2018-10-25 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob00ba522e43942e2614ae440807430815f402814d
1 2018-10-25  Richard Biener  <rguenther@suse.de>
3         * tree-if-conv.c: Include tree-ssa-sccvn.h.
4         (tree_if_conversion): Run CSE on the if-converted loop body.
6 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
8         PR bootstrap/87747
9         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
10         (RTX_CODE_HWINT_P): New macro.
11         (rtx_code_size): Use RTX_CODE_HWINT_P ().
13 2018-10-25  Jan Hubicka  <jh@suse.cz>
15         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
16         is anonymous.
18 2018-10-25  Richard Biener  <rguenther@suse.de>
20         PR tree-optimization/87665
21         PR tree-optimization/87745
22         * tree-vectorizer.h (get_earlier_stmt): Remove.
23         (get_later_stmt): Pick up UID from the original non-pattern stmt.
25 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
27         * options.texi (Deprecated): Move list to Var section.
29 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
31         PR rtl-optimization/87720
32         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
34 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
36         * gimple-ssa-isolate-paths.c
37         (find_implicit_erroneous_behavior): Do not change code if the
38         pass is running for warnings only.
39         (find_explicit_erroneous_behavior): Likewise.
41 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
43         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
44         Define as rs6000_mangle_decl_assembler_name.
45         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
46         long double to IEEE long double, switch the names of the long
47         double built-in functions to be <func>f128 instead of <func>l.
49 2018-10-24  Martin Sebor  <msebor@redhat.com>
51         * doc/extend.texi (nonnull): List no-argument form.  Reference
52         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
54 2018-10-24  Richard Biener  <rguenther@suse.de>
56         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
58 2018-10-24  Martin Liska  <mliska@suse.cz>
60         PR tree-optimization/84436
61         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
62         Remove.
63         (switch_conversion::contains_linear_function_p): New.
64         (switch_conversion::build_one_array): Support linear
65         transformation on input.
66         * tree-switch-conversion.h (struct switch_conversion): Add
67         contains_linear_function_p declaration.
69 2018-10-24  Richard Biener  <rguenther@suse.de>
71         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
72         if its argument is CONSTANT_CLASS_P.
74 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
76         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
77         it is wrong for forward declarations.
79 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
81         * config/s390/s390.c (s390_check_qrst_address): Add the missing
82         SYMBOL_REF_P () check.
84 2018-10-24  Richard Biener  <rguenther@suse.de>
86         PR tree-optimization/87105
87         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
88         dump classification.
89         (vect_analyze_data_ref_accesses): Handle duplicate loads and
90         stores by splitting the affected group after the fact.
91         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
92         fail the SLP build because of size constraints.
94 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
96         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
97         * configure: Regenerate.
98         * config.in: Regenerate.
99         * varasm.c (mergeable_string_section): Use readonly_data_section
100         if linker doesn't support SHF_MERGE with alignment > 8.
101         (mergeable_constant_section): Likewise.
103 2018-10-24  Richard Biener  <rguenther@suse.de>
105         PR tree-optimization/84013
106         * tree-ssa-structalias.c (struct msdi_data): New struct for
107         marshalling data to walk_stmt_load_store_ops.
108         (maybe_set_dependence_info): Refactor as callback for
109         walk_stmt_load_store_ops.
110         (compute_dependence_clique): Set restrict info on all stmt kinds.
112 2018-10-24  Martin Liska  <mliska@suse.cz>
114         * cgraph.c (cgraph_node::dump):
115         Remove reduntant dumps and make tp_first_run dump more compact.
117 2018-10-24  Richard Biener  <rguenther@suse.de>
119         PR tree-optimization/87665
120         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
121         to reflect reality.
123 2018-10-12  Jeff Law  <law@redhat.com>
125         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
126         for H8/S.
128 2018-10-23  Richard Biener  <rguenther@suse.de>
130         * tree-vrp.c (add_assert_info): Guard dump_printf with
131         dump_enabled_p.
132         * gimple-ssa-evrp-analyze.c
133         (evrp_range_analyzer::record_ranges_from_incoming_edge):
134         Use value_range::ignore_equivs_equal_p.
136 2018-10-23  Richard Biener  <rguenther@suse.de>
138         PR tree-optimization/87105
139         PR tree-optimization/87608
140         * passes.def (pass_all_early_optimizations): Add early phi-opt
141         after dce.
142         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
143         addition to debug stmts.
144         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
145         and abs replacement early.
146         * tree-cfg.c (gimple_empty_block_p): Likewise.
148 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
150         PR target/86383
151         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
152         specified to configure.
153         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
155 2018-10-23  Richard Biener  <rguenther@suse.de>
157         PR tree-optimization/87700
158         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
160 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
162         PR target/87674
163         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
164         second argument from __mmask16 to __mmask8.
165         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
166         _mm_mask_packs_epi32): Likewise.
167         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
168         Likewise.
169         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
171 2018-10-23  Richard Biener  <rguenther@suse.de>
173         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
175 2018-10-23  Richard Biener  <rguenther@suse.de>
177         PR tree-optimization/86144
178         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
179         over simd attribute.
181 2018-10-23  Richard Biener  <rguenther@suse.de>
183         PR tree-optimization/87693
184         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
185         the case we do not find the taken edge.
187 2018-10-22  Paul Koning  <ni1d@arrl.net>
189         * symtab.c (symtab_node::increase_alignment): Correct max
190         alignment check.
192 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
194         PR tree-optimization/87633
195         * match.pd: Do not generate unordered integer comparisons.
197 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
199         PR rtl-optimization/87600
200         * combine.c: Add include of expr.h.
201         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
202         register to a pseudo.
203         (make_more_copies): New function, add a copy to a new pseudo after
204         the moves from hard registers into pseudos.
205         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
206         later.  Call make_more_copies.
208 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
210         * lra-constraints.c (process_alt_operands): New local array,
211         matching_early_clobber.  Check matching_early_clobber before
212         decrementing reject, and set matching_early_clobber after.
214 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
216         PR target/87598
217         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
218         call output_addr_const and hope for the best.
220 2018-10-22  Richard Biener  <rguenther@suse.de>
222         * gimple-ssa-evrp-analyze.c
223         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
224         smarter about what ranges to use.
225         * tree-vrp.c (add_assert_info): Dump here.
226         (register_edge_assert_for_2): Instead of here at multiple but
227         not all places.
229         * gcc.dg/tree-ssa/evrp12.c: New testcase.
230         * gcc.dg/predict-6.c: Adjust.
231         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
232         * gcc.dg/tree-ssa/vrp02.c: Likewise.
233         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
235 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
236         Richard Biener  <rguenther@suse.de>
238         * bitmap.h: Update data structure documentation, including a
239         description of bitmap views as either linked-lists or splay trees.
240         (struct bitmap_element_def): Update comments for splay tree bitmaps.
241         (struct bitmap_head_def): Likewise.
242         (bitmap_list_view, bitmap_tree_view): New prototypes.
243         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
244         tree_form fields.
245         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
246         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
247         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
248         released bitmap element here.
249         (bitmap_element_free): Remove.
250         (bitmap_elt_clear_from): Work on splay tree bitmaps.
251         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
252         this function similar ones such that linked-list bitmap implementation
253         functions are grouped.
254         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
255         and moved for grouping.
256         (bitmap_list_insert_element_after): Renamed from
257         bitmap_elt_insert_after, and moved for grouping.
258         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
259         (bitmap_tree_link_left, bitmap_tree_link_right,
260         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
261         bitmap_tree_link_element, bitmap_tree_unlink_element,
262         bitmap_tree_find_element): New functions for splay-tree bitmap
263         implementation.
264         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
265         Renamed and moved, see above entries.
266         (bitmap_tree_listify_from): New function to convert part of a splay
267         tree bitmap to a linked-list bitmap.
268         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
269         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
270         (bitmap_find_bit): Remove.
271         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
272         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
273         Handle splay tree bitmaps.
274         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
275         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
276         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
277         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
278         bitmap_intersect_compl_p, bitmap_ior_and_compl,
279         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
280         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
281         corresponding changes to use linked-list specific bitmap_element
282         manipulation functions as applicable for efficiency.
283         (bitmap_tree_to_vec): New function.
284         (debug_bitmap_elt_file): New function split out from ...
285         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
286         (bitmap_print): Likewise.
288         PR tree-optimization/63155
289         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
290         SSA edge worklists.
291         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
292         in tree-view.
294 2018-10-22  Martin Liska  <mliska@suse.cz>
296         PR tree-optimization/87686
297         Revert
298         2018-08-29  Martin Liska  <mliska@suse.cz>
300         * tree-switch-conversion.c (switch_conversion::expand):
301         Strenghten assumption about gswitch statements.
303 2018-10-22  Martin Liska  <mliska@suse.cz>
305         * ipa-icf.c (sem_item::compare_attributes): Remove.
306         (sem_item::compare_referenced_symbol_properties): Use
307         attribute_list_equal instead.
308         (sem_function::equals_wpa): Likewise.
309         * ipa-icf.h: Remove compare_attributes.
311 2018-10-22  Richard Biener  <rguenther@suse.de>
313         PR middle-end/87682
314         * mem-stats.h (mem_usage::operator==): Fix pasto.
316 2018-10-22  Richard Biener  <rguenther@suse.de>
318         PR tree-optimization/87640
319         * tree-vrp.c (set_value_range_with_overflow): Decompose
320         incomplete result.
321         (extract_range_from_binary_expr_1): Adjust.
323 2018-10-22  Martin Jambor  <mjambor@suse.cz>
325         * tree-eh.h (stmt_could_throw_p): Add function parameter.
326         (stmt_can_throw_external): Likewise.
327         (stmt_can_throw_internal): Likewise.
328         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
329         (lower_eh_constructs_2): Likewise.
330         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
331         (stmt_can_throw_external): Likewise.
332         (stmt_can_throw_internal): Likewise.
333         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
334         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
335         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
336         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
337         (pass_lower_eh_dispatch::execute): Pass cfun to
338         stmt_can_throw_external.
339         (cleanup_empty_eh): Likewise.
340         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
341         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
342         stmt_can_throw_external instead of pushing it to cfun.
343         (symbol_table::create_edge): Likewise.
344         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
345         stmt_can_throw_internal.
346         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
347         to stmt_could_throw_p.
348         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
349         stmt_can_throw_internal.
350         (pass_store_merging::execute): Likewise.
351         * gimple-ssa-strength-reduction.c
352         (find_candidates_dom_walker::before_dom_children): Pass cfun to
353         stmt_could_throw_p.
354         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
355         stmt_can_throw_internal.
356         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
357         to stmt_can_throw_external.
358         (check_stmt): Pass cfun to stmt_could_throw_p.
359         (check_stmt): Pass cfun to stmt_can_throw_external.
360         (pass_nothrow::execute): Likewise.
361         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
362         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
363         stmt_can_throw_internal.
364         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
365         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
366         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
367         * tree-complex.c (expand_complex_libcall): Pass cfun to
368         stmt_could_throw_p and to stmt_can_throw_internal.
369         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
370         * tree-inline.c (copy_edges_for_bb): Likewise.
371         (maybe_move_debug_stmts_to_successors): Likewise.
372         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
373         stmt_could_throw_p.
374         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
375         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
376         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
377         stmt_can_throw_internal.
378         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
379         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
380         stmt_could_throw_p.
381         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
382         stmt_can_throw_internal.
383         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
384         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
385         stmt_could_throw_p.
386         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
387         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
388         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
389         (convert_mult_to_fma_1): Likewise.
390         (convert_to_divmod): Likewise.
391         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
392         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
393         * tree-ssa-propagate.c
394         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
395         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
396         (maybe_optimize_range_tests): Likewise.
397         (linearize_expr_tree): Likewise.
398         (reassociate_bb): Likewise.
399         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
400         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
401         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
402         (handle_char_store): Likewise.
403         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
404         stmt_can_throw_internal.
405         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
406         stmt_could_throw_p.
407         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
408         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
409         (vectorizable_simd_clone_call): Likewise.
410         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
411         (gimple_stringop_fixed_value): Likewise.
413 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
415         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
416         literal pool references.
417         (s390_check_qrst_address): Adapt to the new behavior of
418         s390_loadrelative_operand_p ().
420 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
422         PR target/72782
423         * config/i386/sse.md (*andnot<mode>3_bcst): New.
425 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
427         PR target/72782
428         * config/i386/sse.md (*<code><mode>3_bcst): New.
430 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
432         PR target/72782
433         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
434         V4DI, V16SI and V8DI.
435         (*sub<mode>3<mask_name>_bcst): New.
436         (*add<mode>3<mask_name>_bcst): Likewise.
438 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
440         PR target/72782
441         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
442         __builtin_ia32_vfnmsubpd512_mask.
443         (_mm512_mask_fnmsub_round_pd): Likewise.
444         (_mm512_fnmsub_pd): Likewise.
445         (_mm512_mask_fnmsub_pd): Likewise.
446         (_mm512_maskz_fnmsub_round_pd): Use
447         __builtin_ia32_vfnmsubpd512_maskz.
448         (_mm512_maskz_fnmsub_pd): Likewise.
449         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
450         (_mm512_mask_fnmsub_round_ps): Likewise.
451         (_mm512_fnmsub_ps): Likewise.
452         (_mm512_mask_fnmsub_ps): Likewise.
453         (_mm512_maskz_fnmsub_round_ps): Use
454         __builtin_ia32_vfnmsubps512_maskz.
455         (_mm512_maskz_fnmsub_ps): Likewise.
456         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
457         __builtin_ia32_vfnmsubpd256_mask.
458         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
459         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
460         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
461         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
462         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
463         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
464         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
465         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
466         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
467         __builtin_ia32_vfnmsubpd.
468         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
469         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
470         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
471         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
472         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
473         * config/i386/i386-builtin.def: Add
474         __builtin_ia32_vfnmsubpd256_mask,
475         __builtin_ia32_vfnmsubpd256_maskz,
476         __builtin_ia32_vfnmsubpd128_mask,
477         __builtin_ia32_vfnmsubpd128_maskz,
478         __builtin_ia32_vfnmsubps256_mask,
479         __builtin_ia32_vfnmsubps256_maskz,
480         __builtin_ia32_vfnmsubps128_mask,
481         __builtin_ia32_vfnmsubps128_maskz,
482         __builtin_ia32_vfnmsubpd512_mask,
483         __builtin_ia32_vfnmsubpd512_maskz,
484         __builtin_ia32_vfnmsubps512_mask,
485         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
486         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
487         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
488         __builtin_ia32_vfnmsubpd256.
489         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
490         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
491         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
492         Likewise.
493         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
494         Likewise.
495         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
496         Likewise.
497         (fmai_vmfnmsub_<mode><round_name>): Likewise.
499 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
501         PR target/72782
502         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
503         __builtin_ia32_vfnmaddpd512_mask.
504         (_mm512_mask_fnmadd_round_pd): Likewise.
505         (_mm512_fnmadd_pd): Likewise.
506         (_mm512_mask_fnmadd_pd): Likewise.
507         (_mm512_maskz_fnmadd_round_pd): Use
508         __builtin_ia32_vfnmaddpd512_maskz.
509         (_mm512_maskz_fnmadd_pd): Likewise.
510         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
511         (_mm512_mask_fnmadd_round_ps): Likewise.
512         (_mm512_fnmadd_ps): Likewise.
513         (_mm512_mask_fnmadd_ps): Likewise.
514         (_mm512_maskz_fnmadd_round_ps): Use
515         __builtin_ia32_vfnmaddps512_maskz.
516         (_mm512_maskz_fnmadd_ps): Likewise.
517         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
518         __builtin_ia32_vfnmaddpd256_mask.
519         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
520         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
521         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
522         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
523         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
524         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
525         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
526         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
527         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
528         __builtin_ia32_vfnmaddpd.
529         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
530         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
531         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
532         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
533         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
534         * config/i386/i386-builtin.def: Add
535         __builtin_ia32_vfnmaddpd256_mask,
536         __builtin_ia32_vfnmaddpd256_maskz,
537         __builtin_ia32_vfnmaddpd128_mask,
538         __builtin_ia32_vfnmaddpd128_maskz,
539         __builtin_ia32_vfnmaddps256_mask,
540         __builtin_ia32_vfnmaddps256_maskz,
541         __builtin_ia32_vfnmaddps128_mask,
542         __builtin_ia32_vfnmaddps128_maskz,
543         __builtin_ia32_vfnmaddpd512_mask,
544         __builtin_ia32_vfnmaddpd512_maskz,
545         __builtin_ia32_vfnmaddps512_mask,
546         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
547         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
548         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
549         __builtin_ia32_vfnmaddpd256.
550         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
551         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
552         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
553         Likewise.
554         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
555         Likewise.
556         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
557         Likewise.
558         (fmai_vmfnmadd_<mode><round_name>): Likewise.
560 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
562         PR target/72782
563         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
564         __builtin_ia32_vfmsubpd512_mask.
565         (_mm512_mask_fmsub_round_pd): Likewise.
566         (_mm512_fmsub_pd): Likewise.
567         (_mm512_mask_fmsub_pd): Likewise.
568         (_mm512_maskz_fmsub_round_pd): Use
569         __builtin_ia32_vfmsubpd512_maskz.
570         (_mm512_maskz_fmsub_pd): Likewise.
571         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
572         (_mm512_mask_fmsub_round_ps): Likewise.
573         (_mm512_fmsub_ps): Likewise.
574         (_mm512_mask_fmsub_ps): Likewise.
575         (_mm512_maskz_fmsub_round_ps): Use
576         __builtin_ia32_vfmsubps512_maskz.
577         (_mm512_maskz_fmsub_ps): Likewise.
578         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
579         __builtin_ia32_vfmsubpd256_mask.
580         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
581         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
582         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
583         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
584         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
585         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
586         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
587         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
588         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
589         __builtin_ia32_vfmsubpd.
590         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
591         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
592         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
593         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
594         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
595         * config/i386/i386-builtin.def: Add
596         __builtin_ia32_vfmsubpd256_mask,
597         __builtin_ia32_vfmsubpd256_maskz,
598         __builtin_ia32_vfmsubpd128_mask,
599         __builtin_ia32_vfmsubpd128_maskz,
600         __builtin_ia32_vfmsubps256_mask,
601         __builtin_ia32_vfmsubps256_maskz,
602         __builtin_ia32_vfmsubps128_mask,
603         __builtin_ia32_vfmsubps128_maskz,
604         __builtin_ia32_vfmsubpd512_mask,
605         __builtin_ia32_vfmsubpd512_maskz,
606         __builtin_ia32_vfmsubps512_mask,
607         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
608         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
609         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
610         __builtin_ia32_vfmsubpd256.
611         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
612         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
613         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
614         Likewise.
615         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
616         Likewise.
617         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
618         Likewise.
619         (fmai_vmfmsub_<mode><round_name>): Likewise.
621 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
623         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
624         Remove plus.  Renamed to ...
625         (*sub<mode>3<mask_name>_bcst): This.
626         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
627         (*add<mode>3<mask_name>_bcst): This.
629 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
631         PR target/72782
632         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
634 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
636         PR target/87662
637         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
638         (_mm_or_epi32): Likewise.
639         (_mm256_xor_epi32): Likewise.
640         (_mm_xor_epi32): Likewise.
641         (_mm256_or_epi64): Likewise.
642         (_mm_or_epi64): Likewise.
643         (_mm256_xor_epi64): Likewise.
644         (_mm_xor_epi64): Likewise.
646 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
648         PR target/72782
649         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
651 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
653         PR middle-end/87647
654         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
656 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
658         * doc/ux.texi: Move @section directly after @node.
660 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
662         PR middle-end/85488
663         PR middle-end/87649
664         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
665         depend closely nested inside of loop with ordered clause with
666         a parameter.
668 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
670         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
671         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
672         * doc/ux.texi: New file.
674 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
676         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
677         be the first CR field allocated.
679 2018-10-19  Richard Biener  <rguenther@suse.de>
681         PR target/87657
682         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
683         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
685 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
687         PR target/72782
688         * config/i386/sse.md
689         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
690         (*add<mode>3<mask_name>_bcst_2): Likewise.
692 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
694         * config/i386/sse.md
695         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
696         Replace nonimmediate_operand with register_operand.
697         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
698         Likewise.
699         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
700         Likewise.
702 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
704         PR rtl-optimization/87596
705         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
706         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
707         for instructions in FROM..TO range.
709 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
711         * cfgexpand.c (expand_one_var): Use specific wording in error message
712         for non-local frame variables.
713         * stor-layout.c (layout_decl): Do not issue a warning for them.
715 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
717         * haifa-sched.c (priority): Add force_recompute parameter.
718         (apply_replacement): Call priority () with force_recompute = true.
719         (restore_pattern): Likewise.
721 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
723         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
724         HOST_BITS_PER_WIDE_INT.
725         (test_vector_ops_duplicate): Likewise.
727 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
729         PR target/72782
730         * config/i386/sse.md (VF_AVX512): New.
731         (avx512bcst): Likewise.
732         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
733         Likewise.
734         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
735         Likewise.
736         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
737         Likewise.
739 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
741         * doc/invoke.texi (-dumpversion): Improve grammar.
742         (-dumpfullversion): Make more consistent with -dumpversion.
744 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
746         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
747         Set JUMP_LABEL to the jump insn.
748         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
749         Predict emitted jump and add label to jump insn.
751 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
753         PR tree-optimization/87562
754         * input.c (get_substring_ranges_for_loc): Use
755         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
756         getting the linemap for the endpoint.  Verify that it's either
757         in the same linemap as the start point's spelling location, or
758         at least in the same file.
760 2018-10-18  Richard Biener  <rguenther@suse.de>
762         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
763         feed width-specific load/store costs through ix86_vec_cost.
764         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
765         (k8_cost): Likewise.
766         (bdver_cost): Likewise.
767         (znver1_cost): Likewise.
768         (btver1_cost): Likewise.
769         (btver2_cost): Likewise.
771 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
773         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
774         to simplify subreg of vec_merge.
776 2018-10-18  Richard Biener  <rguenther@suse.de>
778         * config/i386/i386.c: Fix costing of vector FMA.
780 2018-10-18  Richard Biener  <rguenther@suse.de>
782         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
783         and argument.
784         (ix86_builtin_vectorization_cost): For vec_construct properly
785         cost insertion into SSE regs.
786         (...): Adjust calls to ix86_vec_cost.
788 2018-10-18  Richard Biener  <rguenther@suse.de>
790         PR middle-end/87087
791         Revert
792         2018-02-07  Richard Biener  <rguenther@suse.de>
794         PR tree-optimization/84204
795         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
796         this place.
798 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
800         PR target/87537
801         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
802         of vec_duplicate.
803         (test_vector_ops_duplicate): Add test for a scalar subreg of a
804         VEC_MERGE of a VEC_DUPLICATE.
806 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
808         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
809         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
810         * doc/standards.texi (C Language): Document C2X.
811         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
812         (rl78_option_override): Handle "GNU C2X" language name.
814 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
816         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
817         Document C17 as published in 2018.
819 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
821         PR middle-end/87623
822         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
823         bail out if both sides do not have the same storage order.
825 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
827         * bitmap.c (bitmap_head::dump): New.
828         * bitmap.h (bitmap_head): Add dump().
829         * gimple-ssa-evrp-analyze.c
830         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
831         (evrp_range_analyzer::set_ssa_range_info): Same.
832         (evrp_range_analyzer::record_ranges_from_phis): Same.
833         (evrp_range_analyzer::record_ranges_from_stmt): Same.
834         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
835         * gimple-ssa-sprintf.c (get_int_range): Same.
836         (format_integer): Same.
837         (sprintf_dom_walker::handle_gimple_call): Same.
838         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
839         (ipcp_vr_lattice::top_p): Same.
840         (ipcp_vr_lattice::bottom_p): Same.
841         (ipcp_vr_lattice::set_to_bottom): Same.
842         (ipa_vr_operation_and_type_effects): Same.
843         (propagate_vr_across_jump_function): Same.
844         (ipcp_store_vr_results): Same.
845         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
846         (ipa_print_node_jump_functions_for_edge): Same.
847         (ipa_get_value_range): Same.
848         (ipa_compute_jump_functions_for_edge): Same.
849         (ipa_write_jump_function): Same.
850         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
851         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
852         Same.
853         * vr-values.c (set_value_range_to_nonnegative): Same.
854         (set_value_range_to_truthvalue): Same.
855         (vr_values::get_value_range): Same.
856         (vr_values::set_defs_to_varying): Same.
857         (vr_values::update_value_range): Same.
858         (symbolic_range_based_on_p): Same.
859         (vr_values::op_with_boolean_value_range_p): Same.
860         (vr_values::extract_range_for_var_from_comparison_expr): Same.
861         (vr_values::extract_range_from_ssa_name): Same.
862         (vr_values::extract_range_from_binary_expr): Same.
863         (vr_values::extract_range_from_unary_expr): Same.
864         (vr_values::extract_range_from_cond_expr): Same.
865         (vr_values::extract_range_from_comparison): Same.
866         (vr_values::check_for_binary_op_overflow): Same.
867         (vr_values::extract_range_basic): Same.
868         (vr_values::extract_range_from_assignment): Same.
869         (compare_ranges): Same.
870         (compare_range_with_value): Same.
871         (vr_values::adjust_range_with_scev): Same.
872         (vrp_valueize): Same.
873         (vrp_valueize_1): Same.
874         (vr_values::get_vr_for_comparison): Same.
875         (vr_values::compare_name_with_value): Same.
876         (vr_values::compare_names): Same.
877         (vr_values::vrp_evaluate_conditional): Same.
878         (find_case_label_ranges): Same.
879         (vr_values::vrp_visit_switch_stmt): Same.
880         (vr_values::extract_range_from_phi_node): Same.
881         (vr_values::simplify_div_or_mod_using_ranges): Same.
882         (vr_values::simplify_bit_ops_using_ranges): Same.
883         (test_for_singularity): Same.
884         (range_fits_type_p): Same.
885         (vr_values::simplify_cond_using_ranges_1): Same.
886         (vr_values::simplify_switch_using_ranges): Same.
887         (vr_values::simplify_float_conversion_using_ranges): Same.
888         (vr_values::two_valued_val_range_p): Same.
889         (vr_values::add_equivalence): Move to value_range::equiv_add.
890         * vr-values.h (vr_values::add_equivalence): Remove.
891         (VR_INITIALIZER): Remove.
892         * tree-vrp.c (value_range::set): New.
893         (value_range::equiv_add): New.
894         (value_range::value_range): New.
895         (value_range::deep_copy): New.
896         (value_range::check): New.
897         (value_range::equal_p): New.
898         (value_range::ignore_equivs_equal_p): New.
899         (value_range::operator==): New.
900         (value_range::operator!=): New.
901         (value_range::symbolic_p): New.
902         (value_range::numeric_p): New.
903         (value_range::set_undefined): New.
904         (value_range::set_varying): New.
905         (value_range::may_contain_p): New.
906         (value_range::equiv_clear): New.
907         (value_range::singleton_p): New.
908         (value_range::intersect): New.
909         (value_range::dump): New.
910         (value_range::set_and_canonicalize): New.
911         (set_value_range): Adjust for value_range API.
912         (set_value_range_to_undefined): Same.
913         (set_value_range_to_varying): Same.
914         (set_and_canonicalize_value_range): Same.
915         (set_value_range_to_nonnull): Same.
916         (set_value_range_to_null): Same.
917         (range_is_null): Same.
918         (range_is_nonnull): Same.
919         (range_int_cst_p): Same.
920         (range_int_cst_singleton_p): Same.
921         (symbolic_range_p): Same.
922         (range_includes_zero_p): Same.
923         (value_range_constant_singleton): Same.
924         (vrp_set_zero_nonzero_bits): Same.
925         (ranges_from_anti_range): Same.
926         (extract_range_into_wide_ints): Same.
927         (extract_range_from_multiplicative_op): Same.
928         (set_value_range_with_overflow): Same.
929         (extract_range_from_binary_expr_1): Same.
930         (extract_range_from_unary_expr): Same.
931         (dump_value_range): Same.
932         (debug_value_range): Same.
933         (vrp_prop::check_array_ref): Same.
934         (vrp_prop::check_mem_ref): Same.
935         (vrp_prop::vrp_initialize): Same.
936         (vrp_prop::visit_stmt): Same.
937         (intersect_ranges): Same.
938         (vrp_prop::visit_phi): Same.
939         (vrp_prop::vrp_finalize): Same.
940         (determine_value_range_1): Same.
941         (determine_value_range): Same.
942         (vrp_intersect_ranges_1): Rename to...
943         (vrp_intersect_1): this.
944         (vrp_intersect_ranges): Rename to...
945         (value_range::intersect_helper): ...this.
946         (vrp_meet_1): Rename to...
947         (value_range::union_helper): ...this.
948         (vrp_meet): Rename to...
949         (value_range::union_): ...this.
950         (copy_value_range): Remove.
951         * tree-vrp.h (struct value_range): Rewrite into a proper class.
952         (value_range::vrtype): New.
953         (value_range::type): New.
954         (value_range::equiv): New.
955         (value_range::min): New.
956         (value_range::max): New.
957         (value_range::varying_p): New.
958         (value_range::undefined_p): New.
959         (value_range::null_p): New.
960         (value_range::equiv_add): New.
961         (copy_value_range): Remove.
963 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
965         * Makefile.in (SELFTEST_TARGETS): New.
966         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
967         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
968         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
969         c/Make-lang.in.
970         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
971         (selftest-c++-gdb, selftest-c++-valgrind): Move to
972         cp/Make-lang.in.
973         * configure: Regenerate.
974         * configure.ac (selftest_languages): New.
976 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
978         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
979         overflow wraps argument.
980         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
981         wide_int_range_multiplicative_op.
982         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
983         overflow wraps argument.
984         (wide_int_range_multiplicative_op): Same.
985         (wide_int_range_lshift): Same.
986         (wide_int_range_div): Same.
987         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
988         (wide_int_range_lshift): Same.
989         (wide_int_range_div): Same.
991 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
993         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
994         use sign as argument.
995         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
996         wide_int_range_shift_undefined_p.
998 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1000         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
1001         Rename to...
1002         (@despeculate_copy<ALLI_TI:mode>): ... This.
1003         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
1004         switch statement.
1006 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1008         * config.gcc: Obsolete *-*-solaris2.10*.
1009         * doc/install.texi (Specific, *-*-solaris2*): Document it.
1011 2018-10-12  Jeff Law  <law@redhat.com>
1013         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
1014         reg + sym +- const_int addressing modes.
1016 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1018         * common.opt (fdiagnostics-minimum-margin-width=): New option.
1019         * diagnostic-show-locus.c (layout::layout): Apply the minimum
1020         margin width.
1021         (layout::start_annotation_line): Only print up to 3 of the
1022         margin character, to avoid touching the left-hand side.
1023         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
1024         minimum margin width, as set by test_diagnostic_context's ctor.
1025         (selftest::test_fixit_insert_containing_newline): Likewise.
1026         (selftest::test_fixit_insert_containing_newline_2): Likewise.
1027         (selftest::test_line_numbers_multiline_range): Clear
1028         dc.min_margin_width.
1029         * diagnostic.c (diagnostic_initialize): Initialize
1030         min_margin_width.
1031         * diagnostic.h (struct diagnostic_context): Add field
1032         "min_margin_width".
1033         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
1034         * opts.c (common_handle_option): Handle
1035         OPT_fdiagnostics_minimum_margin_width_.
1036         * selftest-diagnostic.c
1037         (selftest::test_diagnostic_context::test_diagnostic_context):
1038         Initialize min_margin_width to 6.
1039         * toplev.c (general_init): Initialize global_dc->min_margin_width.
1041 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1043         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
1044         Fix usage of "error_at_rich_loc" in the comment.
1046 2018-10-15  Renlin Li  <renlin.li@arm.com>
1048         PR target/87563
1049         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
1050         if-conversioned loop when it contains ifn with types not
1051         supported by backend.
1052         * internal-fn.c (expand_direct_optab_fn): Add an assert.
1053         (direct_internal_fn_supported_p): New helper function.
1054         * internal-fn.h (direct_internal_fn_supported_p): Declare.
1056 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
1058         PR target/87572
1059         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1060         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1061         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1063 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
1065         PR tree-optimization/87022
1066         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1067         bits in dist vector rather than the first one.
1069 2018-10-15  Richard Biener  <rguenther@suse.de>
1071         PR middle-end/87610
1072         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
1073         (visit_loadstore): When a used restrict tag escaped verify that
1074         the points-to solution of "other" pointers do not include
1075         escaped.
1076         (compute_dependence_clique): If a used restrict tag escaped
1077         communicated that down to visit_loadstore.
1079 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
1081         * config/s390/s390.c (s390_expand_vec_init): Force vector element
1082         into reg if it isn't a general operand.
1084 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1086         PR target/87599
1087         * config/i386/sse.md (*vec_dupv2di): Add register source to
1088         movddup.
1090 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1092         PR target/87572
1093         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1094         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
1095         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1096         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1098 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
1100         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
1101         (notice_args_size): Set it in the current trace if no insn that can
1102         throw internally has been seen yet.
1103         (connect_traces): When connecting args_size between traces, allow the
1104         incoming values not to match if there is an insn setting it before the
1105         first insn that can throw internally; in that case, force the creation
1106         of a CFI note on this latter insn.
1108 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
1110         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
1111         the base class, because using the injected-class-name was not clearly
1112         specified until DR 176.
1114 2018-10-12  Paul Koning  <ni1d@arrl.net>
1116         * config/pdp11/pdp11.md (doloop_end): New expander.
1117         (doloop_end_insn): renamed from "doloop_end".
1118         (addqi3): New pattern.
1119         (subqi3): New pattern.
1120         * config/pdp11/predicates.md (incdec_operand): New predicate.
1122 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
1124         PR middle-end/81376
1125         * real.c (format_helper::can_represent_integral_type_p): New function
1126         * real.h (format_helper::can_represent_integral_type_p): Ditto.
1127         * match.pd: New pattern.
1129 2018-10-12  Alexandre Oliva <oliva@adacore.com>
1131         * configure.ac: Introduce --enable-large-address-aware
1132         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
1133         * doc/install.texi: Document it.
1134         * configure, config.in: Rebuilt.
1135         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
1136         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
1137         (LINK_SPEC): Insert it.
1138         * config/i386/mingw-264.h: Likewise.
1140         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
1142 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
1144         PR rtl-optimization/87600
1145         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
1147 2018-10-12  Paul Koning  <ni1d@arrl.net>
1149         * doc/md.texi (doloop_end): Document that the pattern code may
1150         need to check operand mode.
1152 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
1154         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
1155         to zero-extend between int and floating-point registers.
1156         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
1157         ldp into floating-point registers.  Add type and arch attributes.
1158         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
1159         Use f_loads for type attribute.
1161 2018-10-11  Martin Sebor  <msebor@redhat.com>
1163         * doc/extend.texi (attribute packed): Correct typos.
1165 2018-10-11  Martin Sebor  <msebor@redhat.com>
1167         * doc/extend.texi (attribute flatten): Mention interaction with
1168         noinline.
1170 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1172         PR target/87156
1173         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1174         Set new_decl virtual flag to zero.
1176 2018-10-11  Martin Sebor  <msebor@redhat.com>
1178         PR middle-end/87593
1179         * doc/extend.texi (attribute format_arg): Discuss using multiple
1180         attributes on a single function.
1182 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1184         PR tree-optimization/86829
1185         * match.pd (sin (atan (x))): New simplification rules.
1186         (cos (atan (x))): Likewise.
1187         * real.c (build_sinatan_real): New function.
1188         * real.h (build_sinatan_real): Prototype.
1190 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
1192         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
1193         function.
1194         (fold_mergeeo_helper): New helper function.
1195         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
1196         intrinsics.  Correct some whitespace indentation issues.
1198 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
1200         PR target/87511
1201         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
1202         Use HOST_WIDE_INT_1U for shift.
1204 2018-10-11  Doug Rupp  <rupp@adacore.com>
1205             Olivier Hainque  <hainque@adacore.com>
1207         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
1208         Pass --relax to the linker for RTPs.
1209         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
1211 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
1212             Jan Hubicka  <jh@suse.cz>
1213             Martin Jambor  <mjambor@suse.cz>
1215         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
1216         the same elements are repeated rather than printing all of them.
1217         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
1218         "repeated" elements.
1219         * read-rtl-function.c (test_loading_repeat): New function.
1220         (read_rtl_function_c_tests): Call test_loading_repeat.
1221         * rtl-tests.c (test_dumping_repeat): New function.
1222         (rtl_tests_c_tests): Call test_dumping_repeat.
1224 2018-10-11  Richard Biener  <rguenther@suse.de>
1226         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
1227         bdver?_cost): Unify to ...
1228         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
1229         * config/i386/i386.c (processor_cost_table): Adjust.
1231 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
1233         PR middle-end/87574
1234         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1235         the thunk when expanding to GIMPLE.
1237 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1239         * varasm.c (mergeable_string_section): Don't try to move zero-length
1240         strings to the merge section.
1242 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
1244         PR target/87573
1245         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1247 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
1249         PR target/87550
1250         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1251         to special_args set.
1253 2018-10-10  Richard Biener  <rguenther@suse.de>
1255         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1256         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1257         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1258         and recursing and pattern terminating the recursion on SSE
1259         vector width using ix86_expand_reduc.
1260         (reduc_sminmax_scal_<mode>): Split into part reducing to half
1261         width and recursing and SSE2 vector variant doing the final
1262         reduction with ix86_expand_reduc.
1263         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1264         with terminating the recursion at AVX level, splitting that
1265         to SSE there.
1267 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
1269         * genmatch.c (error_cb): Rename to...
1270         (diagnostic_cb): ...this, converting int params to enums.
1271         (fatal_at): Update for renaming.
1272         (warning_at): Likewise.
1273         (main): Likewise.
1274         * input.c (selftest::ebcdic_execution_charset::apply):
1275         Update for renaming of...
1276         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1277         to...
1278         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1279         converting level and reason to enums.
1280         (class selftest::lexer_error_sink): Rename to...
1281         (class selftest::lexer_test_options): ...this, renaming field
1282         "m_errors" to "m_diagnostics".
1283         (selftest::lexer_test_options::apply): Update for renaming of...
1284         (selftest::lexer_test_options::on_error): ...this, renaming to...
1285         (selftest::lexer_test_options::on_diagnostic): ...this
1286         converting level and reason to enums.
1287         (selftest::test_lexer_string_locations_raw_string_unterminated):
1288         Update for renamings.
1289         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
1290         "reason".
1292 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
1294         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1295         * config/rs6000/pmmintrin.h: New file.
1297 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
1299         PR tree-optimization/86659
1300         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1302 2018-10-09  Richard Biener  <rguenther@suse.de>
1304         PR tree-optimization/63155
1305         * tree-ssa-structalias.c: Include tree-ssa.h.
1306         (get_constraint_for_ssa_var): For undefs return nothing_id.
1307         (find_func_aliases): Cleanup PHI handling.
1309 2018-10-09  Richard Biener  <rguenther@suse.de>
1311         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1312         replacements.
1314 2018-10-09  Martin Liska  <mliska@suse.cz>
1316         * asan.c (asan_emit_stack_protection): If a stack variable
1317         is located in a same file as current function, then emit
1318         line info into variable definition string.
1320 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1322         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1323         information.
1325 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1327         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1328         on the thunk.
1330 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1332         PR tree-optimization/86659
1333         * gimple-match.h (struct gimple_match_op): Add reverse field.
1334         (gimple_match_op::set_op): New overloaded method.
1335         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1336         the REF_REVERSE_STORAGE_ORDER flag on the value.
1337         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
1338         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
1339         
1340 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
1342         PR middle-end/63155
1343         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
1344         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
1346 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
1348         PR target/87517
1349         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
1350         Defined with __builtin_ia32_vfmaddsubpd512_mask.
1352 2018-10-08  Richard Biener  <rguenther@suse.de>
1354         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
1355         cost the same as AVX128 ones.
1357 2018-10-08  Paul Koning  <ni1d@arrl.net>
1359         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
1360         (expand_block_move): New function.
1361         * config/pdp11/pdp11.c (output_block_move): Remove.
1362         (expand_block_move): New function.
1363         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
1364         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
1365         (*movmemhi1): Remove.
1367 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
1369         * config/s390/2827.md: Increase latencies for some FP instructions.
1371 2018-10-08  Richard Biener  <rguenther@suse.de>
1373         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1374         Open a dump scope.
1375         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
1376         * tree-vectorizer.h (dump_stmt_cost): Adjust.
1377         (add_stmt_cost): Dump return value of the hook.
1379 2018-10-08  Richard Biener  <rguenther@suse.de>
1381         PR tree-optimization/63155
1382         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
1383         (ssa_propagation_engine::ssa_propagate): Remove redundant
1384         bitmap bit clearing.
1386 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
1388         PR rtl-optimization/86939
1389         PR rtl-optimization/87479
1390         * ira.h (non_conflicting_reg_copy_p): New prototype.
1391         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
1392         (make_hard_regno_dead): Don't add conflicts for register
1393         ignore_reg_for_conflicts.
1394         (make_object_dead): Likewise.
1395         (non_conflicting_reg_copy_p): New function.
1396         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
1397         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
1398         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
1399         (make_hard_regno_dead): Don't add conflicts for register
1400         ignore_reg_for_conflicts.  Remove special conflict handling of
1401         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
1402         check_pic_pseudo_p and update callers.
1403         (mark_pseudo_dead): Don't add conflicts for register
1404         ignore_reg_for_conflicts.
1405         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
1407 2018-10-05  Andrew Waterman  <andrew@sifive.com>
1408             Jim Wilson  <jimw@sifive.com>
1410         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
1411         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
1412         new pattern using HONOR_SNANS that emits one extra instruction.
1414 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1416         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
1417         patterns): Merge SI and DI patterns to a GPR pattern.
1418         (unnamed define_insn and define_split for record form of that): Merge
1419         to a single define_insn_and_split pattern.
1421 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
1423         PR c++/56856
1424         * input.c (expand_location_to_spelling_point): Add param "aspect"
1425         and use rather than hardcoding LOCATION_ASPECT_CARET.
1426         (get_substring_ranges_for_loc): Handle the case of a single token
1427         within a macro expansion.
1428         * input.h (expand_location_to_spelling_point): Add "aspect" param,
1429         defaulting to LOCATION_ASPECT_CARET.
1431 2018-10-05  Paul Koning  <ni1d@arrl.net>
1433         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
1434         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
1435         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1436         (pdp11_guard_type): New function.
1438 2018-10-05  Paul Koning  <ni1d@arrl.net>
1440         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
1441         * config/pdp11/pdp11.opt (mfloat32): Remove.
1442         (mfloat64): Remove.
1443         * doc/invoke.texi (pdp11 -mfloat32): Remove:
1444         (pdp11 -mfloat64): Remove.
1446 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1448         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
1449         (*cmp<mode>_cc_i387): Ditto.
1450         (*cmpu<mode>_cc_i387): Ditto.
1451         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1452         * config/i386/i386.c (ix86_expand_fp_compare): Remove
1453         "scratch" argument.
1454         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
1455         Emit x86_sahf_1 pattern.
1456         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
1457         (ix86_expand_carry_flag_compare): Ditto.
1459 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1461         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
1462         to reg_or_0_operand.  Add "C" constraint.
1463         (*cmpxf_cc_i387): Ditto.
1464         (*cmp<mode>_i387): Change operand 2 predicate
1465         to nonimm_or_0_operand.  Add "C" constraint.
1466         (*cmp<mode>_cc_i387): Ditto.
1467         (*cmp<mode>_0_i387): Remove insn pattern.
1468         (*cmp<mode>_0_cc_i387): Ditto.
1470 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1472         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
1473         * config/i386/predicates.md (nonimm_or_0_operand): Rename
1474         from vector_move_operand.  Update all uses.
1476 2018-10-05  Martin Sebor  <msebor@redhat.com>
1478         PR tree-optimization/87490
1479         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
1480         consistently.
1482 2018-10-05  Richard Biener  <rguenther@suse.de>
1484         PR tree-optimization/63155
1485         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
1486         vertical space in dumpfiles.
1487         * tree-ssa-propagate.h
1488         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
1489         * tree-ssa-propagate.c (cfg_blocks_back): New global.
1490         (ssa_edge_worklist_back): Likewise.
1491         (curr_order): Likewise.
1492         (cfg_blocks_get): Remove abstraction.
1493         (cfg_blocks_add): Likewise.
1494         (cfg_blocks_empty_p): Likewise.
1495         (add_ssa_edge): Add to current or next worklist based on
1496         RPO index.
1497         (add_control_edge): Likewise.
1498         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
1499         into ...
1500         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
1501         iteration from CFG and SSA edge worklist so we process
1502         everything in RPO order, prioritizing forward progress
1503         over iteration.
1504         (ssa_prop_init): Allocate new worklists, do not dump
1505         immediate uses.
1506         (ssa_prop_fini): Free new worklists.
1508 2018-10-05  Richard Biener  <rguenther@suse.de>
1510         * tree-core.h (tree_block::abstract_flag): Remove.
1511         (tree_block::block_num): Make full 32bits.
1512         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
1513         * tree.h (BLOCK_ABSTRACT): Remove.
1514         * dwarf2out.c (gen_lexical_block_die): Remove dead code
1515         resulting from BLOCK_ABSTRACT being always false.
1516         (gen_inlined_subroutine_die): Likewise.
1517         (gen_block_die): Likewise.
1518         * tree.c (block_ultimate_origin): Likewise.
1519         * tree-pretty-print.c (dump_block_node): Remove code dealing
1520         with BLOCK_ABSTRACT.
1521         * tree-ssa-live.c (dump_scope_block): Likewise.
1522         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
1523         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
1525 2018-10-05   Richard Biener  <rguenther@suse.de>
1527         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
1528         is asked for initialize mode to the component mode of the
1529         vector type.
1531 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
1533         PR target/87522
1534         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
1535         assembler for -mavx.
1536         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
1538 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1540         PR target/87509
1541         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
1542         RS6000_BTM_DFP.
1543         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
1544         to be DImode.  When using mffscrn, force the operand to a register.
1546 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
1548         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
1549         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
1550         X87MODEF mode iterator.
1551         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
1552         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
1553         X87MODEF mode iterator.
1555 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
1557         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
1558         -Wno-prio-ctor-dtor.
1560 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1562         * Makefile.in (OBJS): Add opt-problem.o.
1563         * dump-context.h: Include "selftest.h.
1564         (selftest::temp_dump_context): New forward decl.
1565         (class dump_context): Make friend of class
1566         selftest::temp_dump_context.
1567         (dump_context::dump_loc_immediate): New decl.
1568         (class dump_pretty_printer): Move here from dumpfile.c.
1569         (class temp_dump_context): Move to namespace selftest.
1570         (temp_dump_context::temp_dump_context): Add param
1571         "forcibly_enable_dumping".
1572         (selftest::verify_dumped_text):
1573         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
1574         (selftest::verify_item):
1575         (ASSERT_IS_TEXT): Move here from dumpfile.c.
1576         (ASSERT_IS_TREE): Likewise.
1577         (ASSERT_IS_GIMPLE): Likewise.
1578         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
1579         to...
1580         (dump_context::dump_loc_immediate): ...this new function.
1581         (class dump_pretty_printer): Move to dump-context.h.
1582         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
1583         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
1584         (temp_dump_context::temp_dump_context): Move to "selftest"
1585         namespace.  Add param "forcibly_enable_dumping", and use it to
1586         conditionalize the use of m_pp;
1587         (selftest::verify_dumped_text): Make non-static.
1588         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
1589         (selftest::verify_item): Make non-static.
1590         (ASSERT_IS_TEXT): Move to dump-context.h.
1591         (ASSERT_IS_TREE): Likewise.
1592         (ASSERT_IS_GIMPLE): Likewise.
1593         (selftest::test_capture_of_dump_calls): Pass "true" for new
1594         param of temp_dump_context.
