md (crypto_sm3): New type attribute.
32442         (crypto_sm4): Ditto.
32443         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
32444         (vsm3tt1aq_u32): Ditto.
32445         (vsm3tt1bq_u32): Ditto.
32446         (vsm3tt2aq_u32): Ditto.
32447         (vsm3tt2bq_u32): Ditto.
32448         (vsm3partw1q_u32): Ditto.
32449         (vsm3partw2q_u32): Ditto.
32450         (vsm4eq_u32): Ditto.
32451         (vsm4ekeyq_u32): Ditto.
32452         (doc/invoke.texi): Document new sm4 option.
32454 2018-01-10  Michael Collison  <michael.collison@arm.com>
32456         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
32457         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
32458         (AARCH64_FL_FOR_ARCH8_4): New.
32459         (AARCH64_FL_V8_4): New flag.
32460         (doc/invoke.texi): Document new armv8.4-a option.
32462 2018-01-10  Michael Collison  <michael.collison@arm.com>
32464         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
32465         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
32466         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
32467         * config/aarch64/aarch64-option-extension.def: Add
32468         AARCH64_OPT_EXTENSION of 'sha2'.
32469         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
32470         (crypto): Disable sha2 and aes if crypto disabled.
32471         (crypto): Enable aes and sha2 if enabled.
32472         (simd): Disable sha2 and aes if simd disabled.
32473         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
32474         New flags.
32475         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
32476         (TARGET_SHA2): New feature flag for sha2.
32477         (TARGET_AES): New feature flag for aes.
32478         * config/aarch64/aarch64-simd.md:
32479         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
32480         conditional on TARGET_AES.
32481         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
32482         (aarch64_crypto_sha1hsi): Make pattern conditional
32483         on TARGET_SHA2.
32484         (aarch64_crypto_sha1hv4si): Ditto.
32485         (aarch64_be_crypto_sha1hv4si): Ditto.
32486         (aarch64_crypto_sha1su1v4si): Ditto.
32487         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
32488         (aarch64_crypto_sha1su0v4si): Ditto.
32489         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
32490         (aarch64_crypto_sha256su0v4si): Ditto.
32491         (aarch64_crypto_sha256su1v4si): Ditto.
32492         (doc/invoke.texi): Document new aes and sha2 options.
32494 2018-01-10  Martin Sebor  <msebor@redhat.com>
32496         PR tree-optimization/83781
32497         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
32498         as string arrays.
32500 2018-01-11  Martin Sebor  <msebor@gmail.com>
32501             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32503         PR tree-optimization/83501
32504         PR tree-optimization/81703
32506         * tree-ssa-strlen.c (get_string_cst): Rename...
32507         (get_string_len): ...to this.  Handle global constants.
32508         (handle_char_store): Adjust.
32510 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
32511             Jim Wilson  <jimw@sifive.com>
32513         * config/riscv/riscv-protos.h (riscv_output_return): New.
32514         * config/riscv/riscv.c (struct machine_function): New naked_p field.
32515         (riscv_attribute_table, riscv_output_return),
32516         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
32517         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
32518         (riscv_compute_frame_info): Only compute frame->mask if not a naked
32519         function.
32520         (riscv_expand_prologue): Add early return for naked function.
32521         (riscv_expand_epilogue): Likewise.
32522         (riscv_function_ok_for_sibcall): Return false for naked function.
32523         (riscv_set_current_function): New.
32524         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
32525         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
32526         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
32527         * doc/extend.texi (RISC-V Function Attributes): New.
32529 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
32531         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
32532         check for 128-bit long double before checking TCmode.
32533         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
32534         128-bit long doubles before checking TFmode or TCmode.
32535         (FLOAT128_IBM_P): Likewise.
32537 2018-01-10  Martin Sebor  <msebor@redhat.com>
32539         PR tree-optimization/83671
32540         * builtins.c (c_strlen): Unconditionally return zero for the empty
32541         string.
32542         Use -Warray-bounds for warnings.
32543         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
32544         for non-constant array indices with COMPONENT_REF, arrays of
32545         arrays, and pointers to arrays.
32546         (gimple_fold_builtin_strlen): Determine and set length range for
32547         non-constant character arrays.
32549 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
32551         PR middle-end/81897
32552         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
32553         empty blocks.
32555 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
32557         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
32559 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
32561         PR target/83399
32562         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
32563         VECTOR_MEM_ALTIVEC_OR_VSX_P.
32564         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
32565         indexed_or_indirect_operand predicate.
32566         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
32567         (*vsx_le_perm_load_v8hi): Likewise.
32568         (*vsx_le_perm_load_v16qi): Likewise.
32569         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
32570         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
32571         (*vsx_le_perm_store_v8hi): Likewise.
32572         (*vsx_le_perm_store_v16qi): Likewise.
32573         (eight unnamed splitters): Likewise.
32575 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
32577         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
32578         * config/rs6000/emmintrin.h: Likewise.
32579         * config/rs6000/mmintrin.h: Likewise.
32580         * config/rs6000/xmmintrin.h: Likewise.
32582 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
32584         PR c++/43486
32585         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
32586         "public_flag".
32587         * tree.c (tree_nop_conversion): Return true for location wrapper
32588         nodes.
32589         (maybe_wrap_with_location): New function.
32590         (selftest::check_strip_nops): New function.
32591         (selftest::test_location_wrappers): New function.
32592         (selftest::tree_c_tests): Call it.
32593         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
32594         (maybe_wrap_with_location): New decl.
32595         (EXPR_LOCATION_WRAPPER_P): New macro.
32596         (location_wrapper_p): New inline function.
32597         (tree_strip_any_location_wrapper): New inline function.
32599 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
32601         PR target/83735
32602         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
32603         stack_realign_offset for the largest alignment of stack slot
32604         actually used.
32605         (ix86_find_max_used_stack_alignment): New function.
32606         (ix86_finalize_stack_frame_flags): Use it.  Set
32607         max_used_stack_alignment if we don't realign stack.
32608         * config/i386/i386.h (machine_function): Add
32609         max_used_stack_alignment.
32611 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
32613         * config/arm/arm.opt (-mbranch-cost): New option.
32614         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
32615         account.
32617 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
32619         PR target/83629
32620         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
32621         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
32623 2018-01-10  Richard Biener  <rguenther@suse.de>
32625         PR debug/83765
32626         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
32627         early out so it also covers the case where we have a non-NULL
32628         origin.
32630 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
32632         PR tree-optimization/83753
32633         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
32634         for non-strided grouped accesses if the number of elements is 1.
32636 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
32638         PR target/81616
32639         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
32640         * i386.h (TARGET_USE_GATHER): Define.
32641         * x86-tune.def (X86_TUNE_USE_GATHER): New.
32643 2018-01-10  Martin Liska  <mliska@suse.cz>
32645         PR bootstrap/82831
32646         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
32647         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
32648         partitioning.
32649         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
32650         CLEANUP_NO_PARTITIONING is not set.
32652 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
32654         * doc/rtl.texi: Remove documentation of (const ...) wrappers
32655         for vectors, as a partial revert of r254296.
32656         * rtl.h (const_vec_p): Delete.
32657         (const_vec_duplicate_p): Don't test for vector CONSTs.
32658         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
32659         * expmed.c (make_tree): Likewise.
32661         Revert:
32662         * common.md (E, F): Use CONSTANT_P instead of checking for
32663         CONST_VECTOR.
32664         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
32665         checking for CONST_VECTOR.
32667 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
32669         PR middle-end/83575
32670         * predict.c (force_edge_cold): Handle in more sane way edges
32671         with no prediction.
32673 2018-01-09  Carl Love  <cel@us.ibm.com>
32675         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
32676         V4SI, V4SF types.
32677         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
32678         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
32679         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
32680         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
32681         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
32682         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
32683         * config/rs6000/rs6000-protos.h: Add extern defition for
32684         rs6000_generate_float2_double_code.
32685         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
32686         function.
32687         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
32688         (float2_v2df): Add define_expand.
32690 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
32692         PR target/83628
32693         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
32694         op_mode in the force_to_mode call.
32696 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
32698         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
32699         instead of checking each element individually.
32700         (aarch64_evpc_uzp): Likewise.
32701         (aarch64_evpc_zip): Likewise.
32702         (aarch64_evpc_ext): Likewise.
32703         (aarch64_evpc_rev): Likewise.
32704         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
32705         instead of checking each element individually.  Return true without
32706         generating rtl if
32707         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
32708         whether all selected elements come from the same input, instead of
32709         checking each element individually.  Remove calls to gen_rtx_REG,
32710         start_sequence and end_sequence and instead assert that no rtl is
32711         generated.
32713 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
32715         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
32716         order of HIGH and CONST checks.
32718 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
32720         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
32721         if the destination isn't an SSA_NAME.
32723 2018-01-09  Richard Biener  <rguenther@suse.de>
32725         PR tree-optimization/83668
32726         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
32727         move prologue...
32728         (canonicalize_loop_form): ... here, renamed from ...
32729         (canonicalize_loop_closed_ssa_form): ... this and amended to
32730         swap successor edges for loop exit blocks to make us use
32731         the RPO order we need for initial schedule generation.
32733 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
32735         PR tree-optimization/64811
32736         * match.pd: When optimizing comparisons with Inf, avoid
32737         introducing or losing exceptions from comparisons with NaN.
32739 2018-01-09  Martin Liska  <mliska@suse.cz>
32741         PR sanitizer/82517
32742         * asan.c (shadow_mem_size): Add gcc_assert.
32744 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
32746         Don't save registers in main().
32748         PR target/83738
32749         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
32750         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
32751         * config/avr/avr.c (avr_set_current_function): Don't error if
32752         naked, OS_task or OS_main are specified at the same time.
32753         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
32754         OS_main.
32755         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
32756         attribute.
32757         * common/config/avr/avr-common.c (avr_option_optimization_table):
32758         Switch on -mmain-is-OS_task for optimizing compilations.
32760 2018-01-09  Richard Biener  <rguenther@suse.de>
32762         PR tree-optimization/83572
32763         * graphite.c: Include cfganal.h.
32764         (graphite_transform_loops): Connect infinite loops to exit
32765         and remove fake edges at the end.
32767 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
32769         * ipa-inline.c (edge_badness): Revert accidental checkin.
32771 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
32773         PR ipa/80763
32774         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
32775         symbols; not inline clones.
32777 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
32779         PR target/83507
32780         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
32781         hard registers.  Formatting fixes.
32783         PR preprocessor/83722
32784         * gcc.c (try_generate_repro): Pass
32785         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
32786         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
32787         do_report_bug.
32789 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
32790             Kito Cheng  <kito.cheng@gmail.com>
32792         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
32793         (riscv_leaf_function_p): Delete.
32794         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
32796 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32798         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
32799         function.
32800         (do_ifelse): New function.
32801         (do_isel): New function.
32802         (do_sub3): New function.
32803         (do_add3): New function.
32804         (do_load_mask_compare): New function.
32805         (do_overlap_load_compare): New function.
32806         (expand_compare_loop): New function.
32807         (expand_block_compare): Call expand_compare_loop() when appropriate.
32808         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
32809         option description.
32810         (-mblock-compare-inline-loop-limit): New option.
32812 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32814         PR target/83677
32815         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
32816         Reverse order of second and third operands in first alternative.
32817         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
32818         of first and second elements in UNSPEC_VPERMR vector.
32819         (altivec_expand_vec_perm_le): Likewise.
32821 2018-01-08  Jeff Law  <law@redhat.com>
32823         PR rtl-optimizatin/81308
32824         * tree-switch-conversion.c (cfg_altered): New file scoped static.
32825         (process_switch): If group_case_labels makes a change, then set
32826         cfg_altered.
32827         (pass_convert_switch::execute): If a switch is converted, then
32828         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
32830         PR rtl-optimization/81308
32831         * recog.c (split_all_insns): Conditionally cleanup the CFG after
32832         splitting insns.
32834 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
32836         PR target/83663 - Revert r255946
32837         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
32838         generation for cases where splatting a value is not useful.
32839         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
32840         across a vec_duplicate and a paradoxical subreg forming a vector
32841         mode to a vec_concat.
32843 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32845         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
32846         -march=armv8.3-a variants.
32847         * config/arm/t-multilib: Likewise.
32848         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
32850 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32852         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
32853         to generate rtl.
32854         (cceq_ior_compare_complement): Give it a name so I can use it, and
32855         change boolean_or_operator predicate to boolean_operator so it can
32856         be used to generate a crand.
32857         (eqne): New code iterator.
32858         (bd/bd_neg): New code_attrs.
32859         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
32860         a single define_insn.
32861         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
32862         decrement (bdnzt/bdnzf/bdzt/bdzf).
32863         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
32864         with the new names of the branch decrement patterns, and added the
32865         names of the branch decrement conditional patterns.
32867 2018-01-08  Richard Biener  <rguenther@suse.de>
32869         PR tree-optimization/83563
32870         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
32871         cache.
32873 2018-01-08  Richard Biener  <rguenther@suse.de>
32875         PR middle-end/83713
32876         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
32878 2018-01-08  Richard Biener  <rguenther@suse.de>
32880         PR tree-optimization/83685
32881         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
32882         references to abnormals.
32884 2018-01-08  Richard Biener  <rguenther@suse.de>
32886         PR lto/83719
32887         * dwarf2out.c (output_indirect_strings): Handle empty
32888         skeleton_debug_str_hash.
32889         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
32891 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
32893         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
32894         (emit_store_direct): Likewise.
32895         (arc_trampoline_adjust_address): Likewise.
32896         (arc_asm_trampoline_template): New function.
32897         (arc_initialize_trampoline): Use asm_trampoline_template.
32898         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
32899         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
32900         * config/arc/arc.md (flush_icache): Delete pattern.
32902 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
32904         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
32905         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
32906         munaligned-access.
32908 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
32910         PR target/83681
32911         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
32912         by not USED_FOR_TARGET.
32913         (make_pass_resolve_sw_modes): Likewise.
32915 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
32917         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
32918         USED_FOR_TARGET.
32920 2018-01-08  Richard Biener  <rguenther@suse.de>
32922         PR middle-end/83580
32923         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
32925 2018-01-08  Richard Biener  <rguenther@suse.de>
32927         PR middle-end/83517
32928         * match.pd ((t * 2) / 2) -> t): Add missing :c.
32930 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
32932         PR middle-end/81897
32933         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
32934         basic blocks with a small number of successors.
32935         (convert_control_dep_chain_into_preds): Improve handling of
32936         forwarder blocks.
32937         (dump_predicates): Split apart into...
32938         (dump_pred_chain): ...here...
32939         (dump_pred_info): ...and here.
32940         (can_one_predicate_be_invalidated_p): Add debugging printfs.
32941         (can_chain_union_be_invalidated_p): Improve check for invalidation
32942         of paths.
32943         (uninit_uses_cannot_happen): Avoid unnecessary if
32944         convert_control_dep_chain_into_preds yielded nothing.
32946 2018-01-06  Martin Sebor  <msebor@redhat.com>
32948         PR tree-optimization/83640
32949         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
32950         subtracting negative offset from size.
32951         (builtin_access::overlap): Adjust offset bounds of the access to fall
32952         within the size of the object if possible.
32954 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
32956         PR rtl-optimization/83699
32957         * expmed.c (extract_bit_field_1): Restrict the vector usage of
32958         extract_bit_field_as_subreg to cases in which the extracted
32959         value is also a vector.
32961         * lra-constraints.c (process_alt_operands): Test for the equivalence
32962         substitutions when detecting a possible reload cycle.
32964 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
32966         PR debug/83480
32967         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
32968         by default if flag_selective_schedling{,2}.  Formatting fixes.
32970         PR rtl-optimization/83682
32971         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
32972         if it has non-VECTOR_MODE element mode.
32973         (vec_duplicate_p): Likewise.
32975         PR middle-end/83694
32976         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
32977         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
32979 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
32981         PR target/83604
32982         * config/i386/i386-builtin.def
32983         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
32984         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
32985         Require also OPTION_MASK_ISA_AVX512F in addition to
32986         OPTION_MASK_ISA_GFNI.
32987         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
32988         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
32989         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
32990         to OPTION_MASK_ISA_GFNI.
32991         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
32992         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
32993         OPTION_MASK_ISA_AVX512BW.
32994         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
32995         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
32996         addition to OPTION_MASK_ISA_GFNI.
32997         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
32998         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
32999         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
33000         to OPTION_MASK_ISA_GFNI.
33001         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
33002         a requirement for all ISAs rather than any of them with a few
33003         exceptions.
33004         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
33005         processing.
33006         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
33007         bitmasks to be enabled with 3 exceptions, instead of requiring any
33008         enabled ISA with lots of exceptions.
33009         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
33010         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
33011         Change avx512bw in isa attribute to avx512f.
33012         * config/i386/sgxintrin.h: Add license boilerplate.
33013         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
33014         to __AVX512F__ and __AVX512VL to __AVX512VL__.
33015         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
33016         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
33017         defined.
33018         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
33019         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
33020         temporarily sse2 rather than sse if not enabled already.
33022         PR target/83604
33023         * config/i386/sse.md (VI248_VLBW): Rename to ...
33024         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
33025         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
33026         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
33027         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
33028         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
33029         mode iterator instead of VI248_VLBW.
33031 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
33033         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
33034         (record_modified): Skip clobbers; add debug output.
33035         (param_change_prob): Use sreal frequencies.
33037 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
33039         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
33040         punt for user-aligned variables.
33042 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
33044         * tree-chrec.c (chrec_contains_symbols): Return true for
33045         POLY_INT_CST.
33047 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
33049         PR target/82439
33050         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
33051         of (x|y) == x for BICS pattern.
33053 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
33055         PR tree-optimization/83605
33056         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
33057         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
33058         can throw.
33060 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
33062         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
33063         * config/epiphany/rtems.h: New file.
33065 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33066             Uros Bizjak  <ubizjak@gmail.com>
33068         PR target/83554
33069         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
33070         QIreg_operand instead of register_operand predicate.
33071         * config/i386/i386.c (ix86_rop_should_change_byte_p,
33072         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
33073         comments instead of -fmitigate[-_]rop.
33075 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33077         PR bootstrap/81926
33078         * cgraphunit.c (symbol_table::compile): Switch to text_section
33079         before calling assembly_start debug hook.
33080         * run-rtl-passes.c (run_rtl_passes): Likewise.
33081         Include output.h.
33083 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33085         * tree-vrp.c (extract_range_from_binary_expr_1): Check
33086         range_int_cst_p rather than !symbolic_range_p before calling
33087         extract_range_from_multiplicative_op_1.
33089 2018-01-04  Jeff Law  <law@redhat.com>
33091         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
33092         redundant test in assertion.
33094 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33096         * doc/rtl.texi: Document machine_mode wrapper classes.
33098 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33100         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
33101         using tree_to_uhwi.
33103 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33105         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
33106         the VEC_PERM_EXPR fold to fail.
33108 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33110         PR debug/83585
33111         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
33112         to switched_sections.
33114 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33116         PR target/83680
33117         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
33118         test for d.testing.
33120 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
33122         PR target/83387
33123         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
33124         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
33126 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33128         PR debug/83666
33129         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
33130         is BLKmode and bitpos not zero or mode change is needed.