1595         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
1596         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
1597         TDF_COMPARE_DEBUG.
1598         * opt-problem.cc: New file.
1599         * opt-problem.h: New file.
1600         * optinfo-emit-json.cc
1601         (selftest::test_building_json_from_dump_calls): Pass "true" for
1602         new param of temp_dump_context.
1603         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
1604         (optinfo::emit_for_opt_problem): New function.
1605         (optinfo::emit): Clarity which emit_item is used.
1606         * optinfo.h (optinfo::get_dump_location): New accessor.
1607         (optinfo::emit_for_opt_problem): New decl.
1608         (optinfo::emit): Make const.
1609         * selftest-run-tests.c (selftest::run_tests): Call
1610         selftest::opt_problem_cc_tests.
1611         * selftest.h (selftest::opt_problem_cc_tests): New decl.
1612         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
1613         bool to opt_result, converting fprintf messages to
1614         opt_result::failure_at calls.  Add "stmt" param for use by the
1615         failure_at calls.
1616         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
1617         (runtime_alias_check_p): Convert return type from bool to
1618         opt_result, converting dump_printf calls to
1619         opt_result::failure_at, using the statement DDR_A for their
1620         location.
1621         (find_data_references_in_stmt): Convert return type from bool to
1622         opt_result, converting "return false" to opt_result::failure_at
1623         with a new message.
1624         * tree-data-ref.h: Include "opt-problem.h".
1625         (dr_analyze_innermost): Convert return type from bool to opt_result,
1626         and add a const gimple * param.
1627         (find_data_references_in_stmt): Convert return type from bool to
1628         opt_result.
1629         (runtime_alias_check_p): Likewise.
1630         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
1631         dr_analyze_innermost.
1632         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
1633         Convert return type from bool to opt_result, adding a message for
1634         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
1635         (vect_analyze_data_ref_dependence): Convert return type from bool
1636         to opt_result.  Change sense of return type from "false"
1637         effectively meaning "no problems" to "false" meaning a problem,
1638         so that "return false" becomes "return opt_result::success".
1639         Convert "return true" calls to opt_result::failure_at, using
1640         the location of statement A rather than vect_location.
1641         (vect_analyze_data_ref_dependences): Convert return type from bool
1642         to opt_result.
1643         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
1644         calls to opt_result::failure_at, using the stmt location rather
1645         than vect_location.
1646         (vect_verify_datarefs_alignment): Convert return type from bool
1647         to opt_result.
1648         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
1649         into multiple more-tightly-scoped copies.
1650         (vect_analyze_data_refs_alignment): Convert return type from bool
1651         to opt_result.
1652         (vect_analyze_data_ref_accesses): Likewise, converting a
1653         "return false" to a "return opt_result::failure_at", adding a
1654         new message.
1655         (vect_prune_runtime_alias_test_list): Convert return type from
1656         bool to opt_result, converting dump_printf_loc to
1657         opt_result::failure_at.  Add a %G to show the pertinent statement,
1658         and use the stmt's location rather than vect_location.
1659         (vect_find_stmt_data_reference): Convert return type from
1660         bool to opt_result, converting dump_printf_loc to
1661         opt_result::failure_at, using stmt's location.
1662         (vect_analyze_data_refs):  Convert return type from bool to
1663         opt_result.  Convert "return false" to "return
1664         opt_result::failure_at", adding messages as needed.
1665         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
1666         type from bool to opt_result.
1667         (vect_determine_vf_for_stmt): Likewise.
1668         (vect_determine_vectorization_factor): Likewise, converting
1669         dump_printf_loc to opt_result::failure_at, using location of phi
1670         rather than vect_location.
1671         (vect_analyze_loop_form_1): Convert return type from bool to
1672         opt_result, converting dump_printf_loc calls, retaining the use of
1673         vect_location.
1674         (vect_analyze_loop_form): Convert return type from loop_vec_info
1675         to opt_loop_vec_info.
1676         (vect_analyze_loop_operations): Convert return type from bool to
1677         opt_result, converting dump_printf_loc calls, using the location
1678         of phi/stmt rather than vect_location where available.  Convert
1679         various "return false" to "return opt_result::failure_at" with
1680         "unsupported phi" messages.
1681         (vect_get_datarefs_in_loop): Convert return type from bool to
1682         opt_result.  Add a message for the
1683         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
1684         (vect_analyze_loop_2): Convert return type from bool to
1685         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
1686         each "goto again;", adding new messages where needed.
1687         Add "unsupported grouped {store|load}" messages.
1688         (vect_analyze_loop): Convert return type from loop_vec_info to
1689         opt_loop_vec_info.
1690         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
1691         bool to opt_result.
1692         * tree-vect-stmts.c (process_use): Likewise, converting
1693         dump_printf_loc call and using stmt location, rather than
1694         vect_location.
1695         (vect_mark_stmts_to_be_vectorized): Likeise.
1696         (vect_analyze_stmt): Likewise, adding a %G.
1697         (vect_get_vector_types_for_stmt): Convert return type from bool to
1698         opt_result, converting dump_printf_loc calls and using stmt
1699         location, rather than vect_location.
1700         (vect_get_mask_type_for_stmt): Convert return type from tree to
1701         opt_tree, converting dump_printf_loc calls and using stmt location.
1702         * tree-vectorizer.c: Include "opt-problem.h.
1703         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
1704         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
1705         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
1706         enabled, use it to report at the top level "couldn't vectorize
1707         loop" followed by the problem.
1708         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
1709         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
1710         to opt_result.
1711         (vect_analyze_stmt): Likewise.
1712         (vect_get_vector_types_for_stmt): Likewise.
1713         (tree vect_get_mask_type_for_stmt): Likewise.
1714         (vect_analyze_data_ref_dependences): Likewise.
1715         (vect_enhance_data_refs_alignment): Likewise.
1716         (vect_analyze_data_refs_alignment): Likewise.
1717         (vect_verify_datarefs_alignment): Likewise.
1718         (vect_analyze_data_ref_accesses): Likewise.
1719         (vect_prune_runtime_alias_test_list): Likewise.
1720         (vect_find_stmt_data_reference): Likewise.
1721         (vect_analyze_data_refs): Likewise.
1722         (vect_analyze_loop): Convert return type from loop_vec_info to
1723         opt_loop_vec_info.
1724         (vect_analyze_loop_form): Likewise.
1725         (vect_analyze_slp): Convert return type from bool to opt_result.
1727 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1729         * doc/invoke.texi (-fopt-info): Document new "internals"
1730         sub-option.
1731         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
1732         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
1733         MSG_ALL_KINDS.
1734         (optinfo_verbosity_options): Add "internals".
1735         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
1736         (dump_context::apply_dump_filter_p): New member function.
1737         (dump_context::dump_loc): Use apply_dump_filter_p rather than
1738         explicitly masking the dump_kind.
1739         (dump_context::begin_scope): Increment the scope depth first.  Use
1740         apply_dump_filter_p rather than explicitly masking the dump_kind.
1741         (dump_context::emit_item): Use apply_dump_filter_p rather than
1742         explicitly masking the dump_kind.
1743         (dump_dec): Likewise.
1744         (dump_hex): Likewise.
1745         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
1746         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
1747         (opt_info_switch_p): Update handling of default
1748         MSG_OPTIMIZED_LOCATIONS to cope with default of
1749         MSG_PRIORITY_USER_FACING.
1750         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
1751         masking the dump_kind.
1752         (selftest::test_capture_of_dump_calls): Update test_dump_context
1753         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
1754         than MSG_ALL.  Generalize scope test to be run at all four
1755         combinations of with/without MSG_PRIORITY_USER_FACING and
1756         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
1757         for each of the two values.
1758         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
1759         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
1760         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
1761         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
1762         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
1763         with MSG_PRIORITY_*.
1764         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
1765         dump messages as MSG_PRIORITY_USER_FACING.
1766         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
1767         about the interaction with MSG_PRIORITY_*.
1769 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1771         * varasm.c (output_constant): Add new parameter merge_strings.
1772         Make strings properly zero terminated in merge string sections.
1773         (mergeable_string_section): Don't fail if the last char is non-zero.
1774         (assemble_variable_contents): Handle merge string sections.
1775         (assemble_variable): Likewise.
1776         (assemble_constant_contents): Likewise.
1777         (output_constant_def_contents): Likewise.
1778         (output_constructor_array_range,
1779         output_constructor_regular_field): Adjust call to output_constant.
1780         (output_object_block): Adjust call to assemble_constant_contents
1781         and assemble_variable_contents.
1783 2018-10-04  Martin Liska  <mliska@suse.cz>
1785         PR c/87483
1786         * cgraphunit.c (process_function_and_variable_attributes):
1787         Warn about a function with alias attribute and a body.
1789 2018-10-04  Martin Liska  <mliska@suse.cz>
1791         PR ipa/82625
1792         * multiple_target.c (redirect_to_specific_clone): New function.
1793         (ipa_target_clone): Use it.
1794         * tree-inline.c: Fix comment.
1796 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1798         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
1799         fields.
1800         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
1801         (gcc::dump_manager::register_pass): New member function, adapted
1802         from loop body in gcc::pass_manager::register_pass, adding a
1803         call to update_dfi_for_opt_info.
1804         (gcc::dump_manager::opt_info_enable_passes): Store the
1805         -fopt-info options into the new fields.  Move the loop
1806         bodies into...
1807         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
1808         function.
1809         * dumpfile.h (struct opt_pass): New forward decl.
1810         (gcc::dump_manager::register_pass): New decl.
1811         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
1812         (class gcc::dump_manager): Add fields "m_optgroup_flags",
1813         "m_optinfo_flags", and "m_optinfo_filename".
1814         * passes.c (gcc::pass_manager::register_pass): Move all of the
1815         dump-handling code to gcc::dump_manager::register_pass.
1817 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
1819         PR rtl-optimization/87466
1820         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
1821         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
1822         * doc/tm.texi: Regenerate.
1823         * ira-lives.c (process_bb_node_lives): Use the new target hook.
1824         * lra-lives.c (process_bb_lives): Likewise.
1825         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
1826         Define.
1828 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
1830         * params.c (add_params): Fix initialization.
1832 2018-10-04  Martin Liska  <mliska@suse.cz>
1834         PR gcov-profile/84107
1835         * tree-profile.c (init_ic_make_global_vars):
1836         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
1837         Come up with new ic_tuple* variables.  Emit
1838         __gcov_indirect_call{,_topn} variables.
1839         (gimple_gen_ic_profiler): Access the variable
1840         and emit gimple.
1841         (gimple_gen_ic_func_profiler): Access
1842         __gcov_indirect_call.callee field.
1843         (gimple_init_gcov_profiler): Use ptr_type_node.
1844         * value-prof.c (gimple_ic): Use ptr_type_node.
1846 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1848         PR tree-optimization/85787
1849         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
1850         into this function and add support for detecting multiple phis.
1851         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
1853 2018-10-04  Martin Liska  <mliska@suse.cz>
1855         PR ipa/87491
1856         * ipa-inline.c (inline_to_all_callers_1):
1857         Call ultimate_alias_target for node being inlined.
1859 2018-10-03  Jeff Law  <law@redhat.com>
1861         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
1862         target's wchar_t.
1863         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
1864         * tree.h (get_typenode_from_name): Prototype.
1866 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
1868         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
1869         Change operand 2 predicate to nonimmediate_operand.
1870         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1872 2018-10-03  Martin Sebor  <msebor@redhat.com>
1873             Jeff Law  <law@redhat.com>
1875         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
1876         initialize it.
1877         (get_string_length): Detect unterminated arrays.
1878         (format_string): Same.
1879         (format_directive): Warn about unterminated arrays.
1880         (handle_gimple_call): Mark statements with no_warning as needed.
1882 2018-10-03  Jim Wilson  <jimw@sifive.com>
1884         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
1885         also define __riscv_abi_rve.  Delete trailing white space.
1887 2018-10-03  Paul Koning  <ni1d@arrl.net>
1889         Enable LRA register allocator for PDP11.
1890         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
1891         (R): Likewise.
1892         (D): Likewise.
1893         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
1894         * config/pdp11/pdp11.opt (-mlra): New option.
1895         * doc/invoke.texi (PDP-11 Options): Document -mlra.
1897 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
1899         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
1900         (*<absneg:code>extend<mode>xf2): Ditto.
1902 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
1904         PR tree-optimization/87415
1905         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
1906         precision fields.
1908 2018-10-02  Jeff Law  <law@redhat.com>
1910         * gimple-fold.c (get_range_strlen): Only set *nonstr when
1911         an unterminated string is discovered.  Bubble up range
1912         even for unterminated strings.
1913         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
1914         indicates the string was not terminated via NONSTR.
1916 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
1918         * tree-vrp.c (extract_range_from_unary_expr): Special case all
1919         pointer conversions.
1920         Do not do anything special for anti-ranges.
1922 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
1924         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
1925         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
1927 2018-10-03  Martin Liska  <mliska@suse.cz>
1929         PR gcov-profile/86109
1930         * coverage.c (coverage_begin_function): Do not
1931         mark lambdas as artificial.
1932         * tree-core.h (struct GTY): Remove tm_clone_flag
1933         and introduce new lambda_function.
1934         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
1936 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
1938         PR target/87474
1939         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
1940         P8_VECTOR and VSX are enabled.
1942 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
1944         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
1945         0x3907 as CPU model number.
1947 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
1949         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
1950         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
1951         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
1952         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
1953         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
1954         * config/s390/s390.md: Likewise. Rename also the cpu attribute
1955         value from arch12 to z14.
1957 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
1959         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
1960         (isinfxf2): Ditto.
1961         (isinf<mode>2): Ditto.
1963 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
1965         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
1966         before emitting fxam.  Perform calculations in XFmode.
1968 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
1970         * match.pd (((X /[ex] A) +- B) * A): New transformation.
1972 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
1974         PR middle-end/87319
1975         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
1976         * tree.c (signed_or_unsigned_type_for): Handle complex.
1978 2018-10-02  Jeff Law  <law@redhat.com>
1980         * gimple-fold.c (get_range_strlen): Remove dead code.
1982 2018-10-02  Martin Sebor  <msebor@redhat.com>
1983             Jeff Law  <law@redhat.com>
1985         * builtins.c (unterminated_array): Add new arguments.
1986         If argument is not terminated, bubble up size and exact
1987         state to callers.
1988         (expand_builtin_strnlen): Detect, avoid expanding
1989         and diagnose unterminated arrays.
1990         (c_strlen): Fill in offset of start of unterminated strings.
1991         * builtins.h (unterminated_array): Update prototype.
1993 2018-10-02  Richard Biener  <rguenther@suse.de>
1995         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
1996         of haddv4df, first reduce to SSE width and exploit the fact
1997         that we only need element zero with the reduction result.
1998         (reduc_plus_scal_v2df): Likewise.
2000 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2002         * dojump.h (do_jump): Delete.
2003         (do_jump_1): Likewise.
2004         (split_comparison): Move around.
2005         * dojump.c (do_jump): Make static.
2006         (do_jump_1): Likewise.
2007         (jumpifnot): Move around.
2008         (jumpifnot_1): Likewise.
2009         (jumpif): Likewise.
2010         (jumpif_1): Likewise.
2011         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
2013 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2015         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
2016         insns in the delay slot and add_insn_after for the jump insn.
2018 2018-10-02  Richard Biener  <rguenther@suse.de>
2020         * tree-inline.c (expand_call_inline): Use the location of
2021         the callee declaration for the inline-entry marker.
2022         * final.c (notice_source_line): Remove special-casing of
2023         NOTE_INSN_INLINE_ENTRY.
2025 2018-10-01  Carl Love  <cel@us.ibm.com>
2027         PR 69431
2028         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
2029         (__builtin_mtfsb0): New.
2030         (__builtin_mtfsb1): New.
2031         ( __builtin_set_fpscr_rn): New.
2032         (__builtin_set_fpscr_drn): New.
2033         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
2034         (rs6000_expand_set_fpscr_rn_builtin): Add.
2035         (rs6000_expand_set_fpscr_drn_builtin): Add.
2036         (rs6000_expand_builtin): Add case statement entries for
2037         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
2038         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
2039         RS6000_BUILTIN_MFFSL.
2040         (rs6000_init_builtins): Add ftype initialization and def_builtin
2041         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
2042         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
2043         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
2044         rs6000_mffscdrn): Add define_insn.
2045         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
2046         * doc/extend.texi: Add documentation for the builtins.
2048 2018-10-01  Richard Biener  <rguenther@suse.de>
2050         PR tree-optimization/87465
2051         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2052         causing branch miscounts.
2054 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2056         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
2057         aarch64_option_default_param):  New.
2058         (params.h): Include.
2059         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
2060         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
2061         stack-clash protection validation code.
2063 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2065         * params.c (validate_param): New.
2066         (add_params): Use it.
2067         (set_param_value): Refactor param validation into validate_param.
2068         (diagnostic.h): Include.
2069         * diagnostic.h (diagnostic_ready_p): New.
2071 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2073         * params.c (set_param_value):
2074         Add index of parameter being validated.
2075         * common/common-target.def (option_validate_param): New.
2076         * common/common-targhooks.h (default_option_validate_param): New.
2077         * common/common-targhooks.c (default_option_validate_param): New.
2078         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
2079         * doc/tm.texi: Regenerate.
2081 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2083         PR target/86486
2084         * config/aarch64/aarch64.c (aarch64_override_options_internal):
2085         Add validation for stack-clash parameters and set defaults.
2087 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2089         PR target/86486
2090         * configure.ac: Add stack-clash-protection-guard-size.
2091         * doc/install.texi: Document it.
2092         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
2093         * params.def: Update comment for guard-size.
2094         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
2095         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
2096         * configure: Regenerate.
2098 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2100         PR target/86486
2101         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
2102         STACK_DYNAMIC_OFFSET): New.
2103         * config/aarch64/aarch64.c (aarch64_layout_frame):
2104         Update outgoing args size.
2105         (aarch64_stack_clash_protection_alloca_probe_range,
2106         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2108 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2110         PR target/86486
2111         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
2112         probe ranges.
2113         * target.def (stack_clash_protection_alloca_probe_range): New.
2114         (stack_clash_protection_final_dynamic_probe): Remove.
2115         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
2116         (default_stack_clash_protection_final_dynamic_probe): Remove.
2117         * targhooks.c: Likewise.
2118         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2119         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
2120         * doc/tm.texi: Regenerate.
2122 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2124         PR target/86486
2125         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
2126         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
2127         aarch64_clamp_to_uimm12_shift): New.
2128         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
2129         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
2131 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2133         PR target/86486
2134         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
2136 2018-10-01  Jeff Law  <law@redhat.com>
2137             Richard Sandiford <richard.sandiford@linaro.org>
2138             Tamar Christina  <tamar.christina@arm.com>
2140         PR target/86486
2141         * config/aarch64/aarch64.md
2142         (probe_stack_range): Add k (SP) constraint.
2143         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
2144         STACK_CLASH_MAX_UNROLL_PAGES): New.
2145         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
2146         stack probes for stack clash.
2147         (aarch64_allocate_and_probe_stack_space): New.
2148         (aarch64_expand_prologue): Use it.
2149         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
2150         (aarch64_sub_sp): Add emit_move_imm optional param.
2152 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
2154         PR tree-optimization/87261
2155         * match.pd: Remove trailing whitespace.
2156         Add (x & y) | ~(x | y) -> ~(x ^ y),
2157         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
2159 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2161         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
2162         constraints.
2163         (ashlsi3_insn): Update instruction constraints.
2164         (ashrsi3_insn): Likewise.
2165         (rotrsi3): Likewise.
2166         (add_shift): Likewise.
2167         * config/arc/constraints.md (Csz): New 32 bit constraint. It
2168         avoids placing in the limm field small constants which, otherwise,
2169         could end into a small instruction.
2171 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2173         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
2174         destination register is not odd-even.
2175         (umaddsidi4_split): Likewise.
2177 2018-10-01  Richard Biener  <rguenther@suse.de>
2179         * tree-inline.c (expand_call_inline): Store origin of fn
2180         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
2181         * tree.c (block_ultimate_origin): Simplify and do some
2182         checking.
2184 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
2186         * config/i386/mmx.md (EMMS): New int iterator.
2187         (emms): New int attribute.
2188         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
2189         EMMS int iterator.  Explicitly declare clobbers.
2190         (mmx_emms): Remove expander.
2191         (mmx_femms): Ditto.
2192         * config/i386/predicates.md (emms_operation): Remove predicate.
2193         (vzeroall_pattern): New predicate.
2194         (vzeroupper_pattern): Rename from vzeroupper_operation.
2195         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
2196         vzeroupper_pattern and vzeroall_pattern predicates.
2198 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
2200         PR rtl-optimization/86939
2201         * ira-lives.c (make_hard_regno_born): Rename from this...
2202         (make_hard_regno_live): ... to this.  Remove update to conflict
2203         information.  Update function comment.
2204         (make_hard_regno_dead): Add conflict information update.  Update
2205         function comment.
2206         (make_object_born): Rename from this...
2207         (make_object_live): ... to this.  Remove update to conflict information.
2208         Update function comment.
2209         (make_object_dead):  Add conflict information update.  Update function
2210         comment.
2211         (mark_pseudo_regno_live): Call make_object_live.
2212         (mark_pseudo_regno_subword_live): Likewise.
2213         (mark_hard_reg_dead): Update function comment.
2214         (mark_hard_reg_live): Call make_hard_regno_live.
2215         (process_bb_node_lives): Likewise.
2216         * lra-lives.c (make_hard_regno_born): Rename from this...
2217         (make_hard_regno_live): ... to this.  Remove update to conflict
2218         information.  Remove now uneeded check_pic_pseudo_p argument.
2219         Update function comment.
2220         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
2221         to conflict information.  Update function comment.
2222         (mark_pseudo_live): Remove update to conflict information.  Update
2223         function comment.
2224         (mark_pseudo_dead): Add conflict information update.
2225         (mark_regno_live): Call make_hard_regno_live.
2226         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
2227         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
2229 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
2231         PR target/87370
2232         * config/i386/i386.c (construct_container): Use TImode for
2233         BLKmode values in 2 integer registers.
2235 2018-09-29  Jeff Law  <law@redhat.com>
2237         * builtins.c (unterminated_array): Pass in c_strlen_data * to
2238         c_strlen rather than just a tree *.
2239         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2240         Update recursive calls appropriately.  If caller did not provide a
2241         suitable data pointer, create a local one.  When a non-terminated
2242         string is discovered, bubble up information about the string via the
2243         c_strlen_data object.
2244         * builtins.h (c_strlen): Update prototype.
2245         (c_strlen_data): New structure.
2246         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2247         For a type 2 call, if c_strlen indicates a non-terminated string
2248         use the length of the non-terminated string.
2249         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2251 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
2253         PR target/87467
2254         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2255         __m512d type for __A argument rather than __m512.
2257 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
2259         * match.pd (simple_comparison): Don't optimize if either operand is
2260         a function pointer when target needs function pointer canonicalization.
2262 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2264         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2265         power5 .. power9 to remove indirection.
2266         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2267         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2268         ASM_CPU_476_SPEC): Delete.
2269         (ASM_CPU_SPEC): Adjust.
2270         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2271         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2273 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2275         * config.in: Delete HAVE_AS_DCI.
2276         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2277         * config/rs6000/rs6000.h: Ditto.
2278         * configure.ac: Delete HAVE_AS_DCI.
2279         * configure: Regenerate.
2281 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2283         * config.in (HAVE_AS_LWSYNC): Delete.
2284         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2285         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
2286         do it as a .long .
2287         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2288         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
2289         as a .long .
2290         * configure.ac: Delete HAVE_AS_LWSYNC.
2291         * configure: Regenerate.
2293 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2294             Pierre-Marie de Rodat  <derodat@adacore.com>
2296         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2297         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2298         (cgraph_node::create_thunk): Add indirect_offset parameter.
2299         (thunk_adjust): Likewise.
2300         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2301         and initialize the corresponding field with it.
2302         (cgraph_node::dump): Dump indirect_offset field.
2303         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2304         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2305         (thunk_adjust): Add indirect_offset parameter and deal with it.
2306         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2307         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
2308         or if the thunk is external or local.  Fix formatting.  Do not chain
2309         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
2310         if any, in the GIMPLE representation.
2311         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2312         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2313         (input_node): Read indirect_offset field.
2314         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2315         call to thunk_adjust.
2316         * tree-nested.c (struct nesting_info): Add thunk_p field.
2317         (create_nesting_tree): Set it.
2318         (convert_all_function_calls): Copy static chain from targets to thunks.
2319         (finalize_nesting_tree_1): Return early for thunks.
2320         (unnest_nesting_tree_1): Do not finalize thunks.
2321         (gimplify_all_functions): Do not gimplify thunks.
2323 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2325         * opt-suggestions.c (option_proposer::build_option_suggestions):
2326         Release "option_values".
2328 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2330         * coverage.c (get_coverage_counts): Convert problem-reporting dump
2331         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2332         * dumpfile.c (kind_as_string): New function.
2333         (dump_loc): Rather than a hardcoded prefix of "note: ", use
2334         kind_as_string to vary the prefix based on dump_kind.
2335         (selftest::test_capture_of_dump_calls): Update for above.
2337 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2339         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2340         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
2342 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2344         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
2345         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
2346         INVALID_REGNUM instead of FPSR_REG.
2347         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
2348         * config/i386/i386.md: Update comment of FP compares.
2349         (fldenv): Do not clobber FPSR_REG.
2351 2018-09-28  Richard Biener  <rguenther@suse.de>
2353         * tree.h (BLOCK_ORIGIN): New.
2354         * omp-expand.c (grid_expand_target_grid_body): Assign
2355         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
2356         * tree-inline.c (remap_block): Likewise.
2357         * auto-profile.c (get_function_decl_from_block): Simplify
2358         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
2359         * langhooks.c (lhd_print_error_function): Likewise.
2360         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
2361         Likewise.
2362         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2363         * tree.c (block_nonartificial_location): Likewise.
2364         (block_ultimate_origin): Likewise.
2365         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
2366         no longer needed LTO case.
2368 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
2369             Jan Hubicka  <jh@suse.cz>
2370             Martin Jambor  <mjambor@suse.cz>
2372         * simplify-rtx.c (simplify_merge_mask): New function.
2373         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
2374         same masks are used in op1 or op2.
2375         (test_vec_merge): New function.
2376         (test_vector_ops): Call test_vec_merge.
2378 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2380         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
2381         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
2382         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
2383         (cypress_costs): Set it.
2384         (supersparc_costs): Likewise.
2385         (hypersparc_costs): Likewise.
2386         (leon_cost): Likewise.
2387         (leon3_costs): Likewise.
2388         (sparclet_costs): Likewise.
2389         (ultrasparc_costs): Likewise.
2390         (ultrasparc_costs): Likewise.
2391         (niagara_costs): Likewise.
2392         (niagara2_costs): Likewise.
2393         (niagara3_costs): Likewise.
2394         (niagara4_costs): Likewise.
2395         (niagara7_costs): Likewise.
2396         (m8_costs): Likewise.
2397         (TARGET_CAN_FOLLOW_JUMP): Define.
2398         (pass_work_around_errata::gate): Minor tweak.
2399         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
2400         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
2401         Automaitcally clear MASK_FSMULD mask for V7 processors.
2402         (sparc_can_follow_jump): New static function.
2403         (output_ubranch): Deal with CROSSING_JUMP_P.
2404         (sparc_use_sched_lookahead): Rewrite using switch statement.
2405         (sparc_issue_rate): Reorder.
2406         (sparc_branch_cost): New function.
2408 2018-09-27  Martin Sebor  <msebor@redhat.com>
2410         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
2411         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
2412         (int_fits_type_p): Same.
2414 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
2416         * config/i386/i386.md (FPCR_REG): Remove.
2417         (UNSPEC_FLDCW): Remove.
2418         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
2419         (x86_fldcw_1): Remove insn pattern.
2420         (fnstenv): Do not clobber FPCR_REG.
2421         (fldenv): Ditto.
2422         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
2423         (CALL_USED_REGISTERS): Ditto.
2424         (REG_ALLOC_ORDER): Ditto.
2425         (REG_CLASS_CONTENTS): Ditto.
2426         (HI_REGISTER_NAMES): Ditto.
2427         (ADDITIONAL_REGISTER_NAMES): Use defines instead
2428         of numerical constants.
2429         * config/i386/i386.c (regclass_map): Remove fpsr register.
2430         (dbx_register_map): Ditto.
2431         (dbx64_register_map): Ditto.
2432         (svr4_dbx_register_map): Ditto.
2433         (print_reg): Do not handle FPCR_REG.
2435 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
2437         PR target/87149
2438         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
2439         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
2440         Delete, always treat as true.
2441         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2442         Ditto.  Simplify remaining code.
2443         * config/powerpcspe/powerpcspe.h: Ditto.
2444         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
2445         Simplify remaining code.
2446         (rs6000_expand_builtin): Ditto.
2447         * config/rs6000/rs6000.h: Ditto.
2448         * configure.ac: Ditto.
2449         * configure: Regenerate.
2451 2018-09-27  Martin Liska  <mliska@suse.cz>
2453         * coverage.c (get_coverage_counts): Revert the formatting
2454         of missing profile opt info.
2456 2018-09-27  Richard Biener  <rguenther@suse.de>
2458         PR debug/37801
2459         PR debug/87440
2460         * dwarf2out.c (set_block_origin_self): Do not mark outermost
2461         block as we do not output that.
2462         (gen_inlined_subroutine_die): Elide the originally outermost
2463         block, matching what we do for concrete instances.
2464         (decls_for_scope): Add parameter specifying whether to recurse
2465         to subblocks.
2467 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
2468             Tom de Vries  <tom@codesourcery.com>
2470         PR 82089
2472         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
2473         STORE_FLAG_VALUE == 1.
2475 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
2477         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
2478         constant definitions.
2479         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
2480         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
2481         ("speculation_barrier"): New expander definition.
2483 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
2485         PR gcov-profile/86957
2486         * common.opt: New warning option -Wmissing-profile.
2487         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
2488         * doc/invoke.texi: Document -Wmissing-profile.
2490 2018-09-26  Jim Wilson  <jimw@sifive.com>
2492         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
2493         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
2495 2018-09-26  Martin Sebor  <msebor@redhat.com>
2497         * tree.c (zerop): Change return type to bool.
2498         (integer_zerop, integer_onep, integer_each_onep): Same.
2499         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2500         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2501         (real_onep, real_minus_onep, chain_index): Same.
2502         (print_type_hash_statistics, type_list_equal): Same.
2503         * tree.h (zerop): Same.
2504         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
2505         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2506         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2507         (real_onep, real_minus_onep, chain_index): Same.
2508         (print_type_hash_statistics, type_list_equal): Same.
2510 2018-09-26  Jim Wilson  <jimw@sifive.com>
2512         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
2514 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
2516         PR target/87414
2517         * config/i386/i386.c: Include debug.h and dwarf2out.h.
2518         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2519         call.
2521 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
2523         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
2525 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2527         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
2528         and FP_SECOND_SSE_REGS.
2529         (REG_CLASS_NAMES): Ditto.
2530         (REG_CLASS_CONTENTS): Ditto.
2531         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
2532         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
2533         (ix86_preferred_output_reload_class): Ditto.
2534         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
2535         clobber constraint to "=&f".
2536         (fix_truncdi_i387): Ditto.
2537         (lrintxfdi2): Ditto.
2538         (fistdi2_<rounding>): Ditto.
2539         (fpremxf4_i387): Change "=u" constraint to "=f".
2540         (fprem1xf4_i387): Ditto.
2541         (sincosxf3): Ditto.
2542         (fptanxf4_i387): Ditto.
2543         (fxtractxf3_i387): Ditto.
2544         (fscalexf4_i387): Ditto.
2545         (atan2xf3): Change "u" constraint to "f".
2546         (fyl2xxf3_i387): Ditto.
2547         (fyl2xp1xf3_i387): Ditto.
2549 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2551         PR target/87439
2552         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
2553         for removed I387_MASK_PM entity.
2556 2018-09-26  Jeff Law  <law@redhat.com>
2557         Revert
2558         2018-09-26  Alexey Neyman  <stilor@att.net>
2560         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2561         headers are no longer pulled in by <isl/val.h>.
2563 2018-09-26  Richard Biener  <rguenther@suse.de>
2565         PR debug/87443
2566         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
2567         or concrete instance DIE to the tree.  Create abstract origin
2568         attributes also for concrete instances.
2570 2018-09-26  Alexey Neyman  <stilor@att.net>
2572         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2573         headers are no longer pulled in by <isl/val.h>.
2575 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
2577         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
2578         Use new helper functions.
2579         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
2580         Use new helper functions.
2581         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
2582         aarch_mm_needs_release): New declarations.
2583         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
2584         aarch_mm_needs_release): New.
2586 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
2588         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
2589         (arm32_output_mi_thunk): Deal with long calls.
2591 2018-09-26  Richard Biener  <rguenther@suse.de>
2593         PR debug/87428
2594         PR debug/87362
2595         * tree-inline.c (expand_call_inline): When the location
2596         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2597         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2598         the inserted BLOCK to make inlined_function_outer_scope_p
2599         recognize it.
2600         * dwarf2out.c (add_call_src_coords_attributes): Do not add
2601         coords for reserved locations.
2603 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2605         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
2606         (*call_indirect_nonlocal_sysv<mode>): Ditto.
2607         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
2608         (*sibcall_nonlocal_sysv<mode>): Ditto.
2609         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
2610         (<bd>_<mode>): Ditto.
2611         (<bd>tf_<mode>): Ditto.
2613 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2615         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
2616         control string as a list of templates instead of as C code.
2617         (*altivec_movti): Ditto.
2618         * config/rs6000/darwin.md (movdf_low_di): Ditto.
2620 2018-09-25  Jim Wilson  <jimw@sifive.com>
2622         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
2623         when target symbol is weak.
2625 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2627         PR c/87387
2628         * builtins.c (unterminated_array): Simplify.
2629         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
2630         where pointer arithmetic is safe.
2632 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2634         PR target/86987
2635         * config/rs6000/altivec.md (altivec_vspltb): Use
2636         const_0_to_15_operand instead of u5bit_cint_operand.
2637         (*altivec_vspltb_internal): Ditto.
2638         (altivec_vspltb_direct): Ditto.
2639         (altivec_vsplth): Use const_0_to_7_operand instead of
2640         u5bit_cint_operand.
2641         (*altivec_vsplth_internal): Ditto.
2642         (altivec_vsplth_direct): Ditto.
2643         (altivec_vspltw): Use const_0_to_3_operand instead of
2644         u5bit_cint_operand.
2645         (*altivec_vspltw_internal): Ditto.
2646         (altivec_vspltw_direct): Ditto.
2647         (altivec_vspltsf): Ditto.
2648         (*altivec_vspltsf_internal): Ditto.
2649         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
2650         various splats with the proper size immediate.  Reorder the various
2651         cases by ascending size of immediate, and put all such together.
2653 2018-09-25  Richard Biener  <rguenther@suse.de>
2655         PR debug/83941
2656         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
2657         GC-ification.
2658         (maybe_create_die_with_external_ref): Do not create
2659         DW_TAG_imported_unit here.
2660         (add_abstract_origin_attribute): Handle external BLOCK refs.
2661         (dwarf2out_abstract_function): Simplify LTO case.
2662         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
2663         rather than using maybe_create_die_with_external_ref.
2665 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
2667         PR target/71278
2668         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
2669         (frndintxf2_mask_pm_i387): Ditto.
2670         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
2671         Enable for !flag_trapping_math.
2672         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
2673         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
2674         Change operand 1 predicate to nonimmediate_operand.
2675         (attr "i387_cw"): Remove mask_pm.
2676         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
2677         (enum ix86_entity): Remove I387_MASK_PM.
2678         * config/i386/i386.c (ix86_i387_mode_needed): Do not
2679         handle I387_MASK_PM.
2680         (ix86_mode_needed): Ditto.
2681         (ix86_mode_after): Ditto.
2682         (ix86_mode_entry): Ditto.
2683         (ix86_mode_exit): Ditto.
2684         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
2686 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
2688         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
2689         to_update_switch_stmts to vNULL instead of calling create on them
2690         immediately.
2692 2018-09-25  Richard Biener  <rguenther@suse.de>
2694         PR tree-optimization/87402
2695         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
2696         (visit_phi): Re-instantiate handling of supposed to be VARYING
2697         but non-VARYING backedge value.
2699 2018-09-25  Richard Biener  <rguenther@suse.de>
2701         PR debug/83941
2702         * dwarf2out.c (struct sym_off_pair): New.
2703         (external_die_map): New global.
2704         (lookup_decl_die): When in LTO create DIEs lazily from the
2705         external_die_map.
2706         (lookup_block_die): New function, create DIEs lazily in LTO.
2707         (equate_block_to_die): New function.
2708         (dwarf2out_die_ref_for_decl): During WPA get the association
2709         from the external DIE map.
2710         (dwarf2out_register_external_die): Record mapping into the
2711         external DIE map.
2712         (maybe_create_die_with_external_ref): New function split out from
2713         DIE generation part of old dwarf2out_register_external_die.
2714         (add_abstract_origin_attribute): Do not return the DIE.  When
2715         in LTO reference externals directly.
2716         (dwarf2out_abstract_function): When in LTO ignore calls for
2717         decls with external DIEs (already present abstract instances).
2718         (gen_call_site_die): Adjust.
2719         (add_high_low_attributes): Likewise.
2720         (gen_lexical_block_die): Likewise.
2721         (gen_inlined_subroutine_die): Likewie.
2722         (gen_block_die): Likewise.
2723         (dwarf2out_inline_entry): Likewise.
2724         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
2725         DIEs.
2727 2018-09-25  Martin Liska  <mliska@suse.cz>
2729         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
2730         integers and not by a float value.
2732 2018-09-25  Martin Liska  <mliska@suse.cz>
2734         PR fortran/87394
2735         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
2736         instead of NULL.
2737         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
2738         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
2739         * godump.c (go_format_type): Remove extra parenthesis.
2741 2018-09-25  Martin Liska  <mliska@suse.cz>
2743         * alias.c (set_dest_equal_p): Remove unused function.
2744         * config/i386/i386.c (def_builtin_pure2): Likewise.
2745         * diagnostic-show-locus.c (class layout): Remove
2746         unused field.
2747         (layout::layout): Likewise here.
2748         * dump-context.h (class temp_dump_context): Likewise.
2749         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
2750         (add_AT_loclistsptr): Likewise.
2751         (add_AT_offset): Likewise.
2752         (get_AT_hi_pc): Likewise.
2753         (is_comdat_die): Likewise.
2754         (type_is_enum): Likewise.
2755         (ceiling): Likewise.
2756         (add_AT_vms_delta): Likewise.
2757         (is_class_die): Likewise.
2758         * edit-context.c (class line_event): Remove unused field.
2759         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
2760         unused function.
2761         * ipa-cp.c (ipa_get_vr_lat): Likewise.
2762         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
2763         (ok_for_base_p_nonstrict): Likewise.
2764         * tree-chrec.c (is_not_constant_evolution): Likewise.
2765         (chrec_fold_poly_cst): Likewise.
2766         * tree-if-conv.c (has_pred_critical_p): Likewise.
2767         * tree-ssa-coalesce.c (print_exprs): Likewise.
2768         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
2769         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
2770         * tree-vrp.c (value_ranges_intersect_p): Likewise.
2771         (value_range_nonnegative_p): Likewise.
2773 2018-09-25  Martin Liska  <mliska@suse.cz>
2775         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2776         Do not handle "GNU Pascal".
2777         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2778         Likewise.
2779         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
2780         from documentation. Likewise.
2781         * dbxout.c (dbxout_range_type): Likewise.
2782         * doc/cpp.texi: Likewise.
2783         * doc/extend.texi: Likewise.
2784         * doc/frontends.texi: Likewise.
2785         * doc/invoke.texi: Remove Pascal entry.
2786         * tree.def (CLEANUP_POINT_EXPR): Likewise.
2787         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
2789 2018-09-25  Martin Liska  <mliska@suse.cz>
2791         PR middle-end/86078
2792         * doc/invoke.texi: Document all parameters and remove default
2793         of the parameters.
2795 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2797         PR bootstrap/87417
2798         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
2799         contains HOST_WIDE_INTs when computing its size.
2801 2018-09-24  Jim Wilson  <jimw@sifive.com>
2803         PR target/87391
2804         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
2805         not TARGET_RVE.
2806         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
2808 2018-09-24  Andrew Pinski  <apinski@marvell.com>
2810         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2811         access prev before checking it for NULLness in the
2812         AARCH64_FUSE_CMP_BRANCH case.
2814 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
2816         PR target/82699
2817         * config/i386/i386.c (rest_of_insert_endbranch): Set
2818         endbr_queued_at_entrance to true and don't insert ENDBR if
2819         x86_function_profiler will be called.
2820         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2821         is true.
2822         * config/i386/i386.h (machine_function): Add
2823         endbr_queued_at_entrance.
2825 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2827         * genattrtab.c (mk_attr_alt): Use alternative_mask.
2828         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
2829         types.
2830         (check_attr_test): Use alternative_mask.
2831         (get_attr_value): Likewise.
2832         (compute_alternative_mask): Use alternative_mask and XWINT.
2833         (make_alternative_compare): Use alternative_mask.
2834         (attr_alt_subset_p): Use XWINT.
2835         (attr_alt_subset_of_compl_p): Likewise.
2836         (attr_alt_intersection): Use alternative_mask and XWINT.
2837         (attr_alt_union): Likewise.
2838         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
2839         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
2840         (simplify_test_exp): Use alternative_mask and XWINT.
2841         (write_test_expr): Use alternative_mask and XWINT, adjust bit
2842         number calculation to support 64 bits.  Generate code that
2843         checks 64-bit masks.
2844         (main): Use alternative_mask.
2845         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
2847 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2849         PR target/80080
2850         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
2851         RETURN+USE when returning via %r14.
2853 2018-09-24  Martin Liska  <mliska@suse.cz>
2855         * gcov.c (output_lines): Print colorization legend
2856         for both flag_use_colors and flag_use_hotness_colors.
2857         Reword the help.
2859 2018-09-24  Martin Liska  <mliska@suse.cz>
2861         * coverage.c (get_coverage_counts): Use warning_at
2862         with current_function_decl location. Use %qD in warning
2863         message.
2865 2018-09-24  Martin Liska  <mliska@suse.cz>
2867         * memory-block.h (memory_block_pool::release): Annotate with
2868         valgrind that the memory is not accessible.
2870 2018-09-24  Martin Liska  <mliska@suse.cz>
2872         PR sanitizer/85774
2873         * asan.c: Make asan_handled_variables extern.
2874         * asan.h: Likewise.
2875         * cfgexpand.c (expand_stack_vars): Make sure
2876         a representative is unpoison if another
2877         variable in the partition is handled by
2878         use-after-scope sanitization.
2880 2018-09-24  Richard Biener  <rguenther@suse.de>
2882         PR tree-optimization/63155
2883         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2884         the worklist when the edge of the respective argument isn't
2885         executable.
2887 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2889         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
2890         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
2891         (MASK_CLASS_P): Update for rename.
2892         (MAYBE_MASK_CLASS_P): Ditto.
2893         (REG_CLASS_NAMES): Update.
2894         (REG_CLASS_CONTENT): Update.
2895         * config/i386/i386.c (regclass_map): Update for MASK_REG
2896         and ALL_MASK_REGS rename.
2897         * config/i386/constraints.md (Yk): Update for rename.
2898         (k): Ditto.
2900 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2902         * config/i386/i386.h (enum reg_class): Remove
2903         EVEX_SSE_REGS and MOD4_SSE_REGS.
2904         (REG_CLASS_NAMES): Update.
2905         (REG_CLASS_CONTENT): Update.
2906         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
2907         registers as ALL_SSE_REGS.
2908         (ix86_additional_allocno_class_p): Remove.
2909         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
2910         (ix86_register_priority): Lower priority of EVEX SSE registers.
2911         Use IN_RANGE macro where appropriate.
2912         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
2913         AVX-5124VNNIW checks.
2914         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
2915         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
2916         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
2917         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
2918         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
2919         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
2920         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
2921         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
2922         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
2923         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
2924         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
2925         * config/i386/constraints.md (Yh): Remove.
2927 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2929         * config/i386/i386.c (regclass_map): Declare integer REX registers
2930         as GENERAL_REGS.
2932 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
2934         * doc/service.texi (Service): Switch the fsf.org link to https.
2936 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2938         PR target/86798
2939         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2940         Define to speculation_safe_value_not_needed.
2942 2018-09-21  Florian Weimer  <fweimer@redhat.com>
2944         PR middle-end/81035
2945         * doc/extend.texi (Common Function Attributes): Mention that
2946         noreturn suppresses tail call optimization.
2948 2018-09-21  Jeff Law  <law@redhat.com>
2950         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
2951         vr_values::cleanup_edges_and_switches.
2952         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
2953         vr_values class.
2954         (identify_jump_threads): Remove EDGE_IGNORE handling.
2955         (execute_vrp): Move handling of to_remove_edges and
2956         to_update_switch_stmts into vr_values class member functions.
2957         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
2958         (to_update_switch_stmts): Likewise.
2959         * vr-values.c: Include cfghooks.h.
2960         (vr_values::vr_values): Initialize to_remove_edges and
2961         to_update_switch_stmts.
2962         (vr_values::~vr_values): Verify to_remove_edges and
2963         to_update_switch_stmts are empty.
2964         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
2965         (vr_values::cleanup_edges_and_switches): New member function.
2966         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
2967         function.  Add new data members.
2969 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
2971         PR tree-optimization/87309
2972         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
2973         calls with pflags and alt_flags.
2974         (selftest::test_capture_of_dump_calls): Add test of interaction of
2975         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
2977 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2979         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
2981 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2983         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
2985 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2987         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
2988         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
2989         (VXWORKS_LIBS_RTP): Minor reordering.
2991 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2993         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
2994         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
2996 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2998         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
2999         (PTRDIFF_TYPE): Likewise.
3001 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3003         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
3004         triplet, similar to support for VxWorks7.
3005         * config/vxworks-dummy.h: Provide a default definition
3006         of TARGET_VXWORKS64 to 0.
3008 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3010         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
3011         * config/vxworks-dummy.h: here.
3013 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3015         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
3017 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
3018             Bo Zhou  <zbo.zhou@hisilicon.com>
3020         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
3021         * config/aarch64/aarch64-tune.md: Regenerated.
3022         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
3023         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
3024         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
3026 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
3027             Julian Brown  <julian@codesourcery.com>
3029         * builtins.c (get_builtin_sync_mem): Handle address spaces.
3031 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3033         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
3034         if the call takes a static chain.
3036 2018-09-21  Martin Liska  <mliska@suse.cz>
3038         * auto-profile.c (autofdo_source_profile::read): Do not
3039         set sum_all.
3040         (read_profile): Do not add working sets.
3041         (read_autofdo_file): Remove sum_all.
3042         (afdo_callsite_hot_enough_for_early_inline): Remove const
3043         qualifier.
3044         * coverage.c (struct counts_entry): Remove gcov_summary.
3045         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
3046         do not support GCOV_TAG_PROGRAM_SUMMARY.
3047         (get_coverage_counts): Remove summary and expected
3048         arguments.
3049         * coverage.h (get_coverage_counts): Likewise.
3050         * doc/gcov-dump.texi: Remove -w option.
3051         * gcov-dump.c (dump_working_sets): Remove.
3052         (main): Do not support '-w' option.
3053         (print_usage): Likewise.
3054         (tag_summary): Likewise.
3055         * gcov-io.c (gcov_write_summary): Do not dump
3056         histogram.
3057         (gcov_read_summary): Likewise.
3058         (gcov_histo_index): Remove.
3059         (gcov_histogram_merge): Likewise.
3060         (compute_working_sets): Likewise.
3061         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
3062         it not obsolete.
3063         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
3064         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
3065         (GCOV_HISTOGRAM_SIZE): Remove.
3066         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
3067         (struct gcov_summary): Simplify rapidly just
3068         to runs and sum_max fields.
3069         (gcov_histo_index): Remove.
3070         (NUM_GCOV_WORKING_SETS): Likewise.
3071         (compute_working_sets): Likewise.
3072         * gcov-tool.c (print_overlap_usage_message): Remove
3073         trailing empty line.
3074         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
3075         (output_lines): Remove program related line.
3076         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
3077         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
3078         histogram.
3079         (input_profile_summary): Do not read it.
3080         (merge_profile_summaries): And do not merge it.
3081         (input_symtab): Do not call removed function.
3082         * modulo-sched.c (sms_schedule): Do not print sum_max.
3083         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
3084         removed when histogram method was invented.
3085         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
3086         mode.
3087         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
3088         GCOV coding style.
3089         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
3090         and dump selected value.
3091         * profile.c (add_working_set): Remove.
3092         (get_working_sets): Likewise.
3093         (find_working_set): Likewise.
3094         (get_exec_counts): Do not work with working sets.
3095         (read_profile_edge_counts): Do not inform as sum_max is removed.
3096         (compute_branch_probabilities): Likewise.
3097         (compute_value_histograms): Remove argument for call of
3098         get_coverage_counts.
3099         * profile.h: Do not make gcov_summary const.
3101 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
3103         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
3105 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3107         PR tree-optimization/86990
3108         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
3109         Check that the entire merged store group is made of constants only for
3110         overlapping stores.
3112 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
3114         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
3115         (VTABLE_VERIFICATION_SPEC): Likewise.
3116         (SANITIZER_EARLY_SPEC): Likewise.
3117         (SANITIZER_SPEC): Likewise.
3118         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
3119         * doc/invoke.texi (Link Options): Document -r.
3121 2018-09-20  Richard Biener <rguenther@suse.de>
3123         PR middle-end/87054
3124         * gimplify.c (gimplify_expr): Retain alignment of
3125         addressable lvalue in dereference.
3127 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
3129         PR bootstrap/87013
3130         * configure.ac: Check for .loc is_stmt support.
3131         * configure, config.in: Rebuilt.
3132         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
3133         if not supported.
3135 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3137         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
3138         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
3139         -misel=no.
3141 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3143         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3144         VECTOR_OTHER.
3145         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
3146         case VECTOR_OTHER.
3148 2018-09-20  Marek Polacek  <polacek@redhat.com>
3150         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
3152 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3154         PR tree-optimization/87288
3155         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
3156         into account when determining PEELING_FOR_NITERS.
3158 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3160         PR tree-optimization/86877
3161         * tree-vect-loop.c (vect_analyze_loop_2): Call
3162         vect_verify_datarefs_alignment.
3164 2018-09-19  Marek Polacek  <polacek@redhat.com>
3166         * doc/invoke.texi: Document -Wclass-conversion.
3168 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
3170         * config/pa/pa.c (pa_adjust_priority): Delete.
3171         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
3173         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3174         (atomic_storehi): Likewise.
3175         (atomic_storesi): Likewise.
3176         (atomic_loaddi): Restore compare and swap exchange loop code.
3178 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
3180         PR rtl-optimization/86902
3181         * combine.c (try_combine): When changing the CC mode used, don't change
3182         an unrelated mode in other_insn to that new CC mode.
3184 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
3186         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
3187         with %T in place of calls to dump_generic_expr.
3188         (prune_runtime_alias_test_list): Likewise.
3189         (create_runtime_alias_checks): Likewise.
3190         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
3191         (vect_analyze_data_ref_dependence): Likewise.
3192         (vect_slp_analyze_data_ref_dependence): Likewise.
3193         (vect_record_base_alignment): Likewise.  Use %G in place of call
3194         to dump_gimple_stmt.
3195         (vect_compute_data_ref_alignment): Likewise.
3196         (verify_data_ref_alignment): Likewise.
3197         (vect_find_same_alignment_drs): Likewise.
3198         (vect_analyze_group_access_1): Likewise.
3199         (vect_analyze_data_ref_accesses): Likewise.
3200         (dependence_distance_ge_vf): Likewise.
3201         (dump_lower_bound): Likewise.
3202         (vect_prune_runtime_alias_test_list): Likewise.
3203         (vect_find_stmt_data_reference): Likewise.
3204         (vect_analyze_data_refs): Likewise.
3205         (vect_create_addr_base_for_vector_ref): Likewise.
3206         (vect_create_data_ref_ptr): Likewise.
3207         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
3208         (vect_can_advance_ivs_p): Likewise.
3209         (vect_update_ivs_after_vectorizer): Likewise.
3210         (vect_gen_prolog_loop_niters): Likewise.
3211         (vect_prepare_for_masked_peels): Likewise.
3212         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3213         (vect_determine_vectorization_factor): Likewise.
3214         (vect_is_simple_iv_evolution): Likewise.
3215         (vect_analyze_scalar_cycles_1): Likewise.
3216         (vect_analyze_loop_operations): Likewise.
3217         (report_vect_op): Likewise.
3218         (vect_is_slp_reduction): Likewise.
3219         (check_reduction_path): Likewise.
3220         (vect_is_simple_reduction): Likewise.
3221         (vect_create_epilog_for_reduction): Likewise.
3222         (vect_finalize_reduction:): Likewise.
3223         (vectorizable_induction): Likewise.
3224         (vect_transform_loop_stmt): Likewise.
3225         (vect_transform_loop): Likewise.
3226         (optimize_mask_stores): Likewise.
3227         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
3228         (vect_split_statement): Likewise.
3229         (vect_recog_over_widening_pattern): Likewise.
3230         (vect_recog_average_pattern): Likewise.
3231         (vect_determine_min_output_precision_1): Likewise.
3232         (vect_determine_precisions_from_range): Likewise.
3233         (vect_determine_precisions_from_users): Likewise.
3234         (vect_mark_pattern_stmts): Likewise.
3235         (vect_pattern_recog_1): Likewise.
3236         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3237         (vect_record_max_nunits): Likewise.
3238         (vect_build_slp_tree_1): Likewise.
3239         (vect_build_slp_tree_2): Likewise.
3240         (vect_print_slp_tree): Likewise.
3241         (vect_analyze_slp_instance): Likewise.
3242         (vect_detect_hybrid_slp_stmts): Likewise.
3243         (vect_detect_hybrid_slp_1): Likewise.
3244         (vect_slp_analyze_operations): Likewise.
3245         (vect_slp_analyze_bb_1): Likewise.
3246         (vect_transform_slp_perm_load): Likewise.
3247         (vect_schedule_slp_instance): Likewise.
3248         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3249         (vect_mark_stmts_to_be_vectorized): Likewise.
3250         (vect_init_vector_1): Likewise.
3251         (vect_get_vec_def_for_operand): Likewise.
3252         (vect_finish_stmt_generation_1): Likewise.
3253         (vect_check_load_store_mask): Likewise.
3254         (vectorizable_call): Likewise.
3255         (vectorizable_conversion): Likewise.
3256         (vectorizable_operation): Likewise.
3257         (vectorizable_load): Likewise.
3258         (vect_analyze_stmt): Likewise.
3259         (vect_is_simple_use): Likewise.
3260         (vect_get_vector_types_for_stmt): Likewise.
3261         (vect_get_mask_type_for_stmt): Likewise.
3262         * tree-vectorizer.c (increase_alignment): Likewise.
3264 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
3266         * doc/rtl.texi: Adjust vec_select description.
3267         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3268         non-constant selectors.
3270 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
3272         * config/aarch64/aarch64-protos.h
3273         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3274         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3275         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3276         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3277         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3278         (AARCH64_FL_PROFILE): Move index so flags are ordered.
3279         (AARCH64_ISA_RCPC8_4): New flag.
3280         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
3281         to aarch64_offset_9bit_signed_unscaled_p.
3282         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
3283         and use stlur.
3284         * config/aarch64/constraints.md (Ust): New constraint.
3285         * config/aarch64/predicates.md.
3286         (aarch64_9bit_offset_memory_operand): New predicate.
3287         (aarch64_rcpc_memory_operand): New predicate.
3289 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
3291         PR rtl-optimization/87361
3292         * rtlanal.c (nonzero_bits1): Revert accidental change.
3294 2018-09-19  Richard Biener  <rguenther@suse.de>
3296         PR tree-optimization/87349
3297         PR tree-optimization/87342
3298         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3300 2018-09-18  Marek Polacek  <polacek@redhat.com>
3302         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3303         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3305 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3307         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3309 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3311         PR rtl-optimization/86882
3312         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3314 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
3316         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3317         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3319 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
3321         PR other/87353
3322         * doc/invoke.texi (Link Options): Fix formatting and grammar.
3324 2018-09-18  Richard Biener  <rguenther@suse.de>
3326         PR middle-end/63155
3327         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3328         (compute_samebase_partition_bases): Likewise.
3329         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3330         (gimple_can_coalesce_p): Simplify.
3332 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
3334         Handle a library implementation of ffs calling __builtin_ffs.
3335         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3336         (mmix_init_libfuncs): New function: make __builtin_ffs expand
3337         to __ffsdi2.
3339 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3341         * diagnostic-show-locus.c (class layout_range): Add field
3342         "m_original_idx".
3343         (layout_range::layout_range): Add "original_idx" param and use it
3344         to initialize new field.
3345         (make_range): Use 0 for original_idx.
3346         (layout::layout): Pass in index to calls to
3347         maybe_add_location_range.
3348         (layout::maybe_add_location_range): Add param "original_idx" and
3349         pass it on to layout_range.
3350         (layout::print_any_labels): Pass on range->m_original_idx to
3351         get_text call.
3352         (gcc_rich_location::add_location_if_nearby): Use 0 for
3353         original_idx.
3354         * gcc-rich-location.h (text_range_label::get_text): Update for new
3355         param.
3356         (range_label_for_type_mismatch::get_text): Likewise.
3358 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3360         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
3362 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3364         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
3365         format_string_diagnostic_t.
3366         (fmtwarn_n): Likewise.
3367         * substring-locations.c
3368         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
3369         (format_warning_n_va): Convert to...
3370         (format_string_diagnostic_t::emit_warning_n_va): ...this.
3371         (format_warning_va): Convert to...
3372         (format_string_diagnostic_t::emit_warning_va): ...this.
3373         (format_warning_at_substring): Convert to...
3374         (format_string_diagnostic_t::emit_warning): ...this.
3375         (format_warning_at_substring_n): Convert to...
3376         (format_string_diagnostic_t::emit_warning_n): ...this.
3377         * substring-locations.h (class format_string_diagnostic_t): New
3378         class.
3379         (format_warning_va): Convert to
3380         format_string_diagnostic_t::emit_warning_va.
3381         (format_warning_n_va): Convert to
3382         format_string_diagnostic_t::emit_warning_n_va.
3383         (format_warning_at_substring): Convert to
3384         format_string_diagnostic_t::emit_warning.
3385         (format_warning_at_substring_n): Convert to
3386         format_string_diagnostic_t::emit_warning_n.
3388 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
3389             Bernd Schmidt <bernds_cb1@t-online.de>
3391         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
3392         SImode args.
3394 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3396         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
3397         operand 0 predicate to nonimmediate operand.
3398         (rint<mode>2_frndint): Remove insn pattern.
3399         (rint<mode>2): Change operand 1 predicate to general_operand.
3400         Extend operand 1 to XFmode and generate rintxf2 insn.
3401         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
3402         Do not use X87MODEF mode macro.
3403         (frndintxf2_<rounding>_i387): Rename from
3404         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
3405         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
3406         to XFmode and generate significandxf3 insn.
3408 2018-09-17  Richard Biener  <rguenther@suse.de>
3410         PR tree-optimization/87328
3411         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
3412         visiting unexecutable backedges when not iterating.
3413         (do_rpo_vn): Mark all edges not executable even when not
3414         iterating.
3416 2018-09-17  Martin Jambor  <mjambor@suse.cz>
3418         PR c/63886
3419         * doc/invoke.texi (Warning Options): Likewise.
3421 2018-09-17  Richard Biener  <rguenther@suse.de>
3423         PR tree-optimization/87301
3424         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
3425         clean EH info from leftover copy assignments.
3427 2018-09-17  Martin Liska  <mliska@suse.cz>
3429         PR gcov-profile/85871
3430         * gcov.c (output_intermediate_file): Fix out of bounds
3431         access.
3433 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
3435         * config/arc/arc.c: Object attributes for core4 not reflected
3436         correctly.
3437         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
3438         core3).
3440 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
3442         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
3444 2018-09-17  Martin Liska  <mliska@suse.cz>
3446         * doc/gcov.texi: Document new option --use-hotness-colors.
3447         * gcov.c (struct source_info): Declare new field.
3448         (source_info::source_info): Set default for maximum_count.
3449         (print_usage): Add new -q option.
3450         (process_args): Process it.
3451         (accumulate_line_info): Save src->maximum_count.
3452         (output_line_beginning): Make color line number if
3453         flag_use_hotness_colors is set.
3454         (output_line_details): Pass default argument value.
3455         (output_lines): Pass src->maximum_count.
3457 2018-09-17  Martin Liska  <mliska@suse.cz>
3459         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
3460         Use processor_names table.
3461         * config/i386/i386.c (ix86_default_align): Use
3462         processor_cost_table for alignment values.
3463         (ix86_option_override_internal): Use processor_names.
3464         (ix86_function_specific_print): Likewise.
3465         * config/i386/i386.h (struct processor_costs):
3466         Add alignment values.
3467         (struct ptt): Remove and replace with const char *.
3468         * config/i386/x86-tune-costs.h (struct processor_costs):
3469         Declare default alignments for all costs.
3471 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
3473         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
3474         symbolics or VR_VARYING ranges for ABS_EXPR.
3475         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
3476         when range will wrap.
3478 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
3480         PR middle-end/86864
3481         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
3482         before and after a JUMP_TABLE_DATA.
3484 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
3486         PR middle-end/87188
3487         * dojump.c (do_compare_and_jump): Canonicalize function pointers
3488         when one operand is a function pointer.  Use POINTER_TYPE_P and
3489         FUNC_OR_METHOD_TYPE_P.
3490         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3491         * fold-const.c (build_range_check): Likewise.
3492         * match.pd (simple_comparison): Likewise.
3494 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
3496         PR c/82967
3497         * spellcheck.c (get_edit_distance_cutoff): New function.
3498         (selftest::test_edit_distance_unit_test_oneway): Rename to...
3499         (selftest::test_get_edit_distance_one_way): ...this.
3500         (selftest::test_get_edit_distance_unit): Rename to...
3501         (selftest::test_get_edit_distance_both_ways): ...this.
3502         (selftest::test_edit_distances): Move tests to this new function,
3503         and test some more pairs of strings.  Update for above renaming.
3504         (selftest::get_old_cutoff): New function.
3505         (selftest::test_get_edit_distance_cutoff): New function.
3506         (selftest::assert_suggested_for): New function.
3507         (ASSERT_SUGGESTED_FOR): New macro.
3508         (selftest::assert_not_suggested_for): New function.
3509         (ASSERT_NOT_SUGGESTED_FOR): New macro.
3510         (selftest::test_suggestions): New function.
3511         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
3512         tests to selftest::test_edit_distances and call it.  Add calls to
3513         selftest::test_get_edit_distance_cutoff and
3514         selftest::test_suggestions.
3515         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
3516         (best_match::consider): Replace hard-coded cutoff calculation with
3517         a call to...
3518         (best_match::get_cutoff): New declaration.
3519         (best_match::get_best_meaningful_candidate): Likewise.
3521 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3523         * builtins.c (fold_builtin_strlen): Remove TODO comment.
3525 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3527         revert:
3528         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3530         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3531         terminated string literal.
3533 2018-09-14  Martin Sebor  <msebor@redhat.com>
3535         * builtins.c (unterminated_array): Handle ARRAY_REF.
3536         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
3537         * builtins.h (unterminated_array): Declare extern.
3538         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
3539         arrays.
3540         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
3541         calls.
3543 2018-09-14  Martin Sebor  <msebor@redhat.com>
3544             Jeff Law  <law@redhat.com>
3546         * builtins.c (unterminated_array): New.
3547         (expand_builtin_strcpy): Adjust.
3548         (expand_builtin_strcpy_args): Detect unterminated arrays.
3549         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
3550         unterminated arrays.
3551         * gimple-fold.h (get_maxval_strlen): Add argument.
3552         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
3554         * gimple-fold.c (get_range_strlen): Add argument.
3555         (get_maxval_strlen): Adjust.
3556         * gimple-fold.h (get_range_strlen): Add argument.
3558 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
3560         * config/i386/movdirintrin.h: Fix copyright year.
3562 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
3564         * reg-stack.c: Include regs.h.
3565         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
3566         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
3567         FIRST_STACK_REG, not DFmode.
3568         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
3569         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
3570         (change stack): Default register mode to the reg_raw_mode of
3571         FIRST_STACK_REG, not DFmode.
3572         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
3573         (*swapxf): Rename from swapxf.
3575 2018-09-14  Carl Love  <cel@us.ibm.com>
3577         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
3578         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
3580 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
3582         PR target/87224
3583         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
3584         alternatives.
3586 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
3588         PR target/85628
3589         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
3591 2018-09-14  Jason Merrill  <jason@redhat.com>
3593         Fix --enable-gather-detailed-mem-stats.
3594         * hash-table.c (hash_table_usage): Change from variable to function.
3595         * hash-table.h: Adjust.
3596         * Makefile.in: Add missing dependencies on hash-table.h.
3598 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3600         PR tree-optimization/87259
3601         PR lto/87283
3602         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
3603         execute_cse_reciprocals_1 has tried transforming.
3605 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
3607         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
3608         VR_VARYING for PLUS/MINUS_EXPR.
3610 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
3612         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
3613         formatting.
3615 2018-09-14  Richard Biener  <rguenther@suse.de>
3617         PR middle-end/63155
3618         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
3619         bits for the merged partition.
3621 2018-09-13  Martin Sebor  <msebor@redhat.com>
3622             Bernd Edlinger  <bernd.edlinger@hotmail.de>
3624         * builtins.h (c_srlen): Add argument.
3625         * builtins.c (warn_string_no_nul): New function.
3626         (c_strlen): Add argument and use it.  Update recursive calls.
3627         Pass DECL argument to string_constant to get info on non
3628         terminated strings.  Update *NONSTR as needed.
3629         (fold_builtin_strlen): Add argument to calls to c_strlen.
3630         Warn for unterminated arrays.
3631         (warn_string_no_null): Add prototype.
3632         * expr.c (string_constant): Update arguments.  Update recursive
3633         calls appropriately.  Detect missing NUL terminator and outermost
3634         declaration its missing in.
3635         Improve checks for arrays with nonzero lower bound or elements
3636         that are not a single byte.  Simplify offset computation.
3637         Simplify checks for non-NUL terminated strings.
3638         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
3639         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
3641 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3643         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
3644         correctly.
3645         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
3646         argument.  Fix range checks.
3647         * fold-const.h (c_getstr): Adjust protoype.
3648         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
3649         string is constant but contains no NUL byte.
3651         * expr.c (string_constant): Adjust function comment.
3652         Remove bogus check for zero termination.
3654         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
3656         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
3657         (get_constant_size): Don't make STRING_CSTs larger than they are.
3658         (check_string_literal): New check function for STRING_CSTs.
3659         (output_constant): Use it.
3661 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
3663         PR target/86812
3664         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
3666 2018-09-13  Richard Biener  <rguenther@suse.de>
3668         PR tree-optimization/87263
3669         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
3670         (struct unwind_state): Add max_rpo field.
3671         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
3672         Compute max_rpo, the max RPO number a block can be backwards reached
3673         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
3674         separating it from the iterating mode.
3676 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
3678         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
3679         (rfs_decision): New scheduling decision.
3681 2018-09-13  Richard Biener  <rguenther@suse.de>
3683         PR bootstrap/87134
3684         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
3685         (vn_nary_op_insert_pieces_predicated): Do not write useless
3686         valid_dominated_by_p entry outside of the allocated storage.
3688 2018-09-13  Omar Sandoval  <osandov@osandov.com>
3689             Tom de Vries  <tdevries@suse.de>
3691         PR debug/86985
3692         * dwarf2out.c (is_c): New function.
3693         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
3695 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
3697         PR target/85628
3698         * config/aarch64/aarch64.md (*aarch64_bfxil):
3699         Define.
3700         * config/aarch64/constraints.md (Ulc): Define.
3701         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
3702         Define.
3703         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
3704         New function.
3706 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
3708         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
3709         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
3710         aarch64_layout_frame call.
3711         (aarch64_expand_epilogue): Likewise.
3712         (aarch64_initial_elimination_offset): Likewise.
3713         (aarch64_get_separate_components): Likewise.
3714         (aarch64_use_return_insn_p): Likewise.
3715         (aarch64_layout_frame): Remove unneeded check.
3717 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3719         * configure.ac: Only append
3720         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
3721         gcc_config_arguments if it was never reconfigured or last reconfigure
3722         was with different arguments.
3723         * configure: Regenerated.
3725 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3726             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3728         PR middle-end/87290
3729         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
3730         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
3732 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3734         PR tree-optimization/87287
3735         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
3736         X % C == 0 to X % (unsigned) C == 0 optimization to ...
3737         * match.pd (X % C == 0): ... here.  New optimization.
3739 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3741         PR middle-end/82853
3742         * expr.h (maybe_optimize_mod_cmp): Declare.
3743         * expr.c (mod_inv): New function.
3744         (maybe_optimize_mod_cmp): New function.
3745         (do_store_flag): Use it.
3746         * cfgexpand.c (expand_gimple_cond): Likewise.
3748 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
3749             Julian Brown  <julian@codesourcery.com>
3751         PR middle-end/86336
3752         * gimplify.c (gimplify_scan_omp_clauses): Set
3753         target_firstprivatize_array_bases in OpenACC parallel and kernels
3754         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
3755         OpenACC data regions.
3757 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
3759         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
3760         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
3762 2018-09-12  Richard Biener  <rguenther@suse.de>
3764         PR tree-optimization/87280
3765         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
3766         edge but unreachable target.
3767         (do_rpo_vn): For conservatively handling a PHI only mark
3768         the backedge executable but not the block reachable.
3770 2018-09-12  Richard Biener  <rguenther@suse.de>
3772         PR tree-optimization/87266
3773         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
3774         visited blocks.
3776 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
3778         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
3779         constants.
3780         ("trunc<BFP:mode><DFP_ALL:mode>2")
3781         ("trunc<DFP_ALL:mode><BFP:mode>2")
3782         ("extend<BFP:mode><DFP_ALL:mode>2")
3783         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
3784         according to the target operand type.
3786 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3787             Andreas Krebbel  <krebbel@linux.ibm.com>
3789         PR tree-optimization/86844
3790         * gimple-ssa-store-merging.c
3791         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
3792         there are any overlapping stores in between them, make sure they are
3793         also coalesced or we give up completely.
3795 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3797         PR middle-end/87248
3798         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
3799         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
3801 2018-09-12  Tom de Vries  <tdevries@suse.de>
3803         * common.opt (gdescribe-dies): Add option.
3804         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
3805         attribute for artifical and nameless decls.
3806         (dwarf2out_register_external_die): Add description attribute to
3807         external reference die.
3808         (add_desc_attribute): New functions.
3809         (gen_subprogram_die): Add description attribute to
3810         DW_TAG_call_site_parameter.
3811         * tree-pretty-print.c (print_generic_expr_to_str): New function.
3812         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
3813         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
3814         -gno-describe-dies.
3815         (@item -gdescribe-dies): Add.
3817 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
3819         * tree-vrp.c (vrp_shift_undefined_p): Remove.
3820         (extract_range_from_binary_expr_1: Call
3821         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
3822         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
3823         depend on sign.
3825 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
3827         * gimple-ssa-warn-alloca.c
3828         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
3829         field for ALLOCA_BOUND_*_LARGE.
3831 2018-09-11  Nathan Sidwell  <nathan@acm.org>
3833         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
3835 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3837         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
3838         for clobbers.  Remove obsolete comment.
3840 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3842         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
3843         mpxchk, mpxld and mpxst types.
3844         (define_attr length_immediate): Remove all processing of mpx types.
3845         (define_attr prefix_0f): Ditto.
3846         (define_attr memory): Ditto.
3848 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3850         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
3851         (log<mode>2): Change operand 1 predicate to general_operand.
3852         Extend operand 1 to XFmode and generate logxf3 insn.
3853         (log10<mode>2): Change operand 1 predicate to general_operand.
3854         Extend operand 1 to XFmode and generate log10xf3 insn.
3855         (log2<mode>2): Change operand 1 predicate to general_operand.
3856         Extend operand 1 to XFmode and generate log2xf3 insn.
3857         (fyl2xp1_extend<mode>xf3_i387): Remove.
3858         (log1p<mode>2): Change operand 1 predicate to general_operand.
3859         Extend operand 1 to XFmode and generate log1pxf3 insn.
3860         (fxtract_extend<mode>xf3_i387): Remove.
3861         (logb<mode>2): Change operand 1 predicate to general_operand.
3862         Extend operand 1 to XFmode and generate logbxf3 insn.
3863         (ilogb<mode>2): Change operand 1 predicate to general_operand.
3864         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
3865         (significand<mode>2): Change operand 1 predicate to general_operand.
3866         Extend operand 1 to XFmode and generate significandxf3 insn.
3868 2018-09-11  Nathan Sidwell  <nathan@acm.org>
3870         * gcc.c (perror_with_name, pfatal_with_name): Delete.
3871         (load_specs): Use fatal_error.
3872         (DELETE_IF_ORDINARY, process_command): Use error.
3873         (execute, run_attempt): Use fatal_error.
3875 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
3877         * diagnostic-core.h (sorry_at): New prototype.
3878         * diagnostic.c (sorry_at): New function.
3880 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
3882         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
3883         by zero as VR_UNDEFINED.
3885 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
3887         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
3888         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
3889         (<sincos>mode2): New expander.
3890         (sincos_extend<mode>xf3_i387): Remove insn pattern.
3891         (sincos -> sin, cos splitters): Remove splitter patterns.
3892         (sincos<mode>3): Change operand 2 predicate to general_operand.
3893         Extend operand 2 to XFmode and generate sincosxf3 insn.
3894         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
3895         Change operand 3 predicate to const1_operand.
3896         (fptan_extend<mode>xf4_i387): Remove insn pattern.
3897         (tanxf2): Update operands in the call to fptanxf4_i387.
3898         (tan<mode>2): Change operand 1 predicate to general_operand.
3899         Extend operand 1 to XFmode and generate tanxf3 insn.
3900         (atan2xf3): Rename from *fpatanxf3_i387.
3901         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
3902         (atan2xf3): Remove expander.
3903         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
3904         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
3905         (atan<mode>2): Change operand 1 predicate to general_operand.
3906         Extend operand 1 to XFmode and generate atanxf3 insn.
3908 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
3910         * config/i386/i386.md (x87/SSE constant load splitter): Use
3911         memory_operand instead of nonimmediate_operand for input operand
3912         predicate.
3914 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
3916         * config/i386/i386.md (float partial SSE register stall splitter): Move
3917         splitter near its instruction pattern.
3918         (float_extend partial SSE register stall splitter): Ditto.
3919         (float_truncate partial SSE register stall splitter): Ditto.
3921 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
3923         PR target/86794
3924         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3925         to speculation_safe_value_not_needed.
3927         PR target/85666
3928         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
3929         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
3930         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
3931         leaf_function_p, instead use has_hard_reg_initial_val.
3933 2018-09-09  Nathan Sidwell  <nathan@acm.org>
3935         * gcc.h (pfatal_with_name): Don't declare here.
3936         * gcc.c (pfatal_with_name): Make static.
3938 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3940         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
3941         earlyclobber.
3943 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
3945         PR rtl-optimization/85458
3946         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
3947         priority hook to reduce the priority of EXPR.
3949 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
3951         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
3952         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
3953         with X87_ENABLE_FLOAT.  Remove preparation code.
3954         (*float<SWI48:mode><MODEF:mode>2): Rename from
3955         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
3956         math using "enabled" attribute.
3957         (*floatdi<MODEF:mode>2_i387): Rename from
3958         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
3959         enable for 32bit targets only.
3960         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
3961         splitter.
3962         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
3963         as operand 1 predicate.  Rewrite as define_insn_and_split.
3964         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
3966 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
3968         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
3969         to fallthru to FLOAT case.
3971 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3973         PR target/86731
3974         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
3975         around folding of vec_sl to handle out of range shift values.
3977 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
3979         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
3980         Update callers to gen_fix_trunc<mode>_i387_fisttp
3981         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
3982         nonimmediate_operand.
3983         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
3984         and corresponding splitters.
3985         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
3986         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
3987         (fix_truncdi_i387_with_temp): Remove insn pattern
3988         and corresponding splitters.
3989         (fix_trunc<mode>_i387): Change operand 0 predicate to
3990         nonimmediate_operand.
3991         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
3992         and corresponding splitters.
3993         (*fistdi2_1): Remove.
3994         (fistdi2): Ditto.
3995         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
3996         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
3997         (*fist<mode>2_1): Remove.
3998         (fist<mode>2): Ditto.
3999         (fist<mode>2_with_temp): Remove insn pattern and corresponding
4000         splitters.
4001         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
4002         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
4003         (fistdi2_<rounding>): Change operand 0 predicate to
4004         nonimmediate_operand.
4005         (fistdi2_<rounding>_with_temp): Remove insn pattern
4006         and corresponding splitters.
4007         (fist<mode>2_<rounding>): Change operand 0 predicate to
4008         nonimmediate_operand.
4009         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
4010         and corresponding splitters.
4012         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
4014 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4016         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
4017         the init value.
4019 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4021         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4022         early gimple folding of vec_splat().
4023         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
4024         * gimple-fold.h: Add an extern define for tree_vec_extract().
4026 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4028         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
4029         wrappers around TREE_TYPE comparisons.
4031 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4033         PR target/80080
4034         * config/s390/predicates.md: Add nonsym_memory_operand.
4035         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
4036         contains a SYMBOL_REF, load it into an intermediate pseudo.
4037         (s390_emit_compare_and_swap): Legitimize operand.
4038         * config/s390/s390.md: Use the new nonsym_memory_operand
4039         with UNSPECV_CAS patterns.
4041 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4043         PR target/80080
4044         * config/s390/s390-passes.def: New file.
4045         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
4046         declaration.
4047         (make_pass_s390_early_mach): Add declaration.
4048         * config/s390/s390.c (make_pass_s390_early_mach):
4049         (s390_option_override): Remove dynamic registration.
4050         * config/s390/t-s390: Add s390-passes.def.
4052 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4054         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4055         Remove UNSPEC_LTREL_BASE check.
4056         (annotate_constant_pool_refs): Likewise.
4057         (find_constant_pool_ref): Likewise.
4058         (find_ltrel_base): Removed.
4059         (replace_ltrel_base): Removed.
4060         (s390_mainpool_finish): Remove replace_ltrel_base call.
4061         (s390_chunkify_start): Remove pending LTREL_BASE logic.
4062         (s390_chunkify_finish): Remove replace_ltrel_base call.
4063         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
4065 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
4067         PR target/86779
4068         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4069         to speculation_safe_value_not_needed.
4071 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
4072             Bernd Schmidt  <bernds_cb1@t-online.de>
4074         * config/nvptx/nvptx-opts.h: New file.
4075         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
4076         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
4077         (ASM_SPEC): Define.
4078         (TARGET_SM35): New macro.
4079         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
4080         correct predicate.
4081         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
4082         values.
4083         (misa=): New option.
4084         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
4086 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
4088         * config/i386/i386.md (truncdfsf2): Remove expander.
4089         (truncdfsf2_with_temp): Ditto.
4090         (truncxf<mode>2): Ditto.
4091         (*truncdfsf_fast_mixed): Remove insn pattern.
4092         (*truncdfsf_fast_i387): Ditto.
4093         (*truncdfsf_mixed): Ditto.
4094         (*truncdfsf_i387): Ditto.
4095         (*truncdfsf2_i387_1): Ditto.
4096         (*truncxfsf2_mixed): Ditto.
4097         (*truncxfdf2_mixed): Ditto.
4098         (*truncxf<mode>2_i387_noop): Ditto. Update callers
4099         to call gen_truncxf<mode>2 instead.
4100         (*truncxf<mode>2_i387): Remove.
4101         (reg->reg splitters): Remove splitter pattern.
4102         (reg->mem splitters): Ditto.
4104         (truncdfsf2): New insn pattern.
4105         (truncxf<mode>2): Ditto.
4107 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4109         * tree-ssa-math-opts.c (is_mult_by): New function.
4110         (is_square_of): Use the above.
4111         (optimize_recip_sqrt): New function.
4112         (pass_cse_reciprocals::execute): Use the above.
4114 2018-09-05  Richard Biener  <rguenther@suse.de>
4116         PR bootstrap/87134
4117         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
4118         to zero-init the emplaced vec.
4120 2018-09-05  Martin Liska  <mliska@suse.cz>
4122         PR tree-optimization/87205
4123         * tree-switch-conversion.c (pass_lower_switch::execute):
4124         Group cases for switch statements.
4126 2018-09-05  Richard Biener  <rguenther@suse.de>
4128         PR tree-optimization/87217
4129         * tree-ssa-sccvn.c (vuse_valueize): New.
4130         (vn_reference_lookup_pieces): Use it.
4131         (vn_reference_lookup): Likewise.
4133 2018-09-05  Nathan Sidwell  <nathan@acm.org>
4135         PR c++/87137
4136         * stor-layout.c (place_field): Scan forwards to check last
4137         bitfield when ms_bitfield_placement is in effect.
4139 2018-09-05  Richard Biener  <rguenther@suse.de>
4141         PR bootstrap/87225
4142         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
4143         return.
4145 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
4146             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
4148         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
4149         * config.gcc (extra_objs): Build it.
4150         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
4151         Likewise.
4152         * config/aarch64/aarch64-passes.def
4153         (pass_tag_collision_avoidance): New pass.
4154         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
4155         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
4156         (aarch64_classify_address): Remove static qualifier.
4157         (aarch64_address_info, aarch64_address_type): Move to...
4158         * config/aarch64/aarch64-protos.h: ... here.
4159         (make_pass_tag_collision_avoidance): New function.
4160         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
4161         New tuning flag.
4163 2018-09-05  Martin Liska  <mliska@suse.cz>
4165         * doc/gcov.texi: Update documentation of humar
4166         readable mode.
4167         * gcov.c (format_count): Print one decimal place, it provides
4168         more fine number of situations like '1G' vs. '1.4G'.
4170 2018-09-05  Martin Liska  <mliska@suse.cz>
4172         PR target/87164
4173         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
4174         * optc-gen.awk: Allow 'Var' for Deprecated options in order
4175         to generate a MASK value.
4177 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
4179         PR debug/86593
4180         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
4181         if frame pointer isn't used.
4182         (compute_frame_pointer_to_fb_displacement): Likewise.
4184 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
4186         PR target/87198
4187         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
4188         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
4189         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
4190         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
4191         and OPTION_MASK_ISA_XSAVEC_UNSET.
4193 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
4195         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
4196         XOR operations in NAND case.
4198 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4200         * wide-int-range.cc (wide_int_range_convert): New.
4201         * wide-int-range.h (wide_int_range_convert): New.
4202         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
4203         code into wide_int_range_convert.
4204         (extract_range_into_wide_ints): Do not munge anti range constants
4205         into the entire domain.  Just return the range back.
4207 2018-09-04  Martin Liska  <mliska@suse.cz>
4209         * genmatch.c (output_line_directive): Add new argument
4210         fnargs.
4211         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
4213 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4215         * doc/invoke.texi (Option Summary): Add whitespace.
4217         * doc/invoke.texi (Option Summary): Add -Waligned-new.
4219 2018-09-04  Richard Biener  <rguenther@suse.de>
4221         PR tree-optimization/87211
4222         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
4223         backedge value we're supposed to treat as VARYING also number
4224         the PHI to VARYING in case it got a different value-number already.
4226 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4228         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
4229         (extract_range_from_binary_expr_1): Do not call
4230         vrp_can_optimize_bit_op.
4231         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
4232         static.
4233         (wide_int_range_get_mask_and_bounds): New.
4234         (wide_int_range_optimize_bit_op): New.
4235         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4236         (wide_int_range_bit_and): Same.
4237         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4238         (wide_int_range_optimize_bit_op): New.
4239         (wide_int_range_get_mask_and_bounds): New.
4241 2018-09-04  Richard Biener  <rguenther@suse.de>
4243         PR tree-optimization/87176
4244         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4245         variable.  When value-numbering a virtual PHI node make sure
4246         to not value-number to the backedge value.
4248 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4250         * doc/extend.texi (Long Long, Hex Floats): Document support for
4251         long long and hex floats in more recent versions of ISO C++.
4253 2018-09-03  Richard Biener  <rguenther@suse.de>
4255         PR tree-optimization/87177
4256         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4257         cleanup.
4259 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4261         * bb-reorder.c (edge_order): Convert to C-qsort-style
4262         tri-state comparator.
4263         (reorder_basic_blocks_simple): Change std::stable_sort to
4264         gcc_stablesort.
4266 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4268         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4269         tri-state comparator.
4270         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4272 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4274         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
4275         (mergesort): ... here as maximum count for using netsort.
4276         (gcc_qsort): Set nlim to 3 if stable sort is requested.
4277         (gcc_stablesort): New.
4278         * system.h (gcc_stablesort): Declare.
4280 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4282         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
4283         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
4284         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
4286 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
4288         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
4289         lxsdx and stxsdx alternatives.
4290         (*mov<mode>_hardfloat64): Ditto.
4291         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4293 2018-09-03  Richard Biener  <rguenther@suse.de>
4295         PR tree-optimization/87200
4296         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4297         simplify result.
4299 2018-09-03  Martin Liska  <mliska@suse.cz>
4301         PR tree-optimization/87201
4302         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4303         Fix parenthesis in an expression.
4305 2018-09-03  Richard Biener  <rguenther@suse.de>
4307         PR tree-optimization/87197
4308         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4309         visited.  CSE the VN_INFO hashtable lookup.
4311         PR tree-optimization/87169
4312         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4313         iterating make sure there's no extra backedges from irreducible
4314         regions feeding the header.  Mark the destination block
4315         executable.
4317 2018-09-03  Martin Liska  <mliska@suse.cz>
4319         PR driver/83193
4320         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4321         * common/common-targhooks.c (default_get_valid_option_values):
4322         New function.
4323         * common/common-targhooks.h (default_get_valid_option_values):
4324         Likewise.
4325         * common/config/i386/i386-common.c: Move processor_target_table
4326         from i386.c.
4327         (ix86_get_valid_option_values): New function.
4328         (TARGET_GET_VALID_OPTION_VALUES): New macro.
4329         * config/i386/i386.c (struct ptt): Move to i386-common.c.
4330         (PTA_*): Move all defined masks into i386-common.c.
4331         (ix86_function_specific_restore): Use new processor_cost_table.
4332         * config/i386/i386.h (struct ptt): Moved from i386.c.
4333         (struct pta): Likewise.
4334         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4335         * doc/tm.texi.in: Likewise.
4336         * opt-suggestions.c (option_proposer::suggest_option):
4337         Pass prefix to build_option_suggestions.
4338         (option_proposer::get_completions): Likewise.
4339         (option_proposer::build_option_suggestions): Use the new target
4340         hook.
4341         * opts.c (struct option_help_tuple): New struct.
4342         (print_filtered_help): Use the new target hook.
4344 2018-09-03  Martin Liska  <mliska@suse.cz>
4346         PR middle-end/59521
4347         * predict.c (set_even_probabilities): Add likely_edges
4348         argument and handle cases where we have precisely one
4349         likely edge.
4350         (combine_predictions_for_bb): Catch also likely_edges.
4351         (tree_predict_by_opcode): Handle gswitch statements.
4352         * tree-cfg.h (find_case_label_for_value): New declaration.
4353         (find_taken_edge_switch_expr): Likewise.
4354         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4355         Find pivot in decision tree based on probabily, not by number of
4356         nodes.
4358 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
4360         * doc/standards.texi (Standards): Update Objective-C reference.
4362 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4364         * doc/install.texi (Prerequisites): Update link for MPC.
4366 2018-09-01  Michael Matz  <matz@suse.de>
4368         PR tree-optimization/87074
4369         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
4370         PHIs for outer-loop uses.
4372 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4374         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
4375         * doc/invoke.texi (C Dialect Options): Ditto.
4377 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4379         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
4381 2018-08-31  Richard Biener  <rguenther@suse.de>
4383         PR tree-optimization/87168
4384         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
4385         (rpo_elim::eliminate_avail): When OP was not visited it must
4386         be available.
4388 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
4390         * tree-vrp.c (copy_value_range): Convert param "from" from
4391         "value_range *" to "const value_range *".
4392         (range_is_null): Likewise for param "vr".
4393         (range_int_cst_p): Likewise.
4394         (range_int_cst_singleton_p): Likewise.
4395         (symbolic_range_p): Likewise.
4396         (value_ranges_intersect_p): Likewise for both params.
4397         (value_range_nonnegative_p): Likewise for param "vr".
4398         (value_range_constant_singleton): Likewise.
4399         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
4400         (extract_range_into_wide_ints): Likewise for param "vr".
4401         (extract_range_from_multiplicative_op): Likewise for params "vr0"
4402         and "vr1".
4403         (vrp_can_optimize_bit_op): Likewise.
4404         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
4405         "vr1_".
4406         (extract_range_from_unary_expr): Likewise.
4407         (debug_value_range): Likewise for param "vr".
4408         (value_range::dump): Add "const" qualifier.
4409         (vrp_prop::check_array_ref): Convert local "vr" from
4410         "value_range *" to "const value_range *".
4411         (vrp_prop::check_mem_ref): Likewise.
4412         (vrp_prop::visit_stmt): Likewise for local "old_vr".
4413         (vrp_intersect_ranges_1): Likewise for param "vr_1".
4414         (vrp_intersect_ranges): Likewise.
4415         (simplify_stmt_for_jump_threading): Likewise for local "vr".
4416         (vrp_prop::vrp_finalize): Likewise.
4417         * tree-vrp.h (value_range::dump): Add "const" qualifier.
4418         (vrp_intersect_ranges): Add "const" qualifier to params as above.
4419         (extract_range_from_unary_expr): Likewise.
4420         (value_range_constant_singleton): Likewise.
4421         (symbolic_range_p): Likewise.
4422         (copy_value_range): Likewise.
4423         (extract_range_from_binary_expr_1): Likewise.
4424         (range_int_cst_p): Likewise.
4425         (vrp_set_zero_nonzero_bits): Likewise.
4426         (range_int_cst_singleton_p): Likewise.
4428 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
4430         * config/aarch64/arm_neon.h (vabsd_s64): New.
4431         (vnegd_s64): Likewise.
4433 2018-08-31  Martin Jambor  <mjambor@suse.cz>
4435         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
4437 2018-08-31  Martin Liska  <mliska@suse.cz>
4439         * ipa-icf.c (sem_item::add_type): Use
4440         sem_item::m_type_hash_cache.
4441         * ipa-icf.h: Move the cache from sem_item_optimizer
4442         to sem_item.
4444 2018-08-31  Nathan Sidwell  <nathan@acm.org>
4446         * doc/extend.texi (Backwards Compatibility): Remove implicit
4447         extern C leeway of () being (...).
4449 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4451         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
4453 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
4455         PR target/86684
4456         PR target/87149
4457         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
4459 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
4461         PR middle-end/87138
4462         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
4463         gen_int_mode.  Formatting fixes.
4465 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
4467         * target.def (custom_function_descriptors): Improve documentation.
4468         * doc/tm.texi.in (Trampolines): Expand discussion of function
4469         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
4470         beginning of the section.
4471         * doc/tm.texi: Regenerated.
4473 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
4475         * cfg.h (class auto_edge_flag): Spell out the template-id of the
4476         base class in the initializer list.  This is a workaround for
4477         building with older GCC.
4478         (class auto_bb_flag): Likewise.
4480 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
4482         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
4483         (altivec_vcmpequ<VI_char>_p): Remove star.
4484         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
4485         vector load modes.
4486         (expand_strncmp_vec_sequence): New function.
4487         (emit_final_str_compare_vec): New function.
4488         (expand_strn_compare): Add support for vector strncmp.
4489         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
4490         length specification to bytes.
4491         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
4492         (vcmpnezb_p): New pattern.
4493         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
4494         for option -mstring-compare-inline-limit.
4496 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
4498         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
4499         (PTA_SKYLAKE): Add PTA_AES.
4500         (PTA_GOLDMONT): Likewise.
4502 2018-08-29  Jan Hubicka  <jh@suse.cz>
4504         PR lto/86517
4505         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
4506         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
4508 2018-08-29  Jan Hubicka  <jh@suse.cz>
4510         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
4511         TYPE_STUB_DECL.
4512         (hash_tree): Do not visit TYPE_STUB_DECL.
4513         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
4514         stream TYPE_STUB_DECL.
4515         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
4516         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
4517         after free_lang_data.
4518         (type_in_anonymous_namespace_p): Likewise.
4520 2018-08-29  Jan Hubicka  <jh@suse.cz>
4522         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4523         comment that it has to be even number.
4524         (class sreal): Change m_sig type to int32_t.
4525         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4526         int64_t for temporary calculations.
4527         (sreal_verify_basics): Drop one bit from minimum and maximum.
4529 2018-08-30  Richard Biener  <rguenther@suse.de>
4531         PR tree-optimization/87147
4532         * tree-ssa-sccvn.c (SSA_VISITED): New function.
4533         (visit_phi): When the degenerate result is from the backedge and
4534         we didn't visit its definition yet drop to VARYING.
4535         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
4537 2018-08-29  Jan Hubicka  <jh@suse.cz>
4539         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
4540         DECL_VINDEX.
4541         (hash_tree): Likewise.
4543 2018-08-29  Jan Hubicka  <jh@suse.cz>
4545         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
4546         and TYPE_NEXT_REF_TO.
4548 2018-08-29  Jan Hubicka  <jh@suse.cz>
4550         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4551         comment that it has to be even number.
4552         (class sreal): Change m_sig type to int32_t.
4553         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4554         int64_t for temporary calculations.
4555         (sreal_verify_basics): Drop one bit from minimum and maximum.
4557 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
4559         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
4561 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
4563         PR middle-end/86995
4564         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
4565         if to_add is negative.
4567 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4569         PR middle-end/87053
4570         * builtins.c (c_strlen): Improve range checks.
4572 2018-08-29  Martin Sebor  <msebor@redhat.com>
4573             Jeff Law  <law@redhat.com>
4575         PR tree-optimization/86714
4576         PR tree-optimization/86711
4577         * builtins.c (c_strlen): Add arguments to call to string_constant.
4578         * expr.c (string_constant): Add argument.  Detect missing nul
4579         terminator and outermost declaration it's missing in.
4580         * expr.h (string_constant): Add argument.
4581         * fold-const.c (read_from_constant_string): Add arguments to call to
4582         string_constant.
4583         (c_getstr): Likewise.
4584         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4585         to string_constant.
4586         * tree-ssa-strlen.c (get_stridx): Likewise.
4588 2018-08-29  Jan Hubicka  <jh@suse.cz>
4590         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
4591         Do not stream DECL_VINDEX.
4592         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
4593         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
4594         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
4596 2018-08-29  Richard Biener  <rguenther@suse.de>
4598         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
4599         virtual operands that are not default defs to honor region
4600         boundaries.
4601         (rpo_vn_valueize): Remove ineffective code here.
4603 2018-08-29  Richard Biener  <rguenther@suse.de>
4605         PR tree-optimization/87132
4606         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
4607         when skipping defs reachable over backedges.
4609 2018-08-29  Richard Biener  <rguenther@suse.de>
4611         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
4612         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
4613         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
4614         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
4615         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
4616         (vn_reference_lookup_3): Remove use of const_parms.