33132 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
33134         PR target/83675
33135         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
33136         TARGET_VIS2.
33138 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
33140         PR target/83628
33141         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
33142         instead of MULT rtx.  Update all corresponding splitters.
33143         (*saddl_se): Ditto.
33144         (*ssub<modesuffix>): Ditto.
33145         (*ssubl_se): Ditto.
33146         (*cmp_sadd_di): Update split patterns.
33147         (*cmp_sadd_si): Ditto.
33148         (*cmp_sadd_sidi): Ditto.
33149         (*cmp_ssub_di): Ditto.
33150         (*cmp_ssub_si): Ditto.
33151         (*cmp_ssub_sidi): Ditto.
33152         * config/alpha/predicates.md (const23_operand): New predicate.
33153         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
33154         Look for ASHIFT, not MULT inner operand.
33155         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
33157 2018-01-04  Martin Liska  <mliska@suse.cz>
33159         PR gcov-profile/83669
33160         * gcov.c (output_intermediate_file): Add version to intermediate
33161         gcov file.
33162         * doc/gcov.texi: Document new field 'version' in intermediate
33163         file format. Fix location of '-k' option of gcov command.
33165 2018-01-04  Martin Liska  <mliska@suse.cz>
33167         PR ipa/82352
33168         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
33170 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
33172         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
33174 2018-01-03  Martin Sebor  <msebor@redhat.com>
33176         PR tree-optimization/83655
33177         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
33178         checking calls with invalid arguments.
33180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33182         * tree-vect-stmts.c (vect_get_store_rhs): New function.
33183         (vectorizable_mask_load_store): Delete.
33184         (vectorizable_call): Return false for masked loads and stores.
33185         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
33186         instead of gimple_assign_rhs1.
33187         (vectorizable_load): Handle IFN_MASK_LOAD.
33188         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
33190 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33192         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
33193         split out from..,
33194         (vectorizable_mask_load_store): ...here.
33195         (vectorizable_load): ...and here.
33197 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33199         * tree-vect-stmts.c (vect_build_all_ones_mask)
33200         (vect_build_zero_merge_argument): New functions, split out from...
33201         (vectorizable_load): ...here.
33203 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33205         * tree-vect-stmts.c (vect_check_store_rhs): New function,
33206         split out from...
33207         (vectorizable_mask_load_store): ...here.
33208         (vectorizable_store): ...and here.
33210 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33212         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
33213         split out from...
33214         (vectorizable_mask_load_store): ...here.
33216 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33218         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
33219         (vect_model_store_cost): Take a vec_load_store_type instead of a
33220         vect_def_type.
33221         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
33222         (vect_model_store_cost): Take a vec_load_store_type instead of a
33223         vect_def_type.
33224         (vectorizable_mask_load_store): Update accordingly.
33225         (vectorizable_store): Likewise.
33226         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
33228 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33230         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
33231         IFN_MASK_LOAD calls here rather than...
33232         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
33234 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33235             Alan Hayward  <alan.hayward@arm.com>
33236             David Sherwood  <david.sherwood@arm.com>
33238         * expmed.c (extract_bit_field_1): For vector extracts,
33239         fall back to extract_bit_field_as_subreg if vec_extract
33240         isn't available.
33242 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33243             Alan Hayward  <alan.hayward@arm.com>
33244             David Sherwood  <david.sherwood@arm.com>
33246         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
33247         they are variable or constant sized.
33248         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
33249         slots for constant-sized data.
33251 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33252             Alan Hayward  <alan.hayward@arm.com>
33253             David Sherwood  <david.sherwood@arm.com>
33255         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
33256         handling COND_EXPRs with boolean comparisons, try to find a better
33257         basis for the mask type than the boolean itself.
33259 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33261         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
33262         is calculated and how it can be overridden.
33263         * genmodes.c (max_bitsize_mode_any_mode): New variable.
33264         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
33265         if defined.
33266         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
33267         if nonzero.
33269 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33270             Alan Hayward  <alan.hayward@arm.com>
33271             David Sherwood  <david.sherwood@arm.com>
33273         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
33274         Remove the mode argument.
33275         (aarch64_simd_valid_immediate): Remove the mode and inverse
33276         arguments.
33277         * config/aarch64/iterators.md (bitsize): New iterator.
33278         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
33279         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
33280         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
33281         aarch64_simd_valid_immediate.
33282         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
33283         (aarch64_reg_or_bic_imm): Likewise.
33284         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
33285         with an insn_type enum and msl with a modifier_type enum.
33286         Replace element_width with a scalar_mode.  Change the shift
33287         to unsigned int.  Add constructors for scalar_float_mode and
33288         scalar_int_mode elements.
33289         (aarch64_vect_float_const_representable_p): Delete.
33290         (aarch64_can_const_movi_rtx_p)
33291         (aarch64_simd_scalar_immediate_valid_for_move)
33292         (aarch64_simd_make_constant): Update call to
33293         aarch64_simd_valid_immediate.
33294         (aarch64_advsimd_valid_immediate_hs): New function.
33295         (aarch64_advsimd_valid_immediate): Likewise.
33296         (aarch64_simd_valid_immediate): Remove mode and inverse
33297         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
33298         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
33299         and aarch64_float_const_representable_p on the result.
33300         (aarch64_output_simd_mov_immediate): Remove mode argument.
33301         Update call to aarch64_simd_valid_immediate and use of
33302         simd_immediate_info.
33303         (aarch64_output_scalar_simd_mov_immediate): Update call
33304         accordingly.
33306 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33307             Alan Hayward  <alan.hayward@arm.com>
33308             David Sherwood  <david.sherwood@arm.com>
33310         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
33311         (mode_nunits): Likewise CONST_MODE_NUNITS.
33312         * machmode.def (ADJUST_NUNITS): Document.
33313         * genmodes.c (mode_data::need_nunits_adj): New field.
33314         (blank_mode): Update accordingly.
33315         (adj_nunits): New variable.
33316         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
33317         parameter.
33318         (emit_mode_size_inline): Set need_bytesize_adj for all modes
33319         listed in adj_nunits.
33320         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
33321         listed in adj_nunits.  Don't emit case statements for such modes.
33322         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
33323         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
33324         nothing if adj_nunits is nonnull.
33325         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
33326         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
33327         (emit_mode_fbit): Update use of print_maybe_const_decl.
33328         (emit_move_size): Likewise.  Treat the array as non-const
33329         if adj_nunits.
33330         (emit_mode_adjustments): Handle adj_nunits.
33332 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33334         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
33335         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
33336         (VECTOR_MODES): Use it.
33337         (make_vector_modes): Take the prefix as an argument.
33339 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33340             Alan Hayward  <alan.hayward@arm.com>
33341             David Sherwood  <david.sherwood@arm.com>
33343         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
33344         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
33345         for MODE_VECTOR_BOOL.
33346         * machmode.def (VECTOR_BOOL_MODE): Document.
33347         * genmodes.c (VECTOR_BOOL_MODE): New macro.
33348         (make_vector_bool_mode): New function.
33349         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
33350         MODE_VECTOR_BOOL.
33351         * lto-streamer-in.c (lto_input_mode_table): Likewise.
33352         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
33353         Likewise.
33354         * stor-layout.c (int_mode_for_mode): Likewise.
33355         * tree.c (build_vector_type_for_mode): Likewise.
33356         * varasm.c (output_constant_pool_2): Likewise.
33357         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
33358         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
33359         for MODE_VECTOR_BOOL.
33360         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
33361         of mode class checks.
33362         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
33363         instead of a list of mode class checks.
33364         (expand_vector_scalar_condition): Likewise.
33365         (type_for_widest_vector_mode): Handle BImode as an inner mode.
33367 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33368             Alan Hayward  <alan.hayward@arm.com>
33369             David Sherwood  <david.sherwood@arm.com>
33371         * machmode.h (mode_size): Change from unsigned short to
33372         poly_uint16_pod.
33373         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
33374         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
33375         or if measurement_type is not polynomial.
33376         (fixed_size_mode::includes_p): Check for constant-sized modes.
33377         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
33378         return a poly_uint16 rather than an unsigned short.
33379         (emit_mode_size): Change the type of mode_size from unsigned short
33380         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
33381         (emit_mode_adjustments): Cope with polynomial vector sizes.
33382         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
33383         for GET_MODE_SIZE.
33384         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
33385         for GET_MODE_SIZE.
33386         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
33387         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
33388         * caller-save.c (setup_save_areas): Likewise.
33389         (replace_reg_with_saved_mem): Likewise.
33390         * calls.c (emit_library_call_value_1): Likewise.
33391         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
33392         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
33393         (gen_lowpart_for_combine): Likewise.
33394         * convert.c (convert_to_integer_1): Likewise.
33395         * cse.c (equiv_constant, cse_insn): Likewise.
33396         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
33397         (cselib_subst_to_values): Likewise.
33398         * dce.c (word_dce_process_block): Likewise.
33399         * df-problems.c (df_word_lr_mark_ref): Likewise.
33400         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
33401         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
33402         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
33403         (rtl_for_decl_location): Likewise.
33404         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
33405         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
33406         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
33407         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
33408         (expand_expr_real_1): Likewise.
33409         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
33410         (pad_below): Likewise.
33411         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
33412         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
33413         * ira.c (get_subreg_tracking_sizes): Likewise.
33414         * ira-build.c (ira_create_allocno_objects): Likewise.
33415         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
33416         (ira_sort_regnos_for_alter_reg): Likewise.
33417         * ira-costs.c (record_operand_costs): Likewise.
33418         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
33419         (resolve_simple_move): Likewise.
33420         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
33421         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
33422         (lra_constraints): Likewise.
33423         (CONST_POOL_OK_P): Reject variable-sized modes.
33424         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
33425         (add_pseudo_to_slot, lra_spill): Likewise.
33426         * omp-low.c (omp_clause_aligned_alignment): Likewise.
33427         * optabs-query.c (get_best_extraction_insn): Likewise.
33428         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
33429         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
33430         (expand_mult_highpart, valid_multiword_target_p): Likewise.
33431         * recog.c (offsettable_address_addr_space_p): Likewise.
33432         * regcprop.c (maybe_mode_change): Likewise.
33433         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
33434         * regrename.c (build_def_use): Likewise.
33435         * regstat.c (dump_reg_info): Likewise.
33436         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
33437         (find_reloads, find_reloads_subreg_address): Likewise.
33438         * reload1.c (eliminate_regs_1): Likewise.
33439         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
33440         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
33441         (simplify_binary_operation_1, simplify_subreg): Likewise.
33442         * targhooks.c (default_function_arg_padding): Likewise.
33443         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
33444         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
33445         (verify_gimple_assign_ternary): Likewise.
33446         * tree-inline.c (estimate_move_cost): Likewise.
33447         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33448         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
33449         (get_address_cost_ainc): Likewise.
33450         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
33451         (vect_supportable_dr_alignment): Likewise.
33452         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
33453         (vectorizable_reduction): Likewise.
33454         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
33455         (vectorizable_operation, vectorizable_load): Likewise.
33456         * tree.c (build_same_sized_truth_vector_type): Likewise.
33457         * valtrack.c (cleanup_auto_inc_dec): Likewise.
33458         * var-tracking.c (emit_note_insn_var_location): Likewise.
33459         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
33460         (ADDR_VEC_ALIGN): Likewise.
33462 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33463             Alan Hayward  <alan.hayward@arm.com>
33464             David Sherwood  <david.sherwood@arm.com>
33466         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
33467         unsigned short.
33468         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
33469         or if measurement_type is polynomial.
33470         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
33471         * combine.c (make_extraction): Likewise.
33472         * dse.c (find_shift_sequence): Likewise.
33473         * dwarf2out.c (mem_loc_descriptor): Likewise.
33474         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
33475         (extract_bit_field, extract_low_bits): Likewise.
33476         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
33477         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
33478         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
33479         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
33480         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
33481         * reload.c (find_reloads): Likewise.
33482         * reload1.c (alter_reg): Likewise.
33483         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
33484         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
33485         * tree-if-conv.c (predicate_mem_writes): Likewise.
33486         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
33487         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
33488         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
33489         * valtrack.c (dead_debug_insert_temp): Likewise.
33490         * varasm.c (mergeable_constant_section): Likewise.
33491         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
33493 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33494             Alan Hayward  <alan.hayward@arm.com>
33495             David Sherwood  <david.sherwood@arm.com>
33497         * expr.c (expand_assignment): Cope with polynomial mode sizes
33498         when assigning to a CONCAT.
33500 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33501             Alan Hayward  <alan.hayward@arm.com>
33502             David Sherwood  <david.sherwood@arm.com>
33504         * machmode.h (mode_precision): Change from unsigned short to
33505         poly_uint16_pod.
33506         (mode_to_precision): Return a poly_uint16 rather than an unsigned
33507         short.
33508         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
33509         or if measurement_type is not polynomial.
33510         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
33511         in which the mode is already known to be a scalar_int_mode.
33512         * genmodes.c (emit_mode_precision): Change the type of mode_precision
33513         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
33514         initializer.
33515         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
33516         for GET_MODE_PRECISION.
33517         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
33518         for GET_MODE_PRECISION.
33519         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
33520         as polynomial.
33521         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
33522         (expand_field_assignment, make_extraction): Likewise.
33523         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
33524         (get_last_value): Likewise.
33525         * convert.c (convert_to_integer_1): Likewise.
33526         * cse.c (cse_insn): Likewise.
33527         * expr.c (expand_expr_real_1): Likewise.
33528         * lra-constraints.c (simplify_operand_subreg): Likewise.
33529         * optabs-query.c (can_atomic_load_p): Likewise.
33530         * optabs.c (expand_atomic_load): Likewise.
33531         (expand_atomic_store): Likewise.
33532         * ree.c (combine_reaching_defs): Likewise.
33533         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
33534         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
33535         * tree.h (type_has_mode_precision_p): Likewise.
33536         * ubsan.c (instrument_si_overflow): Likewise.
33538 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33539             Alan Hayward  <alan.hayward@arm.com>
33540             David Sherwood  <david.sherwood@arm.com>
33542         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
33543         polynomial numbers of units.
33544         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
33545         (valid_vector_subparts_p): New function.
33546         (build_vector_type): Remove temporary shim and take the number
33547         of units as a poly_uint64 rather than an int.
33548         (build_opaque_vector_type): Take the number of units as a
33549         poly_uint64 rather than an int.
33550         * tree.c (build_vector_from_ctor): Handle polynomial
33551         TYPE_VECTOR_SUBPARTS.
33552         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
33553         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
33554         (build_vector_from_val): If the number of units is variable,
33555         use build_vec_duplicate_cst for constant operands and
33556         VEC_DUPLICATE_EXPR otherwise.
33557         (make_vector_type): Remove temporary is_constant ().
33558         (build_vector_type, build_opaque_vector_type): Take the number of
33559         units as a poly_uint64 rather than an int.
33560         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
33561         VECTOR_CST_NELTS.
33562         * cfgexpand.c (expand_debug_expr): Likewise.
33563         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
33564         (store_constructor, expand_expr_real_1): Likewise.
33565         (const_scalar_mask_from_tree): Likewise.
33566         * fold-const-call.c (fold_const_reduction): Likewise.
33567         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
33568         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
33569         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
33570         (fold_relational_const): Likewise.
33571         (native_interpret_vector): Likewise.  Change the size from an
33572         int to an unsigned int.
33573         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
33574         TYPE_VECTOR_SUBPARTS.
33575         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
33576         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
33577         duplicating a non-constant operand into a variable-length vector.
33578         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
33579         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
33580         * ipa-icf.c (sem_variable::equals): Likewise.
33581         * match.pd: Likewise.
33582         * omp-simd-clone.c (simd_clone_subparts): Likewise.
33583         * print-tree.c (print_node): Likewise.
33584         * stor-layout.c (layout_type): Likewise.
33585         * targhooks.c (default_builtin_vectorization_cost): Likewise.
33586         * tree-cfg.c (verify_gimple_comparison): Likewise.
33587         (verify_gimple_assign_binary): Likewise.
33588         (verify_gimple_assign_ternary): Likewise.
33589         (verify_gimple_assign_single): Likewise.
33590         * tree-pretty-print.c (dump_generic_node): Likewise.
33591         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33592         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
33593         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
33594         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
33595         (vect_shift_permute_load_chain): Likewise.
33596         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
33597         (expand_vector_condition, optimize_vector_constructor): Likewise.
33598         (lower_vec_perm, get_compute_type): Likewise.
33599         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
33600         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
33601         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
33602         (vect_recog_mask_conversion_pattern): Likewise.
33603         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
33604         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
33605         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33606         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
33607         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
33608         (vectorizable_shift, vectorizable_operation, vectorizable_store)
33609         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
33610         (supportable_widening_operation): Likewise.
33611         (supportable_narrowing_operation): Likewise.
33612         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
33613         Likewise.
33614         * varasm.c (output_constant): Likewise.
33616 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33617             Alan Hayward  <alan.hayward@arm.com>
33618             David Sherwood  <david.sherwood@arm.com>
33620         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
33621         so that both the length == 3 and length != 3 cases set up their
33622         own permute vectors.  Add comments explaining why we know the
33623         number of elements is constant.
33624         (vect_permute_load_chain): Likewise.
33626 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33627             Alan Hayward  <alan.hayward@arm.com>
33628             David Sherwood  <david.sherwood@arm.com>
33630         * machmode.h (mode_nunits): Change from unsigned char to
33631         poly_uint16_pod.
33632         (ONLY_FIXED_SIZE_MODES): New macro.
33633         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
33634         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
33635         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
33636         New typedefs.
33637         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
33638         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
33639         or if measurement_type is not polynomial.
33640         * genmodes.c (ZERO_COEFFS): New macro.
33641         (emit_mode_nunits_inline): Make mode_nunits_inline return a
33642         poly_uint16.
33643         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
33644         Use ZERO_COEFFS when emitting initializers.
33645         * data-streamer.h (bp_pack_poly_value): New function.
33646         (bp_unpack_poly_value): Likewise.
33647         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
33648         for GET_MODE_NUNITS.
33649         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
33650         for GET_MODE_NUNITS.
33651         * tree.c (make_vector_type): Remove temporary shim and make
33652         the real function take the number of units as a poly_uint64
33653         rather than an int.
33654         (build_vector_type_for_mode): Handle polynomial nunits.
33655         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
33656         * emit-rtl.c (const_vec_series_p_1): Likewise.
33657         (gen_rtx_CONST_VECTOR): Likewise.
33658         * fold-const.c (test_vec_duplicate_folding): Likewise.
33659         * genrecog.c (validate_pattern): Likewise.
33660         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
33661         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
33662         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
33663         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
33664         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
33665         * rtlanal.c (subreg_get_info): Likewise.
33666         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33667         (vect_grouped_load_supported): Likewise.
33668         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
33669         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
33670         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
33671         (simplify_const_unary_operation, simplify_binary_operation_1)
33672         (simplify_const_binary_operation, simplify_ternary_operation)
33673         (test_vector_ops_duplicate, test_vector_ops): Likewise.
33674         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
33675         instead of CONST_VECTOR_NUNITS.