4617         (free_rpo_vn): Do not free const_parms.
4618         (do_rpo_vn): Do not call init_const_parms.
4619         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
4620         SSA_NAME_POINTS_TO_READONLY_MEMORY.
4621         (call_may_clobber_ref_p_1): Likewise.
4623 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
4625         PR other/86726
4626         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
4627         (-O): Ditto.
4628         (-ftree-scev-cprop): Document.
4630 2018-08-29  Jan Hubicka  <jh@suse.cz>
4632         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
4633         parameters.
4634         (sreal constructor): Update.
4635         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
4636         sreal:operator/): Update.
4638 2018-08-29  Martin Liska  <mliska@suse.cz>
4640         * tree-switch-conversion.c (switch_conversion::expand):
4641         Strenghten assumption about gswitch statements.
4643 2018-08-29  Richard Biener  <rguenther@suse.de>
4645         PR tree-optimization/87117
4646         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
4647         re-value-number released SSA VDEFs.
4649 2018-08-29  Richard Biener  <rguenther@suse.de>
4651         PR tree-optimization/87126
4652         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
4654 2018-08-28  Jim Wilson  <jimw@sifive.com>
4656         * config/riscv/pic.md: Rewrite.
4657         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
4658         invalid address.
4659         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
4660         (SOFTF, default_load, softload, softstore): New.
4662 2018-08-28  Jeff Law  <law@redhat.com>
4664         * fold-const.c (fold_binary_loc): Remove recently added assert.
4666 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
4668         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
4669         to OP parmeter of generated function.
4671 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
4673         PR tree-optimization/87009
4674         * match.pd: Add boolean optimizations.
4676 2018-08-28  Martin Sebor  <msebor@redhat.com>
4678         PR middle-end/86631
4679         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
4680         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
4681         (pass_walloca::gate): Use it.
4682         (alloca_call_type): Same.
4683         (pass_walloca::execute): Same.
4684         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
4686 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
4688         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
4689         GCC_VERSION for usage of "__gcc_dump_printf__" format from
4690         >= 3005 to >= 9000.
4692 2018-08-28  Richard Biener  <rguenther@suse.de>
4694         PR tree-optimization/87124
4695         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
4696         constants before looking up avail.
4698 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
4700         PR middle-end/87099
4701         * calls.c (maybe_warn_nonstring_arg): Punt early if
4702         warn_stringop_overflow is zero.  Don't call get_range_strlen
4703         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
4704         Swap comparison operands to have constants on rhs.  Only use
4705         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
4706         increment lenrng[0].
4708 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
4710         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
4711         use of tree_to_shwi.  Remove duplicated test for the size being
4712         a whole number of bytes.
4714 2018-08-28  Richard Biener  <rguenther@suse.de>
4716         PR tree-optimization/87117
4717         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
4718         Handle removed stmt without LHS (GIMPLE_NOP).
4720 2018-08-28  Richard Biener  <rguenther@suse.de>
4722         PR tree-optimization/87117
4723         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
4724         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
4726 2018-08-28  Richard Biener  <rguenther@suse.de>
4728         PR tree-optimization/87117
4729         * tree-ssa-pre.c (compute_avail): Do not make expressions
4730         with predicated values available.
4731         (get_expr_value_id): Assert we do not run into predicated value
4732         expressions.
4734 2018-08-28  Richard Biener  <rguenther@suse.de>
4736         PR tree-optimization/87117
4737         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
4738         get virtual operands.
4739         (get_expr_operands): Handle STRING_CST like other decls.
4741 2018-08-28  Martin Liska  <mliska@suse.cz>
4743         * tree.h: Update documentation of fndecl_built_in_p
4744         functions.
4747 2018-08-27  Jeff Law  <law@redhat.com>
4748         PR tree-optimization/87110
4749         * tree-ssa-dse.c (compute_trims): Handle non-constant
4750         TYPE_SIZE_UNIT.
4752 2018-08-27  Martin Sebor  <msebor@redhat.com>
4754         PR tree-optimization/86914
4755         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
4757 2018-08-27  Martin Sebor  <msebor@redhat.com>
4759         PR tree-optimization/87112
4760         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
4761         the type of the bound argument.
4763 2018-08-27  Jeff Law  <law@redhat.com>
4765         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
4766         type does not have a TYPE_SIZE_UNIT.
4768 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
4770         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
4771         with include of backend.h.
4773 2018-08-27  Richard Biener  <rguenther@suse.de>
4775         PR tree-optimization/86927
4776         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
4777         use const cond reduction code.
4779 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
4781         PR tree-optimization/85758
4782         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
4784 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
4786         PR c++/87091
4787         * diagnostic-show-locus.c (class layout_range): Update for
4788         conversion of show_caret_p to a tri-state.
4789         (layout_range::layout_range): Likewise.
4790         (make_range): Likewise.
4791         (layout::maybe_add_location_range): Likewise.
4792         (layout::should_print_annotation_line_p): Don't show annotation
4793         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
4794         (layout::get_state_at_point): Update for conversion of
4795         show_caret_p to a tri-state.  Bail out early for
4796         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
4797         underlining or source colorization.
4798         (gcc_rich_location::add_location_if_nearby): Update for conversion
4799         of show_caret_p to a tri-state.
4800         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
4801         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
4802         Likewise.
4803         (selftest::test_one_liner_labels): Likewise.
4804         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
4805         conversion of show_caret_p to a tri-state.
4806         * pretty-print.c (text_info::set_location): Likewise.
4807         * pretty-print.h (text_info::set_location): Likewise.
4808         * substring-locations.c (format_warning_n_va): Likewise.
4809         * tree-diagnostic.c (default_tree_printer): Likewise.
4810         * tree-pretty-print.c (newline_and_indent): Likewise.
4812 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
4814         PR c++/87091
4815         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
4816         line above for line-insertion fix-it hints.
4817         (selftest::test_fixit_insert_containing_newline): Update the
4818         expected results, and add a test with line-numbering enabled.
4820 2018-08-27  Martin Liska  <mliska@suse.cz>
4822         PR sanitizer/86962
4823         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
4824         params with DECL_HAS_VALUE_EXPR_P.
4826 2018-08-27  Martin Liska  <mliska@suse.cz>
4828         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
4829         selected expansion strategy.
4831 2018-08-27  Martin Liska  <mliska@suse.cz>
4833         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
4834         * builtins.c (is_builtin_fn): Likewise.
4835         * attribs.c (diag_attr_exclusions): Use new function
4836         fndecl_built_in_p and remove check for FUNCTION_DECL if
4837         possible.
4838         (builtin_mathfn_code): Likewise.
4839         (fold_builtin_expect): Likewise.
4840         (fold_call_expr): Likewise.
4841         (fold_builtin_call_array): Likewise.
4842         (fold_call_stmt): Likewise.
4843         (set_builtin_user_assembler_name): Likewise.
4844         (is_simple_builtin): Likewise.
4845         * calls.c (gimple_alloca_call_p): Likewise.
4846         (maybe_warn_nonstring_arg): Likewise.
4847         * cfgexpand.c (expand_call_stmt): Likewise.
4848         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
4849         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
4850         (cgraph_node::verify_node): Likewise.
4851         * cgraphclones.c (build_function_decl_skip_args): Likewise.
4852         (cgraph_node::create_clone): Likewise.
4853         * config/arm/arm.c (arm_insert_attributes): Likewise.
4854         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
4855         * dse.c (scan_insn): Likewise.
4856         * expr.c (expand_expr_real_1): Likewise.
4857         * fold-const.c (operand_equal_p): Likewise.
4858         (fold_binary_loc): Likewise.
4859         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
4860         * gimple-low.c (lower_stmt): Likewise.
4861         * gimple-pretty-print.c (dump_gimple_call): Likewise.
4862         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
4863         Likewise.
4864         * gimple.c (gimple_build_call_from_tree): Likewise.
4865         (gimple_call_builtin_p): Likewise.
4866         (gimple_call_combined_fn): Likewise.
4867         * gimplify.c (gimplify_call_expr): Likewise.
4868         (gimple_boolify): Likewise.
4869         (gimplify_modify_expr): Likewise.
4870         (gimplify_addr_expr): Likewise.
4871         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
4872         * ipa-cp.c (determine_versionability): Likewise.
4873         * ipa-fnsummary.c (compute_fn_summary): Likewise.
4874         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
4875         * ipa-split.c (visit_bb): Likewise.
4876         (split_function): Likewise.
4877         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4878         * lto-cgraph.c (input_node): Likewise.
4879         * lto-streamer-out.c (write_symbol): Likewise.
4880         * omp-low.c (setjmp_or_longjmp_p): Likewise.
4881         (lower_omp_1): Likewise.
4882         * predict.c (strip_predict_hints): Likewise.
4883         * print-tree.c (print_node): Likewise.
4884         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
4885         * trans-mem.c (is_tm_irrevocable): Likewise.
4886         (is_tm_load): Likewise.
4887         (is_tm_simple_load): Likewise.
4888         (is_tm_store): Likewise.
4889         (is_tm_simple_store): Likewise.
4890         (is_tm_abort): Likewise.
4891         (tm_region_init_1): Likewise.
4892         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
4893         * tree-cfg.c (verify_gimple_call): Likewise.
4894         (move_stmt_r): Likewise.
4895         (stmt_can_terminate_bb_p): Likewise.
4896         * tree-eh.c (lower_eh_constructs_2): Likewise.
4897         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
4898         * tree-inline.c (remap_gimple_stmt): Likewise.
4899         (copy_bb): Likewise.
4900         (estimate_num_insns): Likewise.
4901         (fold_marked_statements): Likewise.
4902         * tree-sra.c (scan_function): Likewise.
4903         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
4904         (optimize_stack_restore): Likewise.
4905         (pass_fold_builtins::execute): Likewise.
4906         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
4907         (mark_all_reaching_defs_necessary_1): Likewise.
4908         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
4909         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4910         (pass_forwprop::execute): Likewise.
4911         * tree-ssa-loop-im.c (stmt_cost): Likewise.
4912         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4913         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
4914         * tree-ssa-strlen.c (get_string_length): Likewise.
4915         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
4916         (find_func_aliases_for_call): Likewise.
4917         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
4918         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
4919         * tree-tailcall.c (find_tail_calls): Likewise.
4920         * tree.c (need_assembler_name_p): Likewise.
4921         (free_lang_data_in_decl): Likewise.
4922         (get_call_combined_fn): Likewise.
4923         * ubsan.c (is_ubsan_builtin_p): Likewise.
4924         * varasm.c (incorporeal_function_p): Likewise.
4925         * tree.h (DECL_BUILT_IN): Remove and replace with
4926         fndecl_built_in_p.
4927         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
4928         (fndecl_built_in_p): New.
4930 2018-08-27  Martin Liska  <mliska@suse.cz>
4932         PR tree-optimization/86847
4933         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
4934         Dump also subtree probability.
4935         (switch_decision_tree::do_jump_if_equal): New function.
4936         (switch_decision_tree::emit_case_nodes): Handle special
4937         situations in balanced tree that can be emitted much simpler.
4938         Fix calculation of probabilities that happen in tree expansion.
4939         * tree-switch-conversion.h (struct cluster): Add
4940         is_single_value_p.
4941         (struct simple_cluster): Likewise.
4942         (struct case_tree_node): Add new function has_child.
4943         (do_jump_if_equal): New.
4945 2018-08-27  Martin Liska  <mliska@suse.cz>
4947         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
4948         Add new argument to bit_test_cluster constructor.
4949         (bit_test_cluster::emit): Set bits really number of values
4950         handlel by a test.
4951         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
4952         probability argument.
4953         * tree-switch-conversion.h (struct bit_test_cluster):
4954         Add m_handles_entire_switch.
4956 2018-08-27  Martin Liska  <mliska@suse.cz>
4958         PR tree-optimization/86702
4959         * tree-switch-conversion.c (jump_table_cluster::emit):
4960         Make probabilities even for values in jump table
4961         according to number of cases handled.
4962         (switch_decision_tree::compute_cases_per_edge): Pass
4963         argument to reset_out_edges_aux function.
4964         (switch_decision_tree::analyze_switch_statement): Likewise.
4965         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
4966         Make it static.
4968 2018-08-27  Martin Liska  <mliska@suse.cz>
4970         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
4971         cfun argument explicitly.
4972         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
4973         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
4974         function gimple_switch_default_bb.
4975         (convert_switch_statements):
4976         (expand_builtins):
4977         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
4978         * stmt.c (label_to_block_fn): Use label_to_block and pass
4979         cfun argument explicitly and use gimple_switch_label_bb.
4980         (expand_case): Likewise.
4981         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
4982         cfun argument explicitly. Likewise.
4983         (make_edges_bb): Likewise.
4984         (make_cond_expr_edges): Likewise.
4985         (get_cases_for_edge): Likewise.
4986         (make_gimple_switch_edges): Likewise.
4987         (label_to_block_fn): Likewise.
4988         (label_to_block): Likewise.
4989         (make_goto_expr_edges): Likewise.
4990         (make_gimple_asm_edges): Likewise.
4991         (main_block_label): Likewise.
4992         (group_case_labels_stmt): Likewise.
4993         (find_taken_edge_computed_goto): Likewise.
4994         (find_taken_edge_switch_expr): Likewise.
4995         (gimple_verify_flow_info): Likewise.
4996         (gimple_redirect_edge_and_branch): Likewise.
4997         (gimple_switch_label_bb): New function.
4998         (gimple_switch_default_bb): Likewise.
4999         (gimple_switch_edge): Likewise.
5000         (gimple_switch_default_edge): Likewise.
5001         * tree-cfg.h (label_to_block_fn): Remove and replace ...
5002         (label_to_block): ... with this.
5003         (gimple_switch_label_bb): New.
5004         (gimple_switch_default_bb): Likewise.
5005         (gimple_switch_edge): Likewise.
5006         (gimple_switch_default_edge): Likewise.
5007         * tree-cfgcleanup.c (convert_single_case_switch): Use
5008         new gimple functions and pass new argument to label_to_block.
5009         (cleanup_control_flow_bb):
5010         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
5011         cfun argument explicitly.
5012         (make_eh_edges): Likewise.
5013         (redirect_eh_dispatch_edge): Likewise.
5014         (lower_resx): Likewise.
5015         (lower_eh_dispatch): Likewise.
5016         (maybe_remove_unreachable_handlers): Likewise.
5017         (unsplit_eh): Likewise.
5018         (cleanup_empty_eh): Likewise.
5019         (verify_eh_edges): Likewise.
5020         (verify_eh_dispatch_edge): Likewise.
5021         * tree-ssa-dom.c (record_edge_info): Likewise.
5022         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
5023         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
5024         (thread_through_normal_block): Likewise.
5025         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
5026         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
5027         * tree-switch-conversion.c (switch_conversion::collect): Use new
5028         gimple functions.
5029         (switch_conversion::check_final_bb): Likewise.
5030         (switch_conversion::gather_default_values): Pass new argument
5031         to label_to_block.
5032         (switch_conversion::build_constructors): Likewise.
5033         (switch_decision_tree::compute_cases_per_edge): Use new
5034         gimple_switch_edge function.
5035         (switch_decision_tree::analyze_switch_statement): Pass new argument
5036         to label_to_block.
5037         (switch_decision_tree::try_switch_expansion): Use
5038         gimple_switch_default_edge.
5039         * tree-vrp.c (find_switch_asserts): Pass new argument
5040         to label_to_block.
5041         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
5042         (vr_values::simplify_switch_using_ranges): Likewise.
5044 2018-08-27  Richard Biener  <rguenther@suse.de>
5046         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
5047         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
5049         * tree-ssa-sccvn.h (struct vn_pval): New structure.
5050         (struct vn_nary_op_s): Add unwind_to member.  Add
5051         predicated_values flag and put result into a union together
5052         with a linked list of vn_pval.
5053         (struct vn_ssa_aux): Add name member to make maintaining
5054         a map of SSA name to vn_ssa_aux possible.  Remove no longer
5055         needed info, dfsnum, low, visited, on_sccstack, use_processed
5056         and range_info_anti_range_p members.
5057         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
5058         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
5059         New functions.
5060         (vn_valueize): New global.
5061         (vn_context_bb): Likewise.
5062         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
5063         VN_INFO_PTR_INFO): Remove.
5064         * tree-ssa-sccvn.c: ... (rewrite)
5065         (pass_fre::execute): For -O2+ initialize loops and run
5066         RPO VN in optimistic mode (iterating).  For -O1 and -Og
5067         run RPO VN in non-optimistic mode.
5068         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
5069         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
5070         * doc/invoke.texi (sccvn-max-scc-size): Remove.
5071         (rpo-vn-max-loop-depth): Document.
5072         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
5073         when valuezing the VUSE signals we walked out of the region.
5074         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
5075         (phi_translate): Set VN context block to use for availability
5076         lookup.
5077         (compute_avail): Likewise.
5078         (pre_valueize): New function.
5079         (pass_pre::execute): Adjust to the RPO VN API.
5081         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
5082         (propagate_constants_for_unrolling): Remove.
5083         (tree_unroll_loops_completely): Perform value-numbering
5084         on the unrolled bodies loop parent.
5086 2018-08-27  Richard Biener  <rguenther@suse.de>
5088         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
5089         for partial antic compute.
5091 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
5093         PR rtl-optimization/87065
5094         * combine.c (simplify_if_then_else): Formatting fix.
5095         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
5096         check.
5097         (known_cond): Don't return const_true_rtx for vector modes.  Use
5098         CONST0_RTX instead of const0_rtx.  Formatting fixes.
5100 2018-08-27  Martin Liska  <mliska@suse.cz>
5102         PR gcov-profile/87069
5103         * gcov.c (process_file): Record files already processed
5104         and warn about a file being processed multiple times.
5106 2018-08-27  Martin Liska  <mliska@suse.cz>
5108         PR driver/83193
5109         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5110         Set default values for x_aarch64_*_string strings.
5111         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
5112         prefix.  For -mabi do not print '=ABI' in help and use
5113         <option_value> format for -msve-vector-bits and -moverride
5114         options.
5116 2018-08-26  Jeff Law  <law@redhat.com>
5118         * config/mips/frame-header-opt.c: Include "backend.h" rather than
5119         "cfg.h"
5121 2018-08-26  Marek Polacek  <polacek@redhat.com>
5123         PR c++/87029, Implement -Wredundant-move.
5124         * doc/invoke.texi: Document -Wredundant-move.
5126 2018-08-25  Martin Sebor  <msebor@redhat.com>
5128         PR tree-optimization/87059
5129         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
5130         to the same type as the other.
5131         * fold-const.c (fold_binary_loc): Assert expectation.
5133 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
5135         * config/darwin.c (machopic_legitimize_pic_address): Clean up
5136         extraneous parentheses, dead code section and formatting.
5138 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
5140         PR c++/87091
5141         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
5142         wide enough for jumps in the line-numbering to be visible.
5143         (layout::print_gap_in_line_numbering): New member function.
5144         (layout::calculate_line_spans): When using line numbering, merge
5145         line spans that are only 1 line apart.
5146         (diagnostic_show_locus): When printing line numbers, show gaps in
5147         line numbering directly, rather than printing headers.
5148         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
5149         line-numbering with multiple line spans.
5150         (selftest::test_fixit_insert_containing_newline_2): Add test of
5151         line-numbering, in which the spans are close enough to be merged.
5153 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5155         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
5156         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
5157         * tree-vrp.c (range_is_nonnull): Remove.
5158         (range_includes_zero_p): Accept value_range instead of min/max.
5159         (extract_range_from_binary_expr_1): Do not early bail on
5160         POINTER_PLUS_EXPR.
5161         Use range_includes_zero_p instead of range_is_nonnull.
5162         (extract_range_from_unary_expr): Use range_includes_zero_p instead
5163         of range_is_nonnull.
5164         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
5165         special case VR_ANTI_RANGE.
5166         (vrp_finalize): Same.
5167         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
5168         instead of min/max.
5169         (range_is_nonnull): Remove.
5170         * vr-values.c (vrp_stmt_computes_nonzero): Use
5171         range_includes_zero_p instead of range_is_nonnull.
5172         (extract_range_basic): Pass value_range to range_includes_zero_p
5173         instead of range_is_nonnull.
5175 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
5177         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
5178         * emit-rtl.h (rtl_data): Remove return_bnd.
5179         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
5180         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
5181         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
5182         (POINTER_BOUNDS_MODE): Remove definition.
5183         (make_pointer_bounds_mode): Remove.
5184         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
5185         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
5186         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
5187         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
5188         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
5189         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
5190         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
5192         * config/i386/i386-modes.def (BND32, BND64): Remove.
5193         * config/i386/i386.c (dbx_register_map): Remove bound registers.
5194         (dbx64_register_map): Ditto.
5195         (svr4_dbx_register_map): Ditto.
5196         (indirect_thunk_bnd_needed): Remove.
5197         (indirect_thunks_bnd_used): Ditto.
5198         (indirect_return_bnd_needed): Ditto.
5199         (indirect_return_via_cx_bnd): Ditto.
5200         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
5201         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
5202         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
5203         (output_indirect_thunk_function): Remove handling of
5204         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
5205         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
5206         (ix86_save_reg): Remove handling of crtl->return_bnd.
5207         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
5208         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
5209         and UNSPEC_BNDLX_ADDR.
5210         (ix86_output_indirect_branch_via_reg): Remove handling of
5211         indirect_thunk_prefix_bnd.
5212         (ix86_output_indirect_branch_via_push): Ditto.
5213         (ix86_output_function_return): Ditto.
5214         (ix86_output_indirect_function_return): Ditto.
5215         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
5216         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
5217         (CALL_USED_REGISTERS): Ditto.
5218         (REG_ALLOC_ORDER): Update for removal of bound registers.
5219         (HI_REGISTER_NAMES): Ditto.
5220         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
5221         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
5222         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
5223         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
5224         (FIRST_PSEUDO_REG): Update.
5225         (BND): Remove mode iterator.
5226         * config/i386/predicates.md (bnd_mem_operator): Remove.
5228 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5230         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
5231         vectors.
5233 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5235         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5236         the case in which the permute needs only a single element and
5237         repeats for every vector of the result.  Extend that case to
5238         handle variable-length vectors.
5239         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5241 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
5243         PR debug/79342
5244         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5245         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5247 2018-08-24  Richard Biener  <rguenther@suse.de>
5249         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5250         bb_flags_allocated members.
5251         (auto_flag): New RAII class for allocating flags.
5252         (auto_edge_flag): New RAII class for allocating edge flags.
5253         (auto_bb_flag): New RAII class for allocating bb flags.
5254         * cfgloop.c (verify_loop_structure): Allocate temporary edge
5255         flag dynamically.
5256         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5257         in favor of temporarily allocated BB flag.
5258         * hsa-brig.c: Re-order includes.
5259         * hsa-dump.c: Likewise.
5260         * hsa-regalloc.c: Likewise.
5261         * print-rtl.c: Likewise.
5262         * profile-count.c: Likewise.
5264 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
5266         PR target/86989
5267         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5268         the TOC register.
5270 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5272         PR 87073/bootstrap
5273         * wide-int-range.cc (wide_int_range_div): Do not ignore result
5274         from wide_int_range_multiplicative_op.
5276 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5278         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5279         "permutaion".
5281 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5283         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
5284         to 'expanded'.
5286 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
5288         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
5289         full GENERIC expression used for replacement.
5291 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
5293         * tree-vrp.c (abs_extent_range): Remove.
5294         (extract_range_into_wide_ints): Pass wide ints by reference.
5295         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5296         Pass wide ints by reference in all calls to
5297         extract_range_into_wide_ints.
5298         * wide-int-range.cc (wide_int_range_div): New.
5299         * wide-int-range.h (wide_int_range_div): New.
5300         (wide_int_range_includes_zero_p): New.
5301         (wide_int_range_zero_p): New.
5303 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
5305         * config/aarch64/aarch64.md (arches): New enum.
5306         (arch): New enum attr.
5307         (arch_enabled): New attr.
5308         (enabled): Now uses arch_enabled only.
5309         (simd, sve, fp16): Removed attribute.
5310         (fp): Attr now defined in terms of 'arch'.
5311         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5312         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5313         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5314         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5315         attributes into 'arch'.
5316         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5317         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5318         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5319         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5320         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5321         'simd' attribute into 'arch'.
5322         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5323         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5324         Convert use of 'fp' attribute to 'arch'.
5325         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5326         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5327         into 'arch'.
5328         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5329         (different modes) Merge 'fp' and 'simd' into 'arch'.
5330         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5331         'simd' into 'arch'.
5333 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
5335         PR rtl-optimization/87026
5336         * expmed.c (canonicalize_comparison): If we can no longer create
5337         pseudoregisters, don't.
5339 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
5341         PR target/86951
5342         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
5343         prototype.
5344         * config/arm/arm.c (speculation_barrier_libfunc): New static
5345         variable.
5346         (arm_init_libfuncs): Initialize it.
5347         (arm_emit_speculation_barrier): New function.
5348         * config/arm/arm.md (speculation_barrier): Call
5349         arm_emit_speculation_barrier for architectures that do not have
5350         DSB or ISB.
5351         (speculation_barrier_insn): Only match on Armv7 or later.
5353 2018-08-23  Richard Biener  <rguenther@suse.de>
5355         PR middle-end/87024
5356         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
5357         calls.
5359 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5361         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
5362         of single-vector TBLs.
5363         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
5364         one input is given.
5366 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5368         PR target/85910
5369         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
5370         aarch64_evpc_tbl guard.
5372 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5374         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
5375         behaviour.
5377 2018-08-22  Martin Sebor  <msebor@redhat.com>
5379         PR middle-end/87052
5380         * tree-pretty-print.c (pretty_print_string): Add argument.
5381         (dump_generic_node): Call to pretty_print_string with string size.
5383 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
5385         PR rtl-optimization/86771
5386         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
5387         of two SETs into those two SETs, one to be placed at i2, if that SETs
5388         destination is modified between i2 and i3.
5390 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5392         PR tree-optimization/86725
5393         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
5394         function.
5395         (vect_analyze_scalar_cycles_1): Check it.
5397 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5399         PR tree-optimization/86725
5400         * tree-vect-loop.c (vect_is_simple_reduction): When treating
5401         an outer loop phi as a double reduction, make sure that the
5402         single user of the phi result is an inner loop phi.
5404 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5406         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
5407         grouped stores with gaps to a strided group.
5409 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5411         * tree-vect-stmts.c (get_group_load_store_type)
5412         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
5413         first statement in a group.
5415 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5417         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
5418         the sequence used in gcc/gcc.c.
5420 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5422         PR other/704
5423         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
5424         building it.
5426 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5428         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
5429         Darwin10-specific unwinder-shim.
5430         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
5431         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
5432         New to cater for Darwin10 Rosetta.
5434 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5436         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
5437         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
5439 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5441         PR bootstrap/81033
5442         PR target/81733
5443         PR target/52795
5444         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
5445         (dwarf2out_switch_text_section): Generate a local label for the
5446         second function sub-section and apply it as the second FDE start
5447         label.
5448         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
5449         second sub-section start.
5451 2018-08-22  Richard Biener  <rguenther@suse.de>
5453         PR tree-optimization/86988
5454         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
5456 2018-08-22  Richard Biener  <rguenther@suse.de>
5458         PR tree-optimization/86945
5459         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
5461 2018-08-22  Alexandre Oliva <oliva@adacore.com>
5463         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
5464         a comment about how uses of r2 for .sdata2 come about.
5466 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
5468         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
5470 2018-08-21  Marek Polacek  <polacek@redhat.com>
5472         PR c++/86981, Implement -Wpessimizing-move.
5473         * doc/invoke.texi: Document -Wpessimizing-move.
5475 2018-08-21  Jan Hubicka  <jh@suse.cz>
5477         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
5478         * tree.h (is_redundant_typedef): Remove.
5479         * dwarf2out.c (is_redundant_typedef): Turn into static function.
5481 2018-08-21  Jan Hubicka  <jh@suse.cz>
5483         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
5484         when possible.
5486 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
5488         * expmed.c (extract_low_bits): Reject invalid subregs early.
5490 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5492         PR middle-end/86121
5493         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
5494         behaviour.
5496 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5498         * config/vxworks.h: Guard vxworks_asm_out_constructor and
5499         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
5500         * config/vxworks.c: Likewise.
5502 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5504         * config/vxworks.c: Set targetm.have_ctors_dtors
5505         if HAVE_INITFINI_ARRAY_SUPPORT.
5506         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
5507         if HAVE_INITFINI_ARRAY_SUPPORT.
5509 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5511         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
5512         default search path for VxWorks < 7.
5514 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5516         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
5517         (get_format_string): Refer to c_getstr.
5519 2018-08-21  Tom de Vries  <tdevries@suse.de>
5521         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5522         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5523         (debuginfo_early_stop): Declare.
5524         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5525         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5526         (debuginfo_early_stop): New function.
5527         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
5528         and debuginfo_early_stop.
5529         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
5530         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
5531         (general_init): Call debuginfo_early_init.
5532         (finalize): Call debuginfo_fini.
5533         (do_compile): Call debuginfo_init.
5534         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
5535         -fdump-early-debug.
5536         (@item -fdump-debug, @item -fdump-earlydebug): Add.
5538 2018-08-21  Tom de Vries  <tdevries@suse.de>
5540         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
5541         flag_dump_noaddr and flag_dump_unnumbered.
5543 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
5545         * wide-int-range.cc (wide_int_range_abs): New.
5546         (wide_int_range_order_set): Rename from wide_int_range_min_max.
5547         * wide-int-range.h (wide_int_range_abs): New.
5548         (wide_int_range_min_max): New.
5549         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
5550         case to call wide_int_range_abs.
5551         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
5552         (extract_range_from_abs_expr): Delete.
5554 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
5556         PR target/87033
5557         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
5558         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
5559         for indexed loads.
5561 2018-08-20  Nathan Sidwell  <nathan@acm.org>
5562             Jeff Law <law@redhat.com>
5564         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
5565         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
5567 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
5569         PR other/84889
5570         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
5571         (decl_attributes): Likewise.
5572         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
5573         instance.
5574         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5575         * diagnostic-core.h (class auto_diagnostic_group): New class.
5576         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
5577         (diagnostic_report_diagnostic): Handle the first diagnostics within
5578         a group.
5579         (emit_diagnostic): Add auto_diagnostic_group instance.
5580         (inform): Likewise.
5581         (inform_n): Likewise.
5582         (warning): Likewise.
5583         (warning_at): Likewise.
5584         (warning_n): Likewise.
5585         (pedwarn): Likewise.
5586         (permerror): Likewise.
5587         (error): Likewise.
5588         (error_n): Likewise.
5589         (error_at): Likewise.
5590         (sorry): Likewise.
5591         (fatal_error): Likewise.
5592         (internal_error): Likewise.
5593         (internal_error_no_backtrace): Likewise.
5594         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
5595         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
5596         * diagnostic.h (struct diagnostic_context): Add fields
5597         "diagnostic_group_nesting_depth",
5598         "diagnostic_group_emission_count", "begin_group_cb",
5599         "end_group_cb".
5600         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
5601         Add auto_diagnostic_group instance(s).
5602         (find_explicit_erroneous_behavior): Likewise.
5603         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
5604         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
5605         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
5606         (gimplify_va_arg_expr): Likewise.
5607         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
5608         (HSA_SORRY_AT): Likewise.
5609         * ipa-devirt.c (compare_virtual_tables): Likewise.
5610         (warn_odr): Likewise.
5611         * multiple_target.c (expand_target_clones): Likewise.
5612         * opts-common.c (cmdline_handle_error): Likewise.
5613         * reginfo.c (globalize_reg): Likewise.
5614         * substring-locations.c (format_warning_n_va): Likewise.
5615         * tree-inline.c (expand_call_inline): Likewise.
5616         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
5617         * tree-ssa-loop-niter.c
5618         (do_warn_aggressive_loop_optimizations): Likewise.
5619         * tree-ssa-uninit.c (warn_uninit): Likewise.
5620         * tree.c (warn_deprecated_use): Likewise.
5622 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
5624         PR target/87014
5625         * config/i386/i386.md (eh_return): Always update EH return
5626         address in word_mode.
5628 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
5630         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
5631         TARGET_SPLIT_COMPLEX_ARG is defined.
5633 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5635         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
5637 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5639         PR target/86984
5640         * expr.c (expand_assignment): Assert that bitpos is positive.
5641         (store_field): Likewise
5642         (expand_expr_real_1): Make sure that bitpos is positive.
5643         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
5644         integer overflow.
5646 2018-08-20  Nathan Sidwell  <nathan@acm.org>
5648         * Makefile.in (CPP_ID_DATA_H): Delete.
5649         (CPP_INTERNAL_H): Don't add it.
5650         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
5651         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
5653 2018-08-20  Richard Biener  <rguenther@suse.de>
5655         PR tree-optimization/78655
5656         * tree-vrp.c (extract_range_from_binary_expr_1): Make
5657         pointer + offset nonnull if either operand is nonnull work.
5659 2018-08-20  Tom de Vries  <tdevries@suse.de>
5661         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
5662         unless the referenced die describes the added property using
5663         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
5664         Otherwise, add a DW_AT_location to the referenced die.
5666 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
5668         PR target/86994
5669         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
5670         register_operand when calling ix86_set_reg_reg_cost.
5671         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
5672         Set *total to 0 for operands that satisfy x86_64_immediate_operand
5673         predicate and to 1 otherwise.
5675 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
5677         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
5678         emit a diagnostic that it is not supported and reset the option.
5679         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
5680         supported and consume the option.  (ASM_FINAL_SPEC): New.
5682 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
5684         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
5685         a sentence.
5687 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
5689         C-SKY port: Documentation
5691         * doc/extend.texi (C-SKY Function Attributes): New section.
5692         * doc/invoke.texi (Option Summary): Add C-SKY options.
5693         (C-SKY Options): New section.
5694         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
5696 2018-08-17  Jojo  <jijie_rong@c-sky.com>
5697             Huibin Wang  <huibin_wang@c-sky.com>
5698             Sandra Loosemore  <sandra@codesourcery.com>
5699             Chung-Lin Tang  <cltang@codesourcery.com>
5701         C-SKY port: Backend implementation
5703         * config/csky/*: New.
5704         * common/config/csky/*: New.
5706 2018-08-17  Jojo  <jijie_rong@c-sky.com>
5707             Huibin Wang  <huibin_wang@c-sky.com>
5708             Sandra Loosemore  <sandra@codesourcery.com>
5709             Chung-Lin Tang  <cltang@codesourcery.com>
5710             Andrew Jenner  <andrew@codesourcery.com>
5712         C-SKY port: Configury
5714         * config.gcc (csky-*-*): New.
5715         * configure.ac: Add csky to targets for dwarf2 debug_line support.
5716         * configure: Regenerated.
5718 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
5720         * dump-context.h: Include "dumpfile.h".
5721         (dump_context::dump_printf_va): Convert final param from va_list
5722         to va_list *.  Convert from ATTRIBUTE_PRINTF to
5723         ATTRIBUTE_GCC_DUMP_PRINTF.
5724         (dump_context::dump_printf_loc_va): Likewise.
5725         * dumpfile.c: Include "stringpool.h".
5726         (make_item_for_dump_printf_va): Delete.
5727         (make_item_for_dump_printf): Delete.
5728         (class dump_pretty_printer): New class.
5729         (dump_pretty_printer::dump_pretty_printer): New ctor.
5730         (dump_pretty_printer::emit_items): New member function.
5731         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
5732         function.
5733         (dump_pretty_printer::emit_item): New member function.
5734         (dump_pretty_printer::stash_item): New member function.
5735         (dump_pretty_printer::format_decoder_cb): New member function.
5736         (dump_pretty_printer::decode_format): New member function.
5737         (dump_context::dump_printf_va): Reimplement in terms of
5738         dump_pretty_printer.
5739         (dump_context::dump_printf_loc_va): Convert final param from va_list
5740         to va_list *.
5741         (dump_context::begin_scope): Reimplement call to
5742         make_item_for_dump_printf.
5743         (dump_printf): Update for change to dump_printf_va.
5744         (dump_printf_loc): Likewise.
5745         (selftest::test_capture_of_dump_calls): Convert "stmt" from
5746         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
5747         with %T, %E, and %G.
5748         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
5749         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
5750         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
5751         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
5752         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
5753         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
5754         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
5755         within a dump_printf_loc call to "%wu".
5756         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
5757         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
5758         missing space after "=".
5759         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
5760         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
5761         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
5762         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
5763         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
5764         duplicate "vectorized" from message.
5766 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
5768         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
5769         polyNxK_t element's TYPE_STRING_FLAG.
5771 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
5773         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
5774         (they were unnamed before).  Fix comments.
5776 2018-08-17  Nathan Sidwell  <nathan@acm.org>
5778         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
5780 2018-08-17  Richard Biener  <rguenther@suse.de>
5782         PR tree-optimization/86841
5783         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
5785 2018-08-17  Martin Liska  <mliska@suse.cz>
5787         * common.opt: Remove Warn, Init and Report for options with
5788         Ignore/Deprecated flag. Warning is done automatically for
5789         Deprecated flags.
5790         * config/i386/i386.opt: Likewise.
5791         * config/ia64/ia64.opt: Likewise.
5792         * config/rs6000/rs6000.opt: Likewise.
5793         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5794         Remove usage of flag_check_pointer_bounds.
5795         * lto-wrapper.c (merge_and_complain): Do not handle
5796         OPT_fcheck_pointer_bounds.
5797         (append_compiler_options): Likewise.
5798         * opt-functions.awk: Do not handle Deprecated.
5799         * optc-gen.awk: Check that Var, Report and Init are not
5800         used for an option with Ignore/Deprecated flag.
5801         * opts-common.c (decode_cmdline_option): Do not report
5802         CL_ERR_DEPRECATED.
5803         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
5804         options.
5805         * opts.h (struct cl_option): Remove cl_deprecated flag.
5806         (CL_ERR_DEPRECATED): Remove error enum value.
5808 2018-08-17  Richard Biener  <rguenther@suse.de>
5810         PR middle-end/86505
5811         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
5812         across a va-arg-pack using call adjust its return value accordingly.
5814 2018-08-16  Martin Sebor  <msebor@redhat.com>
5816         PR tree-optimization/86853
5817         * gimple-ssa-sprintf.c (struct format_result): Rename member.
5818         (struct fmtresult): Add member and initialize it in ctors.
5819         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
5820         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
5821         (format_directive): Set POSUNDER4K when MAYFAIL is set.
5822         (parse_directive): Handle %C same as %c.
5823         (sprintf_dom_walker::compute_format_length): Adjust.
5824         (is_call_safe): Adjust.
5826 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5828         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
5829         for determining how to count the elements.
5830         * builtins.h (c_strlen): Adjust prototype.
5831         * expr.c (string_constant): Add new parameter mem_size.
5832         Set *mem_size appropriately.
5833         * expr.h (string_constant): Adjust protoype.
5834         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
5835         * gimple-fold.h (get_range_strlen): Adjust prototype.
5836         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
5837         (format_string): Call get_string_length with eltsize.
5839 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
5841         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
5842         to emit the span, rather than setting it as the prefix.
5844 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
5846         * diagnostic-show-locus.c (layout::start_annotation_line): Add
5847         "margin_char" parameter, defaulting to space.  Use it in place
5848         of pp_space for the initial part of the margin.
5849         (layout::print_leading_fixits): Use '+' when filling the margin
5850         of line-insertion fix-it hints.
5852 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5854         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
5855         Delete.
5857 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5859         * config/rs6000/altivec.md: Don't set length attribute to the default
5860         value.
5861         * config/rs6000/darwin.md: Ditto.
5862         * config/rs6000/dfp.md: Ditto.
5863         * config/rs6000/htm.md: Ditto.
5864         * config/rs6000/rs6000.md: Ditto.
5865         * config/rs6000/sync.md: Ditto.
5866         * config/rs6000/vsx.md: Ditto.
5868 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5870         * config/rs6000/altivec.md: Don't set length attribute to the default
5871         value, for branch instructions.
5872         * config/rs6000/darwin.md: Ditto.
5873         * config/rs6000/rs6000.md: Ditto.
5875 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5877         * config/rs6000/rs6000.md (length): Always define as const_int 4.
5878         (unnamed conditional branch define_insn): Set length to 4 or 8
5879         depending on offset.
5880         (<bd>_<mode>): Similar, for alternative 0.
5881         (<bd>tf_<mode>): Ditto.
5883 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5885         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
5887 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
5889         * doc/rtl.texi: Replace old RTX class names with new names.
5892 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
5894         * expmed.h (canonicalize_comparison): New declaration.
5895         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
5896         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
5897         * optabs.c (prepare_cmp_insn): Likewise.
5898         * rtl.h (unsigned_condition_p): New function which checks if a
5899         comparison operator is unsigned.
5901 2018-08-16  Nathan Sidwell  <nathan@acm.org>
5903         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
5904         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
5906 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5908         PR target/84711
5909         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
5910         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
5911         (mov<mov>): ..this and enable unconditionally.
5913 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5915         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
5917 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
5919         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
5920         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
5921         "Common" with "Target".
5923 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
5925         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
5926         * doc/invoke.texi (mmitigate-rop): Remove.
5927         * config/i386/i386.c: Do not include "regrename.h".
5928         (ix86_rop_should_change_byte_p, reg_encoded_number)
5929         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
5930         Remove.
5931         (ix86_reorg): Remove call to ix86_mitigate_rop.
5932         * config/i386/i386.md (attr "modrm_class"): Remove.
5933         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
5934         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
5935         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
5937 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
5939         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
5940         allow folding of mergeh() and mergel() for the float and double types.
5941         (fold_mergehl_helper): Rework to handle building a permute tree
5942         for float vectors.
5944 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
5946         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
5947         for TARGET_SSE.
5949 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
5951         * common.opt (fdiagnostics-show-labels): New option.
5952         * diagnostic-show-locus.c (class layout_range): Add field
5953         "m_label".
5954         (class layout): Add field "m_show_labels_p".
5955         (layout_range::layout_range): Add param "label" and use it to
5956         initialize m_label.
5957         (make_range): Pass in NULL for new "label" param of layout_range's
5958         ctor.
5959         (layout::layout): Initialize m_show_labels_p.
5960         (layout::maybe_add_location_range): Pass in loc_range->m_label
5961         when constructing layout_range instances.
5962         (struct line_label): New struct.
5963         (layout::print_any_labels): New member function.
5964         (layout::print_line): Call it if label-printing is enabled.
5965         (selftest::test_one_liner_labels): New test.
5966         (selftest::test_diagnostic_show_locus_one_liner): Call it.
5967         * diagnostic.c (diagnostic_initialize): Initialize
5968         context->show_labels_p.
5969         * diagnostic.h (struct diagnostic_context): Add field
5970         "show_labels_p".
5971         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5972         -fno-diagnostics-show-labels.
5973         * dwarf2out.c (gen_producer_string): Add
5974         OPT_fdiagnostics_show_labels to the ignored options.
5975         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
5976         param.
5977         (gcc_rich_location::maybe_add_expr): Likewise.
5978         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
5979         label" param, defaulting to NULL.
5980         (gcc_rich_location::add_expr): Add "label" param.
5981         (gcc_rich_location::maybe_add_expr): Likewise.
5982         (class text_range_label): New class.
5983         (class range_label_for_type_mismatch): New class.
5984         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
5985         of format_warning_va.
5986         (fmtwarn_n): Likewise for new params of format_warning_n_va.
5987         * lto-wrapper.c (merge_and_complain): Add
5988         OPT_fdiagnostics_show_labels to the "pick one setting" options.
5989         (append_compiler_options): Likewise to the dropped options.
5990         (append_diag_options): Likewise to the passed-on options.
5991         * opts.c (common_handle_option): Handle the new option.
5992         * selftest-diagnostic.c
5993         (test_diagnostic_context::test_diagnostic_context): Enable
5994         show_labels_p.
5995         * substring-locations.c: Include "gcc-rich-location.h".
5996         (format_warning_n_va): Add "fmt_label" and "param_label" params
5997         and use them as appropriate.
5998         (format_warning_va): Add "fmt_label" and "param_label" params,
5999         passing them on to format_warning_n_va.
6000         (format_warning_at_substring): Likewise.
6001         (format_warning_at_substring_n): Likewise.
6002         * substring-locations.h (format_warning_va): Add "fmt_label" and
6003         "param_label" params.
6004         (format_warning_n_va): Likewise.
6005         (format_warning_at_substring): Likewise.
6006         (format_warning_at_substring_n): Likewise.
6007         * toplev.c (general_init): Initialize global_dc->show_labels_p.
6009 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
6011         PR testsuite/86519
6012         * builtins.c (expand_builtin_memcmp): Do not expand the call
6013         when overflow is detected.
6015 2018-08-15  Martin Sebor  <msebor@redhat.com>
6017         PR tree-optimization/71625
6018         * config/aarch64/aarch64-builtins.c
6019         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
6021 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
6023         * config/s390/s390.c (s390_reorg): Remove loop.
6025 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6027         * config/darwin.c
6028          (darwin_function_switched_text_sections): Delete.
6029         * gcc/config/darwin.h
6030          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
6032 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6034         PR target/81685
6035         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
6036         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
6037         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
6039 2018-08-15  Martin Liska  <mliska@suse.cz>
6041         PR tree-optimization/86925
6042         * predict.c (expr_expected_value_1): When taking
6043         later predictor, assign also probability.
6044         Use fold_build2_initializer_loc in order to fold
6045         the expression in -frounding-math.
6047 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
6049         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
6050         patterns.
6051         (expand_vec_perm_1): Try the new method.
6053 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6055         PR target/86547
6056         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
6057         Check whether lra_live_max_point is 0 before dividing.
6059 2018-08-14  Martin Sebor  <msebor@redhat.com>
6061         PR tree-optimization/86650
6062         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
6063         (vrp_prop::check_mem_ref): Same.
6065 2018-08-13  Liu Hao <lh_mouse@126.com>
6067         * pretty-print.c (eat_esc_sequence): Swap the foreground and
6068         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
6069         and clear it thereafter, as it only works for DBCS.
6071 2018-08-13  Liu Hao <lh_mouse@126.com>
6073         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
6074         handle returned by _get_osf_handle().
6076 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6078         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
6079         for folding vec_perm.
6081 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6083         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
6084         Add support for gimple-folding of vec_pack() and vec_unpack()
6085         intrinsics.
6087 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
6089         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
6090         vec_xst variants to the list.
6091         (rs6000_gimple_fold_builtin): Add support for folding unaligned
6092         vector loads and stores.
6094 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
6096         * config.gcc (rs6000-ibm-aix4.x): Delete.
6097         (rs6000-ibm-aix5.1): Delete.
6098         (rs6000-ibm-aix5.2): Delete.
6099         (rs6000-ibm-aix5.3): Delete.
6100         * config/rs6000/aix43.h: Delete.
6101         * config/rs6000/aix51.h: Delete.
6102         * config/rs6000/aix52.h: Delete.
6103         * config/rs6000/t-aix43: Delete.
6105 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
6107         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6108         New function.
6109         (s390_decompose_address): Factor out constant pool ref
6110         decomposition.
6112 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6114         * config/nds32/nds32-predicates.c
6115         (nds32_can_use_bclr_p): Change return type as bool.
6116         (nds32_can_use_bset_p): Ditto.
6117         (nds32_can_use_btgl_p): Ditto.
6118         (nds32_can_use_bitci_p): Ditto.
6119         * config/nds32/nds32-protos.h
6120         (nds32_can_use_bclr_p): Change declaration.
6121         (nds32_can_use_bset_p): Ditto.
6122         (nds32_can_use_btgl_p): Ditto.
6123         (nds32_can_use_bitci_p): Ditto.
6125 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6127         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
6128         Support -msched-prolog-epilog option.
6129         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
6131 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6133         * common/config/nds32/nds32-common.c
6134         (nds32_option_optimization_table): Enalbe -malways-align.