33676         * varasm.c (output_constant_pool_2): Likewise.
33677         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
33678         explicit-encoded elements in the XVEC for variable-length vectors.
33680 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33682         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
33684 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33685             Alan Hayward  <alan.hayward@arm.com>
33686             David Sherwood  <david.sherwood@arm.com>
33688         * coretypes.h (fixed_size_mode): Declare.
33689         (fixed_size_mode_pod): New typedef.
33690         * builtins.h (target_builtins::x_apply_args_mode)
33691         (target_builtins::x_apply_result_mode): Change type to
33692         fixed_size_mode_pod.
33693         * builtins.c (apply_args_size, apply_result_size, result_vector)
33694         (expand_builtin_apply_args_1, expand_builtin_apply)
33695         (expand_builtin_return): Update accordingly.
33697 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33699         * cse.c (hash_rtx_cb): Hash only the encoded elements.
33700         * cselib.c (cselib_hash_rtx): Likewise.
33701         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
33702         CONST_VECTOR encoding.
33704 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33705             Jeff Law  <law@redhat.com>
33707         PR target/83641
33708         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
33709         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
33710         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
33711         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
33713         PR target/83641
33714         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
33715         explicitly probe *sp in a noreturn function if there were any callee
33716         register saves or frame pointer is needed.
33718 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33720         PR debug/83621
33721         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
33722         BLKmode for ternary, binary or unary expressions.
33724         PR debug/83645
33725         * var-tracking.c (delete_vta_debug_insn): New inline function.
33726         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
33727         insns from get_insns () to NULL instead of each bb separately.
33728         Use delete_vta_debug_insn.  No longer static.
33729         (vt_debug_insns_local, variable_tracking_main_1): Adjust
33730         delete_vta_debug_insns callers.
33731         * rtl.h (delete_vta_debug_insns): Declare.
33732         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
33733         instead of variable_tracking_main.
33735 2018-01-03  Martin Sebor  <msebor@redhat.com>
33737         PR tree-optimization/83603
33738         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
33739         arguments past the endof the argument list in functions declared
33740         without a prototype.
33741         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
33742         Avoid checking when arguments are null.
33744 2018-01-03  Martin Sebor  <msebor@redhat.com>
33746         PR c/83559
33747         * doc/extend.texi (attribute const): Fix a typo.
33748         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
33749         issuing -Wsuggest-attribute for void functions.
33751 2018-01-03  Martin Sebor  <msebor@redhat.com>
33753         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
33754         offset_int::from instead of wide_int::to_shwi.
33755         (maybe_diag_overlap): Remove assertion.
33756         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
33757         * gimple-ssa-sprintf.c (format_directive): Same.
33758         (parse_directive): Same.
33759         (sprintf_dom_walker::compute_format_length): Same.
33760         (try_substitute_return_value): Same.
33762 2018-01-03  Jeff Law  <law@redhat.com>
33764         PR middle-end/83654
33765         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
33766         non-constant residual for zero at runtime and avoid probing in
33767         that case.  Reorganize code for trailing problem to mirror handling
33768         of the residual.
33770 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33772         PR tree-optimization/83501
33773         * tree-ssa-strlen.c (get_string_cst): New.
33774         (handle_char_store): Call get_string_cst.
33776 2018-01-03  Martin Liska  <mliska@suse.cz>
33778         PR tree-optimization/83593
33779         * tree-ssa-strlen.c: Include tree-cfg.h.
33780         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
33781         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
33782         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
33783         to false.
33784         (strlen_dom_walker::before_dom_children): Call
33785         gimple_purge_dead_eh_edges. Dump tranformation with details
33786         dump flags.
33787         (strlen_dom_walker::before_dom_children): Update call by adding
33788         new argument cleanup_eh.
33789         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
33791 2018-01-03  Martin Liska  <mliska@suse.cz>
33793         PR ipa/83549
33794         * cif-code.def (VARIADIC_THUNK): New enum value.
33795         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
33796         thunks.
33798 2018-01-03  Jan Beulich  <jbeulich@suse.com>
33800         * sse.md (mov<mode>_internal): Tighten condition for when to use
33801         vmovdqu<ssescalarsize> for TI and OI modes.
33803 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33805         Update copyright years.
33807 2018-01-03  Martin Liska  <mliska@suse.cz>
33809         PR ipa/83594
33810         * ipa-visibility.c (function_and_variable_visibility): Skip
33811         functions with noipa attribure.
33813 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
33815         * gcc.c (process_command): Update copyright notice dates.
33816         * gcov-dump.c (print_version): Ditto.
33817         * gcov.c (print_version): Ditto.
33818         * gcov-tool.c (print_version): Ditto.
33819         * gengtype.c (create_file): Ditto.
33820         * doc/cpp.texi: Bump @copying's copyright year.
33821         * doc/cppinternals.texi: Ditto.
33822         * doc/gcc.texi: Ditto.
33823         * doc/gccint.texi: Ditto.
33824         * doc/gcov.texi: Ditto.
33825         * doc/install.texi: Ditto.
33826         * doc/invoke.texi: Ditto.
33828 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33830         * vector-builder.h (vector_builder::m_full_nelts): Change from
33831         unsigned int to poly_uint64.
33832         (vector_builder::full_nelts): Update prototype accordingly.
33833         (vector_builder::new_vector): Likewise.
33834         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
33835         (vector_builder::operator ==): Likewise.
33836         (vector_builder::finalize): Likewise.
33837         * int-vector-builder.h (int_vector_builder::int_vector_builder):
33838         Take the number of elements as a poly_uint64 rather than an
33839         unsigned int.
33840         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
33841         from unsigned int to poly_uint64.
33842         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
33843         (vec_perm_indices::new_vector): Likewise.
33844         (vec_perm_indices::length): Likewise.
33845         (vec_perm_indices::nelts_per_input): Likewise.
33846         (vec_perm_indices::input_nelts): Likewise.
33847         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
33848         number of elements per input as a poly_uint64 rather than an
33849         unsigned int.  Use the original encoding for variable-length
33850         vectors, rather than clamping each individual element.
33851         For the second and subsequent elements in each pattern,
33852         clamp the step and base before clamping their sum.
33853         (vec_perm_indices::series_p): Handle polynomial element counts.
33854         (vec_perm_indices::all_in_range_p): Likewise.
33855         (vec_perm_indices_to_tree): Likewise.
33856         (vec_perm_indices_to_rtx): Likewise.
33857         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
33858         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
33859         (tree_vector_builder::new_binary_operation): Handle polynomial
33860         element counts.  Return false if we need to know the number
33861         of elements at compile time.
33862         * fold-const.c (fold_vec_perm): Punt if the number of elements
33863         isn't known at compile time.
33865 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33867         * vec-perm-indices.h (vec_perm_builder): Change element type
33868         from HOST_WIDE_INT to poly_int64.
33869         (vec_perm_indices::element_type): Update accordingly.
33870         (vec_perm_indices::clamp): Handle polynomial element_types.
33871         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
33872         (vec_perm_indices::all_in_range_p): Likewise.
33873         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
33874         than shwi trees.
33875         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
33876         polynomial vec_perm_indices element types.
33877         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
33878         * fold-const.c (fold_vec_perm): Likewise.
33879         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
33880         * tree-vect-generic.c (lower_vec_perm): Likewise.
33881         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
33882         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
33883         element type to HOST_WIDE_INT.
33885 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33886             Alan Hayward  <alan.hayward@arm.com>
33887             David Sherwood  <david.sherwood@arm.com>
33889         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
33890         rather than an int.  Use plus_constant.
33891         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
33892         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
33894 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33895             Alan Hayward  <alan.hayward@arm.com>
33896             David Sherwood  <david.sherwood@arm.com>
33898         * calls.c (emit_call_1, expand_call): Change struct_value_size from
33899         a HOST_WIDE_INT to a poly_int64.
33901 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33902             Alan Hayward  <alan.hayward@arm.com>
33903             David Sherwood  <david.sherwood@arm.com>
33905         * calls.c (load_register_parameters): Cope with polynomial
33906         mode sizes.  Require a constant size for BLKmode parameters
33907         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
33908         forces a parameter to be padded at the lsb end in order to
33909         fill a complete number of words, require the parameter size
33910         to be ordered wrt UNITS_PER_WORD.
33912 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33913             Alan Hayward  <alan.hayward@arm.com>
33914             David Sherwood  <david.sherwood@arm.com>
33916         * reload1.c (spill_stack_slot_width): Change element type
33917         from unsigned int to poly_uint64_pod.
33918         (alter_reg): Treat mode sizes as polynomial.
33920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33921             Alan Hayward  <alan.hayward@arm.com>
33922             David Sherwood  <david.sherwood@arm.com>
33924         * reload.c (complex_word_subreg_p): New function.
33925         (reload_inner_reg_of_subreg, push_reload): Use it.
33927 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33928             Alan Hayward  <alan.hayward@arm.com>
33929             David Sherwood  <david.sherwood@arm.com>
33931         * lra-constraints.c (process_alt_operands): Reject matched
33932         operands whose sizes aren't ordered.
33933         (match_reload): Refer to this check here.
33935 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33936             Alan Hayward  <alan.hayward@arm.com>
33937             David Sherwood  <david.sherwood@arm.com>
33939         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
33940         that the mode size is in the set {1, 2, 4, 8, 16}.
33942 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33943             Alan Hayward  <alan.hayward@arm.com>
33944             David Sherwood  <david.sherwood@arm.com>
33946         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
33947         Use plus_constant instead of gen_rtx_PLUS.
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         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
33954         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
33955         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
33956         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
33957         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
33958         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
33959         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
33960         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
33961         * config/i386/i386.c (ix86_push_rounding): ...this new function.
33962         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
33963         a poly_int64.
33964         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
33965         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
33966         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
33967         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
33968         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
33969         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
33970         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
33971         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
33972         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
33973         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
33974         function.
33975         * expr.c (emit_move_resolve_push): Treat the input and result
33976         of PUSH_ROUNDING as a poly_int64.
33977         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
33978         (emit_push_insn): Likewise.
33979         * lra-eliminations.c (mark_not_eliminable): Likewise.
33980         * recog.c (push_operand): Likewise.
33981         * reload1.c (elimination_effects): Likewise.
33982         * rtlanal.c (nonzero_bits1): Likewise.
33983         * calls.c (store_one_arg): Likewise.  Require the padding to be
33984         known at compile time.
33986 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33987             Alan Hayward  <alan.hayward@arm.com>
33988             David Sherwood  <david.sherwood@arm.com>
33990         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
33991         Use plus_constant instead of gen_rtx_PLUS.
33993 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33994             Alan Hayward  <alan.hayward@arm.com>
33995             David Sherwood  <david.sherwood@arm.com>
33997         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
33998         rather than an int.
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 (expand_expr_real_1): Use tree_to_poly_uint64
34005         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
34006         via stack temporaries.  Treat the mode size as polynomial too.
34008 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34009             Alan Hayward  <alan.hayward@arm.com>
34010             David Sherwood  <david.sherwood@arm.com>
34012         * expr.c (expand_expr_real_2): When handling conversions involving
34013         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
34014         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
34015         as a poly_uint64 too.
34017 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34018             Alan Hayward  <alan.hayward@arm.com>
34019             David Sherwood  <david.sherwood@arm.com>
34021         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
34023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34024             Alan Hayward  <alan.hayward@arm.com>
34025             David Sherwood  <david.sherwood@arm.com>
34027         * combine.c (can_change_dest_mode): Handle polynomial
34028         REGMODE_NATURAL_SIZE.
34029         * expmed.c (store_bit_field_1): Likewise.
34030         * expr.c (store_constructor): Likewise.
34031         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
34032         and polynomial REGMODE_NATURAL_SIZE.
34033         (gen_lowpart_common): Likewise.
34034         * reginfo.c (record_subregs_of_mode): Likewise.
34035         * rtlanal.c (read_modify_subreg_p): Likewise.
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         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
34042         numbers of elements.
34044 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34045             Alan Hayward  <alan.hayward@arm.com>
34046             David Sherwood  <david.sherwood@arm.com>
34048         * match.pd: Cope with polynomial numbers of vector elements.
34050 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34051             Alan Hayward  <alan.hayward@arm.com>
34052             David Sherwood  <david.sherwood@arm.com>
34054         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
34055         in a POINTER_PLUS_EXPR.
34057 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34058             Alan Hayward  <alan.hayward@arm.com>
34059             David Sherwood  <david.sherwood@arm.com>
34061         * omp-simd-clone.c (simd_clone_subparts): New function.
34062         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
34063         (ipa_simd_modify_function_body): Likewise.
34065 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34066             Alan Hayward  <alan.hayward@arm.com>
34067             David Sherwood  <david.sherwood@arm.com>
34069         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
34070         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
34071         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
34072         (expand_vector_condition, vector_element): Likewise.
34073         (subparts_gt): New function.
34074         (get_compute_type): Use subparts_gt.
34075         (count_type_subparts): Delete.
34076         (expand_vector_operations_1): Use subparts_gt instead of
34077         count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
34084         (vect_compile_time_alias): ...this new function.  Do the calculation
34085         on poly_ints rather than trees.
34086         (vect_prune_runtime_alias_test_list): Update call accordingly.
34088 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34089             Alan Hayward  <alan.hayward@arm.com>
34090             David Sherwood  <david.sherwood@arm.com>
34092         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
34093         numbers of units.
34094         (vect_schedule_slp_instance): Likewise.
34096 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34097             Alan Hayward  <alan.hayward@arm.com>
34098             David Sherwood  <david.sherwood@arm.com>
34100         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
34101         constant and extern definitions for variable-length vectors.
34102         (vect_get_constant_vectors): Note that the number of units
34103         is known to be constant.
34105 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34106             Alan Hayward  <alan.hayward@arm.com>
34107             David Sherwood  <david.sherwood@arm.com>
34109         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
34110         of units as polynomial.  Choose between WIDE and NARROW based
34111         on multiple_p.
34113 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34114             Alan Hayward  <alan.hayward@arm.com>
34115             David Sherwood  <david.sherwood@arm.com>
34117         * tree-vect-stmts.c (simd_clone_subparts): New function.
34118         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
34120 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34121             Alan Hayward  <alan.hayward@arm.com>
34122             David Sherwood  <david.sherwood@arm.com>
34124         * tree-vect-stmts.c (vectorizable_call): Treat the number of
34125         vectors as polynomial.  Use build_index_vector for
34126         IFN_GOMP_SIMD_LANE.
34128 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34129             Alan Hayward  <alan.hayward@arm.com>
34130             David Sherwood  <david.sherwood@arm.com>
34132         * tree-vect-stmts.c (get_load_store_type): Treat the number of
34133         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
34134         for variable-length vectors.
34135         (vectorizable_mask_load_store): Treat the number of units as
34136         polynomial, asserting that it is constant if the condition has
34137         already been enforced.
34138         (vectorizable_store, vectorizable_load): Likewise.
34140 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34141             Alan Hayward  <alan.hayward@arm.com>
34142             David Sherwood  <david.sherwood@arm.com>
34144         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
34145         of units as polynomial.  Punt if we can't tell at compile time
34146         which vector contains the final result.
34148 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34149             Alan Hayward  <alan.hayward@arm.com>
34150             David Sherwood  <david.sherwood@arm.com>
34152         * tree-vect-loop.c (vectorizable_induction): Treat the number
34153         of units as polynomial.  Punt on SLP inductions.  Use an integer
34154         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
34155         cast of such a series for variable-length floating-point
34156         reductions.
34158 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34159             Alan Hayward  <alan.hayward@arm.com>
34160             David Sherwood  <david.sherwood@arm.com>
34162         * tree.h (build_index_vector): Declare.
34163         * tree.c (build_index_vector): New function.
34164         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
34165         of units as polynomial, forcibly converting it to a constant if
34166         vectorizable_reduction has already enforced the condition.
34167         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
34168         to create a {1,2,3,...} vector.
34169         (vectorizable_reduction): Treat the number of units as polynomial.
34170         Choose vectype_in based on the largest scalar element size rather
34171         than the smallest number of units.  Enforce the restrictions
34172         relied on above.
34174 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34175             Alan Hayward  <alan.hayward@arm.com>
34176             David Sherwood  <david.sherwood@arm.com>
34178         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
34179         number of units as polynomial.
34181 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34182             Alan Hayward  <alan.hayward@arm.com>
34183             David Sherwood  <david.sherwood@arm.com>
34185         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
34186         * target.def (autovectorize_vector_sizes): Return the vector sizes
34187         by pointer, using vector_sizes rather than a bitmask.
34188         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
34189         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
34190         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
34191         Likewise.
34192         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
34193         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
34194         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
34195         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
34196         * omp-general.c (omp_max_vf): Likewise.
34197         * omp-low.c (omp_clause_aligned_alignment): Likewise.
34198         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
34199         * tree-vect-loop.c (vect_analyze_loop): Likewise.
34200         * tree-vect-slp.c (vect_slp_bb): Likewise.
34201         * doc/tm.texi: Regenerate.
34202         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
34203         to a poly_uint64.
34204         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
34205         the vector size as a poly_uint64 rather than an unsigned int.
34206         (current_vector_size): Change from an unsigned int to a poly_uint64.
34207         (get_vectype_for_scalar_type): Update accordingly.
34208         * tree.h (build_truth_vector_type): Take the size and number of
34209         units as a poly_uint64 rather than an unsigned int.
34210         (build_vector_type): Add a temporary overload that takes
34211         the number of units as a poly_uint64 rather than an unsigned int.
34212         * tree.c (make_vector_type): Likewise.
34213         (build_truth_vector_type): Take the number of units as a poly_uint64
34214         rather than an unsigned int.
34216 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34217             Alan Hayward  <alan.hayward@arm.com>
34218             David Sherwood  <david.sherwood@arm.com>
34220         * target.def (get_mask_mode): Take the number of units and length
34221         as poly_uint64s rather than unsigned ints.
34222         * targhooks.h (default_get_mask_mode): Update accordingly.
34223         * targhooks.c (default_get_mask_mode): Likewise.
34224         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
34225         * doc/tm.texi: Regenerate.
34227 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34228             Alan Hayward  <alan.hayward@arm.com>
34229             David Sherwood  <david.sherwood@arm.com>
34231         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
34232         * omp-general.c (omp_max_vf): Likewise.
34233         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
34234         (expand_omp_simd): Handle polynomial safelen.
34235         * omp-low.c (omplow_simd_context): Add a default constructor.
34236         (omplow_simd_context::max_vf): Change from int to poly_uint64.
34237         (lower_rec_simd_input_clauses): Update accordingly.
34238         (lower_rec_input_clauses): Likewise.
34240 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34241             Alan Hayward  <alan.hayward@arm.com>
34242             David Sherwood  <david.sherwood@arm.com>
34244         * tree-vectorizer.h (vect_nunits_for_cost): New function.
34245         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
34246         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
34247         (vect_analyze_slp_cost): Likewise.
34248         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
34249         (vect_model_load_cost): Likewise.
34251 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34252             Alan Hayward  <alan.hayward@arm.com>
34253             David Sherwood  <david.sherwood@arm.com>
34255         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
34256         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
34257         from an unsigned int * to a poly_uint64_pod *.