6136 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6138         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
6139         extra_headers.
6140         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
6141         OPT_misr_secure_ case.
6142         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
6143         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
6144         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
6145         secure attribute.
6146         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
6147         (nds32_isr_info): New field security_level.
6148         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
6149         * config/nds32/nds32.md (return_internal): Consider critical attribute.
6150         * config/nds32/nds32.opt (misr-secure): New option.
6151         * config/nds32/nds32_init.inc: New file.
6152         * config/nds32/nds32_isr.h: New file.
6154 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
6156         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
6157         Update comment for atomic instructions.
6158         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
6159         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
6160         Remove.
6161         (atomic_loaddi): Revise fence expansion to only emit fence prior to
6162         load for __ATOMIC_SEQ_CST model.
6163         (atomic_loaddi_1): Remove float register target.
6164         (atomic_storedi): Handle CONST_INT values.
6165         (atomic_storedi_1): Remove float register source.  Add special case
6166         for zero value.
6167         (memory_barrier): New expander and insn.
6169 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
6171         PR tree-optimization/86835
6172         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
6173         new_stmt after def_gsi, make sure to insert new_square_stmt after
6174         that stmt, not 2 stmts before it.
6176 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
6178         PR target/82418
6179         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
6180         instead of SWI48.
6182 2018-08-10  Martin Liska  <mliska@suse.cz>
6184         PR target/83610
6185         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
6186         function type.
6187         * builtins.c (expand_builtin_expect_with_probability):
6188         New function.
6189         (expand_builtin_expect_with_probability): New function.
6190         (build_builtin_expect_predicate): Add new argumnet probability
6191         for BUILT_IN_EXPECT_WITH_PROBABILITY.
6192         (fold_builtin_expect):
6193         (fold_builtin_2):
6194         (fold_builtin_3):
6195         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
6196         * builtins.h (fold_builtin_expect): Set new argument.
6197         * doc/extend.texi: Document __builtin_expect_with_probability.
6198         * doc/invoke.texi: Likewise.
6199         * gimple-fold.c (gimple_fold_call): Pass new argument.
6200         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
6201         also BUILT_IN_EXPECT_WITH_PROBABILITY.
6202         * predict.c (get_predictor_value): New function.
6203         (expr_expected_value): Add new argument probability. Assume
6204         that predictor and probability are always non-null.
6205         (expr_expected_value_1): Likewise.  For __builtin_expect and
6206         __builtin_expect_with_probability set probability.  Handle
6207         combination in binary expressions.
6208         (tree_predict_by_opcode): Simplify code by simply calling
6209         get_predictor_value.
6210         (pass_strip_predict_hints::execute): Add handling of
6211         BUILT_IN_EXPECT_WITH_PROBABILITY.
6212         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
6213         new predictor.
6214         * tree.h (DECL_BUILT_IN_P): New function.
6216 2018-08-10  Martin Liska  <mliska@suse.cz>
6218         PR tree-optimization/85799
6219         * passes.def: Add argument for pass_strip_predict_hints.
6220         * predict.c (class pass_strip_predict_hints): Add new argument
6221         early_p.
6222         (strip_predictor_early): New function.
6223         (pass_strip_predict_hints::execute): Call the function to
6224         strip predictors.
6225         (strip_predict_hints): New function.
6226         * predict.def: Fix comment.
6228 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
6230         * Makefile.in: Clarify which tm.texi to copy over to assert the
6231         right to grant a GFDL license for all.
6233 2018-08-09  Jeff Law  <law@redhat.com>
6235         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6236         unused variable.
6238 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
6240         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6241         prototype.
6243 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6245         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6246         reductions for variable-length vectors.
6248 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
6250         PR other/84889
6251         * common.opt (fdiagnostics-show-line-numbers): New option.
6252         * diagnostic-show-locus.c (class layout): Add fields
6253         "m_show_line_numbers_p" and "m_linenum_width";
6254         (num_digits): New function.
6255         (test_num_digits): New function.
6256         (layout::layout): Initialize new fields.  Update m_x_offset
6257         logic to handle any left margin.
6258         (layout::print_source_line): Print line number when requested.
6259         (layout::start_annotation_line): New member function.
6260         (layout::print_annotation_line): Call it.
6261         (layout::print_leading_fixits): Likewise.
6262         (layout::print_trailing_fixits): Likewise.  Update calls to
6263         move_to_column for new parameter.
6264         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6265         it to potentially call start_annotation_line.
6266         (layout::show_ruler): Call start_annotation_line.
6267         (selftest::test_line_numbers_multiline_range): New selftest.
6268         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6269         and selftest::test_line_numbers_multiline_range.
6270         * diagnostic.c (diagnostic_initialize): Initialize
6271         show_line_numbers_p.
6272         * diagnostic.h (struct diagnostic_context): Add field
6273         "show_line_numbers_p".
6274         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6275         -fno-diagnostics-show-line-numbers.
6276         * dwarf2out.c (gen_producer_string): Add
6277         OPT_fdiagnostics_show_line_numbers to the ignored options.
6278         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6279         one setting" options.
6280         (append_compiler_options): Likewise to the dropped options.
6281         (append_diag_options): Likewise to the passed-on options.
6282         * opts.c (common_handle_option): Handle the new option.
6283         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
6285 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6287         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
6288         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
6289         third argument of type "const signed char" to descriptions of
6290         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
6291         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
6292         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6293         __builtin_bcdsub_ov functions.
6295 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6297         PR tree-optimization/86858
6298         * tree-vect-loop.c (vect_is_simple_reduction): Restore
6299         flow_bb_inside_loop_p calls.
6301 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6303         PR tree-optimization/86871
6304         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6305         instead of gimple_assign_lhs.
6307 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
6309         PR target/86887
6310         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6311         register constraint to operand 0.
6312         (add<mode>3_carryinC): Likewise.
6313         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6315 2018-08-09  Martin Liska  <mliska@suse.cz>
6317         PR c/86895
6318         * common.opt: Remove extra line.
6320 2018-08-09  Martin Liska  <mliska@suse.cz>
6322         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6323         at the end of a line, make first letter capital and end up
6324         a sentence with a dot.
6325         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6326         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6327         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6328         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6329         (PARAM_MAX_ISL_OPERATIONS): Likewise.
6330         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6331         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6332         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6333         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6334         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6335         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6336         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6337         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
6338         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
6339         (PARAM_TREE_REASSOC_WIDTH): Likewise.
6340         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
6341         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
6342         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
6344 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
6346         PR target/84332
6347         * config/s390/s390.c (s390_option_override_internal): Reduce the
6348         stack-clash-protection-probe-interval param if it would be too big
6349         for z900.
6351 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
6353         PR target/46179
6354         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
6355         * config/m68k/m68k.c (handle_move_double): Don't call
6356         m68k_final_prescan_insn.
6357         (m68k_adjust_decorated_operand): Renamed from
6358         m68k_final_prescan_insn, remove first and third operand and
6359         simplify.
6360         (print_operand): Call it.
6361         (print_operand_address): Call it.
6363 2018-08-08  Nathan Sidwell  <nathan@acm.org>
6365         * diagnostic.c (diagnostic_report_current_module): Use
6366         linemap_included_from & linemap_included_from_linemap.
6368 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
6370         * config/aarch64/aarch64-cores.def: Add phecda core.
6371         * config/aarch64/aarch64-tune.md: Regenerate.
6372         * doc/invoke.texi: Add phecda core.
6374 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6376         PR target/85295
6377         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
6378         definitions.
6379         * config/s390/s390.md ("movti"): Add more alternatives for
6380         constant to GPR copies.
6382 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6384         * config/s390/s390.c: Fix whitespace damage throughout the file.
6385         * config/s390/s390.h: Likewise.
6386         * config/s390/tpf.h: Likewise.
6388 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6390         * config/s390/s390.c (s390_loadrelative_operand_p):
6391         Remove TARGET_CPU_ZARCH usages.
6392         (s390_rtx_costs): Likewise.
6393         (s390_legitimate_constant_p): Likewise.
6394         (s390_cannot_force_const_mem): Likewise.
6395         (legitimate_reload_constant_p): Likewise.
6396         (s390_preferred_reload_class): Likewise.
6397         (legitimize_pic_address): Likewise.
6398         (legitimize_tls_address): Likewise.
6399         (s390_split_branches): Removed.
6400         (s390_add_execute): Removed.
6401         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
6402         (s390_mainpool_start): Likewise.
6403         (s390_mainpool_finish): Likewise.
6404         (s390_mainpool_cancel): Removed.
6405         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
6406         (s390_chunkify_cancel): Likewise.
6407         (s390_return_addr_rtx): Likewise.
6408         (s390_register_info): Remove split_branches_pending_p uages.
6409         (s390_optimize_register_info): Likewise.
6410         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
6411         split_branches_pending_p usages.
6412         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
6413         (s390_load_got): Likewise.
6414         (s390_expand_split_stack_prologue): Likewise.
6415         (output_asm_nops): Likewise.
6416         (s390_function_profiler): Likewise.
6417         (s390_emit_call): Likewise.
6418         (s390_conditional_register_usage): Likewise.
6419         (s390_optimize_prologue): Likewise.
6420         (s390_reorg): Remove TARGET_CPU_ZARCH and
6421         split_branches_pending_p usages.
6422         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
6423         usages.
6424         (s390_output_indirect_thunk_function): Likewise.
6425         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
6426         (TARGET_CPU_ZARCH_P): Removed.
6427         (struct machine_function): Remove split_branches_pending_p.
6428         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
6430 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6432         * common/config/s390/s390-common.c (processor_flags_table):
6433         Remove flags.
6434         * config.gcc: Remove with_arch/with_tune support.
6435         * config/s390/2064.md: Remove cpu attribute comparisons.
6436         * config/s390/driver-native.c (s390_host_detect_local_cpu):
6437         Remove MTN.
6438         * config/s390/linux.h (ASM_SPEC):
6439         Remove -march support.
6440         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
6441         Use a table to get an arch level.
6442         * config/s390/s390-opts.h (enum processor_type):
6443         Remove enum values.
6444         * config/s390/s390.c
6445         (processor_table): Remove entries, add arch_level values.
6446         (s390_issue_rate): Remove cases.
6447         (s390_option_override): Adjust
6448         s390_option_override_internal() call.
6449         (s390_option_override_internal): Remove deprecation warning.
6450         (s390_valid_target_attribute_tree): Adjust
6451         s390_option_override_internal() call.
6452         * config/s390/s390.h (struct s390_processor):
6453         Share with s390-c.c, add arch_level field.
6454         * config/s390/s390.md:
6455         Remove occurrences in cpu attribute.
6456         * config/s390/s390.opt: Remove -march/-mtune support.
6457         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
6458         * doc/invoke.texi: Remove deprecation warning.
6460 2018-08-08  Luis Machado  <luis.machado@linaro.org>
6462         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
6463         global.
6464         (qdf24xx_tunings): Set vector cost structure to
6465         qdf24xx_vector_cost.
6467         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
6468         <register_sextend>: Set to 3.
6470 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
6472         PR target/86838
6473         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
6474         * config/aarch64/aarch64-simd.md
6475         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
6476         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
6477         (aarch64_frecpx<mode>): ...this new pattern.
6478         * config/aarch64/aarch64-simd-builtins.def: Remove comment
6479         about aarch64_frecp<FRECP:frecp_suffix><mode>.
6481 2018-08-07  Martin Liska  <mliska@suse.cz>
6483         PR middle-end/83023
6484         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
6485         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
6486         * predict.def (PRED_MALLOC_NONNULL): New predictor.
6487         * doc/extend.texi: Document that malloc attribute adds
6488         hit to compiler.
6490 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
6492         PR target/86785
6493         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6494         Define to speculation_safe_value_not_needed.
6496 2018-08-06  Jeff Law  <law@redhat.com>
6498         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
6499         the vr_values instance to cprop_into_stmt.
6500         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
6501         (cprop_operand): Also query EVRP to determine if OP is a constant.
6503 2018-08-06  Nathan Sidwell  <nathan@acm.org>
6505         * diagnostic.c (diagnostic_report_current_module): Reroll
6506         included-at loop.  Translate text.
6508 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
6510         * function-tests.c (selftest::test_expansion_to_rtl): Call
6511         free_after_compilation.
6513 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6515         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
6517 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
6519         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
6520         loops with memory block operations from getting unrolled.
6522 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
6524         PR target/86807
6525         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6526         Define to speculation_safe_value_not_needed.
6528 2018-08-06  Jeff Law  <law@redhat.com>
6530         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
6531         assert.
6533 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6535         PR target/86662
6536         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
6537         with all enabled __intN types.
6539         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
6541 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6543         * alias.c (record_set): Check for clobber high.
6544         * cfgexpand.c (expand_gimple_stmt): Likewise.
6545         * combine-stack-adj.c (single_set_for_csa): Likewise.
6546         * combine.c (find_single_use_1): Likewise.
6547         (set_nonzero_bits_and_sign_copies): Likewise.
6548         (get_combine_src_dest): Likewise.
6549         (is_parallel_of_n_reg_sets): Likewise.
6550         (try_combine): Likewise.
6551         (record_dead_and_set_regs_1): Likewise.
6552         (reg_dead_at_p_1): Likewise.
6553         (reg_dead_at_p): Likewise.
6554         * dce.c (deletable_insn_p): Likewise.
6555         (mark_nonreg_stores_1): Likewise.
6556         (mark_nonreg_stores_2): Likewise.
6557         * df-scan.c (df_find_hard_reg_defs): Likewise.
6558         (df_uses_record): Likewise.
6559         (df_get_call_refs): Likewise.
6560         * dwarf2out.c (mem_loc_descriptor): Likewise.
6561         * haifa-sched.c (haifa_classify_rtx): Likewise.
6562         * ira-build.c (create_insn_allocnos): Likewise.
6563         * ira-costs.c (scan_one_insn): Likewise.
6564         * ira.c (equiv_init_movable_p): Likewise.
6565         (rtx_moveable_p): Likewise.
6566         (interesting_dest_for_shprep): Likewise.
6567         * jump.c (mark_jump_label_1): Likewise.
6568         * postreload-gcse.c (record_opr_changes): Likewise.
6569         * postreload.c (reload_cse_simplify): Likewise.
6570         (struct reg_use): Add source expr.
6571         (reload_combine): Check for clobber high.
6572         (reload_combine_note_use): Likewise.
6573         (reload_cse_move2add): Likewise.
6574         (move2add_note_store): Likewise.
6575         * print-rtl.c (print_pattern): Likewise.
6576         * recog.c (decode_asm_operands): Likewise.
6577         (store_data_bypass_p): Likewise.
6578         (if_test_bypass_p): Likewise.
6579         * regcprop.c (kill_clobbered_value): Likewise.
6580         (kill_set_value): Likewise.
6581         * reginfo.c (reg_scan_mark_refs): Likewise.
6582         * reload1.c (maybe_fix_stack_asms): Likewise.
6583         (eliminate_regs_1): Likewise.
6584         (elimination_effects): Likewise.
6585         (mark_not_eliminable): Likewise.
6586         (scan_paradoxical_subregs): Likewise.
6587         (forget_old_reloads_1): Likewise.
6588         * reorg.c (find_end_label): Likewise.
6589         (try_merge_delay_insns): Likewise.
6590         (redundant_insn): Likewise.
6591         (own_thread_p): Likewise.
6592         (fill_simple_delay_slots): Likewise.
6593         (fill_slots_from_thread): Likewise.
6594         (dbr_schedule): Likewise.
6595         * resource.c (update_live_status): Likewise.
6596         (mark_referenced_resources): Likewise.
6597         (mark_set_resources): Likewise.
6598         * rtl.c (copy_rtx): Likewise.
6599         * rtlanal.c (reg_referenced_p): Likewise.
6600         (single_set_2): Likewise.
6601         (noop_move_p): Likewise.
6602         (note_stores): Likewise.
6603         * sched-deps.c (sched_analyze_reg): Likewise.
6604         (sched_analyze_insn): Likewise.
6606 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6608         * cse.c (invalidate_reg): New function extracted from...
6609         (invalidate): ...here.
6610         (canonicalize_insn): Check for clobber high.
6611         (invalidate_from_clobbers): invalidate clobber highs.
6612         (invalidate_from_sets_and_clobbers): Likewise.
6613         (count_reg_usage): Check for clobber high.
6614         (insn_live_p): Likewise.
6615         * cselib.c (cselib_expand_value_rtx_1):Likewise.
6616         (cselib_invalidate_regno): Check for clobber in setter.
6617         (cselib_invalidate_rtx): Pass through setter.
6618         (cselib_invalidate_rtx_note_stores):
6619         (cselib_process_insn): Check for clobber high.
6620         * cselib.h (cselib_invalidate_rtx): Add operand.
6622 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6624         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
6625         (mark_not_eliminable): Likewise.
6626         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
6627         * lra-lives.c (process_bb_lives): Check for clobber high.
6628         * lra.c (new_insn_reg): Remember clobber highs.
6629         (collect_non_operand_hard_regs): Check for clobber high.
6630         (lra_set_insn_recog_data): Likewise.
6631         (add_regs_to_insn_regno_info): Likewise.
6632         (lra_update_insn_regno_info): Likewise.
6634 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6636         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
6637         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
6639 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6641         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
6642         (copy_insn_1): Likewise.
6643         (gen_hard_reg_clobber_high): New gen function.
6644         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
6645         * genemit.c (gen_exp): Likewise.
6646         (gen_emit_seq): Pass through info.
6647         (gen_insn): Check for CLOBBER_HIGH.
6648         (gen_expand): Pass through info.
6649         (gen_split): Likewise.
6650         (output_add_clobbers): Likewise.
6651         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
6652         (remove_clobbers): Likewise.
6653         * rtl.h (gen_hard_reg_clobber_high): New declaration.
6655 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6657         * doc/rtl.texi (clobber_high): Add.
6658         (parallel): Add in clobber high
6659         * rtl.c (rtl_check_failed_code3): Add function.
6660         * rtl.def (CLOBBER_HIGH): Add expression.
6661         * rtl.h (RTL_CHECKC3): Add macro.
6662         (rtl_check_failed_code3): Add declaration.
6663         (XC3EXP): Add macro.
6665 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
6667         PR target/86386
6668         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
6669         cfun->machine->max_used_stack_alignment if needed.
6671 2018-08-04  Martin Sebor  <msebor@redhat.com>
6673         PR tree-optimization/86571
6674         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
6675         NaN output to 4.
6677 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
6679         PR target/86799
6680         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6681         Define.
6683 2018-08-03  Jeff Law  <law@redhat.com>
6685         PR target/86795
6686         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6687         Define to speculation_safe_value_not_needed.
6689 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
6691         * doc/gcov.texi (-x): Remove duplicate "to".
6692         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
6693         (-Wif-not-aligned): Remove duplicate "is".
6694         (-flto): Remove duplicate "the".
6695         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
6696         duplicate "v5.00.b".
6697         (MSP430 Options): Remove duplicate "and" from the description
6698         of "-mgprel-sec=regexp".
6699         (x86 Options): Remove duplicate copies of "vmldLog102" and
6700         vmlsLog104 from description of "-mveclibabi=type".
6702 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
6704         * internal-fn.h (first_commutative_argument): Declare.
6705         * internal-fn.c (first_commutative_argument): New function.
6706         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
6707         restrictions for pattern statements.  Use first_commutative_argument
6708         to look for commutative operands in calls to internal functions.
6710 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
6712         * Makefile.in (wide-int-range.o): New.
6713         * tree-vrp.c: Move all the wide_int_* functions to...
6714         * wide-int-range.cc: ...here.
6715         * tree-vrp.h: Move all the wide_int_* prototypes to...
6716         * wide-int-range.h: ...here.
6718 2018-08-03  Tom de Vries  <tdevries@suse.de>
6720         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
6721         UI_NONE.
6722         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
6723         * except.c (output_function_exception_table): Do early exit if
6724         targetm_common.except_unwind_info (&global_options) == UI_NONE.
6726 2018-08-03  Martin Liska  <mliska@suse.cz>
6728         * predict.c (dump_prediction): Change to 2 digits
6729         in fraction part.
6731 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
6733         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
6734         neon_dup_q to...
6735         (falkor_am_1_gtov_gtov): ... a new insn reservation.
6737 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
6739         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
6740         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
6741         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
6742         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
6743         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
6745 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
6747         * diagnostic-show-locus.c (diagnostic_show_locus): Use
6748         pp_take_prefix when saving the existing prefix.
6749         * diagnostic.c (diagnostic_append_note): Likewise.
6750         * langhooks.c (lhd_print_error_function): Likewise.
6751         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
6752         param's type.  Free the existing prefix.
6753         (pp_take_prefix): New function.
6754         (pretty_printer::pretty_printer): Drop the prefix parameter.
6755         Rename the length parameter to match the comment.
6756         (pretty_printer::~pretty_printer): Free the prefix.
6757         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
6758         parameter.
6759         (struct pretty_printer): Drop the "const" from "prefix" field's
6760         type and clarify memory management.
6761         (pp_set_prefix): Drop the "const" from the 2nd param.
6762         (pp_take_prefix): New decl.
6764 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
6766         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
6767         for word_mode_ok here instead of passing as argument.
6768         (expand_block_compare): Change select_block_compare_mode() call.
6769         (expand_strncmp_gpr_sequence): New function.
6770         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
6772 2018-08-02  Jeff Law  <law@redhat.com>
6774         PR target/86790
6775         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6776         Define to speculation_safe_value_not_needed.
6778         PR target/86784
6779         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6780         Define to speculation_safe_value_not_needed.
6782 2018-08-02  Tom de Vries  <tdevries@suse.de>
6784         PR target/86660
6785         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
6786         function.  Return UI_TARGET unconditionally.
6787         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
6788         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
6790 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6792         * genemit.c (print_overload_test): Fix typo.
6794 2018-08-02  Richard Biener  <rguenther@suse.de>
6796         PR tree-optimization/86816
6797         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
6798         which checks for value availability before querying it.
6799         (gvn_uses_equal): Use it.
6800         (same_succ_hash): Likewise.
6801         (gimple_equal_p): Likewise.
6803 2018-08-02  Nick Clifton  <nickc@redhat.com>
6805         PR target/86813
6806         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6807         Define to speculation_safe_value_not_needed.
6809         PR target/86810
6810         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6811         Define to speculation_safe_value_not_needed.
6813         PR target/86810
6814         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6815         Define to speculation_safe_value_not_needed.
6817         PR target/86803
6818         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6819         Define to speculation_safe_value_not_needed.
6821         PR target/86797
6822         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6823         Define to speculation_safe_value_not_needed.
6825         PR target/86791
6826         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6827         Define to speculation_safe_value_not_needed.
6829         PR target/86789
6830         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6831         Define to speculation_safe_value_not_needed.
6833         PR target/86787
6834         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6835         Define to speculation_safe_value_not_needed.
6837         PR target/86782
6838         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6839         speculation_safe_value_not_needed.
6841         PR target/86781
6842         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
6843         to speculation_safe_value_not_needed.
6845 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6847         * doc/md.texi: Expand the documentation of instruction names
6848         to mention port-local uses.  Document '@' in pattern names.
6849         * read-md.h (overloaded_instance, overloaded_name): New structs.
6850         (mapping): Declare.
6851         (md_reader::handle_overloaded_name): New member function.
6852         (md_reader::get_overloads): Likewise.
6853         (md_reader::m_first_overload): New member variable.
6854         (md_reader::m_next_overload_ptr): Likewise.
6855         (md_reader::m_overloads_htab): Likewise.
6856         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
6857         m_next_overload_ptr and m_overloads_htab.
6858         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
6859         (get_mode_token, get_code_token, get_int_token): New functions.
6860         (map_attr_string): Add an optional argument that passes back
6861         the associated iterator.
6862         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
6863         (md_reader::handle_overloaded_name, add_overload_instance): New
6864         functions.
6865         (apply_iterators): Handle '@' names.  Report an error if '@'
6866         is used without iterators.
6867         (initialize_iterators): Initialize the new iterator_group fields.
6868         * genopinit.c (handle_overloaded_code_for)
6869         (handle_overloaded_gen): New functions.
6870         (main): Use them to print declarations of maybe_code_for_* and
6871         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
6872         * genemit.c (print_overload_arguments, print_overload_test)
6873         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
6874         (main): Use it to print definitions of maybe_code_for_* and
6875         maybe_gen_* functions.
6876         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
6877         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
6878         instead of explicit mode checks.
6879         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
6880         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
6881         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
6882         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
6883         (aarch64_expand_compare_and_swap): Likewise
6884         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
6885         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
6886         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
6887         (aarch64_constant_pool_reload_icode): Delete.
6888         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
6889         instead of aarch64_constant_pool_reload_icode.  Use
6890         code_for_aarch64_reload_mov instead of explicit mode checks.
6891         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
6892         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
6893         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
6894         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
6895         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
6896         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
6897         (aarch64_atomic_load_op_code): Delete.
6898         (aarch64_emit_atomic_load_op): Likewise.
6899         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
6900         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
6901         instead of aarch64_emit_atomic_load_op.
6902         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
6903         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
6904         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
6905         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
6906         character before the pattern name.
6907         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
6908         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
6909         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
6910         (aarch64_frecps<mode>): Likewise.
6911         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
6912         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
6913         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
6914         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
6915         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
6917 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6919         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
6920         Allow HFmode constants if TARGET_FP_F16INST.
6922 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
6924         PR target/86014
6925         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
6926         No longer check last store for clobber of address register.
6928 2018-08-02  Martin Liska  <mliska@suse.cz>
6930         PR gcov-profile/86817
6931         * gcov.c (process_all_functions): New function.
6932         (main): Call it.
6933         (process_file): Move functions processing to
6934         process_all_functions.
6936 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
6938         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
6939         "const" to the "gimple *" and "rtx_insn *" parameters.
6940         * dumpfile.h (dump_user_location_t::dump_user_location_t):
6941         Likewise.
6942         (dump_location_t::dump_location_t): Likewise.
6944 2018-08-01  Martin Sebor  <msebor@redhat.com>
6946         PR tree-optimization/86650
6947         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
6948         rather than a "gcall *".  Directly pass the data of interest
6949         to percent_K_format, rather than building a temporary CALL_EXPR
6950         to hold it.
6951         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
6952         (gimple_fold_builtin_strncat): Adjust.
6953         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
6954         gcall* argument with gimple*.
6955         * gimple-ssa-warn-restrict.c (check_call): Same.
6956         (wrestrict_dom_walker::before_dom_children): Same.
6957         (builtin_access::builtin_access): Same.
6958         (check_bounds_or_overlap): Same
6959         (maybe_diag_overlap): Same.
6960         (maybe_diag_offset_bounds): Same.
6961         * tree-diagnostic.c (default_tree_printer): Move usage of
6962         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
6963         to this callsite.
6964         * tree-pretty-print.c (percent_K_format): Add argument.
6965         * tree-pretty-print.h: Add argument.
6966         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
6967         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
6968         (maybe_diag_stxncpy_trunc): Same.
6969         (handle_builtin_stxncpy): Same.
6970         (handle_builtin_strcat): Same.
6972 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6974         * match.pd: Optimise pointer range checks.
6976 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6978         PR tree-optimization/86758
6979         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
6980         to remove pattern statements.
6982 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6984         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
6985         result of dfs_enumerate_from when constructing stmt_vec_infos,
6986         instead of additionally calling get_loop_body.
6988 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6990         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
6991         parameter.
6992         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
6993         When creating an iv, assert that the step is not known to be zero.
6994         (vect_setup_realignment): Update call accordingly.
6995         * tree-vect-stmts.c (vectorizable_store): Likewise.
6996         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
6998 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7000         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
7001         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
7002         (vectorizable_reduction): Likewise.
7003         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7004         (vect_detect_hybrid_slp_stmts): Likewise.
7005         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7007 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
7009         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
7010         (wide_int_set_zero_nonzero_bits): ...this.
7011         (zero_nonzero_bits_from_vr): Rename to...
7012         (vrp_set_zero_nonzero_bits): ...this.
7013         (extract_range_from_multiplicative_op_1): Abstract wide int
7014         code...
7015         (wide_int_range_multiplicative_op): ...here.
7016         (extract_range_from_binary_expr_1): Extract wide int binary
7017         operations into their own functions.
7018         (wide_int_range_lshift): New.
7019         (wide_int_range_can_optimize_bit_op): New.
7020         (wide_int_range_shift_undefined_p): New.
7021         (wide_int_range_bit_xor): New.
7022         (wide_int_range_bit_ior): New.
7023         (wide_int_range_bit_and): New.
7024         (wide_int_range_trunc_mod): New.
7025         (extract_range_into_wide_ints): New.
7026         (vrp_shift_undefined_p): New.
7027         (extract_range_from_multiplicative_op): New.
7028         (vrp_can_optimize_bit_op): New.
7029         * tree-vrp.h (value_range::dump): New.
7030         (wide_int_range_multiplicative_op): New.
7031         (wide_int_range_lshift):New.
7032         (wide_int_range_shift_undefined_p): New.
7033         (wide_int_range_bit_xor): New.
7034         (wide_int_range_bit_ior): New.
7035         (wide_int_range_bit_and): New.
7036         (wide_int_range_trunc_mod): New.
7037         (zero_nonzero_bits_from_bounds): Rename to...
7038         (wide_int_set_zero_nonzero_bits): ...this.
7039         (zero_nonzero_bits_from_vr): Rename to...
7040         (vrp_set_zero_nonzero_bits): ...this.
7041         (range_easy_mask_min_max): Rename to...
7042         (wide_int_range_can_optimize_bit_op): this.
7043         * vr-values.c (simplify_bit_ops_using_ranges): Rename
7044         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
7046 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7048         * tree-vectorizer.h (vect_orig_stmt): New function.
7049         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
7050         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
7051         (vect_create_epilog_for_reduction): Likewise.
7052         (vectorizable_live_operation): Likewise.
7053         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7054         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
7055         * tree-vect-stmts.c (vectorizable_call): Likewise.
7056         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
7058 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7060         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
7061         argument.
7062         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
7063         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
7064         (vect_transform_loop): Likewise.
7065         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7067 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7069         * tree-vectorizer.h (vect_schedule_slp): Return void.
7070         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7071         (vect_schedule_slp): Likewise.
7073 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7075         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
7076         argument.
7077         (vect_transform_loop): Update calls accordingly.  Schedule SLP
7078         instances before the main loop, if any exist.
7080 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7082         PR tree-optimization/86749
7083         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
7084         If the lhs is used in a COND_EXPR, check that it is being used
7085         as the "then" or "else" value.
7087 2018-08-01  Tom de Vries  <tdevries@suse.de>
7089         PR target/86800
7090         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7091         speculation_safe_value_not_needed.
7093 2018-08-01  Richard Biener  <rguenther@suse.de>
7095         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
7096         as base and offset.
7098 2018-08-01  Martin Liska  <mliska@suse.cz>
7100         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
7101         format how successful transformation is dumped.
7102         (gimple_mod_pow2_value_transform): Likewise.
7103         (gimple_mod_subtract_transform): Likewise.
7104         (gimple_stringops_transform): Likewise.
7106 2018-08-01  Martin Liska  <mliska@suse.cz>
7108         PR value-prof/35543
7109         * value-prof.c (interesting_stringop_to_profile_p):
7110         Simplify the code and add BUILT_IN_MEMMOVE.
7111         (gimple_stringops_transform): Likewise.
7113 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
7115         * config/aarch64/aarch64-simd.md
7116         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
7117         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
7118         use GPI iterator instead of SI mode.
7120 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7122         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
7123         rs6000_speculation_barrier.
7124         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
7125         new barrier pattern name.
7127 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7129         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
7130         (speculation_barrier): New insn.
7132 2018-08-01  Richard Biener  <rguenther@suse.de>
7134         PR bootstrap/86724
7135         * graphite.h: Include isl/id.h and isl/space.h to allow build
7136         with ISL 0.20.
7138 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
7140         PR target/86651
7141         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7142         mode for COFF targets.
7143         * defaults.h (TARGET_COFF): Define.
7144         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
7145         TARGET_COFF): Define.
7146         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
7147         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
7148         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
7150 2018-07-31  Alexandre Oliva <oliva@adacore.com>
7152         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
7153         * gimple-streamer-out.c (output_bb): Save it.
7154         * lto-streamer-in.c (input_struct_function_base): Restore
7155         instance discriminator if available.  Create map on demand.
7156         * lto-streamer-out.c (output_struct_function_base): Save it if
7157         available.
7158         * final.c (decl_to_instance_map): Document LTO strategy.
7160 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
7161             Olivier Hainque  <hainque@adacore.com>
7163         * debug.h (decl_to_instance_map_t): New type.
7164         (decl_to_instance_map): Declare.
7165         (maybe_create_decl_to_instance_map): New inline function.
7166         * final.c (bb_discriminator, last_bb_discriminator): New statics,
7167         to track basic block discriminators.
7168         (final_start_function_1): Initialize them.
7169         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
7170         bb_discriminator.
7171         (decl_to_instance_map): New variable.
7172         (map_decl_to_instance, maybe_set_discriminator): New functions.
7173         (notice_source_line): Set discriminator.
7175 2018-07-31  Ian Lance Taylor  <iant@golang.org>
7177         * targhooks.c (default_have_speculation_safe_value): Add
7178         ATTRIBUTE_UNUSED.
7180 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7182         * dump-context.h: Include "pretty-print.h".
7183         (dump_context::refresh_dumps_are_enabled): New decl.
7184         (dump_context::emit_item): New decl.
7185         (class dump_context): Add fields "m_test_pp" and
7186         "m_test_pp_flags".
7187         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7188         (temp_dump_context::get_dumped_text): New decl.
7189         (class temp_dump_context): Add field "m_pp".
7190         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
7191         (dump_context::refresh_dumps_are_enabled): ...and add a test for
7192         m_test_pp.
7193         (set_dump_file): Update for above change.
7194         (set_alt_dump_file): Likewise.
7195         (dump_loc): New overload, taking a pretty_printer *.
7196         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
7197         to any test pretty-printer.
7198         (make_item_for_dump_gimple_stmt): New function, adapted from
7199         optinfo::add_gimple_stmt.
7200         (dump_context::dump_gimple_stmt): Call it, and use the result,
7201         eliminating the direct usage of dump_file and alt_dump_file in
7202         favor of indirectly using them via emit_item.
7203         (make_item_for_dump_gimple_expr): New function, adapted from
7204         optinfo::add_gimple_expr.
7205         (dump_context::dump_gimple_expr): Call it, and use the result,
7206         eliminating the direct usage of dump_file and alt_dump_file in
7207         favor of indirectly using them via emit_item.
7208         (make_item_for_dump_generic_expr): New function, adapted from
7209         optinfo::add_tree.
7210         (dump_context::dump_generic_expr): Call it, and use the result,
7211         eliminating the direct usage of dump_file and alt_dump_file in
7212         favor of indirectly using them via emit_item.
7213         (make_item_for_dump_printf_va): New function, adapted from
7214         optinfo::add_printf_va.
7215         (make_item_for_dump_printf): New function.
7216         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
7217         and use the result, eliminating the direct usage of dump_file and
7218         alt_dump_file in favor of indirectly using them via emit_item.
7219         (make_item_for_dump_dec): New function.
7220         (dump_context::dump_dec): Call it, and use the result,
7221         eliminating the direct usage of dump_file and alt_dump_file in
7222         favor of indirectly using them via emit_item.
7223         (make_item_for_dump_symtab_node): New function, adapted from
7224         optinfo::add_symtab_node.
7225         (dump_context::dump_symtab_node): Call it, and use the result,
7226         eliminating the direct usage of dump_file and alt_dump_file in
7227         favor of indirectly using them via emit_item.
7228         (dump_context::begin_scope): Reimplement, avoiding direct usage
7229         of dump_file and alt_dump_file in favor of indirectly using them
7230         via emit_item.
7231         (dump_context::emit_item): New member function.
7232         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7233         Set up test pretty-printer on the underlying context.  Call
7234         refresh_dumps_are_enabled.
7235         (temp_dump_context::~temp_dump_context): Call
7236         refresh_dumps_are_enabled.
7237         (temp_dump_context::get_dumped_text): New member function.
7238         (selftest::verify_dumped_text): New function.
7239         (ASSERT_DUMPED_TEXT_EQ): New macro.
7240         (selftest::test_capture_of_dump_calls): Run all tests twice, with
7241         and then without optinfo enabled.  Add uses of
7242         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
7243         * dumpfile.h: Update comment for the dump_* API.
7244         * optinfo-emit-json.cc
7245         (selftest::test_building_json_from_dump_calls): Update for new
7246         param for temp_dump_context ctor.
7247         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7248         and "m_owned" field.
7249         (optinfo_item::~optinfo_item): Likewise.
7250         (optinfo::add_item): New member function.
7251         (optinfo::emit): Update comment.
7252         (optinfo::add_string): Delete.
7253         (optinfo::add_printf): Delete.
7254         (optinfo::add_printf_va): Delete.
7255         (optinfo::add_gimple_stmt): Delete.
7256         (optinfo::add_gimple_expr): Delete.
7257         (optinfo::add_tree): Delete.
7258         (optinfo::add_symtab_node): Delete.
7259         (optinfo::add_dec): Delete.
7260         * optinfo.h (class dump_context): New forward decl.
7261         (optinfo::add_item): New decl.
7262         (optinfo::add_string): Delete.
7263         (optinfo::add_printf): Delete.
7264         (optinfo::add_printf_va): Delete.
7265         (optinfo::add_gimple_stmt): Delete.
7266         (optinfo::add_gimple_expr): Delete.
7267         (optinfo::add_tree): Delete.
7268         (optinfo::add_symtab_node): Delete.
7269         (optinfo::add_dec): Delete.
7270         (optinfo::add_poly_int): Delete.
7271         (optinfo_item::optinfo_item): Remove "owned" param.
7272         (class optinfo_item): Remove field "m_owned".
7274 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7276         PR middle-end/86705
7277         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7278         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7279         requested variable alignment.
7280         (expand_one_ssa_partition): Likewise.
7281         (expand_one_var): Likewise.
7283 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7285         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7286         to speculation_safe_value_not_needed.
7288 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7290         * targhooks.h (speculation_safe_value_not_needed): New prototype.
7291         * targhooks.c (speculation_safe_value_not_needed): New function.
7292         * target.def (have_speculation_safe_value): Update documentation.
7293         * doc/tm.texi: Regenerated.
7295 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7297         * config/aarch64/iterators.md (ALLI_TI): New iterator.
7298         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7299         expand.
7300         (despeculate_copy<ALLI:mode>_insn): New insn.
7301         (despeculate_copyti_insn): New insn.
7302         (despeculate_simple<ALLI:mode>): New insn
7303         (despeculate_simpleti): New insn.
7304         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7305         function.
7306         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7307         aarch64_speculation_safe_value.
7308         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7310 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7312         * config/aarch64/aarch64-speculation.cc: New file.
7313         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7314         before pass_reorder_blocks.
7315         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7316         prototype.
7317         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7318         X14 and X15 when tracking speculation.
7319         * config/aarch64/aarch64.md (register name constants): Add
7320         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7321         (unspec): Add UNSPEC_SPECULATION_TRACKER.
7322         (speculation_barrier): New insn attribute.
7323         (cmp<mode>): Allow SP in comparisons.
7324         (speculation_tracker): New insn.
7325         (speculation_barrier): Add speculation_barrier attribute.
7326         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7327         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7328         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7330 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7332         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7333         aarch64_track_speculation is true.
7334         (tb<optab><mode>1): Likewise.
7335         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7336         generate CB[N]Z when tracking speculation.
7337         (aarch64_split_compare_and_swap): Likewise.
7338         (aarch64_split_atomic_op): Likewise.
7340 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7342         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
7344 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7346         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
7347         (speculation_barrier): New insn.
7349 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7351         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
7352         * config/arm/arm.md (speculation_barrier): New expand.
7353         (speculation_barrier_insn): New pattern.
7355 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7357         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
7358         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
7359         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
7360         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
7361         list.
7362         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
7363         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
7364         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
7365         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
7366         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
7367         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
7368         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
7369         * builtins.c (expand_speculation_safe_value): New function.
7370         (expand_builtin): Call it.
7371         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
7372         * doc/extend.texi: Document __builtin_speculation_safe_value.
7373         * doc/md.texi: Document "speculation_barrier" pattern.
7374         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
7375         TARGET_HAVE_SPECULATION_SAFE_VALUE.
7376         * doc/tm.texi: Regenerated.
7377         * target.def (have_speculation_safe_value, speculation_safe_value): New
7378         hooks.
7379         * targhooks.c (default_have_speculation_safe_value): New function.
7380         (default_speculation_safe_value): New function.
7381         * targhooks.h (default_have_speculation_safe_value): Add prototype.
7382         (default_speculation_safe_value): Add prototype.
7384 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7386         * dump-context.h (dump_context::dump_loc): New decl.
7387         * dumpfile.c (dump_context::dump_loc): New member function.
7388         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
7389         and dump_gimple_stmt.
7390         (dump_context::dump_gimple_expr_loc): Likewise, using
7391         dump_gimple_expr.
7392         (dump_context::dump_generic_expr_loc): Likewise, using
7393         dump_generic_expr.
7394         (dump_context::dump_printf_loc_va): Likewise, using
7395         dump_printf_va.
7396         (dump_context::begin_scope): Explicitly using the global function
7397         "dump_loc", rather than the member function.
7399 2018-07-31  Martin Sebor  <msebor@redhat.com>
7401         PR tree-optimization/86741
7402         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
7404 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
7406         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
7407         depend on whether prefetch instructions will be emitted or not.
7408         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
7409         will be emitted or not.
7410         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
7411         (TARGET_SETMEM_PFD): New macros.
7413 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7415         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
7416         (NULL_STMT_VEC_INFO): Delete.
7417         (stmt_vec_info::operator*): Likewise.
7418         (stmt_vec_info::operator gimple *): Likewise.
7419         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
7420         of NULL_STMT_VEC_INFO.
7421         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7422         (vect_reassociating_reduction_p): Likewise.
7423         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7424         (vectorizable_store): Likewise.
7425         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
7426         (vec_info::free_stmt_vec_infos): Likewise.
7428 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7430         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
7431         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
7433 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7435         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
7436         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7437         (vec_info::free_stmt_vec_info): New private member functions.
7438         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
7439         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
7440         * tree-parloops.c (gather_scalar_reductions): Remove calls to
7441         set_stmt_vec_info_vec and free_stmt_vec_infos.
7442         * tree-vect-loop.c (_loop_vec_info): Remove call to
7443         set_stmt_vec_info_vec.
7444         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
7445         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
7446         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
7447         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7448         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
7449         assignments in {vec_info::,}new_stmt_vec_info that are redundant
7450         with the clearing in the xcalloc.
7451         (stmt_vec_info_vec): Delete.
7452         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
7453         (vectorize_loops): Likewise.
7454         (vec_info::~vec_info): Remove argument from call to
7455         free_stmt_vec_infos.
7456         (vec_info::add_stmt): Remove vinfo argument from call to
7457         new_stmt_vec_info.
7459 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7461         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
7462         rather than a gimple stmt.
7463         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
7464         information for pattern statements when passed the original
7465         statement; instead wait to be passed the pattern statement itself.
7466         Don't call set_vinfo_for_stmt here.
7467         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
7468         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
7469         stmt_vec_infos here.
7470         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
7471         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
7472         stmt_vec_infos entry.
7474 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7476         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
7477         * tree-vectorizer.c (vec_info::replace_stmt): New function.
7478         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
7479         * tree-vect-stmts.c (vectorizable_call): Likewise.
7480         (vectorizable_simd_clone_call): Likewise.
7482 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7484         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
7485         * tree-vectorizer.c (vec_info::remove_stmt): New function.
7486         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
7487         * tree-vect-loop.c (vect_transform_loop): Likewise.
7488         * tree-vect-slp.c (vect_schedule_slp): Likewise.
7489         * tree-vect-stmts.c (vect_remove_stores): Likewise.
7491 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7493         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
7494         (vect_dr_stmt): Delete.
7495         * tree-vectorizer.c (vec_info::lookup_dr): New function.
7496         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
7497         of DR_VECT_AUX.
7498         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7499         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
7500         (vect_verify_datarefs_alignment, vect_peeling_supportable)
7501         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7502         (vect_analyze_data_refs): Likewise.
7503         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
7504         argument.
7505         (vect_find_same_alignment_drs): Likewise.
7506         (vect_slp_analyze_node_dependences): Update calls accordingly.
7507         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7508         instead of DR_VECT_AUX.
7509         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
7510         of a vector data references.  Use vec_info::lookup_dr instead of
7511         DR_VECT_AUX.
7512         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
7513         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7514         instead of DR_VECT_AUX.
7516 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7518         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
7519         dr_vec_info.
7520         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7521         accordingly.
7522         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7523         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7524         (vect_gen_prolog_loop_niters): Likewise.
7526 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7528         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
7529         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
7530         (vect_known_alignment_in_bytes, vect_dr_behavior)
7531         (vect_get_scalar_dr_size): Take references as dr_vec_infos
7532         instead of data_references.  Update calls to other routines for
7533         which the same change has been made.
7534         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
7535         dr_vec_infos instead of stmt_vec_infos.
7536         (vect_analyze_data_ref_dependence): Update call accordingly.
7537         (vect_slp_analyze_data_ref_dependence)
7538         (vect_record_base_alignments): Use DR_VECT_AUX.
7539         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
7540         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
7541         (vector_alignment_reachable_p, vect_get_data_access_cost)
7542         (vect_peeling_supportable, vect_analyze_group_access_1)
7543         (vect_analyze_group_access, vect_analyze_data_ref_access)
7544         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
7545         (vect_compile_time_alias, vect_small_gap_p)
7546         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
7547         (vect_supportable_dr_alignment): Take references as dr_vec_infos
7548         instead of data_references.  Update calls to other routines for
7549         which the same change has been made.
7550         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
7551         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7552         (vect_slp_analyze_and_verify_node_alignment)
7553         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7554         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7555         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
7556         above changes.
7557         (_vect_peel_info::dr): Replace with...
7558         (_vect_peel_info::dr_info): ...this new field.
7559         (vect_peeling_hash_get_most_frequent)
7560         (vect_peeling_hash_choose_best_peeling): Update accordingly.
7561         (vect_peeling_hash_get_lowest_cost):
7562         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
7563         routines for which the same change has been made.
7564         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
7565         data_reference.
7566         * tree-vect-loop-manip.c (get_misalign_in_elems)
7567         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
7568         above changes.
7569         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7570         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
7571         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
7572         (get_group_load_store_type, get_negative_load_store_type)
7573         (vect_get_data_ptr_increment, vectorizable_store)
7574         (vectorizable_load): Likewise.
7575         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
7576         Update calls to other routines for which the same change has been made.
7578 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7580         * tree-vectorizer.h (vec_info::move_dr): New member function.
7581         (dataref_aux): Rename to...
7582         (dr_vec_info): ...this and add "dr" and "stmt" fields.
7583         (_stmt_vec_info::dr_aux): Update accordingly.
7584         (_stmt_vec_info::data_ref_info): Delete.
7585         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
7586         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
7587         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
7588         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
7589         of data_ref.
7590         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
7591         (STMT_VINFO_DR_INFO): New macro.
7592         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
7593         (set_dr_misalignment): Update after rename of dataref_aux.
7594         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
7595         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
7596         initialization of STMT_VINFO_DATA_REF.
7597         * tree-vectorizer.c (vec_info::move_dr): New function.
7598         * tree-vect-patterns.c (vect_recog_bool_pattern)
7599         (vect_recog_mask_conversion_pattern)
7600         (vect_recog_gather_scatter_pattern): Use it.
7601         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
7602         the "dr" and "stmt" fields of dr_vec_info instead of
7603         STMT_VINFO_DATA_REF.
7605 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7607         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
7608         (is_pattern_stmt_p): Use it.
7609         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
7610         on pattern statements.
7612 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7614         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
7615         original stmt as a stmt_vec_info rather than a gimple stmt.