34258         (calculate_unrolling_factor): New function.
34259         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
34261 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34262             Alan Hayward  <alan.hayward@arm.com>
34263             David Sherwood  <david.sherwood@arm.com>
34265         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
34266         from an unsigned int to a poly_uint64.
34267         (_loop_vec_info::slp_unrolling_factor): Likewise.
34268         (_loop_vec_info::vectorization_factor): Change from an int
34269         to a poly_uint64.
34270         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
34271         (vect_get_num_vectors): New function.
34272         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
34273         (vect_get_num_copies): Use vect_get_num_vectors.
34274         (vect_analyze_data_ref_dependences): Change max_vf from an int *
34275         to an unsigned int *.
34276         (vect_analyze_data_refs): Change min_vf from an int * to a
34277         poly_uint64 *.
34278         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
34279         than an unsigned HOST_WIDE_INT.
34280         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
34281         (vect_analyze_data_ref_dependence): Change max_vf from an int *
34282         to an unsigned int *.
34283         (vect_analyze_data_ref_dependences): Likewise.
34284         (vect_compute_data_ref_alignment): Handle polynomial vf.
34285         (vect_enhance_data_refs_alignment): Likewise.
34286         (vect_prune_runtime_alias_test_list): Likewise.
34287         (vect_shift_permute_load_chain): Likewise.
34288         (vect_supportable_dr_alignment): Likewise.
34289         (dependence_distance_ge_vf): Take the vectorization factor as a
34290         poly_uint64 rather than an unsigned HOST_WIDE_INT.
34291         (vect_analyze_data_refs): Change min_vf from an int * to a
34292         poly_uint64 *.
34293         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
34294         vfm1 as a poly_uint64 rather than an int.  Make the same change
34295         for the returned bound_scalar.
34296         (vect_gen_vector_loop_niters): Handle polynomial vf.
34297         (vect_do_peeling): Likewise.  Update call to
34298         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
34299         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
34300         be constant.
34301         * tree-vect-loop.c (vect_determine_vectorization_factor)
34302         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
34303         (vect_get_known_peeling_cost): Likewise.
34304         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
34305         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
34306         (vect_transform_loop): Likewise.  Use the lowest possible VF when
34307         updating the upper bounds of the loop.
34308         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
34309         rather than an int.
34310         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
34311         polynomial unroll factors.
34312         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
34313         (vect_make_slp_decision): Likewise.
34314         (vect_supported_load_permutation_p): Likewise, and polynomial
34315         vf too.
34316         (vect_analyze_slp_cost): Handle polynomial vf.
34317         (vect_slp_analyze_node_operations): Likewise.
34318         (vect_slp_analyze_bb_1): Likewise.
34319         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
34320         than an unsigned HOST_WIDE_INT.
34321         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
34322         (vectorizable_load): Handle polynomial vf.
34323         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
34324         a poly_uint64.
34325         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
34327 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34328             Alan Hayward  <alan.hayward@arm.com>
34329             David Sherwood  <david.sherwood@arm.com>
34331         * match.pd: Handle bit operations involving three constants
34332         and try to fold one pair.
34334 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
34336         * tree-vect-loop-manip.c: Include gimple-fold.h.
34337         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
34338         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
34339         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
34340         Add a path that uses a step of VF instead of 1, but disable it
34341         for now.
34342         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
34343         and niters_no_overflow parameters.  Update calls to
34344         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
34345         Create a new SSA name if the latter choses to use a ste other
34346         than zero, and return it via niters_vector_mult_vf_var.
34347         * tree-vect-loop.c (vect_transform_loop): Update calls to
34348         vect_do_peeling, vect_gen_vector_loop_niters and
34349         slpeel_make_loop_iterate_ntimes.
34350         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
34351         (vect_gen_vector_loop_niters): Update declarations after above changes.
34353 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
34355         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
34356         128-bit round to integer instructions.
34357         (ceil<mode>2): Likewise.
34358         (btrunc<mode>2): Likewise.
34359         (round<mode>2): Likewise.
34361 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34363         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
34364         unaligned VSX load/store on P8/P9.
34365         (expand_block_clear): Allow the use of unaligned VSX
34366         load/store on P8/P9.
34368 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34370         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
34371         New function.
34372         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
34373         swap associated with both a load and a store.
34375 2018-01-02  Andrew Waterman  <andrew@sifive.com>
34377         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
34378         * config/riscv/riscv.md (clear_cache): Use it.
34380 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
34382         * web.c: Remove out-of-date comment.
34384 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34386         * expr.c (fixup_args_size_notes): Check that any existing
34387         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
34388         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
34389         (emit_single_push_insn): ...here.
34391 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34393         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
34394         (const_vector_encoded_nelts): New function.
34395         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
34396         (const_vector_int_elt, const_vector_elt): Declare.
34397         * emit-rtl.c (const_vector_int_elt_1): New function.
34398         (const_vector_elt): Likewise.
34399         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
34400         of CONST_VECTOR_ELT.
34402 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34404         * expr.c: Include rtx-vector-builder.h.
34405         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
34406         directly on the tree encoding.
34407         (const_vector_from_tree): Likewise.
34408         * optabs.c: Include rtx-vector-builder.h.
34409         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
34410         sequence of "u" values.
34411         * vec-perm-indices.c: Include rtx-vector-builder.h.
34412         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
34413         directly on the vec_perm_indices encoding.
34415 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34417         * doc/rtl.texi (const_vector): Describe new encoding scheme.
34418         * Makefile.in (OBJS): Add rtx-vector-builder.o.
34419         * rtx-vector-builder.h: New file.
34420         * rtx-vector-builder.c: Likewise.
34421         * rtl.h (rtx_def::u2): Add a const_vector field.
34422         (CONST_VECTOR_NPATTERNS): New macro.
34423         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
34424         (CONST_VECTOR_DUPLICATE_P): Likewise.
34425         (CONST_VECTOR_STEPPED_P): Likewise.
34426         (CONST_VECTOR_ENCODED_ELT): Likewise.
34427         (const_vec_duplicate_p): Check for a duplicated vector encoding.
34428         (unwrap_const_vec_duplicate): Likewise.
34429         (const_vec_series_p): Check for a non-duplicated vector encoding.
34430         Say that the function only returns true for integer vectors.
34431         * emit-rtl.c: Include rtx-vector-builder.h.
34432         (gen_const_vec_duplicate_1): Delete.
34433         (gen_const_vector): Call gen_const_vec_duplicate instead of
34434         gen_const_vec_duplicate_1.
34435         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
34436         (gen_const_vec_duplicate): Use rtx_vector_builder.
34437         (gen_const_vec_series): Likewise.
34438         (gen_rtx_CONST_VECTOR): Likewise.
34439         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
34440         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
34441         Build a new vector rather than modifying a CONST_VECTOR in-place.
34442         (handle_special_swappables): Update call accordingly.
34443         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
34444         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
34445         Build a new vector rather than modifying a CONST_VECTOR in-place.
34446         (handle_special_swappables): Update call accordingly.
34448 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34450         * simplify-rtx.c (simplify_const_binary_operation): Use
34451         CONST_VECTOR_ELT instead of XVECEXP.
34453 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34455         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
34456         the selector elements to be different from the data elements
34457         if the selector is a VECTOR_CST.
34458         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
34459         ssizetype for the selector.
34461 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34463         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
34464         before testing each element individually.
34465         * tree-vect-generic.c (lower_vec_perm): Likewise.
34467 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34469         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
34470         * selftest-run-tests.c (selftest::run_tests): Call it.
34471         * vector-builder.h (vector_builder::operator ==): New function.
34472         (vector_builder::operator !=): Likewise.
34473         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
34474         (vec_perm_indices::all_from_input_p): New function.
34475         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
34476         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
34477         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
34478         instead of reading the VECTOR_CST directly.  Detect whether both
34479         vector inputs are the same before constructing the vec_perm_indices,
34480         and update the number of inputs argument accordingly.  Use the
34481         utility functions added above.  Only construct sel2 if we need to.
34483 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34485         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
34486         the broadcast of the low byte.
34487         (expand_mult_highpart): Use an explicit encoding for the permutes.
34488         * optabs-query.c (can_mult_highpart_p): Likewise.
34489         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
34490         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34491         (vectorizable_bswap): Likewise.
34492         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
34493         explicit encoding for the power-of-2 permutes.
34494         (vect_permute_store_chain): Likewise.
34495         (vect_grouped_load_supported): Likewise.
34496         (vect_permute_load_chain): Likewise.
34498 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34500         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
34501         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
34502         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
34503         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
34504         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
34505         (vect_gen_perm_mask_any): Likewise.
34507 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34509         * int-vector-builder.h: New file.
34510         * vec-perm-indices.h: Include int-vector-builder.h.
34511         (vec_perm_indices): Redefine as an int_vector_builder.
34512         (auto_vec_perm_indices): Delete.
34513         (vec_perm_builder): Redefine as a stand-alone class.
34514         (vec_perm_indices::vec_perm_indices): New function.
34515         (vec_perm_indices::clamp): Likewise.
34516         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
34517         (vec_perm_indices::new_vector): New function.
34518         (vec_perm_indices::new_expanded_vector): Update for new
34519         vec_perm_indices class.
34520         (vec_perm_indices::rotate_inputs): New function.
34521         (vec_perm_indices::all_in_range_p): Operate directly on the
34522         encoded form, without computing elided elements.
34523         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
34524         encoding.  Update for new vec_perm_indices class.
34525         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
34526         the given vec_perm_builder.
34527         (expand_vec_perm_var): Update vec_perm_builder constructor.
34528         (expand_mult_highpart): Use vec_perm_builder instead of
34529         auto_vec_perm_indices.