7616         (vect_pattern_recog_1): Take the statement directly as a
7617         stmt_vec_info, rather than via a gimple_stmt_iterator.
7618         Update call to vect_mark_pattern_stmts.
7619         (vect_pattern_recog): Update calls accordingly.
7621 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7623         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
7624         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
7625         a vect_def_type for the first argument.
7626         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
7627         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
7628         operand if it isn't defined by a vectorized statement.
7629         (vect_build_gather_load_calls): Remove the mask_dt argument and
7630         update calls to vect_get_vec_def_for_stmt_copy.
7631         (vectorizable_bswap): Likewise the dt argument.
7632         (vectorizable_call): Update calls to vectorizable_bswap and
7633         vect_get_vec_def_for_stmt_copy.
7634         (vectorizable_simd_clone_call, vectorizable_assignment)
7635         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
7636         (vectorizable_comparison): Update calls to
7637         vect_get_vec_def_for_stmt_copy.
7638         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
7639         vect_is_simple_use.
7640         (vect_get_loop_based_defs): Remove dt argument and update call
7641         to vect_get_vec_def_for_stmt_copy.
7642         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
7643         and vect_get_vec_def_for_stmt_copy.
7644         (vectorizable_load): Update calls to vect_build_gather_load_calls
7645         and vect_get_vec_def_for_stmt_copy.
7646         * tree-vect-loop.c (vect_create_epilog_for_reduction)
7647         (vectorizable_reduction, vectorizable_live_operation): Update calls
7648         to vect_get_vec_def_for_stmt_copy.
7650 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7652         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
7653         and gimple stmt arguments with a stmt_vec_info.
7654         (vect_record_base_alignments): Update calls accordingly.
7655         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
7656         and gimple stmt arguments with a stmt_vec_info.
7657         (vect_build_slp_tree_1): Remove vinfo argument and update call
7658         to vect_record_max_nunits.
7659         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
7660         and vect_record_max_nunits.
7662 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7664         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
7665         file and take a stmt_vec_info instead of a gimple stmt.
7666         (supportable_widening_operation, vect_finish_replace_stmt)
7667         (vect_finish_stmt_generation, vect_get_store_rhs)
7668         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
7669         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
7670         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
7671         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
7672         (vect_create_data_ref_ptr, bump_vector_ptr)
7673         (vect_permute_store_chain, vect_setup_realignment)
7674         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
7675         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
7676         (vectorizable_reduction, vectorizable_induction)
7677         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
7678         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
7679         than gimple stmts as arguments.
7680         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
7681         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
7682         (can_group_stmts_p, vect_check_gather_scatter)
7683         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7684         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
7685         (vect_permute_load_chain, vect_shift_permute_load_chain)
7686         (vect_transform_grouped_load)
7687         (vect_record_grouped_load_vectors): Likewise.
7688         * tree-vect-loop.c (vect_fixup_reduc_chain)
7689         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
7690         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
7691         (vectorizable_reduction, vectorizable_induction)
7692         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
7693         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
7694         (vect_get_load_store_mask): Likewise.
7695         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
7696         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
7697         * tree-vect-stmts.c (vect_mark_relevant)
7698         (is_simple_and_all_uses_invariant)
7699         (exist_non_indexing_operands_for_use_p, process_use)
7700         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
7701         (vect_get_vec_def_for_operand, vect_get_vec_defs)
7702         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
7703         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
7704         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
7705         (get_negative_load_store_type, get_load_store_type)
7706         (vect_check_load_store_mask, vect_check_store_rhs)
7707         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
7708         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7709         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7710         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7711         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
7712         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
7713         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
7714         (supportable_widening_operation): Likewise.
7716 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7718         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
7719         a stmt_vec_info instead of a gcall.
7720         (vect_check_gather_scatter): Update call accordingly.
7721         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
7722         of a gphi.
7723         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
7724         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
7725         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
7726         instead of a gimple stmt.
7727         (vect_transform_loop): Update calls accordingly.
7728         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
7729         stmt_vec_infos instead of gimple stmts.
7730         (vect_analyze_slp_instance): Update use accordingly.
7731         * tree-vect-stmts.c (read_vector_array, write_vector_array)
7732         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
7733         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
7734         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
7735         (vect_gen_widened_results_half, vect_get_loop_based_defs)
7736         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
7737         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
7738         down to subroutines.
7740 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7742         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
7743         of the worklist from a vector of gimple stmts to a vector of
7744         stmt_vec_infos.
7745         * tree-vect-stmts.c (vect_mark_relevant, process_use)
7746         (vect_mark_stmts_to_be_vectorized): Likewise
7748 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7750         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
7751         statement before passing it to vect_analyze_stmt.
7752         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
7753         the chain of phi vector definitions.  Track the exit phi via its
7754         stmt_vec_info.
7755         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
7756         STMT_VINFO_REDUC_DEF.
7757         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
7758         stmt_vec_infos to handle the statement chains.
7759         (vect_get_slp_defs): Record the first statement in the node
7760         using a stmt_vec_info.
7761         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
7762         statements here and pass their stmt_vec_info down to subroutines.
7763         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
7764         down to vect_finish_stmt_generation.
7765         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
7766         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
7767         stmt_vec_infos to subroutines.
7768         (vect_remove_stores): Use stmt_vec_infos to handle the statement
7769         chains.
7771 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7773         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
7774         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
7775         (vect_permute_store_chain, vect_setup_realignment)
7776         (vect_permute_load_chain, vect_shift_permute_load_chain)
7777         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
7778         stmts internally, and when passing values to other vectorizer routines.
7779         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7780         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
7781         (vect_analyze_loop_operations, get_initial_def_for_reduction)
7782         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
7783         (vectorizable_reduction, vectorizable_induction)
7784         (vectorizable_live_operation, vect_transform_loop_stmt)
7785         (vect_transform_loop): Likewise.
7786         * tree-vect-patterns.c (vect_reassociating_reduction_p)
7787         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
7788         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
7789         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7790         (vect_slp_analyze_node_operations_1): Likewise.
7791         * tree-vect-stmts.c (vect_mark_relevant, process_use)
7792         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
7793         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
7794         (vect_finish_stmt_generation_1, get_group_load_store_type)
7795         (get_load_store_type, vect_build_gather_load_calls)
7796         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7797         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7798         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7799         (vectorizable_store, vectorizable_load, vectorizable_condition)
7800         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
7801         (supportable_widening_operation): Likewise.
7802         (vect_get_vector_types_for_stmt): Likewise.
7803         * tree-vectorizer.h (vect_dr_behavior): Likewise.
7805 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7807         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7808         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
7809         (vect_permute_store_chain, vect_permute_load_chain)
7810         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
7811         repeated stmt_vec_info lookups.
7812         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7813         (vect_update_ivs_after_vectorizer): Likewise.
7814         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
7815         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7816         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
7817         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7818         (vect_bb_slp_scalar_cost): Likewise.
7819         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
7821 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7823         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
7824         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
7825         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
7826         to get gassigns and gcalls, rather than operating on generc gimple
7827         stmts.
7828         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
7829         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
7830         (vectorizable_load, vect_analyze_stmt): Likewise.
7831         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
7833 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7835         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
7836         return stmt_vec_infos rather than gimple stmts.  Do not accept
7837         null arguments.
7838         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
7839         of a gimple stmt.
7840         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7841         Update use of get_later_stmt.
7842         (vect_get_constant_vectors): Update call accordingly.
7843         (vect_schedule_slp_instance): Likewise
7844         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
7845         (vect_slp_analyze_instance_dependence): Likewise.
7846         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
7848 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7850         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
7851         (stmt_info_for_cost::stmt_info): ...this new field.
7852         (add_stmt_costs): Update accordingly.
7853         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7854         (vect_get_known_peeling_cost): Likewise.
7855         (vect_estimate_min_profitable_iters): Likewise.
7856         * tree-vect-stmts.c (record_stmt_cost): Likewise.
7858 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7860         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
7861         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7862         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7863         accordingly.
7864         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
7866 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7868         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
7869         a gimple stmt to a stmt_vec_info.
7870         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7872 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7874         * tree-vectorizer.h (vec_info::grouped_stores): Change from
7875         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7876         (_loop_vec_info::reduction_chains): Likewise.
7877         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
7878         accordingly.
7879         * tree-vect-slp.c (vect_analyze_slp): Likewise.
7881 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7883         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
7884         a gimple stmt to a stmt_vec_info.
7885         (_stmt_vec_info::next_element): Likewise.
7886         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
7887         (vect_slp_analyze_and_verify_node_alignment)
7888         (vect_analyze_group_access_1, vect_analyze_group_access)
7889         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
7890         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
7891         (vect_supportable_dr_alignment): Update accordingly.
7892         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
7893         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
7894         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
7895         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7896         * tree-vect-slp.c (vect_build_slp_tree_1)
7897         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
7898         (vect_split_slp_store_group, vect_analyze_slp_instance)
7899         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
7900         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
7901         (get_group_load_store_type, get_load_store_type)
7902         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
7903         (vect_transform_stmt, vect_remove_stores): Likewise.
7905 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7907         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
7908         than a gimple stmt.
7909         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7910         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
7911         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
7912         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
7913         (vector_alignment_reachable_p, vect_get_data_access_cost)
7914         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
7915         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
7916         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7917         (vect_analyze_group_access_1, vect_analyze_group_access)
7918         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
7919         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
7920         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
7921         result of vect_dr_stmt and use the stmt_vec_info instead of
7922         the associated gimple stmt.
7923         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7924         (vect_gen_prolog_loop_niters): Likewise.
7925         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7927 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7929         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
7930         to a vec<stmt_vec_info>.
7931         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
7932         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
7933         vec<stmt_vec_info>.
7934         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
7935         to a vec<stmt_vec_info>.
7936         (bst_traits::value_type, bst_traits::value_type): Likewise.
7937         (bst_traits::hash): Update accordingly.
7938         (vect_get_and_check_slp_defs): Change the stmts parameter from
7939         a vec<gimple *> to a vec<stmt_vec_info>.
7940         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
7941         (vect_build_slp_tree): Likewise.
7942         (vect_build_slp_tree_2): Likewise.  Update uses of
7943         SLP_TREE_SCALAR_STMTS.
7944         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
7945         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
7946         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
7947         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
7948         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
7949         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
7950         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
7951         (vect_get_constant_vectors, vect_get_slp_defs)
7952         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
7953         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
7954         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
7955         instead of gimple stmts.
7956         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
7957         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
7958         (vect_slp_analyze_instance_dependence): Update uses of
7959         SLP_TREE_SCALAR_STMTS.
7960         (vect_slp_analyze_and_verify_node_alignment): Likewise.
7961         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
7962         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
7963         (get_initial_defs_for_reduction): Likewise.
7964         (vect_create_epilog_for_reduction): Likewise.
7965         (vectorize_fold_left_reduction): Likewise.
7966         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
7967         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
7968         (can_vectorize_live_stmts): Likewise.
7970 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7972         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
7973         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7974         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
7975         than gimple stmts.
7976         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
7977         of a gimple stmt.
7978         (gather_scalar_reductions): Update after above interface changes.
7979         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
7980         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
7981         than gimple stmts.
7982         (vect_force_simple_reduction): Likewise.
7983         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
7984         LOOP_VINFO_REDUCTIONS.
7985         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7987 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7989         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
7990         a gimple stmt to a stmt_vec_info.
7991         * tree-vect-loop.c (vect_active_double_reduction_p)
7992         (vect_force_simple_reduction, vectorizable_reduction): Update
7993         accordingly.
7995 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7997         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
7998         vec<gimple *> to a vec<stmt_vec_info>.
7999         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
8000         the reduction_phis argument from a vec<gimple *> to a
8001         vec<stmt_vec_info>.
8002         (vectorizable_reduction): Likewise the phis local variable that
8003         is passed to vect_create_epilog_for_reduction.  Update for new type
8004         of SLP_TREE_VEC_STMTS.
8005         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
8006         (vectorizable_live_operation): Likewise.
8007         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
8008         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
8010 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8012         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
8013         a gimple stmt to a stmt_vec_info.
8014         (vectorizable_condition, vectorizable_live_operation)
8015         (vectorizable_reduction, vectorizable_induction): Pass back the
8016         vectorized statement as a stmt_vec_info.
8017         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
8018         use of STMT_VINFO_VEC_STMT.
8019         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
8020         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
8021         as stmt_vec_infos rather than gimple stmts.
8022         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
8023         to a stmt_vec_info.
8024         (vectorizable_live_operation): Likewise.
8025         (vectorizable_reduction, vectorizable_induction): Likewise,
8026         updating use of STMT_VINFO_VEC_STMT.
8027         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
8028         of STMT_VINFO_VEC_STMT.
8029         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
8030         (vectorizable_simd_clone_call, vectorizable_conversion)
8031         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8032         (vectorizable_store, vectorizable_load, vectorizable_condition)
8033         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
8034         from a gimple stmt to a stmt_vec_info.
8035         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
8036         pointer to a stmt_vec_info to the vectorizable_* routines.
8038 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8040         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
8041         a gimple stmt to a stmt_vec_info.
8042         (is_pattern_stmt_p): Update accordingly.
8043         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
8044         (vect_record_grouped_load_vectors): Likewise.
8045         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8046         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
8047         (vect_model_reduction_cost): Likewise.
8048         (vect_create_epilog_for_reduction): Likewise.
8049         (vectorizable_reduction, vectorizable_induction): Likewise.
8050         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8051         Return the stmt_vec_info for the pattern statement.
8052         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
8053         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
8054         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8055         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
8056         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8057         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
8058         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
8059         (free_stmt_vec_info, vect_is_simple_use): Likewise.
8061 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8063         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
8064         (vect_finish_stmt_generation): Likewise.
8065         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8066         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
8067         (vect_build_gather_load_calls): Use the return value of the above
8068         functions instead of a separate call to vinfo_for_stmt.  Use narrow
8069         scopes for the input gimple stmt and wider scopes for the associated
8070         stmt_vec_info.  Use vec_info::lookup_def when setting these
8071         stmt_vec_infos from an SSA_NAME definition.
8072         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8073         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8074         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8075         (vectorizable_store, vectorizable_load, vectorizable_condition)
8076         (vectorizable_comparison): Likewise.
8077         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
8078         (vectorizable_reduction): Likewise.
8080 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8082         * tree-vectorizer.h (vect_is_simple_use): Add an optional
8083         stmt_vec_info * parameter before the optional gimple **.
8084         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8085         (process_use, vect_get_vec_def_for_operand_1): Update callers.
8086         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
8087         * tree-vect-loop.c (vectorizable_reduction): Likewise.
8088         (vectorizable_live_operation): Likewise.
8089         * tree-vect-patterns.c (type_conversion_p): Likewise.
8090         (vect_look_through_possible_promotion): Likewise.
8091         (vect_recog_rotate_pattern): Likewise.
8092         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8094 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8096         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
8097         a typedef to a wrapper class.
8098         (NULL_STMT_VEC_INFO): New macro.
8099         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
8100         (stmt_vec_info::operator*): New function.
8101         (stmt_vec_info::operator gimple *): Likewise.
8102         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
8103         (add_stmt_costs): Likewise.
8104         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
8105         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8106         (vect_get_known_peeling_cost): Likewise.
8107         (vect_estimate_min_profitable_iters): Likewise.
8108         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8109         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
8110         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8111         (vectorizable_store, free_stmt_vec_infos): Likewise.
8112         (new_stmt_vec_info): Change return type of xcalloc to
8113         _stmt_vec_info *.
8115 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8117         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
8118         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
8119         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
8120         a single_imm_use-based sequence.
8121         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
8123 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8125         * tree-vectorizer.h (vec_info::lookup_def): Declare.
8126         * tree-vectorizer.c (vec_info::lookup_def): New function.
8127         * tree-vect-patterns.c (vect_get_internal_def): Use it.
8128         (vect_widened_op_tree): Likewise.
8129         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8130         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
8131         (vectorizable_reduction): Likewise.
8132         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
8133         of a gimple *.
8134         (vect_is_slp_reduction): Update calls accordingly.  Use
8135         vec_info::lookup_def.
8136         (vect_is_simple_reduction): Likewise
8137         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
8139 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8141         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
8142         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
8143         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
8144         of vinfo_for_stmt.
8145         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
8146         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
8147         (vect_update_vf_for_slp, vect_analyze_loop_operations)
8148         (vect_is_slp_reduction, vectorizable_induction)
8149         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
8150         * tree-vect-patterns.c (vect_init_pattern_stmt):
8151         (vect_determine_min_output_precision_1, vect_determine_precisions)
8152         (vect_pattern_recog): Likewise.
8153         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
8154         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
8155         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
8156         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8157         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
8158         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
8159         info field from a loop to a loop_vec_info.
8161 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8163         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
8164         (vec_info::add_stmt): Declare.
8165         * tree-vectorizer.c (vec_info::add_stmt): New function.
8166         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
8167         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
8168         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8169         (vectorizable_induction): Likewise.
8170         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
8171         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8172         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
8173         (vectorizable_load): Likewise.
8174         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8175         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
8176         (vect_recog_gather_scatter_pattern): Likewise.
8177         (append_pattern_def_seq): Likewise.  Remove a check that is
8178         performed by add_stmt itself.
8180 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8182         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
8183         which make_ssa_name was called with new_stmt before new_stmt
8184         had been created.
8186 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8188         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
8189         split out from...
8190         (vect_is_slp_reduction): ...here...
8191         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
8192         that are already known to be false.
8194 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8196         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
8197         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
8198         STMT_VINFO_NUM_SLP_USES when it's true.
8199         (vect_free_slp_instance): Add a final_p parameter and pass it to
8200         vect_free_slp_tree.
8201         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
8202         (vect_analyze_slp_instance): Likewise.
8203         (vect_slp_analyze_operations): Likewise.
8204         (vect_slp_analyze_bb_1): Likewise.
8205         * tree-vectorizer.c (vec_info): Likewise.
8206         * tree-vect-loop.c (vect_transform_loop): Likewise.
8208 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8210         * tree-vect-loop.c (vectorizable_reduction): Assert that the
8211         function is not called for second and subsequent members of
8212         a reduction group.
8214 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8216         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
8217         cases for nested loops from here to ...
8218         (vect_create_epilog_for_reduction): ...here.  Only call
8219         vect_is_simple_use for inner-loop reductions.
8221 2018-07-31  Martin Liska  <mliska@suse.cz>
8223         PR gcov-profile/85338
8224         PR gcov-profile/85350
8225         PR gcov-profile/85372
8226         * profile.c (struct location_triplet): New.
8227         (struct location_triplet_hash): Likewise.
8228         (output_location): Do not output a BB that
8229         is already recorded for a line.
8230         (branch_prob): Use streamed_locations.
8232 2018-07-31  Martin Liska  <mliska@suse.cz>
8234         PR gcov-profile/85370
8235         * coverage.c (coverage_begin_function): Do not mark target
8236         clones as artificial functions.
8238 2018-07-31  Martin Liska  <mliska@suse.cz>
8240         PR gcov-profile/83813
8241         PR gcov-profile/84758
8242         PR gcov-profile/85217
8243         PR gcov-profile/85332
8244         * profile.c (branch_prob): Do not record GOTO expressions
8245         for GIMPLE statements which locations are already streamed.
8247 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8249         * gcc.c (handle_spec_function): Accept a soft_matched_part
8250         argument, as do_spec_1.  Pass it down to ...
8251         (eval_spec_function): Accept a soft_matched_part argument,
8252         and pass it down to ...
8253         (do_spec_2): Accept a soft_matched_part argument, and pass
8254         it down to do_spec_1.
8255         (do_spec_1): Pass soft_matched_part to handle_spec_function.
8256         (handle_braces): Update call to handle_spec_function.
8257         (driver::set_up_specs): Update calls to do_spec_2.
8258         (compare_debug_dump_opt_spec_function): Likewise.
8259         (compare_debug_self_opt_spec_function): Likewise.
8261 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8263         * common.opt (nolibc): New option.
8264         * doc/invoke.texi (Link Options): Document it.
8265         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8266         * config/alpha/linux.h: Likewise.
8267         * config/arc/elf.h: Likewise.
8268         * config/arm/uclinux-elf.h: Likewise.
8269         * config/arm/unknown-elf.h: Likewise.
8270         * config/avr/avrlibc.h: Likewise.
8271         * config/bfin/bfin.h: Likewise.
8272         * config/bfin/linux.h: Likewise.
8273         * config/bfin/uclinux.h: Likewise.
8274         * config/darwin.h: Likewise.
8275         * config/darwin10.h: Likewise.
8276         * config/darwin12.h: Likewise.
8277         * config/gnu-user.h: Likewise.
8278         * config/lm32/uclinux-elf.h: Likewise.
8279         * config/pa/pa-hpux11.h: Likewise.
8280         * config/pa/pa64-hpux.h: Likewise.
8281         * config/sparc/sparc.h: Likewise.
8283 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8285         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
8286         undefined variables.
8288 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8290         PR target/86640
8291         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8292         instead of GEN_INT.
8294 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8296         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8297         terminated string literal.
8299 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8301         PR rtl-optimization/85160
8302         * combine.c (is_just_move): New function.
8303         (try_combine): Allow combining two instructions into two if neither of
8304         the original instructions was a move.
8306 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
8308         PR target/86673
8309         * doc/extend.texi (Global Register Variables): Discourage use of type
8310         qualifiers.
8311         (Local Register Variables): Likewise.
8313 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
8315         PR tree-optimization/86506
8316         * hwint.h (ceil_log2): Resync with hwint.c implementation.
8318 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
8320         PR target/86547
8321         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8322         hard_regno, make sure no insn between `from` and `to` clobbers it.
8324 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
8325             Tom de Vries  <tdevries@suse.de>
8327         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8328         (PTX_DEFAULT_RUNTIME_DIM): ... this.
8329         (nvptx_goacc_validate_dims): Set default worker and gang dims to
8330         PTX_DEFAULT_RUNTIME_DIM.
8331         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8333 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
8335         * config/pa/pa.c (pa_output_addr_vec): Align address table.
8336         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8337         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
8339 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
8341         * config/rs6000/constraints.md (wG constraint): Delete, no longer
8342         used.
8343         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
8344         predicate to reflect toc fusion has been deleted.
8345         (toc_fusion_mem_raw): Delete, no longer used.
8346         (toc_fusion_mem_wrapped): Likewise.
8347         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
8348         fusion mask bit.
8349         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
8350         Delete, no longer used.
8351         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
8352         meant to be used for toc fusion.
8353         (rs6000_debug_print_mode): Delete toc fusion debugging.
8354         (rs6000_debug_reg_global): Likewise.
8355         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
8356         fusion and secondary reload support that were never used.
8357         (rs6000_option_override_internal): Delete TOC fusion, that was only
8358         partially defined, and it did not work unless you also used the
8359         -mcmodel= switch.
8360         (rs6000_legitimate_address_p): Delete TOC fusion support.
8361         (rs6000_opt_masks): Likewise.
8362         (fusion_wrap_memory_address): Delete function, no longer used.
8363         (fusion_split_address); Delete TOC fusion support.
8364         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
8365         longer used with toc fusion being deleted.
8366         (TARGET_TOC_FUSION_FP): Likewise.
8367         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
8368         UNSPEC.
8369         (toc fusion spliter): Delete TOC fusion support.
8370         (toc_fusionload_<mode>): Likewise.
8371         (toc_fusionload_di): Likewise.
8372         (fusion_gpr_load_<mode>): Delete generator function, this insn no
8373         longer needs to be named.  Rename predicate to delete TOC fusion.
8374         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8375         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8376         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8377         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8378         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
8380 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8382         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
8383         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
8384         __int128 in built-in function prototypes.
8385         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
8386         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
8388 2018-07-27  Martin Sebor  <msebor@redhat.com>
8390         PR tree-optimization/86696
8391         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
8392         types, including enums.
8393         (handle_char_store): Be prepared for the above function to fail.
8395 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
8397         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
8398         when optimization level is lower than 2 or optimize for size.
8400 2018-07-26  Martin Sebor  <msebor@redhat.com>
8402         PR tree-optimization/86043
8403         PR tree-optimization/86042
8404         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
8405         (get_string_cst_length): Rename...
8406         (get_min_string_length): ...to this.  Add argument.
8407         (handle_char_store): Extend to handle multi-character stores by
8408         MEM_REF.
8409         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
8410         * tree.h (initializer_zerop): Add argument.
8412 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
8414         PR middle-end/86660
8415         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
8416         declare target to variables if they have always,{to,from,tofrom} map
8417         kinds.
8419 2018-07-26  Martin Liska  <mliska@suse.cz>
8421         PR lto/86548
8422         * lto-wrapper.c: Add linker_output as prefix
8423         for ltrans_output_file.
8425 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
8427         PR rtl-optimization/85805
8428         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
8429         value for hard registers if that was written in the same mode.
8431 2018-07-26  Martin Liska  <mliska@suse.cz>
8433         PR gcov-profile/86536
8434         * gcov.c (format_gcov): Use printf format %.*f directly
8435         and do not handle special values.
8437 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8439         * common/config/arc/arc-common.c (arc_option_optimization_table):
8440         Update default optimizations for size.
8442 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8444         * config/arc/arc.md (movsf_insn): Add short instruction selection.
8445         * config/arc/constraints.md (CfZ): New constraint.
8446         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
8447         (subsf3_fpu): Likewise.
8448         (cmpsf_fpu): Likewise.
8449         (cmpsf_fpu_uneq): Likewise.
8451 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8453         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
8454         accesses as well.
8455         (arc_is_uncached_mem_p): uncached applies to both the variable and
8456         the pointer.
8458 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8460         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
8461         register names.
8463 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
8465         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
8466         field "m_scopes" from vec to auto_vec.
8468 2018-07-25  Martin Liska  <mliska@suse.cz>
8470         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
8471         return type.
8473 2018-07-25  Richard Biener  <rguenther@suse.de>
8475         PR debug/86654
8476         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
8477         special wrt context_die late.
8478         (gen_subprogram_die): Re-use DIEs in local scope.
8480 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
8482         PR tree-optimization/86644
8483         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
8485 2018-07-25  Martin Liska  <mliska@suse.cz>
8487         PR middle-end/86645
8488         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
8489         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
8491 2018-07-25  Martin Liska  <mliska@suse.cz>
8493         PR sanitizer/79635
8494         * params.def: Explain ASan abbreviation and provide
8495         a documentation link.
8497 2018-07-24  Martin Sebor  <msebor@redhat.com>
8499         PR tree-optimization/86622
8500         PR tree-optimization/86532
8501         * builtins.h (string_length): Declare.
8502         * builtins.c (c_strlen): Correct handling of non-constant offsets.
8503         (check_access): Be prepared for non-constant length ranges.
8504         (string_length): Make extern.
8505         * expr.c (string_constant): Only handle the minor non-constant
8506         array index.  Use string_constant to compute the length of
8507         a generic string constant.
8509 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
8511         PR tree-optimization/86618
8512         * tree-vect-stmts.c (vectorizable_call): Don't take the address
8513         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
8515 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
8517         PR tree-optimization/86636
8518         * json.cc (json::object::set): Fix comment.  Add assertions.
8519         (json::array::append): Move here from json.h.  Add comment and an
8520         assertion.
8521         (json::string::string): Likewise.
8522         * json.h (json::array::append): Move to json.cc.
8523         (json::string::string): Likewise.
8524         * optinfo-emit-json.cc
8525         (optrecord_json_writer::impl_location_to_json): Assert that we
8526         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
8527         wrapper around it.  Expand the location once, rather than three
8528         times.
8529         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
8530         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
8531         wrappers.
8532         (optrecord_json_writer::optinfo_to_json): Likewise, in four
8533         places.  Fix some overlong lines.
8535 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
8537         * config/aarch64/aarch64-simd.md
8538         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
8539         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
8540         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
8541         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
8542         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
8543         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
8544         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
8545         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
8546         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
8548 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
8550         PR middle-end/86627
8551         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
8552         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
8553         and abs_d == d, do the power of two handling if profitable.
8555 2018-07-24  Richard Biener  <rguenther@suse.de>
8557         * match.pd: Add BIT_FIELD_REF canonicalizations.
8559 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8561         PR c/86617
8562         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
8564 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8566         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
8567         terminated STRING_CST object.
8569 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8571         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
8573 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
8575         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
8576         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
8577         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
8578         the elements into a register.
8579         (rs6000_split_v4si_init_di_reg): Delete.
8580         (rs6000_split_v4si_init): Delete.
8581         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
8582         (vsx_init_v4si): Rewrite as a define_expand.
8584 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
8586         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
8587         zero_extend argument from memory): New.
8589 2018-07-22  Martin Sebor  <msebor@redhat.com>
8591         PR bootstrap/86621
8592         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
8593         diagnosing calls with unknown arguments unless -Walloca-larger-than
8594         is restricted to less than PTRDIFF_MAX bytes.
8596 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
8598         * doc/gcov.texi (Invoking Gcov): Editorial changes.
8600 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
8602         * pretty-print.c (text_info::set_location): Remove redundant
8603         "line_table" parameter from call to rich_location::set_range.
8605 2018-07-20  Martin Sebor  <msebor@redhat.com>
8607         PR middle-end/82063
8608         * builtins.c (expand_builtin_alloca): Adjust.
8609         * calls.c (alloc_max_size): Simplify.
8610         * cgraphunit.c (cgraph_node::expand): Adjust.
8611         * common.opt (larger_than_size, warn_frame_larger_than): Remove
8612         variables.
8613         (frame_larger_than_size): Same.
8614         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
8615         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
8616         Initialize.
8617         * doc/invoke.texi (GCC Command Options): Document option arguments.
8618         Explain byte-size arguments and suffixes.
8619         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
8620         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
8621         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
8622         * doc/options.texi (UInteger): Expand.
8623         (Host_Wide_Int, ByteSize): Document new properties.
8624         * final.c (final_start_function_1): Include sizes in an error message.
8625         * function.c (frame_offset_overflow): Same.
8626         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
8627         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
8628         Diagnose unbounded alloca calls only for limits of less than
8629         PTRDIFF_MAX.
8630         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
8631         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
8632         for alloca(0).
8633         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
8634         only for limits of less than PTRDIFF_MAX.
8635         * langhooks-def.h (lhd_handle_option): Change function argument
8636         to HOST_WIDE_INT.
8637         * langhooks.c (lhd_handle_option): Same.
8638         * langhooks.h (handle_option): Same.
8639         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
8640         ByteSize flags.
8641         (var_type, var_type_struct): Same.
8642         (var_set): Handle ByteSize flag.
8643         * optc-gen.awk: Add comments to output to ease debugging.  Make
8644         use of HOST_WIDE_INT where appropriate.
8645         * opts-gen-save.awk:  Use %lx to format unsigned long.
8646         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
8647         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
8648         arguments.  Parse bytes-size suffixes.
8649         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
8650         (enum_value_to_arg): Same.
8651         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
8652         (handle_option): Adjust.
8653         (generate_option): Change function argument to HOST_WIDE_INT.
8654         (cmdline_handle_error): Adjust.
8655         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
8656         (set_option): Change function argument to HOST_WIDE_INT.
8657         (option_enabled): Handle cl_host_wide_int.
8658         (get_option_state): Handle CLVC_SIZE.
8659         (control_warning_option): Same.
8660         * opts.c (common_handle_option): Change function argument to
8661         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
8662         OPT_Wvla_larger_than_.
8663         * opts.h (enum cl_var_type): Add an enumerator.
8664         * stor-layout.c (layout_decl): Print a more meaningful warning.
8665         * toplev.c (output_stack_usage): Adjust.
8667 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
8669         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
8670         call to inline_expand_builtin_string_cmp.
8671         (expand_builtin_strcmp): Likewise.
8672         (expand_builtin_strncmp): Likewise.
8673         (inline_string_cmp): Delete the last parameter, change char_type_node
8674         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
8675         two operands.
8676         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
8677         the inlining expansion on target where the type of the call has same or
8678         narrower precision than unsigned char.
8680 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
8682         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
8683         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
8684         * common.opt (fsave-optimization-record): New option.
8685         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
8686         * doc/invoke.texi (-fsave-optimization-record): New option.
8687         * dumpfile.c: Include "optinfo-emit-json.h".
8688         (struct kv_pair): Move to coretypes.h.
8689         (optgroup_options): Make non-static.
8690         (dump_context::end_scope): Call
8691         optimization_records_maybe_pop_dump_scope.
8692         * dumpfile.h (optgroup_options): New decl.
8693         * json.cc: New file.
8694         * json.h: New file.
8695         * optinfo-emit-json.cc: New file.
8696         * optinfo-emit-json.h: New file.
8697         * optinfo.cc: Include "optinfo-emit-json.h".
8698         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
8699         (optinfo_enabled_p): Check optimization_records_enabled_p.
8700         (optinfo_wants_inlining_info_p): Likewise.
8701         * optinfo.h: Update comment.
8702         * profile-count.c (profile_quality_as_string): New function.
8703         * profile-count.h (profile_quality_as_string): New decl.
8704         (profile_count::quality): New accessor.
8705         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
8706         and optinfo_emit_json_cc_tests.
8707         * selftest.h (selftest::json_cc_tests): New decl.
8708         (selftest::optinfo_emit_json_cc_tests): New decl.
8709         * toplev.c: Include "optinfo-emit-json.h".
8710         (compile_file): Call optimization_records_finish.
8711         (do_compile): Call optimization_records_start.
8712         * tree-ssa-live.c: Include optinfo.h.
8713         (remove_unused_scope_block_p): Retain inlining information if
8714         optinfo_wants_inlining_info_p returns true.
8716 2018-07-20  Richard Biener  <rguenther@suse.de>
8718         PR debug/86585
8719         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
8720         to cover -flto-partition=none.
8722 2018-07-20  Martin Liska  <mliska@suse.cz>
8724         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
8725         (get_decl_source_range): Remove unused function.
8727 2018-07-20  Richard Biener  <rguenther@suse.de>
8729         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
8730         (struct vn_phi_s): Likewise.
8731         (struct vn_reference_s): Likewise.
8732         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
8733         for searching the slot of an entry known to be in the hash itself.
8734         (vn_phi_hasher::equal): Likewise.
8735         (vn_reference_hasher::equal): Likewise.
8736         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
8737         globals.
8738         (optimistic_info, current_info): Remove, keeping only valid_info.
8739         (vn_reference_lookup_1): Remove fallback lookup.
8740         (vn_reference_lookup_2): Likewise.
8741         (vn_nary_op_lookup_1): Likewise.
8742         (vn_phi_lookup): Likewise.
8743         (vn_nary_build_or_lookup_1): Make sure to not chain the built
8744         hash element.
8745         (vn_reference_insert): Adjust, chain the inserted hash element
8746         at last_inserted_ref.
8747         (vn_reference_insert_pieces): Likewise.
8748         (visit_reference_op_call): Likewise.
8749         (vn_nary_op_insert_into): Chain the inserted hash element at
8750         last_inserted_nary.
8751         (vn_nary_op_insert_pieces): Adjust.
8752         (vn_nary_op_insert): Likewise.
8753         (vn_nary_op_insert_stmt): Likewise.
8754         (vn_phi_insert): Adjust, chain the inserted hash element at
8755         last_inserted_phi.
8756         (process_scc): Remove clearing and copying the optimistic
8757         table.  Instead remove elements inserted during an optimistic
8758         iteration from the single table we maintain.
8759         (init_scc_vn): Adjust.
8760         (free_scc_vn): Likewise.
8761         (sccvn_dom_walker::record_cond): Likewise.
8762         (sccvn_dom_walker::after_dom_children): Likewise.
8764 2018-07-19  Martin Sebor  <msebor@redhat.com>
8766         PR tree-optimization/84047
8767         PR tree-optimization/83776
8768         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
8769         (check_array_bounds): Call it.
8771 2018-07-19  Martin Sebor  <msebor@redhat.com>
8773         * align.h (align_flags): Use member initialization.
8775 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
8777         * Makefile.in (OBJS): Add optinfo.o.
8778         * coretypes.h (class symtab_node): New forward decl.
8779         (struct cgraph_node): New forward decl.
8780         (class varpool_node): New forward decl.
8781         * dump-context.h: New file.
8782         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
8783         "tree-pass.h".
8784         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
8785         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
8786         (set_alt_dump_file): Likewise.
8787         (dump_context::~dump_context): New dtor.
8788         (dump_gimple_stmt): Move implementation to...
8789         (dump_context::dump_gimple_stmt): ...this new member function.
8790         Add the stmt to any pending optinfo, creating one if need be.
8791         (dump_gimple_stmt_loc): Move implementation to...
8792         (dump_context::dump_gimple_stmt_loc): ...this new member function.
8793         Start a new optinfo and add the stmt to it.
8794         (dump_gimple_expr): Move implementation to...
8795         (dump_context::dump_gimple_expr): ...this new member function.
8796         Add the stmt to any pending optinfo, creating one if need be.
8797         (dump_gimple_expr_loc): Move implementation to...
8798         (dump_context::dump_gimple_expr_loc): ...this new member function.
8799         Start a new optinfo and add the stmt to it.
8800         (dump_generic_expr): Move implementation to...
8801         (dump_context::dump_generic_expr): ...this new member function.
8802         Add the tree to any pending optinfo, creating one if need be.
8803         (dump_generic_expr_loc): Move implementation to...
8804         (dump_context::dump_generic_expr_loc): ...this new member
8805         function.  Add the tree to any pending optinfo, creating one if
8806         need be.
8807         (dump_printf): Move implementation to...
8808         (dump_context::dump_printf_va): ...this new member function.  Add
8809         the text to any pending optinfo, creating one if need be.
8810         (dump_printf_loc): Move implementation to...
8811         (dump_context::dump_printf_loc_va): ...this new member function.
8812         Start a new optinfo and add the stmt to it.
8813         (dump_dec): Move implementation to...
8814         (dump_context::dump_dec): ...this new member function.  Add the
8815         value to any pending optinfo, creating one if need be.
8816         (dump_context::dump_symtab_node): New member function.
8817         (dump_context::get_scope_depth): New member function.
8818         (dump_context::begin_scope): New member function.
8819         (dump_context::end_scope): New member function.
8820         (dump_context::ensure_pending_optinfo): New member function.
8821         (dump_context::begin_next_optinfo): New member function.
8822         (dump_context::end_any_optinfo): New member function.
8823         (dump_context::s_current): New global.
8824         (dump_context::s_default): New global.
8825         (dump_scope_depth): Delete global.
8826         (dumpfile_ensure_any_optinfo_are_flushed): New function.
8827         (dump_symtab_node): New function.
8828         (get_dump_scope_depth): Reimplement in terms of dump_context.
8829         (dump_begin_scope): Likewise.
8830         (dump_end_scope): Likewise.
8831         (selftest::temp_dump_context::temp_dump_context): New ctor.
8832         (selftest::temp_dump_context::~temp_dump_context): New dtor.
8833         (selftest::verify_item): New function.
8834         (ASSERT_IS_TEXT): New macro.
8835         (ASSERT_IS_TREE): New macro.
8836         (ASSERT_IS_GIMPLE): New macro.
8837         (selftest::test_capture_of_dump_calls): New test.
8838         (selftest::dumpfile_c_tests): Call it.
8839         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
8840         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
8841         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
8842         descriptive comment.
8843         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
8844         (dump_node, dump_bb): Move these unrelated decls.
8845         (class dump_manager): Add leading comment.
8846         * optinfo.cc: New file.
8847         * optinfo.h: New file.
8849 2018-07-19  Michael Collison  <michael.collison@arm.com>
8850             Richard Henderson <rth@redhat.com>
8852         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
8853         (subti): Handle op1 zero.
8854         (subvti4, usub4ti4): New.
8855         (*sub<GPI>3_compare1_imm): New.
8856         (sub<GPI>3_carryinCV): New.
8857         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
8858         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
8860 2018-07-19  Michael Collison  <michael.collison@arm.com>
8861             Richard Henderson <rth@redhat.com>
8863         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
8864         (addti3): Create simpler code if low part is already known to be 0.
8865         (addvti4, uaddvti4): New.
8866         (*add<GPI>3_compareC_cconly_imm): New.
8867         (*add<GPI>3_compareC_cconly): New.
8868         (*add<GPI>3_compareC_imm): New.
8869         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
8870         handle constants within this pattern..
8871         (*add<GPI>3_compareV_cconly_imm): New.
8872         (*add<GPI>3_compareV_cconly): New.
8873         (*add<GPI>3_compareV_imm): New.
8874         (add<GPI>3_compareV): New.
8875         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
8876         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
8877         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
8878         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
8879         with 'comparison' operator.
8880         (*add<GPI>3_compareV_cconly_imm): Ditto.
8881         (*add<GPI>3_compareV_cconly): Ditto.
8882         (*add<GPI>3_compareV_imm): Ditto.
8883         (add<GPI>3_compareV): Ditto.
8884         (add<mode>3_carryinC): Ditto.
8885         (*add<mode>3_carryinC_zero): Ditto.
8886         (*add<mode>3_carryinC): Ditto.
8887         (add<mode>3_carryinV): Ditto.
8888         (*add<mode>3_carryinV_zero): Ditto.
8889         (*add<mode>3_carryinV): Ditto.
8891 2018-07-19  Michael Collison  <michael.collison@arm.com>
8892             Richard Henderson <rth@redhat.com>
8894         * config/aarch64/aarch64-modes.def (CC_V): New.
8895         * config/aarch64/aarch64-protos.h
8896         (aarch64_addti_scratch_regs): Declare
8897         (aarch64_subvti_scratch_regs): Declare.
8898         (aarch64_expand_subvti): Declare.
8899         (aarch64_gen_unlikely_cbranch): Declare
8900         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
8901         for signed overflow using CC_Vmode.
8902         (aarch64_get_condition_code_1): Handle CC_Vmode.
8903         (aarch64_gen_unlikely_cbranch): New function.
8904         (aarch64_addti_scratch_regs): New function.
8905         (aarch64_subvti_scratch_regs): New function.
8906         (aarch64_expand_subvti): New function.
8908 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8910         * config/aarch64/aarch64-option-extensions.def: New entry for profile
8911         extension.
8912         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
8913         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
8914         extension.
8916 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8918         PR target/83009
8919         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8920         address check not strict.
8922 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8924         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
8925         Umq with Umn.
8926         (store_pair_lanes<mode>): Likewise.
8927         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
8928         enum value 'ADDR_QUERY_LDP_STP_N'.
8929         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
8930         (aarch64_print_address_internal): Add declaration.
8931         (aarch64_print_ldpstp_address): Remove.
8932         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
8933         (aarch64_print_operand): Change printing of 'y'.
8934         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
8935         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
8936         'true' rather than '1'.
8937         * config/aarch64/constraints.md (Uml): Likewise.
8938         (Uml): Rename to Umn.
8939         (Umq): Remove.
8941 2018-07-19  Richard Biener  <rguenther@suse.de>
8943         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
8944         a trailing array.
8945         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
8946         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
8947         (vn_reference_hasher): Likewise.
8948         (struct vn_tables_s): Remove obstack and alloc-pool members.
8949         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
8950         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
8951         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
8952         (vn_reference_insert_pieces): Likewise.
8953         (alloc_vn_nary_op_noinit): Adjust.
8954         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
8955         (vn_phi_eq): Adjust.
8956         (shared_lookup_phiargs): Remove.
8957         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
8958         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
8959         (visit_reference_op_call): Likewise.
8960         (copy_nary, copy_phi, copy_reference): Remove.
8961         (process_scc): Rewind the obstack when iterating.  Do not
8962         copy the elements to valid_info but just move them from one
8963         hashtable to the other.
8964         (allocate_vn_table): Adjust.
8965         (free_vn_table): Likewise.
8966         (init_scc_vn): Likewise.
8967         (free_scc_vn): Likewise.
8969 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
8971         PR target/86560
8972         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
8973         indirect_return as function type attribute.
8974         (ix86_attribute_table): Change indirect_return to function
8975         type attribute.
8976         * doc/extend.texi: Update indirect_return attribute.
8978 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
8980         * wide-int.h (widest2_int): New.
8981         * gimple-fold.c (arith_overflowed_p): Use it.
8982         * tree.h (widest2_int_cst): New.
8983         * tree-vrp.c (wide_int_binop_overflow): Rename from
8984         vrp_int_const_binop.
8985         Rewrite to work on trees.
8986         (extract_range_from_multiplicative_op_1): Abstract code to...
8987         (wide_int_range_min_max): ...here.
8988         (wide_int_range_cross_product): ...and here.
8989         (extract_range_from_binary_expr_1): Abstract overflow code to...
8990         (wide_int_range_mult_wrapping): ...here.
8991         * tree-vrp.h (wide_int_range_cross_product): New.
8992         (wide_int_range_mult_wrapping): New.
8994 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
8995             Julia Koval  <julia.koval@intel.com>
8997         * config/i386/x86-tune-costs.h (skylake_memcpy,
8998         skylake_memset): Replace rep_prefix with unrolling for size 512.
9000 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
9002         PR middle-end/86544
9003         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
9004         comparision with EQ_EXPR in last stmt.
9006 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9008         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
9009         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
9010         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
9011         previously known as "PowerPC AltiVec Built-in Functions".  Move
9012         some material to new subsubsections "PowerPC AltiVec Built-in
9013         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
9014         ISA 2.07".
9015         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
9016         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
9017         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
9018         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
9020 2018-07-18  Richard Biener  <rguenther@suse.de>
9022         PR tree-optimization/86557
9023         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
9024         EXACT_DIV_EXPR.
9026 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
9028         * config/s390/s390.c (s390_function_profiler): Generate CFI.
9030 2018-07-17  Jeff Law  <law@redhat.com>
9032         * config/arm/arm.c (get_label_padding): Update for recent
9033         changes to label_to_alignment.
9035         PR tree-optimization/86010
9036         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
9038         * config/mips/mips.c (vr4130_align_insns): Update for recent
9039         changes to label_to_alignment.
9041         * config/frv/frv.c (frv_label_align): Update for recent changes
9042         to label_to_alignment.
9044         * config/nios2/nios2.c (nios2_label_align): Update for recent
9045         changes which dropped ALIGN_LABELS_LOG.
9047 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
9049         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
9050         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
9052 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
9054         * config/arc/arc.c (arc_label_align): Use align_labels instead of
9055         deprecated align_labels_log.
9057 2018-07-17  Richard Biener  <rguenther@suse.de>
9059         PR lto/86456
9060         * dwarf2out.c (init_sections_and_labels): Always generate
9061         a debug_line_str_section for early LTO debug.
9062         (dwarf2out_finish): Reset debug_line_str_hash output early.
9063         Bump counter for extra dwarf5 .debug_loc labels to not conflict
9064         with fat LTO part.
9065         (dwarf2out_early_finish): Output debug_line_str.
9067 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9069         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
9070         index register on z196 or later.
9072 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9074         * config/s390/s390.c (s390_default_align): Set default function
9075         alignment to 16.
9076         (s390_override_options_after_change): Call s390_default align.
9077         (s390_option_override_internal): Call s390_default_align.
9078         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
9080 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9082         PR middle-end/86542
9083         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
9084         to _looptemp_ clauses, other than the first two.
9086 2018-07-17  Martin Liska  <mliska@suse.cz>
9088         * opts.c: Do not enable OPT_falign_* for -Os.
9090 2018-07-17  Martin Liska  <mliska@suse.cz>
9092         * align.h (MAX_CODE_ALIGN): New.
9093         (MAX_CODE_ALIGN_VALUE): New.
9094         * common/config/i386/i386-common.c (ix86_handle_option):
9095         (MAX_CODE_ALIGN): Moved to align.h.
9096         * final.c (MAX_CODE_ALIGN): Likewise.
9097         * opts.c (parse_and_check_align_values):
9098         (MAX_CODE_ALIGN): Likewise.
9099         (MAX_CODE_ALIGN_VALUE): Likewise.
9101 2018-07-17  Martin Liska  <mliska@suse.cz>
9103         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
9104         in order to fulfil coding style.
9105         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
9106         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9107         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9108         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
9109         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
9110         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
9111         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
9112         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
9113         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9115 2018-07-17  Martin Liska  <mliska@suse.cz>
9117         * align.h: New file.
9118         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
9119         directly.
9120         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
9121         align_flags of label_to_alignment.
9122         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
9123         align_flags class.
9124         * config/m68k/m68k.c: Do not use removed align_labels_value and
9125         align_loops_value.
9126         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
9127         (LOOP_ALIGN): Likewise.
9128         (LABEL_ALIGN): Likewise.
9129         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9130         Remove not used macro.
9131         (rs6000_loop_align): Change return type to align_flags.
9132         (rs6000_loop_align_max_skip): Remove.
9133         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
9134         Change return type to align_flags.
9135         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9136         Remove not used macro.
9137         (rs6000_loop_align):  Change return type to align_flags.
9138         (rs6000_loop_align_max_skip): Remove.
9139         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
9140         * config/rx/rx-protos.h (rx_align_for_label): Make it
9141         static function.
9142         * config/rx/rx.c (rx_align_for_label): Change return type
9143         to align_flags.
9144         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
9145         macro definitions.
9146         into align_flags class.
9147         (LABEL_ALIGN): Likewise.
9148         (LOOP_ALIGN): Likewise.
9149         * config/s390/s390.c (s390_label_align): Use align_flags
9150         class member.
9151         (s390_asm_output_function_label): Likewise.
9152         * config/sh/sh.c (sh_override_options_after_change):
9153         Use align_flags class directly without macros.
9154         (find_barrier): Likewise.
9155         (barrier_align): Likewise.
9156         (sh_loop_align): Likewise.
9157         * config/spu/spu.c (spu_option_override):
9158         Use align_flags_tuple::get_value instead of removed macros.
9159         (spu_sched_init): Likewise.
9160         * config/spu/spu.h (GTY): Likewise.
9161         * config/visium/visium.c (visium_option_override):
9162         Set "8" as default secondary alignment.
9163         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
9164         in order to guarantee secondary alignment of 8.
9165         * coretypes.h: Include align.h header file.
9166         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
9167         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
9168         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
9169         * doc/tm.texi.in: Likewise.
9170         * final.c (struct label_alignment): Remove not used structure.
9171         (LABEL_ALIGN): Change type to align_flags.
9172         (LOOP_ALIGN): Likewise.
9173         (JUMP_ALIGN): Likewise.
9174         (default_loop_align_max_skip): Remove.
9175         (default_label_align_max_skip): Likewise.
9176         (default_jump_align_max_skip): Likewise.
9177         (default_label_align_after_barrier_max_skip):
9178         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
9179         (LABEL_TO_MAX_SKIP): Remove.
9180         (label_to_alignment): Return align_flags type instead of integer.
9181         (label_to_max_skip): Remove.
9182         (align_fuzz): Use align_flags type.
9183         (compute_alignments): Use align_flags type and use align_flags::max
9184         to combine multiple alignments.
9185         (grow_label_align): Grow vec instead of C array.
9186         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
9187         (shorten_branches):  Use align_flags type and use align_flags::max
9188         to combine multiple alignments.
9189         (final_scan_insn_1): Remove usage of secondary alignment that comes
9190         from label alignment, but instead use proper secondary alignment
9191         which is computed in grow_label_align.
9192         * flags.h (struct align_flags_tuple): Move to align.h.
9193         (struct align_flags): Likewise.
9194         (state_align_loops): Rename to align_loops.
9195         (state_align_jumps): Rename to align_jumps.
9196         (state_align_labels): Rename to align_labels.
9197         (state_align_functions): Rename to align_functions.
9198         (align_loops_log): Remove.
9199         (align_jumps_log): Remove.
9200         (align_labels_log): Remove.
9201         (align_functions_log): Remove.
9202         (align_loops_max_skip): Remove.
9203         (align_jumps_max_skip): Remove.
9204         (align_labels_max_skip): Remove.
9205         (align_functions_max_skip): Remove.
9206         (align_loops_value): Remove.
9207         (align_jumps_value): Remove.
9208         (align_labels_value): Remove.
9209         (align_functions_value): Remove.
9210         * output.h (label_to_alignment): Change return type to align_flags.
9211         (label_to_max_skip): Remove.
9212         * target.def: Remove loop_align_max_skip, label_align_max_skip,
9213         jump_align_max_skip macros.
9214         * targhooks.h (default_loop_align_max_skip): Remove.
9215         (default_label_align_max_skip): Likewise.
9216         (default_jump_align_max_skip): Likewise.
9217         (default_label_align_after_barrier_max_skip): Remove.
9218         * toplev.c (read_log_maxskip): Use ::normalize function.
9219         (parse_N_M): Remove not used argument and also call ::normalize.
9220         (parse_alignment_opts): Do not pass unused arguments.
9221         * varasm.c (assemble_start_function): Use directly align_functions
9222         instead of removed macros.
9223         * system.h: Do not poison removed macros.
9225 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9227         PR middle-end/86539
9228         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
9229         and cond temporaries don't have reference type if iterator has
9230         pointer type.  For init use &for_pre_body instead of pre_p if
9231         for_pre_body is non-empty.
9233 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9235         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9236         double-double modes to SFmode directly directly.
9237         (trunc<mode>sf2_fprs): Delete.
9239 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9241         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9242         for conversions between IFmode and the decimal floating point modes.
9243         (init_float128_ieee): Use the correct names for conversions between
9244         KFmode and the decimal floating point modes.
9246 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9248         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9249         for the conversions between TDmode and IFmode.
9250         (init_float128_ieee): Use more correct names for the conversions
9251         between TDmode and KFmode.
9253 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
9255         PR tree-optimization/86526
9256         * builtins.c (expand_builtin_memcmp): Formatting fixes.
9257         (inline_expand_builtin_string_cmp): Likewise.
9258         (inline_string_cmp): Likewise.  Use c_readstr instead of
9259         builtin_memcpy_read_str.  Add unit_mode temporary.
9261 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9263         PR middle-end/86528
9264         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9265         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9267 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9269         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9270         Alphabetize prototypes of built-in functions, separating out
9271         built-in functions that are listed in this section but should be
9272         described elsewhere.
9274 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
9276         PR target/86511
9277         * expmed.c (emit_store_flag): Do not emit setcc followed by a
9278         conditional move when trapping comparison was split to a
9279         non-trapping one (and vice versa).
9281 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9283         * config/s390/s390.c (s390_function_profiler): Generate nops
9284         instead of profiler call sequences.
9285         * config/s390/s390.opt: Add the new option.
9287 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9289         * config/s390/s390.c (s390_function_profiler): Generate
9290         __mcount_loc section.
9291         * config/s390/s390.opt: Add the new option.
9293 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9295         * common.opt: Add the new warning.
9296         * config/s390/s390.c (s390_function_profiler): Emit "brasl
9297         %r0,__fentry__" when -mfentry is specified.
9298         (s390_option_override_internal): Disallow -mfentry for 31-bit
9299         CPUs.
9300         * config/s390/s390.opt: Add the new option.
9302 2018-07-16  Richard Biener  <rguenther@suse.de>
9304         PR lto/86523
9305         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9306         for function-local FUNCTION_DECL and RESULT_DECL immediately.
9308 2018-07-16  Martin Liska  <mliska@suse.cz>
9310         PR ipa/86529
9311         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9312         to ::get_create.
9314 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9316         * config/arc/arcHS.md: Update ARCHS scheduling rules.
9318 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9320         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9321         for ARCHS4x.
9322         * config/arc/arc-cpus.def (hs4x): New cpu.
9323         (hs4xd): Likewise.
9324         * config/arc/arc-tables.opt: Regenerate.
9325         * config/arc/arc.c (arc_sched_issue_rate): New function.
9326         (TARGET_SCHED_ISSUE_RATE): Define.
9327         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9328         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9329         fpu_cvt.
9330         (attr tune): Add ARCHS4x tune values.
9331         (attr tune_dspmpy): Define.
9332         (*tst): Correct instruction type.
9333         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9334         * config/arc/arcHS4x.md: New file.
9335         * config/arc/fpu.md: Update instruction type attributes.
9336         * config/arc/t-multilib: Regenerate.
9338 2018-07-16  Tom de Vries  <tdevries@suse.de>
9340         PR debug/86455
9341         * var-tracking.c (vt_initialize): Fix pre_dec handling.
9343 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9345         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
9346         early clobber.
9348 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
9350         PR tree-optimization/86514
9351         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
9352         conversion to a boolean type from a type with greater precision.
9354 2018-07-16  Tom de Vries  <tdevries@suse.de>
9356         * var-tracking.c (vt_initialize): Print adjusted insn slim if
9357         dump_flags request TDF_SLIM.
9359 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
9361         * fold-const.c (int_const_binop_1): Abstract...
9362         (wide_int_binop): ...wide int code here.
9363         (poly_int_binop): ...poly int code here.
9364         Abstract the rest of int_const_binop_1 into int_const_binop.
9365         * fold-const.h (wide_int_binop): New.
9366         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
9367         Remove useless PLUS/MINUS_EXPR case.
9368         (zero_nonzero_bits_from_vr): Move wide int code...
9369         (zero_nonzero_bits_from_bounds): ...here.
9370         (extract_range_from_binary_expr_1): Move mask optimization code...
9371         (range_easy_mask_min_max): ...here.
9372         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
9373         (range_easy_mask_min_max): New.
9375 2018-07-15  Jeff Law  <law@redhat.com>
9377         PR target/85993
9378         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
9379         block.
9381 2018-07-14  Jim Wilson  <jimw@sifive.com>
9383         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
9385 2018-07-14  Paul Koning  <ni1d@arrl.net>
9387         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
9389 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
9391         * lto-streamer-out.c (copy_function_or_variable): Dump info about
9392         copying section.
9394 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
9395             Steve Munroe  <munroesj52@gmail.com>
9397         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
9398         (_mm_andnot_si128): Likewise.
9399         (_mm_or_si128): Likewise.
9400         (_mm_xor_si128): Likewise.
9402 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
9404         PR middle-end/78809
9405         * builtins.c (expand_builtin_memcmp): Inline the calls first
9406         when result_eq is false.
9407         (expand_builtin_strcmp): Inline the calls first.
9408         (expand_builtin_strncmp): Likewise.
9409         (inline_string_cmp): New routine. Expand a string compare
9410         call by using a sequence of char comparison.
9411         (inline_expand_builtin_string_cmp): New routine. Inline expansion
9412         a call to str(n)cmp/memcmp.
9413         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
9414         New option.
9415         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
9417 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
9419         * config/arm/driver-arm.c: Include arm-native.h.
9420         (host_detect_local_cpu): Use auto-generated data tables.
9421         (vendors, arm_cpu_table): Delete.  Move part information to ...
9422         * config/arm/arm-cpus.in: ... here.
9423         * config/arm/parsecpu.awk (gen_native): New function.
9424         (vendor, part): New CPU fields.
9425         (END): Add support for building the native CPU detection tables.
9426         * config/arm/t-arm (arm-native.h): Add build rule.
9427         (driver-arm.o): Add dependency on arm-native.h.
9429 2018-07-13  Richard Biener  <rguenther@suse.de>
9431         PR middle-end/85974
9432         * match.pd (addr1 - addr2): Allow either of the operand to
9433         have a conversion.
9435 2018-07-13  Tom de Vries  <tdevries@suse.de>
9437         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
9438         in remap_ssa_name.
9440 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
9442         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
9443         arrays instead of numbered variables.
9445 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
9447         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
9448         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
9450 2018-07-13  Richard Biener  <rguenther@suse.de>
9452         PR debug/86452
9453         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
9454         instead of get_context_die.
9456 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
9457             Richard Biener  <rguenther@suse.de>
9459         PR middle-end/86489
9460         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
9461         that the loop latch destination where phi is defined.
9463 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
9465         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
9466         (riscv_expand_epilogue): Add assertion to check interrupt mode.
9467         (riscv_set_current_function): Extract getting interrupt type to new
9468         function.
9469         (riscv_get_interrupt_type): New function.
9470         (riscv_merge_decl_attributes): New function, checking interrupt type is
9471         same.
9472         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
9474 2018-07-12  Paul Koning  <ni1d@arrl.net>
9476         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
9477         directive.
9479 2018-07-12  Paul Koning  <ni1d@arrl.net>
9481         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
9482         zero reference, add doloop_end instead.
9483         * doc/md.texi (decrement_and_branch_until_zero): Remove.
9484         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
9485         detail for doloop_end.
9487 2018-07-12  Martin Sebor  <msebor@redhat.com>
9489         PR c/86453
9490         * attribs.c (decl_attributes): Reject conflicting attributes before
9491         calling attribute handlers.
9493 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9495         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
9496          parameter.
9497         (gcc::dump_manager::get_dump_file_name): likewise.
9498         (dump_begin): Likewise.
9499         * dumpfile.h (dump_begin): Update prototype.
9500         (gcc::dump_manager::get_dump_file_name,
9501         gcc::dump_manager::get_dump_file_name): Update prototype.
9503 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9505         * internal-fn.h (vectorizable_internal_fn_p): New function.
9506         * tree-vect-slp.c (compatible_calls_p): Likewise.
9507         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
9508         to internal functions.
9509         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
9511 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9513         * fold-const.h (inverse_conditions_p): Declare.
9514         * fold-const.c (inverse_conditions_p): New function.
9515         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
9516         that test the inverse condition of a conditional internal function.
9517         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
9518         * internal-fn.c (internal_fn_mask_index): Handle conditional
9519         internal functions.
9520         (vectorized_internal_fn_supported_p): New function.
9521         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
9522         (any_pred_load_store): Replace with...
9523         (need_to_predicate): ...this new variable.
9524         (redundant_ssa_names): New variable.
9525         (ifcvt_can_use_mask_load_store): Move initial checks to...
9526         (ifcvt_can_predicate): ...this new function.  Handle tree codes
9527         for which a conditional internal function exists.
9528         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
9529         instead of ifcvt_can_use_mask_load_store.  Update after variable
9530         name change.
9531         (predicate_load_or_store): New function, split out from
9532         predicate_mem_writes.
9533         (check_redundant_cond_expr): New function.
9534         (value_available_p): Likewise.
9535         (predicate_rhs_code): Likewise.
9536         (predicate_mem_writes): Rename to...
9537         (predicate_statements): ...this.  Use predicate_load_or_store
9538         and predicate_rhs_code.
9539         (combine_blocks, tree_if_conversion): Update after above name changes.
9540         (ifcvt_local_dce): Handle redundant_ssa_names.
9541         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
9542         general conditional functions.
9543         * tree-vect-stmts.c (vectorizable_call): Likewise.
9545 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9546             Alan Hayward  <alan.hayward@arm.com>
9547             David Sherwood  <david.sherwood@arm.com>
9549         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
9550         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
9551         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
9552         plus and minus and convert them into IFN_COND_FMA-based sequences.
9553         (convert_mult_to_fma): Handle conditional plus and minus.
9555 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9557         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
9558         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
9559         (cond_fnms_optab): New optabs.
9560         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
9561         internal functions.
9562         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
9563         * internal-fn.h (get_conditional_internal_fn): Declare.
9564         (get_unconditional_internal_fn): Likewise.
9565         * internal-fn.c (cond_ternary_direct): New macro.
9566         (expand_cond_ternary_optab_fn): Likewise.
9567         (direct_cond_ternary_optab_supported_p): Likewise.
9568         (FOR_EACH_COND_FN_PAIR): Likewise.
9569         (get_conditional_internal_fn): New function.
9570         (get_unconditional_internal_fn): Likewise.
9571         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
9572         (gimple_match_op::gimple_match_op): Add a new overload for 5
9573         operands.
9574         (gimple_match_op::set_op): Likewise.
9575         (gimple_resimplify5): Declare.
9576         * genmatch.c (decision_tree::gen): Generate simplifications for
9577         5 operands.
9578         * gimple-match-head.c (gimple_simplify): Define an overload for
9579         5 operands.  Handle calls with 5 arguments in the top-level overload.
9580         (convert_conditional_op): Handle conversions from unconditional
9581         internal functions to conditional ones.
9582         (gimple_resimplify5): New function.
9583         (build_call_internal): Pass a fifth operand.
9584         (maybe_push_res_to_seq): Likewise.
9585         (try_conditional_simplification): Try converting conditional
9586         internal functions to unconditional internal functions.
9587         Handle 3-operand unconditional forms.
9588         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
9589         Define ternary equivalents of the current rules for binary conditional
9590         internal functions.
9591         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
9592         ternary operations.
9593         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
9594         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
9595         (optab): Handle them.
9596         (SVE_COND_FP_TERNARY): New int iterator.
9597         (sve_fmla_op, sve_fmad_op): New int attributes.
9598         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
9599         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
9600         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
9602 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9604         * target.def (preferred_else_value): New target hook.
9605         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
9606         * doc/tm.texi: Regenerate.
9607         * targhooks.h (default_preferred_else_value): Declare.
9608         * targhooks.c (default_preferred_else_value): New function.
9609         * internal-fn.h (conditional_internal_fn_code): Declare.
9610         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
9611         (get_conditional_internal_fn): Use it.
9612         (conditional_internal_fn_code): New function.
9613         * gimple-match.h (gimple_match_cond): New struct.
9614         (gimple_match_op): Add a cond member function.
9615         (gimple_match_op::gimple_match_op): Update all forms to take a
9616         gimple_match_cond.
9617         * genmatch.c (expr::gen_transform): Use the same condition as res_op
9618         for the suboperation, but don't specify a particular else_value.
9619         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
9620         (visit_nary_op, visit_reference_op_load): Pass
9621         gimple_match_cond::UNCOND to the gimple_match_op constructor.
9622         * gimple-match-head.c: Include tree-eh.h
9623         (convert_conditional_op): New function.
9624         (maybe_resimplify_conditional_op): Likewise.
9625         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
9626         (gimple_resimplify2): Likewise.
9627         (gimple_resimplify3): Likewise.
9628         (gimple_resimplify4): Likewise.
9629         (maybe_push_res_to_seq): Return null for conditional operations.
9630         (try_conditional_simplification): New function.
9631         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
9632         constructor.
9633         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
9634         IFN_COND_* call.
9635         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
9636         function.
9637         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
9639 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9641         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9642         DECL_FCONTEXT
9643         (hash_tree): Do not hash DECL_FCONTEXT
9644         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
9645         Do not stream DECL_FCONTEXT.
9646         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
9647         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
9649 2018-07-12  Richard Biener  <rguenther@suse.de>
9651         PR debug/86462
9652         * dwarf2out.c (gen_block_die): Only output blocks when they have
9653         at least one !DECL_IGNORED_P variable.
9655 2018-07-12  Richard Biener  <rguenther@suse.de>
9657         PR target/84829
9658         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
9659         Remove -mieee-fp handling.
9661 2018-07-12  Richard Biener  <rguenther@suse.de>
9663         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
9664         left-over from last patch.
9666 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
9668         PR tree-optimization/86492
9669         * gimple-ssa-store-merging.c
9670         (imm_store_chain_info::coalesce_immediate_stores): Call
9671         check_no_overlap even for the merge_overlapping case.  Formatting fix.
9673 2018-07-12  Richard Biener  <rguenther@suse.de>
9675         PR middle-end/86479
9676         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
9677         move possibly trapping operations into the conditional.
9679 2018-07-12  Richard Biener  <rguenther@suse.de>
9681         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
9682         (vn_lookup_simplify_result): Remove recursion limit applied
9683         here.
9684         (vn_nary_build_or_lookup_1): Adjust.
9685         (try_to_simplify): Likewise.
9686         * gimple-match-head.c (gimple_resimplify1): Instead apply one
9687         here.
9688         (gimple_resimplify2): Likewise.
9689         (gimple_resimplify3): Likewise.
9690         (gimple_resimplify4): Likewise.
9692 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
9694         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
9695         Use __mmask64 type instead of __mmask8 for __M argument.
9696         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
9697         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
9698         __U argument.
9699         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
9700         __mmask16 for __M argument.
9701         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
9702         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
9703         to __mmask16 instead of __mmask8.
9704         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
9705         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
9706         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
9707         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
9708         instead of __mmask16 for __U argument.
9709         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
9710         __mmask16 instead of __mmask8 for __U argument.
9711         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
9712         __U argument.
9713         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
9714         __mmask16.
9715         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
9716         argument.
9717         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
9718         __U argument.
9719         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
9720         __mmask16.
9721         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
9722         of __mmask16.
9723         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
9724         __U argument.
9725         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
9726         __U argument.
9727         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
9728         __U argument.
9729         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
9730         __U argument.
9731         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
9732         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
9733         return type as well as __M argument type and all casts from __mmask8
9734         to __mmask32.
9735         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
9736         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
9737         return type as well as __M argument type and all casts from __mmask8
9738         to __mmask16.
9739         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
9740         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
9741         return type as well as __M argument type and all casts from __mmask8
9742         to __mmask32.
9743         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
9744         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
9745         return type as well as __M argument type and all casts from __mmask8
9746         to __mmask16.
9747         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
9748         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
9749         __mmask16.
9751 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
9753         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
9754         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
9755         for __U argument.
9757 2018-07-11  Paul Koning  <ni1d@arrl.net>
9759         * doc/md.texi (define_subst): Document how multiple occurrences of
9760         the same argument in the replacement pattern are handled.
9762 2018-07-11  Paul Koning  <ni1d@arrl.net>
9764         * doc/extend.texi (Common Variable Attributes): Move "mode" into
9765         alphabetical order.
9766         (Common Type Attributes): Add "mode" attribute.
9768 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
9770         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
9771         stream DECL_ORIGINAL_TYPE.
9772         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
9773         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
9774         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
9775         Do not walk original type.
9776         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
9777         external decls.
9778         (write_ts_decl_non_common_tree_pointers): Do not stream
9779         DECL_ORIGINAL_TYPE
9780         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
9781         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
9783 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
9785         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
9786         thread twice from the same starting edge.
9788 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
9790         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
9791         * gimple.c (gimple_call_nonnull_result_p): ...here...
9792         (gimple_call_nonnull_arg): ...and here.
9793         * gimple.h (gimple_call_nonnull_result_p): New.
9794         (gimple_call_nonnull_arg): New.
9796 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
9798         * config/arm/arm-cpus.in: Move information from fpu field of each
9799         cpu definition to the isa field.
9800         * config/arm/parsecpu.awk (fpu): Delete match rule.
9801         (gen_comm_data): Don't add bits from the CPU's FPU entry.
9803 2018-07-11  Richard Biener  <rguenther@suse.de>
9805         PR debug/86457
9806         * dwarf2out.c (init_sections_and_labels): Use
9807         output_asm_line_debug_info consistently.
9808         (dwarf2out_early_finish): Likewise.
9809         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
9810         type units.
9812 2018-07-11  Richard Biener  <rguenther@suse.de>
9814         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
9815         Rework father_bb setting in a way to avoid propagating constants
9816         multiple times on a loop body.
9818 2018-07-10  Mark Wielaard  <mark@klomp.org>
9820         PR debug/86459
9821         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
9823 2018-07-10  Richard Biener  <rguenther@suse.de>
9825         * hash-map.h (hash_map::iterator::operator*): Return
9826         references to key and value.
9828 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
9830         PR c++/86443
9831         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
9832         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
9833         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
9834         (gimplify_omp_for): For composite loops, move outer
9835         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
9836         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
9837         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
9838         TREE_LIST for both the original class iterator and the "last" helper
9839         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
9840         loop, remember has_decl_expr from outer composite loops for the
9841         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
9843 2018-07-09  Martin Sebor  <msebor@redhat.com>
9845         PR middle-end/77357
9846         PR middle-end/86428
9847         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
9848         accessing implicitly initialized array elements.
9849         * expr.c (string_constant): Handle string initializers of
9850         character arrays within aggregates.
9851         * gimple-fold.c (fold_array_ctor_reference): Add argument.
9852         Store element offset.  As a special case, handle zero size.
9853         (fold_nonarray_ctor_reference): Same.
9854         (fold_ctor_reference): Add argument.  Store subobject offset.
9855         * gimple-fold.h (fold_ctor_reference): Add argument.
9857 2018-07-09  Paul Koning  <ni1d@arrl.net>
9859         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
9860         (pdp11_insn_cost): New function.
9861         (pdp11_md_asm_adjust): New function.
9862         (TARGET_INVALID_WITHIN_DOLOOP): Define.
9863         (pdp11_rtx_costs): Update to match machine better.
9864         (output_addr_const_pdp11): Correct format mismatch warnings.
9865         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
9866         * config/pdp11/pdp11.md: General change to add base_cost and/or
9867         length attributes for use by new pdp11_insn_cost function.
9868         (MIN_BRANCH): Correct definition.
9869         (MIN_SOB): Ditto.
9870         (doloop_end): Use standard pattern name for looping pattern.
9871         (doloop_end_nocc): New.
9872         (movsf): Add another constraint alternative.
9873         (zero_extendqihi2): Add constraint alternatives for not in place
9874         extend.
9875         (zero_extendhisi2): Remove.
9876         (shift patterns): Add CC handling variants.
9877         (bswaphi2): New.
9878         (bswapsi2): New.
9879         (rothi3): New.
9880         (define_peephole2): New peephole to recognize mov that sets CC for
9881         subsequent test.
9883 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9885         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
9886         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
9887         wi::add.
9889 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
9891         PR c/86420
9892         * real.c (real_nextafter): Return true if result is denormal.
9894 2018-07-09  Martin Liska  <mliska@suse.cz>
9896         * common.opt: Add back wrongly removed attribute.
9898 2018-07-09  Richard Biener  <rguenther@suse.de>
9900         PR debug/86413
9901         * dwarf2out.c (gen_block_die): For an early generated DIE
9902         always output high/low PC attributes.
9904 2018-07-09  Tom de Vries  <tdevries@suse.de>
9906         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
9907         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
9908         onto VAR_DECL with abstract origin.
9910 2018-07-07  Jim Wilson  <jimw@sifive.com>
9912         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
9914 2018-07-07  Tom de Vries  <tdevries@suse.de>
9916         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
9918 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
9920         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
9921         overflow_type.
9922         (combine_bound): Use wide-int overflow calculation instead of
9923         rolling our own.
9924         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
9925         overflow_type.
9926         * fold-const.c (int_const_binop_2): Same.
9927         (extract_muldiv_1): Same.
9928         (fold_div_compare): Same.
9929         (fold_abs_const): Same.
9930         * match.pd: Same.
9931         * poly-int.h (add): Same.
9932         (sub): Same.
9933         (neg): Same.
9934         (mul): Same.
9935         * predict.c (predict_iv_comparison): Same.
9936         * profile-count.c (slow_safe_scale_64bit): Same.
9937         * simplify-rtx.c (simplify_const_binary_operation): Same.
9938         * tree-chrec.c (tree_fold_binomial): Same.
9939         * tree-data-ref.c (split_constant_offset_1): Same.
9940         * tree-if-conv.c (idx_within_array_bound): Same.
9941         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
9942         * tree-ssa-phiopt.c (minmax_replacement): Same.
9943         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
9944         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
9945         * vr-values.c (vr_values::adjust_range_with_scev): Same.
9946         * wide-int.cc (wi::add_large): Same.
9947         (wi::mul_internal): Same.
9948         (wi::sub_large): Same.
9949         (wi::divmod_internal): Same.
9950         * wide-int.h: Change overflow type to overflow_type for neg, add,
9951         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
9952         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
9953         mul_internal, divmod_internal.
9954         (overflow_type): New enum.
9955         (accumulate_overflow): New.
9957 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
9959         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
9960         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
9962 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
9964         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
9965         argument is checked for zero before entering loop, avoid checking again.
9967 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
9969         * gimplify.h (generic_expr_could_trap_p): Set as global function.
9970         * gimplify.h (generic_expr_could_trap_p): Likwise.
9971         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
9973 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
9975         PR tree-optimization/86401
9976         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
9977         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
9978         (fold_bit_and_mask): ... here.  New helper function for match.pd.
9979         * fold-const.h (fold_bit_and_mask): Declare.
9980         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
9982 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
9984         PR target/86324
9985         * target.def (translate_mode_attribute): New hook.
9986         * targhooks.h (default_translate_mode_attribute): Declare.
9987         * targhooks.c (default_translate_mode_attribute): New function.
9988         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
9989         * doc/tm.texi: Regenerate.
9990         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
9991         (rs6000_translate_mode_attribute): New function.
9993 2018-07-06  Paul Koning  <ni1d@arrl.net>
9995         * doc/md.texi (define_split): Document DONE and FAIL.
9996         (define_peephole2): Ditto.
9998 2018-07-05  Jeff Law  <law@redhat.com>
10000         PR tree-optimization/86010
10001         * tree-ssa-dse.c (compute_trims): More aggressively trim at
10002         both the head and tail of mem* and str* calls.
10004 2018-07-05  Jim Wilson  <jimw@sifive.com>
10006         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
10008 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
10010         * config/aarch64/aarch64-simd.md: correct flags text for
10011         MIN_EXPR replacement.
10013 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
10015         * configure: Regenerated.
10017 2018-07-05  Carl Love  <cel@us.ibm.com>
10019         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
10020         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
10021         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
10022         VSX_BUILTIN_DOUBLEL_V4SF.
10024 2018-07-05  Martin Sebor  <msebor@redhat.com>
10026         PR c++/86400
10027         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
10028         than its domain to compute its the upper bound of a char array.
10030 2018-07-05  Nathan Sidwell  <nathan@acm.org>
10032         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
10033         * doc/cpp.texi: Update comment.
10034         * doc/tm.texi: Rebuilt.
10035         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
10036         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
10037         * doc/extend.texi (Backwards Compatibility): Clarify it is system
10038         headers affected by extern "C".
10039         * system.h: Poison NO_IMPLICIT_EXTERN_C.
10040         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
10041         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
10042         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
10043         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
10044         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
10045         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
10046         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
10047         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
10048         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
10049         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
10050         config/sparc/sp64-elf.h, config/spu/spu.h,
10051         config/stormy16/stormy16.h, config/v850/v850.h,
10052         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
10053         define NO_IMPLICIT_EXTERN_C.
10054         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
10056 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10058         PR target/84711
10059         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10060         instead of GET_MODE_SIZE when comparing Units.
10062 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10064         PR target/84711
10065         * rtlanal.c (set_noop_p): Constrain on mode change,
10066         include hard-reg-set.h
10068 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10070         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
10072 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
10074         Revert
10075         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10077         PR sanitizer/84250
10078         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10079         libasan.
10080         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10082 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10084         PR sanitizer/84250
10085         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10086         libasan.
10087         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10089 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
10091         PR middle-end/86380
10092         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
10094 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
10096         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
10097         neg_*_op* variables.
10099 2018-07-04  Martin Liska  <mliska@suse.cz>
10101         * tree-switch-conversion.c: Define
10102         max_ratio_for_speed and max_ratio_for_size constants.
10104 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10105             Martin Liska  <mliska@suse.cz>
10107         PR middle-end/66240
10108         PR target/45996
10109         PR c/84100
10110         * common.opt: Rename align options with 'str_' prefix.
10111         * common/config/i386/i386-common.c (set_malign_value): New
10112         function.
10113         (ix86_handle_option): Use it to set -falign-* options/
10114         * config/aarch64/aarch64-protos.h (struct tune_params): Change
10115         type from int to string.
10116         * config/aarch64/aarch64.c: Update default values from int
10117         to string.
10118         * config/alpha/alpha.c (alpha_override_options_after_change):
10119         Likewise.
10120         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
10121         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10122         max skip conditionally.
10123         * config/i386/freebsd.h (SUBALIGN_LOG): New.
10124         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10125         max skip conditionally.
10126         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10127         max skip conditionally.
10128         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
10129         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10130         max skip conditionally.
10131         * config/i386/i386.c (struct ptt): Change type from int to
10132         string.
10133         (ix86_default_align): Set default values.
10134         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
10135         max skip conditionally.
10136         * config/i386/iamcu.h (SUBALIGN_LOG): New.
10137         (ASM_OUTPUT_MAX_SKIP_ALIGN):
10138         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
10139         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10140         max skip conditionally.
10141         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
10142         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
10143         * config/i386/x86-64.h (SUBALIGN_LOG): New.
10144         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10145         max skip conditionally.
10146         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10147         * config/ia64/ia64.c (ia64_option_override): Set default values
10148         for alignment options.
10149         * config/m68k/m68k.c: Handle new str_align_* options.
10150         * config/mips/mips.c (mips_set_compression_mode): Change
10151         type of constants.
10152         (mips_option_override): Set default values for options.
10153         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
10154         Likewise.
10155         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10156         Likewise.
10157         * config/rx/rx.c (rx_option_override): Likewise.
10158         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
10159         (LABEL_ALIGN): Use align_labels_log.
10160         (LOOP_ALIGN): Use align_loops_align.
10161         * config/s390/s390.c (s390_asm_output_function_label): Use new
10162         macros.
10163         * config/sh/sh.c (sh_override_options_after_change):
10164         Change type of constants.
10165         * config/spu/spu.c (spu_sched_init): Likewise.
10166         * config/sparc/sparc.c (sparc_option_override): Set default
10167         values for options.
10168         * config/visium/visium.c (visium_option_override): Likewise.
10169         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
10170         emit p2align format with last argument if it's not needed.
10171         * doc/invoke.texi: Document extended format of -falign-*.
10172         * final.c: Use align_labels alignment.
10173         * flags.h (struct target_flag_state): Change type to use
10174         align_flags.
10175         (struct align_flags_tuple): New.
10176         (struct align_flags): Likewise.
10177         (align_loops_log): Redefine macro to use new types.
10178         (align_loops_max_skip): Redefine macro to use new types.
10179         (align_jumps_log): Redefine macro to use new types.
10180         (align_jumps_max_skip): Redefine macro to use new types.
10181         (align_labels_log): Redefine macro to use new types.
10182         (align_labels_max_skip): Redefine macro to use new types.
10183         (align_functions_log): Redefine macro to use new types.
10184         (align_loops): Redefine macro to use new types.
10185         (align_jumps): Redefine macro to use new types.
10186         (align_labels): Redefine macro to use new types.
10187         (align_functions): Redefine macro to use new types.
10188         (align_functions_max_skip): Redefine macro to use new types.
10189         (align_loops_value): New macro.
10190         (align_jumps_value): New macro.
10191         (align_labels_value): New macro.
10192         (align_functions_value): New macro.
10193         * function.c (invoke_set_current_function_hook): Propagate
10194         alignment values from flags to global variables default in
10195         topleev.h.
10196         * ipa-icf.c (sem_function::equals_wpa): Use
10197         cl_optimization_option_eq instead of memcmp.
10198         * lto-streamer.h (cl_optimization_stream_out): Support streaming
10199         of string types.
10200         (cl_optimization_stream_in): Likewise.
10201         * optc-save-gen.awk: Support strings in cl_optimization.
10202         * opth-gen.awk: Likewise.
10203         * opts.c (finish_options): Remove error checking of invalid
10204         value ranges.
10205         (MAX_CODE_ALIGN): Remove.
10206         (MAX_CODE_ALIGN_VALUE): Likewise.
10207         (parse_and_check_align_values): New function.
10208         (check_alignment_argument): Likewise.
10209         (common_handle_option): Use check_alignment_argument.
10210         * opts.h (parse_and_check_align_values): Declare.
10211         * toplev.c (init_alignments): Remove.
10212         (read_log_maxskip): New.
10213         (parse_N_M): Likewise.
10214         (parse_alignment_opts): Likewise.
10215         (backend_init_target): Remove usage of init_alignments.
10216         * toplev.h (parse_alignment_opts): Declare.
10217         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
10218         argument.
10219         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
10220         * tree.c (cl_option_hasher::equal): New.
10221         * varasm.c: Use new global macros.
10223 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10225         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
10226         Use a simpler align directive also if MAXSKIP = ALIGN-1.
10227         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10228         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10229         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10230         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10231         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10232         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
10233         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10234         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10235         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10236         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10237         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10239 2018-07-04  Martin Liska  <mliska@suse.cz>
10240             Jonathan Wakely  <jwakely@redhat.com>
10242         * coverage.c: Use correct type.
10243         * doc/invoke.texi: Language correction.
10245 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
10247         PR target/85620
10248         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10249         ENDBRANCH for non-tail call which may return via indirect branch.
10250         * doc/extend.texi: Document indirect_return attribute.
10252 2018-07-03  Martin Sebor  <msebor@redhat.com>
10254         PR tree-optimization/86274
10255         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10256         precondition.
10257         (format_floating): Correct handling of infinities and NaNs.
10259 2018-07-03  Martin Sebor  <msebor@redhat.com>
10261         * print-tree.c (print_real_cst): New function.
10262         (print_node_brief): Call it.
10263         (print_node): Ditto.
10265 2018-07-03  Jeff Law  <law@redhat.com>
10267         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10268         into a single pattern.
10270         * config/h8300/h8300.md (ors code_iterator): New.
10271         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10272         a single pattern and single splitter.
10273         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10274         (iorqi3_1, xorqi3_1): Likewise.
10275         (iorqi3, xorqi3 expanders): Similarly.
10277         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10278         (movmd_internal) into a single pattern using the P mode iterator.
10279         (movmd splitters): Similarly.
10280         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
10281         (movsd splitters): Similarly.
10283         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
10284         ADDB, ADDW and ADDL into a single ADD attribute which selects the
10285         right table based on the size of the operand.
10286         * config/h8300/h8300.md (length_table): Corresponding changes. All
10287         references to "addb", "addw" and "addl" changed to "add".
10288         (btst patterns): Merge two variants into a single pattern.
10289         (tstqi, tsthi): Likewise.
10290         (addhi3_incdec, addsi3_incdec): Likewise.
10291         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
10292         (mulhi3, mulsi3): Likewise.
10293         (udivhi3, udivsi3): Likewise.
10294         (divhi3, divsi3): Likewise.
10295         (andorqi3, andorhi3, andorsi3): Likewise.
10297 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
10299         PR target/85694
10300         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10301         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10303 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10305         PR tree-optimization/85694
10306         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10307         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10308         UNSPEC_URHADD.
10309         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10310         (<u>avg<mode>3_ceil): New patterns.
10312 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
10314         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10315         scan-tree-dump directive.
10316         * gcc.dg/vect/slp-perm-2.c: Likewise.
10317         * gcc.dg/vect/slp-perm-3.c: Likewise.
10318         * gcc.dg/vect/slp-perm-5.c: Likewise.
10319         * gcc.dg/vect/slp-perm-6.c: Likewise.
10320         * gcc.dg/vect/slp-perm-7.c: Likewise.
10321         * gcc.dg/vect/slp-perm-8.c: Likewise.
10323 2018-07-03  Marek Polacek  <polacek@redhat.com>
10325         PR middle-end/86202
10326         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10328 2018-07-03  Richard Biener  <rguenther@suse.de>
10330         PR ipa/86389
10331         * tree-ssa-structalias.c (find_func_clobbers): Properly
10332         handle indirect calls.
10334 2018-07-03  Jeff Law  <law@redhat.com>
10336         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10337         (shifts): New code iterator.
10338         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
10339         expander.  Fix HImode handling on H8/SX.
10340         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
10341         (subqi3, subhi3, subsi3 expanders): Likewise.
10342         (andqi3, andhi3, andsi3 expanders): Likewise.
10343         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
10344         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
10345         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
10346         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
10347         (zero_extendqihi2, zero_extendqisi2): Likewise.
10348         (extendqihi2, extendqisi2): Likewise.
10349         (rotlqi3, rotlhi3, rotlsi3): Likewise.
10350         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
10351         (rotlqi3_1, rotlhi3_1): Likewise.
10352         (logicalhi3_sn, logicalsi3_sn): Likewise.
10353         (logicalhi3, logicalsi3): Likewise.
10355 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10357         * tree-vect-patterns.c (vect_recog_rotate_pattern)
10358         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10359         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
10360         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
10361         type to append_pattern_def_seq instead of creating a stmt_vec_info
10362         directly.
10363         (build_mask_conversion): Likewise.  Remove vinfo argument.
10364         (vect_add_conversion_to_patterm): Likewise, renaming to...
10365         (vect_add_conversion_to_pattern): ...this.
10366         (vect_recog_mask_conversion_pattern): Update call to
10367         build_mask_conversion.  Pass the vector type to
10368         append_pattern_def_seq here too.
10369         (vect_recog_gather_scatter_pattern): Update call to
10370         vect_add_conversion_to_pattern.
10372 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10374         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
10375         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10376         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
10377         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
10378         STMT_VINFO_PATTERN_DEF_SEQ to null here.
10379         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
10380         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10381         append_pattern_def_seq instead of new_pattern_def_seq.
10382         (vect_recog_divmod_pattern): Do both of the above.
10383         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
10384         is null.
10386 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10388         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10389         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
10390         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
10391         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
10392         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
10393         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10394         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
10395         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10396         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10397         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
10398         parameter with a single stmt_vec_info.
10399         (vect_recog_func_ptr): Likewise.
10400         (vect_recog_gather_scatter_pattern): Likewise, folding in...
10401         (vect_try_gather_scatter_pattern): ...this.
10402         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
10403         the stmt_vec_info of the statement to be matched.  Don't clear
10404         STMT_VINFO_RELATED_STMT.
10405         (vect_pattern_recog): Update call accordingly.
10407 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10409         PR tree-optimization/85694
10410         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
10411         (uavgM3_ceil): Document new optabs.
10412         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
10413         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
10414         functions.
10415         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
10416         (savg_ceil_optab): New optabs.
10417         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
10418         (vect_vect_recog_func_ptrs): Add it.
10419         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
10420         constant directly from the associated lhs.
10422 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10424         * tree-vect-patterns.c (vect_split_statement): New function.
10425         (vect_convert_input): Use it to try to split an existing cast.
10427 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10429         * poly-int.h (print_hex): New function.
10430         * dumpfile.h (dump_dec, dump_hex): Declare.
10431         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
10432         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
10433         min_input_precision, operation_precision and operation_sign.
10434         * tree-vect-patterns.c (vect_get_range_info): New function.
10435         (vect_same_loop_or_bb_p, vect_single_imm_use)
10436         (vect_operation_fits_smaller_type): Delete.
10437         (vect_look_through_possible_promotion): Add an optional
10438         single_use_p parameter.
10439         (vect_recog_over_widening_pattern): Rewrite to use new
10440         stmt_vec_info infomration.  Handle one operation at a time.
10441         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
10442         (vect_truncatable_operation_p, vect_set_operation_type)
10443         (vect_set_min_input_precision): New functions.
10444         (vect_determine_min_output_precision_1): Likewise.
10445         (vect_determine_min_output_precision): Likewise.
10446         (vect_determine_precisions_from_range): Likewise.
10447         (vect_determine_precisions_from_users): Likewise.
10448         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
10449         (vect_vect_recog_func_ptrs): Put over_widening first.
10450         Add cast_forwprop.
10451         (vect_pattern_recog): Call vect_determine_precisions.
10453 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10455         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
10456         statements that have been replaced by further pattern statements.
10457         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
10459 2018-07-03  Richard Biener  <rguenther@suse.de>
10461         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10462         always set *dt.  Dump vectype in vectype overload.
10463         * dumpfile.h (dump_gimple_expr): New function.
10464         (dump_gimple_expr_loc): Likewise.
10465         * dumpfile.c (dump_gimple_expr): New function.
10466         (dump_gimple_expr_loc): Likewise.
10468 2018-07-02  Jeff Law  <law@redhat.com>
10470         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
10471         the H8/300, H8/300H and H8/S variants into a single pattern.
10472         (movhi_h8300, movqi_h8300hs): Similarly.
10473         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
10474         (QHI mode iterator): New.
10476         * config/h8300/h8300.md: Remove trailing whitespace.
10478 2018-07-02  Jim Wilson  <jimw@sifive.com>
10480         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
10481         instead of emit_insn for interrupt returns.
10482         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
10483         (riscv_sret, riscv_uret): Likewise.
10485 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10487         * pretty-print.c (selftest::test_pp_format): Move save and restore
10488         of quotes to class auto_fix_quotes, and add an instance.
10489         * selftest.c: Include "intl.h".
10490         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
10491         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
10492         * selftest.h (selftest::auto_fix_quotes): New class.
10494 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
10496         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
10497         (aarch64_sve_prepare_conditional_op): Remove.
10498         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
10499         Allow aarch64_simd_reg_or_zero as select operand; remove
10500         the aarch64_sve_prepare_conditional_op call.
10501         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
10502         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10503         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
10504         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
10505         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
10506         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
10507         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
10508         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
10509         and a splitters to match all of the *_any patterns.
10510         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
10512         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
10513         (SVE_COND_FP_BINARY_REV): Remove.
10514         (sve_int_op_rev, sve_fp_op_rev): New.
10515         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
10516         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
10517         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
10518         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
10519         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
10520         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
10521         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
10522         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
10523         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
10525         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
10526         Remove match_dup 1 from the inner unspec.
10527         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10529         * config/aarch64/aarch64.md (movprfx): New attr.
10530         (length): Default movprfx to 8.
10531         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
10532         (*madd<SVE_I>, *msub<SVE_I): Likewise.
10533         (*<su>mul<SVE_I>3_highpart): Likewise.
10534         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
10535         (*v<ASHIFT><SVE_I>3): Likewise.
10536         (*<su><MAXMIN><SVE_I>3): Likewise.
10537         (*<su><MAXMIN><SVE_F>3): Likewise.
10538         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
10539         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
10540         (*div<SVE_F>4): Likewise.
10542 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
10544         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
10545         in dump string.
10547 2018-07-02  Richard Biener  <rguenther@suse.de>
10549         PR tree-optimization/86363
10550         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
10551         memset argument refers to a non-variable address.
10553 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
10555         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
10556         {PLUS,MINUS}_EXPR code to...
10557         (adjust_symbolic_bound): ...here,
10558         (combine_bound): ...here,
10559         (set_value_range_with_overflow): ...and here.
10561 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
10563         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
10564         code...
10565         (extract_range_from_abs_expr): ...here.
10567 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
10569         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
10570         -fno-omit-frame-pointer when not optimizing.
10572 2018-07-02  Martin Liska  <mliska@suse.cz>
10574         PR ipa/86279
10575         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
10576         (propagate_nothrow): Likewise.
10578 2018-07-02  Martin Liska  <mliska@suse.cz>
10580         PR ipa/86323
10581         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
10583 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10585         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
10586         function in r262149, changing "loc" param from source_location to
10587         const dump_location_t &.
10588         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
10589         declaration, as above.
10591 2018-07-01  Paul Koning  <ni1d@arrl.net>
10593         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10594         -munit-asm, -mgnu-asm, -mdec-asm.
10595         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
10596         (pdp11_output_labelref): New.
10597         (pdp11_output_def): New.
10598         (pdp11_output_addr_vec_elt): New.
10599         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
10600         %# and %@ format codes.
10601         (pdp11_option_override): New.
10602         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
10603         (pdp11_output_ident): New.
10604         (pdp11_asm_named_section): New.
10605         (pdp11_asm_init_sections): New.
10606         (pdp11_file_start): New.
10607         (pdp11_file_end): New.
10608         (output_ascii): Use .ascii/.asciz for -mdec-asm.
10609         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
10610         %o, like %c but octal.
10611         (pdp11_option_override): New.
10612         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
10613         -mdec-asm.
10614         (DATA_SECTION_ASM_OP): Ditto.
10615         (READONLY_DATA_SECTION_ASM_OP): New.
10616         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
10617         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
10618         (ASM_OUTPUT_LABELREF): Ditto.
10619         (ASM_OUTPUT_DEF): Ditto.
10620         (ASM_OUTPUT_EXTERNAL): New.
10621         (ASM_OUTPUT_SOURCE_FILENAME): New.
10622         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
10623         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
10624         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
10625         %# and %@ format codes.
10626         * config/pdp11/pdp11.opt (mgnu-asm): New.
10627         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
10628         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
10629         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
10631 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
10633         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
10634         dereferencing path[] beyond its length.