34530         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
34531         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
34532         or double series encoding as appropriate.
34533         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
34534         vec_perm_indices instead of auto_vec_perm_indices.
34535         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34536         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34537         (vect_permute_store_chain): Likewise.
34538         (vect_grouped_load_supported): Likewise.
34539         (vect_permute_load_chain): Likewise.
34540         (vect_shift_permute_load_chain): Likewise.
34541         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
34542         (vect_transform_slp_perm_load): Likewise.
34543         (vect_schedule_slp_instance): Likewise.
34544         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34545         (vectorizable_mask_load_store): Likewise.
34546         (vectorizable_bswap): Likewise.
34547         (vectorizable_store): Likewise.
34548         (vectorizable_load): Likewise.
34549         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
34550         vec_perm_indices instead of auto_vec_perm_indices.  Use
34551         tree_to_vec_perm_builder to read the vector from a tree.
34552         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
34553         vec_perm_builder instead of a vec_perm_indices.
34554         (have_whole_vector_shift): Use vec_perm_builder and
34555         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
34556         truncation to calc_vec_perm_mask_for_shift.
34557         (vect_create_epilog_for_reduction): Likewise.
34558         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
34559         from auto_vec_perm_indices to vec_perm_indices.
34560         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
34561         instead of changing individual elements.
34562         (aarch64_vectorize_vec_perm_const): Use new_vector to install
34563         the vector in d.perm.
34564         * config/arm/arm.c (expand_vec_perm_d::perm): Change
34565         from auto_vec_perm_indices to vec_perm_indices.
34566         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
34567         instead of changing individual elements.
34568         (arm_vectorize_vec_perm_const): Use new_vector to install
34569         the vector in d.perm.
34570         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
34571         Update vec_perm_builder constructor.
34572         (rs6000_expand_interleave): Likewise.
34573         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
34574         (rs6000_expand_interleave): Likewise.
34576 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34578         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
34579         to qimode could truncate the indices.
34580         * optabs.c (expand_vec_perm_var): Likewise.
34582 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34584         * Makefile.in (OBJS): Add vec-perm-indices.o.
34585         * vec-perm-indices.h: New file.
34586         * vec-perm-indices.c: Likewise.
34587         * target.h (vec_perm_indices): Replace with a forward class
34588         declaration.
34589         (auto_vec_perm_indices): Move to vec-perm-indices.h.
34590         * optabs.h: Include vec-perm-indices.h.
34591         (expand_vec_perm): Delete.
34592         (selector_fits_mode_p, expand_vec_perm_var): Declare.
34593         (expand_vec_perm_const): Declare.
34594         * target.def (vec_perm_const_ok): Replace with...
34595         (vec_perm_const): ...this new hook.
34596         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
34597         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
34598         * doc/tm.texi: Regenerate.
34599         * optabs.def (vec_perm_const): Delete.
34600         * doc/md.texi (vec_perm_const): Likewise.
34601         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
34602         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
34603         expand_vec_perm for constant permutation vectors.  Assert that
34604         the mode of variable permutation vectors is the integer equivalent
34605         of the mode that is being permuted.
34606         * optabs-query.h (selector_fits_mode_p): Declare.
34607         * optabs-query.c: Include vec-perm-indices.h.
34608         (selector_fits_mode_p): New function.
34609         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
34610         is defined, instead of checking whether the vec_perm_const_optab
34611         exists.  Use targetm.vectorize.vec_perm_const instead of
34612         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
34613         fit in the vector mode before using a variable permute.
34614         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
34615         vec_perm_indices instead of an rtx.
34616         (expand_vec_perm): Replace with...
34617         (expand_vec_perm_const): ...this new function.  Take the selector
34618         as a vec_perm_indices rather than an rtx.  Also take the mode of
34619         the selector.  Update call to shift_amt_for_vec_perm_mask.
34620         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
34621         Use vec_perm_indices::new_expanded_vector to expand the original
34622         selector into bytes.  Check whether the indices fit in the vector
34623         mode before using a variable permute.
34624         (expand_vec_perm_var): Make global.
34625         (expand_mult_highpart): Use expand_vec_perm_const.
34626         * fold-const.c: Includes vec-perm-indices.h.
34627         * tree-ssa-forwprop.c: Likewise.
34628         * tree-vect-data-refs.c: Likewise.
34629         * tree-vect-generic.c: Likewise.
34630         * tree-vect-loop.c: Likewise.
34631         * tree-vect-slp.c: Likewise.
34632         * tree-vect-stmts.c: Likewise.
34633         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
34634         Delete.
34635         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
34636         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
34637         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
34638         (aarch64_vectorize_vec_perm_const): ...this new function.
34639         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
34640         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34641         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
34642         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
34643         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
34644         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34645         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
34646         into...
34647         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
34648         check for NEON modes.
34649         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
34650         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
34651         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
34652         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
34653         into...
34654         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
34655         the old VEC_PERM_CONST conditions.
34656         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
34657         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
34658         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
34659         (ia64_vectorize_vec_perm_const_ok): Merge into...
34660         (ia64_vectorize_vec_perm_const): ...this new function.
34661         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
34662         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
34663         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
34664         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
34665         * config/mips/mips.c (mips_expand_vec_perm_const)
34666         (mips_vectorize_vec_perm_const_ok): Merge into...
34667         (mips_vectorize_vec_perm_const): ...this new function.
34668         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
34669         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
34670         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
34671         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
34672         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
34673         (rs6000_expand_vec_perm_const): Delete.
34674         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
34675         Delete.
34676         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34677         (altivec_expand_vec_perm_const_le): Take each operand individually.
34678         Operate on constant selectors rather than rtxes.
34679         (altivec_expand_vec_perm_const): Likewise.  Update call to
34680         altivec_expand_vec_perm_const_le.
34681         (rs6000_expand_vec_perm_const): Delete.
34682         (rs6000_vectorize_vec_perm_const_ok): Delete.
34683         (rs6000_vectorize_vec_perm_const): New function.
34684         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
34685         an element count and rtx array.
34686         (rs6000_expand_extract_even): Update call accordingly.
34687         (rs6000_expand_interleave): Likewise.
34688         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
34689         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
34690         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
34691         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
34692         (rs6000_expand_vec_perm_const): Delete.
34693         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
34694         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34695         (altivec_expand_vec_perm_const_le): Take each operand individually.
34696         Operate on constant selectors rather than rtxes.
34697         (altivec_expand_vec_perm_const): Likewise.  Update call to
34698         altivec_expand_vec_perm_const_le.
34699         (rs6000_expand_vec_perm_const): Delete.
34700         (rs6000_vectorize_vec_perm_const_ok): Delete.
34701         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
34702         reference to the SPE evmerge intructions.
34703         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
34704         an element count and rtx array.
34705         (rs6000_expand_extract_even): Update call accordingly.
34706         (rs6000_expand_interleave): Likewise.
34707         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
34708         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
34709         new function.
34710         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
34712 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34714         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
34715         vector mode and that that mode matches the mode of the data
34716         being permuted.
34717         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
34718         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
34719         directly using expand_vec_perm_1 when forcing selectors into
34720         registers.
34721         (expand_vec_perm_var): New function, split out from expand_vec_perm.
34723 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34725         * optabs-query.h (can_vec_perm_p): Delete.
34726         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
34727         * optabs-query.c (can_vec_perm_p): Split into...
34728         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
34729         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
34730         particular selector is valid.
34731         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34732         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34733         (vect_grouped_load_supported): Likewise.
34734         (vect_shift_permute_load_chain): Likewise.
34735         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
34736         (vect_transform_slp_perm_load): Likewise.
34737         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34738         (vectorizable_bswap): Likewise.
34739         (vect_gen_perm_mask_checked): Likewise.
34740         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
34741         implementations of variable permutation vectors into account
34742         when deciding which selector to use.
34743         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
34744         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
34745         with a false third argument.
34746         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
34747         to test whether the constant selector is valid and can_vec_perm_var_p
34748         to test whether a variable selector is valid.
34750 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34752         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
34753         * optabs-query.c (can_vec_perm_p): Likewise.
34754         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
34755         instead of vec_perm_indices.
34756         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
34757         (vect_gen_perm_mask_checked): Likewise,
34758         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
34759         (vect_gen_perm_mask_checked): Likewise,
34761 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
34763         * optabs-query.h (qimode_for_vec_perm): Declare.
34764         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
34765         (qimode_for_vec_perm): ...this new function.
34766         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
34768 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34770         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
34771         does not have a conditional at the top.
34773 2018-01-02  Richard Biener  <rguenther@suse.de>
34775         * ipa-inline.c (big_speedup_p): Fix expression.
34777 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
34779         PR target/81616
34780         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
34781         for generic 4->6.
34783 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
34785         PR target/81616
34786         Generic tuning.
34787         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
34788         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
34789         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
34790         cond_taken_branch_cost 3->4.
34792 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
34794         PR tree-optimization/83581
34795         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
34796         TODO_cleanup_cfg if any changes have been made.
34798         PR middle-end/83608
34799         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
34800         convert_modes if target mode has the right side, but different mode
34801         class.
34803         PR middle-end/83609
34804         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
34805         last argument when extracting from CONCAT.  If either from_real or
34806         from_imag is NULL, use expansion through memory.  If result is not
34807         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
34808         the parts directly to inner mode, if even that fails, use expansion
34809         through memory.
34811         PR middle-end/83623
34812         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
34813         check for bswap in mode rather than HImode and use that in expand_unop
34814         too.
34816 Copyright (C) 2018 Free Software Foundation, Inc.
34818 Copying and distribution of this file, with or without modification,
34819 are permitted in any medium without royalty provided the copyright
34820 notice and this notice are preserved.