10635         (debug_path): New.
10636         (debug_all_paths): New.
10637         (rewire_first_differing_edge): New.
10638         (adjust_paths_after_duplication): New.
10639         (duplicate_thread_path): Call adjust_paths_after_duplication.
10640         Add new argument.
10641         (thread_through_all_blocks): Add new argument to
10642         duplicate_thread_path.
10644 2018-06-30  Jim Wilson  <jimw@sifive.com>
10646         * config/riscv/predicates.md (p2m1_shift_operand): New.
10647         (high_mask_shift_operand): New.
10648         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
10649         pattern using p2m1_shift_operand.
10650         (lshsi3_zero_extend_3+2): New combiner pattern using
10651         high_mask_shift_operand.
10653 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10655         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
10656         split out from...
10657         (vect_recog_rotate_pattern): ...here.
10658         (vect_convert_input): Try to insert casts of invariants in the
10659         preheader.
10660         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
10661         preheader to be empty.
10663 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10665         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
10666         vector type.  If given, install it in the new statement's
10667         STMT_VINFO_VECTYPE.
10668         (vect_element_precision): New function.
10669         (vect_unpromoted_value): New struct.
10670         (vect_unpromoted_value::vect_unpromoted_value): New function.
10671         (vect_unpromoted_value::set_op): Likewise.
10672         (vect_look_through_possible_promotion): Likewise.
10673         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
10674         (vect_widened_op_tree, vect_convert_input): Likewise.
10675         (vect_convert_inputs, vect_convert_output): Likewise.
10676         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
10677         to handle the optional cast of the multiplication result and
10678         vect_widened_op_tree to detect the widened multiplication itself.
10679         Do not require the input and output of promotion casts to have
10680         the same sign, but base the signedness of the operation on the
10681         input rather than the result.  If the pattern includes two
10682         promotions, check that those promotions have the same sign.
10683         Do not restrict the MULT_EXPR handling to a double-width result;
10684         handle quadruple-width results and wider.  Use vect_convert_inputs
10685         to convert the inputs to the common type.
10686         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
10687         to handle the optional cast of the ABS result.  Also allow a sign
10688         change or a sign extension between the ABS and MINUS.
10689         Use vect_widened_op_tree to detect the widened subtraction and use
10690         vect_convert_inputs to convert the inputs to the common type.
10691         (vect_handle_widen_op_by_const): Delete.
10692         (vect_recog_widen_op_pattern): New function.
10693         (vect_recog_widen_mult_pattern): Use it.
10694         (vect_recog_widen_shift_pattern): Likewise.
10695         (vect_recog_widen_sum_pattern): Use
10696         vect_look_through_possible_promotion to handle the promoted
10697         PLUS_EXPR operand.
10699 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10701         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
10702         the containing gimple_seq *.
10703         * gimple-iterator.h (gsi_for_stmt): Declare it.
10704         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
10705         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
10706         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10707         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10708         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
10709         checks.
10710         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
10711         split out from...
10712         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
10713         statement being replaced is part of an existing pattern
10714         definition sequence, inserting the new pattern statements before
10715         the original one.
10716         (vect_pattern_recog_1): Don't return a bool.  If the statement
10717         is already part of a pattern, instead apply pattern matching
10718         to the pattern definition statements.  Don't clear the
10719         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
10720         (vect_pattern_recog): Don't break after the first match;
10721         continue processing the pattern definition statements instead.
10722         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
10724 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10726         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10727         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10728         (vect_recog_widen_sum_pattern): Use it.
10730 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10732         * tree-vect-loop.c (vectorizable_reduction): Assert that the
10733         phi is not a pattern statement and has not been replaced by
10734         a pattern statement.
10735         * tree-vect-patterns.c (type_conversion_p): Don't check
10736         STMT_VINFO_IN_PATTERN_P.
10737         (vect_recog_vector_vector_shift_pattern): Likewise.
10738         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
10739         the pattern statement rather than the original statement; check
10740         directly for a WIDEN_MULT_EXPR here.
10741         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
10742         vect_is_simple_use to return the pattern statement rather
10743         than the original statement; use is_pattern_stmt_p to check
10744         for such a pattern statement.
10745         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
10746         to return the pattern statement rather than the original statement;
10747         don't do the same transformation here.
10748         (vect_is_simple_use): If the defining statement has been replaced
10749         by a pattern statement, return the pattern statement instead.
10750         Remove the corresponding (local) transformation from the vectype
10751         overload.
10753 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10755         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
10756         end and default to null.
10757         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10758         (vectorizable_reduction): Update calls accordingly, dropping the
10759         gimple ** argument if the passed-back statement isn't needed.
10760         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
10761         (vect_recog_rotate_pattern): Likewise.
10762         (vect_recog_mask_conversion_pattern): Likewise.
10763         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10764         (vect_mask_constant_operand_p): Likewise.
10765         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
10766         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
10767         (get_group_load_store_type, get_load_store_type): Likewise.
10768         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
10769         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
10770         (vectorizable_conversion, vectorizable_assignment): Likewise.
10771         (vectorizable_shift, vectorizable_operation): Likewise.
10772         (vectorizable_store, vect_is_simple_cond): Likewise.
10773         (vectorizable_condition, vectorizable_comparison): Likewise.
10774         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
10775         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
10776         and move it to the end.  Cope with null def_stmt_outs.
10778 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10780         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
10782 2018-06-29  Jeff Law  <law@redhat.com>
10784         * config/v850/v850.c (v850_legitimate_address_p): Handle large
10785         displacements for TARGET_V850E2V3 and newer.
10786         (TARGET_LRA_P): Remove.  Defaults to LRA now.
10787         * config/v850/v850.md (sign23byte_load): Remove.
10788         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
10789         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
10791 2018-06-29  Martin Liska  <mliska@suse.cz>
10793         PR lto/85759
10794         * coverage.c (coverage_init): Mangle full path name.
10795         * doc/invoke.texi: Document the change.
10796         * gcov-io.c (mangle_path): New.
10797         * gcov-io.h (mangle_path): Likewise.
10798         * gcov.c (mangle_name): Use mangle_path for path mangling.
10800 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10802         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
10803         if starting source register is not even.
10805 2018-06-29  Martin Liska  <mliska@suse.cz>
10807         PR tree-optimization/86263
10808         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
10809         Make edge redirection.
10811 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
10813         * dumpfile.c (dump_loc): Add indentation based on scope depth.
10814         (dump_scope_depth): New variable.
10815         (get_dump_scope_depth): New function.
10816         (dump_begin_scope): New function.
10817         (dump_end_scope): New function.
10818         * dumpfile.h (get_dump_scope_depth): New declaration.
10819         (dump_begin_scope): New declaration.
10820         (dump_end_scope): New declaration.
10821         (class auto_dump_scope): New class.
10822         (AUTO_DUMP_SCOPE): New macro.
10823         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
10824         AUTO_DUMP_SCOPE.
10826 2018-06-29  Richard Biener  <rguenther@suse.de>
10828         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
10829         compute_all_dependences succeeds.
10830         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
10831         exceed --param loop-max-datarefs-for-datadeps.
10833 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
10835         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
10837 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
10839         PR target/86348
10840         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
10841         alternative 0 in preferred_for_speed attribute.
10843 2018-06-28  Paul Koning  <ni1d@arrl.net>
10845         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
10846         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
10847         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
10848         * config/pdp11/pdp11.md: Correct "length" attribute calculation
10849         for shift insn patterns.
10851 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
10853         * cgraph.c (cgraph_node::get_body): Replace assignments to
10854         "dump_file" with calls to set_dump_file.
10855         * dumpfile.c (alt_dump_file): Make static, and group with...
10856         (alt_flags): ...this definition.
10857         (dumps_are_enabled): New variable.
10858         (refresh_dumps_are_enabled): New function.
10859         (set_dump_file): New function.
10860         (set_alt_dump_file): New function.
10861         (gcc::dump_manager::dump_start): Replace assignments to
10862         "dump_file" and "alt_dump_file" with calls to set_dump_file and
10863         set_alt_dump_file.
10864         (gcc::dump_manager::dump_finish): Likewise.
10865         * dumpfile.h (alt_dump_file): Delete decl.
10866         (dumps_are_enabled): New variable decl.
10867         (set_dump_file): New function decl.
10868         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
10869         global.
10870         * tree-nested.c (lower_nested_functions): Replace assignments to
10871         "dump_file" with calls to set_dump_file.
10873 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
10875         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
10876         goto_locus of each outgoing edge of each basic block.
10878 2018-06-28  Richard Biener  <rguenther@suse.de>
10880         * dwarf2out.c (decl_scope_table): Remove.
10881         (push_decl_scope): Likewise.
10882         (pop_decl_scope): Likewise.
10883         (gen_type_die_for_member): Do not call push/pop_decl_scope.
10884         (gen_struct_or_union_type_die): Likewise.
10885         (gen_tagged_type_die): Likewise.
10886         (dwarf2out_init): Do not initialize decl_scope_table.
10887         (dwarf2out_c_finalize): Do not free it.
10889 2018-06-28  Richard Biener  <rguenther@suse.de>
10891         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
10892         deciding whether to not re-use a DIE.
10894 2018-06-28  Richard Biener  <rguenther@suse.de>
10896         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
10897         DW_AT_abstract_origin attribute.
10899 2018-06-28  Martin Liska  <mliska@suse.cz>
10901         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10902         Use newly introduced constants.
10903         * tree-switch-conversion.h (struct jump_table_cluster):
10904         Define max_ratio_for_size and max_ratio_for_speed.
10906 2018-06-28  Martin Liska  <mliska@suse.cz>
10908         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10909         Add new checking assert to catch invalid state.
10910         (jump_table_cluster::can_be_handled): Handle single case
10911         clusters.
10912         (jump_table_cluster::is_beneficial): Bail out for such case.
10913         (bit_test_cluster::find_bit_tests):
10914         Add new checking assert to catch invalid state.
10915         (bit_test_cluster::can_be_handled): Handle single case
10916         clusters.
10917         (bit_test_cluster::is_beneficial): Bail out for such case.
10918         (switch_decision_tree::analyze_switch_statement):
10919         Fix comment.
10921 2018-06-28  Martin Liska  <mliska@suse.cz>
10923         * common.opt: Introduce -completion option.
10924         * gcc.c (driver_handle_option): Handle it.
10925         (driver::main): Print completions if completion
10926         is set.
10927         * opt-suggestions.c (option_proposer::get_completions):
10928         New function.
10929         (option_proposer::suggest_completion): Likewise.
10930         (option_proposer::find_param_completions): Likewise.
10931         (verify_autocompletions): Likewise.
10932         (test_completion_valid_options): Likewise.
10933         (test_completion_valid_params): Likewise.
10934         (in_completion_p): Likewise.
10935         (empty_completion_p): Likewise.
10936         (test_completion_partial_match): Likewise.
10937         (test_completion_garbage): Likewise.
10938         (opt_proposer_c_tests): Likewise.
10939         * opt-suggestions.h: Declare new functions.
10940         * opts.c (common_handle_option): Handle OPT__completion_.
10941         * selftest-run-tests.c (selftest::run_tests): Add
10942         opt_proposer_c_tests.
10943         * selftest.c (assert_str_startswith): New.
10944         * selftest.h (assert_str_startswith): Likewise.
10945         (opt_proposer_c_tests): New.
10946         (ASSERT_STR_STARTSWITH): Likewise.
10948 2018-06-28  Martin Liska  <mliska@suse.cz>
10950         * Makefile.in: Add opt-suggestions.o.
10951         * gcc-main.c: Include opt-suggestions.h.
10952         * gcc.c (driver::driver): Likewise.
10953         (driver::~driver): Remove m_option_suggestions.
10954         (driver::build_option_suggestions): Moved to option_proposer.
10955         (driver::suggest_option): Likewise.
10956         (driver::handle_unrecognized_options): Use option_proposer.
10957         * gcc.h (class driver): Add new memver m_option_proposer.
10958         * opt-suggestions.c: New file.
10959         * opt-suggestions.h: New file.
10961 2018-06-28  Martin Liska  <mliska@suse.cz>
10963         * vec.h (class auto_string_vec): New (moved from auto_argvec).
10964         (auto_string_vec::~auto_string_vec): Likewise.
10966 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
10968         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
10969         prevent_decl_creation_for_types fields up and add reset_location field.
10970         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
10971         statement if id->reset_location is true.
10972         (copy_edges_for_bb): Do not set goto_locus on the new edges if
10973         id->reset_location is true.
10974         (copy_phis_for_bb): Force input_location on the arguments if
10975         id->reset_location is true.
10976         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10977         is set on the function to be inlined.
10979 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
10981         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
10983 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
10985         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
10986         registers for Pmode.
10987         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
10988         hard registers for the clobbered pseudo.
10990 2018-06-27  Paul Koning  <ni1d@arrl.net>
10992         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10993         mutually exclusive options.
10994         * config/pdp11/constraints.md (h): New constraint.
10995         (O): Update definition to match shift code generation.
10996         (D): New constraint.
10997         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
10998         (CCFP): Remove.
10999         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
11000         function.
11001         (output_jump): Change arguments.
11002         (pdp11_fixed_cc_regs): New function.
11003         (pdp11_cc_mode): Ditto.
11004         (pdp11_expand_shift): Ditto.
11005         (pdp11_assemble_shift): Ditto.
11006         (pdp11_small_shift): Ditto.
11007         (pdp11_branch_cost): Remove.
11008         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
11009         from output.
11010         (pdp11_register_move_cost): Update for CC registers.
11011         (pdp11_rtx_costs): Add case for LSHIFTRT.
11012         (pdp11_output_jump): Add CCNZ mode conditional branches.
11013         (notice_update_cc_on_set): Remove.
11014         (pdp11_cc_mode): New function.
11015         (simple_memory_operand): Correct pre/post decrement case.
11016         (no_side_effect_operand): New function.
11017         (pdp11_regno_reg_class): Add CC_REGS class.
11018         (pdp11_fixed_cc_regs): New function.
11019         (pdp11_small_shift): New function.
11020         (pdp11_expand_shift): New function to expand shift insns.
11021         (pdp11_assemble_shift): New function to output shifts.
11022         (pdp11_branch_cost): Remove.
11023         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
11024         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
11025         (WCHAR_TYPE): Ditto.
11026         (PTRDIFF_TYPE): Ditto.
11027         (ADJUST_INSN_LENGTH): New macro.
11028         (FIXED_REGISTERS): Add CC registers.
11029         (CALL_USED_REGISTERS): Ditto.
11030         (reg_class): Ditto.
11031         (REG_CLASS_NAMES): Ditto.
11032         (REG_CLASS_CONTENTS): Ditto.
11033         (SELECT_CC_MODE): Use new function.
11034         (TARGET_FLAGS_REGNUM): New macro.
11035         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
11036         (cc0_reg_rtx): Remove.
11037         (CC_STATUS_MDEP): Remove.
11038         (CC_STATUS_MDEFP_INIT): Remove.
11039         (CC_IN_FPU): Remove.
11040         (NOTICE_UPDATE_CC): Remove.
11041         (REGISTER_NAMES): Add CC registers.
11042         (BRANCH_COST): Change to constant 1.
11043         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
11044         handling.
11045         * config/pdp11/pdp11.opt (mbcopy): Remove.
11046         (mbcopy-builtin): Remove.
11047         (mbranch-cheap): Remove.
11048         (mbranch-expensive): Remove.
11049         * config/pdp11/predicates.md (expand_shift_operand): Update to
11050         match shift code generation.
11051         (ccnz_operator): New predicate.
11052         * doc/invoke.texi (PDP-11 Options): Remove deleted options
11053         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
11054         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
11055         exclusive options.
11056         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
11057         description of O constraint.
11059 2018-06-27  Jeff Law  <law@redhat.com>
11060             Austin Law  <austinklaw@gmail.com>
11062         * config/v850/v850.md (addsi3_set_flags): New pattern.
11063         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
11064         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
11065         (zero_extendhisi2_v850_set_flags): Likewise.
11066         (zero_extendqisi2_v850_set_flags): Likewise.
11067         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
11068         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
11069         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
11071         * config/v850/v850-protos.h (notice_update_cc): Remove.
11072         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
11073         (v850_print_operand): Handle 'D' and "d".
11074         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
11075         Add handling of arithmetic/logical operations compared against zero.
11076         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
11077         Do not look at v850_compare_op, instead get mode from last argument.
11078         (v850_gen_compare): Remove
11079         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
11080         after reload for prologue insns.
11081         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
11082         patterns.
11083         (construct_save_jarl): Likewise.
11084         (TARGET_FLAGS_REGNUM): Define.
11085         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
11086         (NOTICE_UPDATE_CC): Remove.
11087         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
11088         than cc0.  Conditionalize on reload_completed.
11089         (cmpsi_insn, setfcc_insn): Likewise.
11090         (tst1 splitter): Turn into define_and_split which sets the flags
11091         after reload.
11092         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
11093         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
11094         (cstoresf4, cstoredf4): Clobber the flags.
11095         (cmpsi, cmpsf, cmpdf): Remove expanders.
11096         (setf_insn): Remove pattern.
11097         (addsi3): Turn into define_and_split which clobbers the flags after
11098         reload and a suitable pattern (addsi3_clobber_flags) for use after
11099         reload.
11100         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
11101         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
11102         (ashrsi3, ashrsi3_v850e2): Likewise.
11103         (bins): Clobber the flags.
11104         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
11105         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
11106         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
11107         (call_value_internal_short, call_value_internal_long): Likewise.
11108         (callt_save_interrupt, callt_return_interrupt): Likewise.
11109         (save_interrupt, return_interrupt): Likewise.
11110         (callt_save_all_interrupt, save_all_interrupt): Likewise.
11111         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
11112         (restore_all_interrupt, _restore_all_interrupt): Likewise.
11113         (All FP comparisons): Only allow after reload has completed.
11114         (trfsr): Likewise.
11115         (divh, divhu): Tweak output template.
11116         (branch_z_normal, branch_z_invert): Remove
11117         (branch_nz_normal, branch_nz_invert): Likewise.
11118         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
11120         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
11121         * config/v850/v850.c (notice_update_cc): Remove.
11122         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
11123         (CC_NO_CARRY): Likewise.
11124         (NOTICE_UPDATE_CC): Define to nothing.
11125         * config/v850/v850.md: Remove block comment on cc0 handling
11126         Remove "cc" attribute from all patterns.  Remove cc_status handling
11127         from all patterns.  Minor formatting fixes.
11129 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11131         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
11132         (cortex-a76.cortex-a55): Likewise.
11133         * config/aarch64/aarch64-tune.md: Regenerate.
11134         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
11135         cortex-a76.cortex-a55.
11137 2018-06-27  Jeff Law  <law@redhat.com>
11139         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
11140         (MULTILIB_DIRNAMES): Similarly.
11142 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
11144         * gimple.h (gimple_return_retbnd): Delete.
11145         (gimple_return_set_retbnd): Likewise.
11146         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11147         gimple_return_set_retbnd.
11148         * gimple-pretty-print.c (dump_gimple_return): Remove call to
11149         gimple_return_retbnd and adjust.
11150         * tree-inline.h (struct copy_body_data): Remove retbnd field.
11151         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
11152         Explicitly return NULL in a couple more cases.  Move assertion
11153         on debug statements and remove unreachable code.
11154         (reset_debug_binding): Do not test id->retbnd.
11155         (expand_call_inline): Do not set it.
11157 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11159         * configure.ac: Add --disable-gcov option.
11160         * configure: Regenerate.
11161         * Makefile.in: Honour @enable_gcov@.
11162         * doc/install.texi: Document --disable-gcov.
11164 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11166         * config/arm/arm-cpus.in (cortex-a76): New entry.
11167         (cortex-a76.cortex-a55): Likewise.
11168         * config/arm/arm-tables.opt: Regenerate.
11169         * config/arm/arm-tune.md: Likewise.
11170         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
11171         * doc/invoke.texi (ARM Options): Document cortex-a76 and
11172         cortex-a76.cortex-a55.
11174 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
11176         PR target/85769
11177         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
11179 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11181         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
11182         comment.
11183         (EPILOGUE_USES): Likewise.
11185 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
11187         * tree-inline.c (remap_location): New function extracted from...
11188         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
11189         (copy_phis_for_bb): ...here.  Call remap_location.
11190         (copy_cfg_body): Adjust call to copy_edges_for_bb.
11192 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
11194         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
11195         unaligned vsx for 16B memset.
11197 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
11199         PR target/86285
11200         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
11201         ieee128_float_type_node to long_double_type_node unless
11202         TARGET_LONG_DOUBLE_128 is set.
11204 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
11206         * cfgloop.c (get_loop_location): Convert return type from
11207         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
11208         by implicit construction from rtx_insn *, and using
11209         dump_user_location_t::from_function_decl for the fallback case.
11210         * cfgloop.h (get_loop_location): Convert return type from
11211         location_t to dump_user_location_t.
11212         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
11213         dump_printf_loc to pass in a dump_location_t rather than a
11214         location_t, via the gimple stmt.
11215         * coverage.c (get_coverage_counts): Update calls to
11216         dump_printf_loc to pass in dump_location_t rather than a
11217         location_t.
11218         * doc/optinfo.texi (Dump types): Convert example of
11219         dump_printf_loc from taking "locus" to taking "insn".  Update
11220         description of the "_loc" calls to cover dump_location_t.
11221         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
11222         "selftest.h".
11223         (dump_user_location_t::dump_user_location_t): New constructors,
11224         from gimple *stmt and rtx_insn *.
11225         (dump_user_location_t::from_function_decl): New function.
11226         (dump_loc): Make static.
11227         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
11228         const dump_location_t &.
11229         (dump_generic_expr_loc): Delete.
11230         (dump_printf_loc): Convert param "loc" from location_t to
11231         const dump_location_t &.
11232         (selftest::test_impl_location): New function.
11233         (selftest::dumpfile_c_tests): New function.
11234         * dumpfile.h: Include "profile-count.h".
11235         (class dump_user_location_t): New class.
11236         (struct dump_impl_location_t): New struct.
11237         (class dump_location_t): New class.
11238         (dump_printf_loc): Convert 2nd param from source_location to
11239         const dump_location_t &.
11240         (dump_generic_expr_loc): Delete.
11241         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11242         const dump_location_t &.
11243         * gimple-fold.c (fold_gimple_assign): Update call to
11244         dump_printf_loc to pass in a dump_location_t rather than a
11245         location_t, via the gimple stmt.
11246         (gimple_fold_call): Likewise.
11247         * gimple-loop-interchange.cc
11248         (loop_cand::analyze_iloop_reduction_var): Update for change to
11249         check_reduction_path.
11250         (tree_loop_interchange::interchange): Update for change to
11251         find_loop_location.
11252         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11253         change in return-type of find_loop_location.
11254         (graphite_regenerate_ast_isl): Likewise.
11255         * graphite-optimize-isl.c (optimize_isl): Likewise.
11256         * graphite.c (graphite_transform_loops): Likewise.
11257         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11258         pass in a dump_location_t rather than a location_t, via the
11259         gimple stmt.
11260         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11261         * ipa.c (walk_polymorphic_call_targets): Likewise.
11262         * loop-unroll.c (report_unroll): Convert "locus" param from
11263         location_t to dump_location_t.
11264         (decide_unrolling): Update for change to get_loop_location's
11265         return type.
11266         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11267         location_t to dump_user_location_t.
11268         (grid_find_single_omp_among_assignments_1): Updates calls to
11269         dump_printf_loc to pass in a dump_location_t rather than a
11270         location_t, via the gimple stmt.
11271         (grid_parallel_clauses_gridifiable): Convert "tloc" from
11272         location_t to dump_location_t.  Updates calls to dump_printf_loc
11273         to pass in a dump_location_t rather than a location_t, via the
11274         gimple stmt.
11275         (grid_inner_loop_gridifiable_p): Likewise.
11276         (grid_dist_follows_simple_pattern): Likewise.
11277         (grid_gfor_follows_tiling_pattern): Likewise.
11278         (grid_target_follows_gridifiable_pattern): Likewise.
11279         (grid_attempt_target_gridification): Convert initialization
11280         of local "grid" from memset to zero-initialization; FIXME: does
11281         this require C++11?  Update call to dump_printf_loc to pass in a
11282         optinfo_location rather than a location_t, via the gimple stmt.
11283         * profile.c (read_profile_edge_counts): Updates call to
11284         dump_printf_loc to pass in a dump_location_t rather than a
11285         location_t
11286         (compute_branch_probabilities): Likewise.
11287         * selftest-run-tests.c (selftest::run_tests): Call
11288         dumpfile_c_tests.
11289         * selftest.h (dumpfile_c_tests): New decl.
11290         * tree-loop-distribution.c (pass_loop_distribution::execute):
11291         Update for change in return type of find_loop_location.
11292         * tree-parloops.c (parallelize_loops): Likewise.
11293         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11294         "locus" from location_t to dump_user_location_t.
11295         (canonicalize_loop_induction_variables): Likewise.
11296         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11297         for change in return type of find_loop_location.
11298         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11299         to dump_printf_loc to pass in a dump_location_t rather than a
11300         location_t, via the stmt.
11301         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11302         Likewise.
11303         * tree-vect-loop-manip.c (find_loop_location): Convert return
11304         type from source_location to dump_user_location_t.
11305         (vect_do_peeling): Update for above change.
11306         (vect_loop_versioning): Update for change in type of
11307         vect_location.
11308         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11309         from location_t to dump_user_location_t.
11310         (vect_estimate_min_profitable_iters): Update for change in type
11311         of vect_location.
11312         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11313         location_t to dump_location_t.
11314         (vect_slp_bb): Update for change in type of vect_location.
11315         * tree-vectorizer.c (vect_location): Convert from source_location
11316         to dump_user_location_t.
11317         (try_vectorize_loop_1): Update for change in vect_location's type.
11318         (vectorize_loops): Likewise.
11319         (increase_alignment): Likewise.
11320         * tree-vectorizer.h (vect_location): Convert from source_location
11321         to dump_user_location_t.
11322         (find_loop_location): Convert return type from source_location to
11323         dump_user_location_t.
11324         (check_reduction_path): Convert 1st param from location_t to
11325         dump_user_location_t.
11326         * value-prof.c (check_counter): Update call to dump_printf_loc to
11327         pass in a dump_user_location_t rather than a location_t; update
11328         call to error_at for change in type of "locus".
11329         (check_ic_target): Update call to dump_printf_loc to
11330         pass in a dump_user_location_t rather than a location_t, via the
11331         call_stmt.
11333 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11335         * config/s390/s390.h (enum processor_flags): Do not use
11336         default tune parameter when -march was specified.
11338 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
11340         PR target/86314
11341         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
11342         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
11343         operands.
11345 2018-06-26  Richard Biener  <rguenther@suse.de>
11347         PR tree-optimization/86287
11348         PR bootstrap/86316
11349         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
11350         (vect_analyze_loop): Initialize n_stmts.
11352 2018-06-26  Richard Biener  <rguenther@suse.de>
11354         PR middle-end/86271
11355         * fold-const.c (fold_convertible_p): Pointer extension
11356         isn't valid.
11358 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
11360         PR debug/86064
11361         * dwarf2out.c (loc_list_has_views): Adjust comments.
11362         (dw_loc_list): Split single cross-partition range with
11363         nonzero locview.
11365 2018-06-25  Jeff Law  <law@redhat.com>
11367         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
11368         on -mbig-switch by default.
11370         * config/v850/predicates.md (const_float_1_operand): Fix match_code
11371         test.
11372         (const_float_0_operand): Remove unused predicate.
11373         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
11374         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
11375         (recipsf2): New expander.  Original pattern now called
11376         (recipsf2_insn).
11377         (recipdf2, recipdf2_insn): Similarly.
11378         (rsqrtsf2, rsqrtsf2_insn): Similarly
11379         (rsqrtdf2, rsqrtdf2_insn): Similarly
11381 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
11383         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
11384         Simplify logic for FreeBSD (twice).
11386 2018-06-25  Martin Sebor  <msebor@redhat.com>
11388         PR tree-optimization/86204
11389         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
11390         a strnlen result if it's less than the length of the string.
11392 2018-06-25  Martin Sebor  <msebor@redhat.com>
11394         PR tree-optimization/85700
11395         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
11396         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
11397         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
11399 2018-06-25  Martin Sebor  <msebor@redhat.com>
11401         * doc/extend.texi (Zero-length arrays): Update and clarify.
11403 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
11405         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
11406         added IEEE/IBM long double multilib support on PowerPC little
11407         endian Linux systems.
11408         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11409         (MULTILIB_DEFAULTS): Likewise.
11410         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11411         Likewise.
11412         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
11413         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
11414         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
11416 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
11418         PR middle-end/86311
11419         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
11420         (REORDER_45): Likewise.
11422 2018-06-25  Jeff Law  <law@redhat.com>
11424         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
11425         dividend to 32 bits.  Adjust length.
11426         (udivmodhi4): Cleanup output template.  Fix length.
11428 2018-06-25  Carl Love  <cel@us.ibm.com>
11430         * config/rs6000/vsx.md: Change word selector to prefered location.
11432 2018-06-25  Richard Biener  <rguenther@suse.de>
11434         PR tree-optimization/86304
11435         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
11436         epilogue-if-converted loops as well.
11438 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
11440         * lto-section-out.c (lto_begin_section): Do not print section
11441         name for noaddr and unnumbered dumps.
11443 2018-06-25  Richard Biener  <rguenther@suse.de>
11445         * tree-vectorizer.h (struct vec_info_shared): New structure
11446         with parts split out from struct vec_info and loop_nest from
11447         struct _loop_vec_info.
11448         (struct vec_info): Adjust accordingly.
11449         (struct _loop_vec_info): Likewise.
11450         (LOOP_VINFO_LOOP_NEST): Adjust.
11451         (LOOP_VINFO_DATAREFS): Likewise.
11452         (LOOP_VINFO_DDRS): Likewise.
11453         (struct _bb_vec_info): Likewise.
11454         (BB_VINFO_DATAREFS): Likewise.
11455         (BB_VINFO_DDRS): Likewise.
11456         (struct _stmt_vec_info): Add dr_aux member.
11457         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
11458         (DR_MISALIGNMENT_UNINITIALIZED): New.
11459         (set_dr_misalignment): Adjust.
11460         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
11461         (vect_analyze_loop): Adjust prototype.
11462         (vect_analyze_loop_form): Likewise.
11463         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11464         Compute dependences lazily.
11465         (vect_record_base_alignments): Use shared datarefs/ddrs.
11466         (vect_verify_datarefs_alignment): Likewise.
11467         (vect_analyze_data_refs_alignment): Likewise.
11468         (vect_analyze_data_ref_accesses): Likewise.
11469         (vect_analyze_data_refs): Likewise.
11470         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
11471         constructor parameter for shared part.
11472         (vect_analyze_loop_form): Pass in shared part and adjust.
11473         (vect_analyze_loop_2): Pass in storage for the number of
11474         stmts.  Move loop nest finding to the caller.  Compute
11475         datarefs lazily.
11476         (vect_analyze_loop): Pass in shared part.
11477         (vect_transform_loop): Verify shared datarefs are unchanged.
11478         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
11479         constructor parameter for shared part.
11480         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
11481         (vect_slp_bb): Verify shared datarefs are unchanged before
11482         transform.
11483         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
11484         change.
11485         (new_stmt_vec_info): Initialize DR_AUX misalignment to
11486         DR_MISALIGNMENT_UNINITIALIZED.
11487         * tree-vectorizer.c (vec_info::vec_info): Add constructor
11488         parameter for shared part.
11489         (vec_info::~vec_info): Adjust.
11490         (vec_info_shared::vec_info_shared): New.
11491         (vec_info_shared::~vec_info_shared): Likewise.
11492         (vec_info_shared::save_datarefs): Likewise.
11493         (vec_info_shared::check_datarefs): Likewise.
11494         (try_vectorize_loop_1): Construct shared part live for analyses
11495         of a single loop for multiple vector sizes.
11496         * tree-parloops.c (gather_scalar_reductions): Adjust.
11498 2018-06-25  Richard Biener  <rguenther@suse.de>
11500         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
11501         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
11502         (vect_analyze_data_refs): Remove similar code from here and
11503         simplify accordingly.
11505 2018-06-25  Richard Biener  <rguenther@suse.de>
11507         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
11508         for reverse storage order accesses rather than asserting
11509         they cannot happen here.
11511 2018-06-25  Tom de Vries  <tdevries@suse.de>
11513         PR debug/86257
11514         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
11515         Use data16 instead of .byte for insn prefix.
11517 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
11519         PR C++/86082
11520         * parser.c (make_char_string_pack): Pass this literal chars
11521         through cpp_interpret_string.
11522         (cp_parser_userdef_numeric_literal): Check the result of
11523         make_char_string_pack.
11525 2018-06-24  Maya Rashish  <coypu@sdf.org>
11527         * ginclude/stddef.h: Simplify conditions around avoiding
11528         re-definition of __size_t.
11530 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11532         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
11533         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
11535 2018-06-22  Maya Rashish  <coypu@sdf.org>
11537         * doc/invoke.texi (mno-fancy-math-387): Update for changes
11538         made to OpenBSD and NetBSD through the years.
11540 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11542         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11543         behavior of vec_pack (vector double, vector double) to match
11544         behavior of vec_float2 (vector double, vector double).
11546 2018-06-22  Olivier Hainque  <hainque@adacore.com>
11548         * gimplify.c (gimplify_function_tree): Prevent creation
11549         of a trampoline for the address of the current function
11550         passed to entry/exit instrumentation hooks.
11552 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
11554         PR target/86222
11555         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
11556         correctly.
11558 2018-06-22  Martin Liska  <mliska@suse.cz>
11560         PR tree-optimization/86263
11561         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11562         Bail out if is_enabled is false.
11563         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
11564         New declaration.
11565         (jump_table_cluster::is_enabled): New function.
11567 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11569         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11570         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
11571         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
11572         (lto_input_ts_binfo_tree_pointers): Likewise.
11573         * tree-streamer-out.c (streamer_write_tree_bitfields,
11574         write_ts_binfo_tree_pointers): Likewise.
11575         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
11577 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11579         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
11581 2018-06-22  Martin Liska  <mliska@suse.cz>
11583         * symbol-summary.h (get): Make it pure and inline move
11584         functionality from ::get function.
11585         (get): Remove and inline into ::get and ::get_create.
11586         (get_create): Move code from ::get function.
11588 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11590         PR target/85994
11591         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
11592         -x assembler-with-cpp.
11594 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11596         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
11597         _FILE_OFFSET_BITS=64 for C++.
11599 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
11601         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
11602         conversion insn that shows up when pr85657-3.c is compiled using
11603         IEEE 128-bit long double.
11604         (neg<mode>2_internal): Use the correct mode to check whether the
11605         mode is IBM extended.
11606         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
11607         multiply and divide external functions from being created more
11608         than once.
11610 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11612         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
11613         functions.
11614         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
11615         the edge can be forwarded.
11616         (cfg_layout_merge_blocks): Likewise.
11618 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11620         * except.c (finish_eh_generation): Commit edge insertions only after
11621         the EH edges have been redirected from post-landing to landing pads.
11623 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11625         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
11626         create_tmp_var_for to create the FRAME decl.
11627         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
11629 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11631         * tree-inline.c (copy_edges_for_bb): Minor tweak.
11632         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
11633         debug statement when resetting its value.
11634         (expand_call_inline): Copy the locus of the call onto the assignment
11635         of the return value, if any.  Use local variable in more cases.
11637 2018-06-21  Martin Liska  <mliska@suse.cz>
11639         * ipa-pure-const.c (propagate_nothrow): Use
11640         funct_state_summaries->get.
11641         (dump_malloc_lattice): Likewise.
11642         (propagate_malloc): Likewise.
11644 2018-06-21  Richard Biener  <rguenther@suse.de>
11646         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
11647         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
11648         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
11649         comment.
11650         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
11651         BLOCK_ABSTRACT_ORIGIN unconditionally.
11653 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
11655         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
11656         deleting it.
11657         * ipa-reference.c (ipa_reference_c_finalize): Delete
11658         ipa_ref_opt_sum_summaries and set it to NULL.
11660 2018-06-21  Tom de Vries  <tdevries@suse.de>
11662         PR tree-optimization/85859
11663         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
11664         test with comment from bb_no_side_effects_p.
11666 2018-06-21  Richard Biener  <rguenther@suse.de>
11668         PR tree-optimization/86232
11669         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
11670         max for constant niter.
11672 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11674         * config/aarch64/aarch64-simd.md
11675         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
11677 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11679         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
11680         Make opernads of the unspec commutative.
11682 2018-06-21  Richard Biener  <rguenther@suse.de>
11684         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
11685         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
11686         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
11687         (vect_analyze_data_ref_dependence): Re-order checks to deal with
11688         NULL DR_STEP.
11689         (vect_record_base_alignments): Do not record base alignment
11690         for gathers or scatters.
11691         (vect_compute_data_ref_alignment): Drop return value that is always
11692         true.  Bail out early for gathers or scatters.
11693         (vect_enhance_data_refs_alignment): Bail out early for gathers
11694         or scatters.
11695         (vect_find_same_alignment_drs): Likewise.
11696         (vect_analyze_data_refs_alignment): Remove dead code.
11697         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11698         (vect_analyze_data_refs): For possible gathers or scatters do
11699         not create an alternate DR, just check their possible validity
11700         and mark them.  Adjust DECL_NONALIASED handling to not rely
11701         on DR_BASE_ADDRESS.
11702         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
11703         update inits of gathers or scatters.
11704         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
11705         Also copy gather/scatter flag to pattern vinfo.
11707 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11709         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11710         behavior of vec_packsu (vector unsigned long long, vector unsigned
11711         long long) to match behavior of vec_packs with same signature.
11713 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
11714             Thomas Schwinge <thomas@codesourcery.com>
11715             Cesar Philippidis  <cesar@codesourcery.com>
11717         * gimplify.c (gimplify_scan_omp_clauses): Add support for
11718         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11719         (gimplify_adjust_omp_clauses): Likewise.
11720         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
11721         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
11722         (gimplify_omp_target_update): Update handling of acc update and
11723         enter/exit data.
11724         * omp-low.c (install_var_field): Remove unused parameter
11725         base_pointers_restrict.
11726         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
11727         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
11728         FINALIZE}
11729         (omp_target_base_pointers_restrict_p): Delete.
11730         (scan_omp_target): Update call to scan_sharing_clauses.
11731         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
11732         FINALIZE}.
11733         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
11734         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11735         (convert_local_omp_clauses): Likewise.
11736         * tree-pretty-print.c (dump_omp_clause): Likewise.
11737         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
11738         FINALIZE}.
11739         (omp_clause_code_name): Likewise.
11741 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
11743         PR debug/86194
11744         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
11745         be narrowed.
11747         PR tree-optimization/86231
11748         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
11749         anti-range don't overwrite *vr0min before using it to compute *vr0max.
11751 2018-06-20  Tom de Vries  <tdevries@suse.de>
11753         PR tree-optimization/86097
11754         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
11755         iv type if signedness of iv type is not the same as that of *nit.
11757 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
11759         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
11760         EDGE_EH edges, verify they are all EDGE_EH.
11762 2018-06-20  Maya Rashish  <coypu@sdf.org>
11764         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
11766 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11768         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
11769         * config/aarch64/aarch64.c (xgene1_tunings): Add
11770         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
11771         (aarch64_mode_valid_for_sched_fusion_p):
11772         Allow 16-byte modes.
11773         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
11774         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
11775         128-bit modes.
11776         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
11777         New pattern.
11778         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
11779         * config/aarch64/iterators.md (VQ2): New mode iterator.
11781 2018-06-20  Martin Liska  <mliska@suse.cz>
11783         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11784         Change default ratio from 10 to 8.
11786 2018-06-20  Martin Liska  <mliska@suse.cz>
11788         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11789         New.
11790         (bit_test_cluster::find_bit_tests): Likewise.
11791         (switch_decision_tree::analyze_switch_statement): Find clusters.
11792         * tree-switch-conversion.h (struct jump_table_cluster): Document
11793         hierarchy.
11795 2018-06-20  Martin Liska  <mliska@suse.cz>
11797         * tree-switch-conversion.c (switch_conversion::collect):
11798         Record m_uniq property.
11799         (switch_conversion::expand): Bail out for special conditions.
11800         (group_cluster::~group_cluster): New.
11801         (group_cluster::group_cluster): Likewise.
11802         (group_cluster::dump): Likewise.
11803         (jump_table_cluster::emit): New.
11804         (switch_decision_tree::fix_phi_operands_for_edges): New.
11805         (struct case_node): Remove struct.
11806         (jump_table_cluster::can_be_handled): New.
11807         (case_values_threshold): Moved to header.
11808         (reset_out_edges_aux): Likewise.
11809         (jump_table_cluster::is_beneficial): New.
11810         (bit_test_cluster::can_be_handled): Likewise.
11811         (add_case_node): Remove.
11812         (bit_test_cluster::is_beneficial): New.
11813         (case_bit_test::cmp): New.
11814         (bit_test_cluster::emit): New.
11815         (expand_switch_as_decision_tree_p): Remove.
11816         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
11817         (fix_phi_operands_for_edge): Likewise.
11818         (switch_decision_tree::analyze_switch_statement): New.
11819         (compute_cases_per_edge): Move ...
11820         (switch_decision_tree::compute_cases_per_edge): ... here.
11821         (try_switch_expansion): Likewise.
11822         (switch_decision_tree::try_switch_expansion): Likewise.
11823         (record_phi_operand_mapping): Likewise.
11824         (switch_decision_tree::record_phi_operand_mapping): Likewise.
11825         (emit_case_decision_tree): Likewise.
11826         (switch_decision_tree::emit): Likewise.
11827         (balance_case_nodes): Likewise.
11828         (switch_decision_tree::balance_case_nodes): Likewise.
11829         (dump_case_nodes): Likewise.
11830         (switch_decision_tree::dump_case_nodes): Likewise.
11831         (emit_jump): Likewise.
11832         (switch_decision_tree::emit_jump): Likewise.
11833         (emit_cmp_and_jump_insns): Likewise.
11834         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
11835         (emit_case_nodes): Likewise.
11836         (switch_decision_tree::emit_case_nodes): Likewise.
11837         (conditional_probability): Remove.
11838         * tree-switch-conversion.h (enum cluster_type): New.
11839         (PRINT_CASE): New.
11840         (struct cluster): Likewise.
11841         (cluster::cluster): Likewise.
11842         (struct simple_cluster): Likewise.
11843         (simple_cluster::simple_cluster): Likewise.
11844         (struct group_cluster): Likewise.
11845         (struct jump_table_cluster): Likewise.
11846         (struct bit_test_cluster): Likewise.
11847         (struct min_cluster_item): Likewise.
11848         (struct case_tree_node): Likewise.
11849         (case_tree_node::case_tree_node): Likewise.
11850         (jump_table_cluster::case_values_threshold): Likewise.
11851         (struct case_bit_test): Likewise.
11852         (struct switch_decision_tree): Likewise.
11853         (struct switch_conversion): Likewise.
11854         (switch_decision_tree::reset_out_edges_aux): Likewise.
11856 2018-06-20  Martin Liska  <mliska@suse.cz>
11858         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
11859         (hoist_edge_and_branch_if_true): Likewise.
11860         (expand_switch_using_bit_tests_p): Likewise.
11861         (struct case_bit_test): Likewise.
11862         (case_bit_test_cmp): Likewise.
11863         (emit_case_bit_tests): Likewise.
11864         (switch_conversion::switch_conversion): New class.
11865         (struct switch_conv_info): Remove old struct.
11866         (collect_switch_conv_info): More to ...
11867         (switch_conversion::collect): ... this.
11868         (check_range): Likewise.
11869         (switch_conversion::check_range): Likewise.
11870         (check_all_empty_except_final): Likewise.
11871         (switch_conversion::check_all_empty_except_final): Likewise.
11872         (check_final_bb): Likewise.
11873         (switch_conversion::check_final_bb): Likewise.
11874         (create_temp_arrays): Likewise.
11875         (switch_conversion::create_temp_arrays): Likewise.
11876         (free_temp_arrays): Likewise.
11877         (gather_default_values): Likewise.
11878         (switch_conversion::gather_default_values): Likewise.
11879         (build_constructors): Likewise.
11880         (switch_conversion::build_constructors): Likewise.
11881         (constructor_contains_same_values_p): Likewise.
11882         (switch_conversion::contains_same_values_p): Likewise.
11883         (array_value_type): Likewise.
11884         (switch_conversion::array_value_type): Likewise.
11885         (build_one_array): Likewise.
11886         (switch_conversion::build_one_array): Likewise.
11887         (build_arrays): Likewise.
11888         (switch_conversion::build_arrays): Likewise.
11889         (gen_def_assigns): Likewise.
11890         (switch_conversion::gen_def_assigns): Likewise.
11891         (prune_bbs): Likewise.
11892         (switch_conversion::prune_bbs): Likewise.
11893         (fix_phi_nodes): Likewise.
11894         (switch_conversion::fix_phi_nodes): Likewise.
11895         (gen_inbound_check): Likewise.
11896         (switch_conversion::gen_inbound_check): Likewise.
11897         (process_switch): Use the newly created class.
11898         (switch_conversion::expand): New.
11899         (switch_conversion::~switch_conversion): New.
11900         * tree-switch-conversion.h: New file.
11902 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11904         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
11905         tree-vect-patterns.c.
11906         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
11907         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
11908         (vect_recog_sad_pattern): Likewise.
11909         (vect_recog_widen_sum_pattern): Likewise.
11910         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
11911         (vect_recog_widen_shift_pattern): Remove the type_in argument.
11912         (vect_recog_rotate_pattern): Likewise.
11913         (vect_recog_mult_pattern): Likewise.
11914         (vect_recog_vector_vector_shift_pattern): Likewise.
11915         (vect_recog_divmod_pattern): Likewise.
11916         (vect_recog_mixed_size_cond_pattern): Likewise.
11917         (vect_recog_bool_pattern): Likewise.
11918         (vect_recog_mask_conversion_pattern): Likewise.
11919         (vect_try_gather_scatter_pattern): Likewise.
11920         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
11921         (vect_recog_over_widening_pattern): Likewise.
11922         (vect_recog_gather_scatter_pattern): Likewise.
11923         (vect_recog_func_ptr): Move from tree-vectorizer.h
11924         (vect_vect_recog_func_ptrs): Move further down the file.
11925         (vect_recog_func): Likewise.  Remove the third argument.
11926         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
11927         (vect_pattern_recog_1): Expect the pattern function to do any
11928         necessary target tests.  Also expect it to provide a vector type.
11929         Remove the type_in handling.
11931 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11933         * tree-vect-patterns.c (vect_pattern_detected): New function.
11934         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
11935         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
11936         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
11937         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
11938         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11939         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11940         (vect_recog_mask_conversion_pattern)
11941         (vect_try_gather_scatter_pattern): Likewise.
11943 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11945         * tree-vect-patterns.c (vect_get_internal_def): New function.
11946         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11947         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
11948         (search_type_for_mask_1): Use it.
11950 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11952         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11953         redundant WIDEN_SUM_EXPR handling.
11954         (vect_recog_sad_pattern): Likewise.
11956 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11958         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11959         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
11960         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
11961         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
11962         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
11964 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11966         * tree-vect-stmts.c (vectorizable_call): Make sure that we
11967         use the stmt_vec_info of the original bb statement for the
11968         new zero assignment, even if the call is part of a pattern.
11970 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11972         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
11973         that the sequence is attached to the original statement rather
11974         than the pattern statement.
11975         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
11976         PATTERN_DEF_SEQ from the original statement rather than
11977         the main pattern statement.
11978         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
11979         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
11980         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
11982 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11984         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
11985         definition statements before the early exit for statements that aren't
11986         live or relevant.
11987         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
11988         split out from...
11989         (vect_transform_loop): ...here.  Process pattern definition
11990         statements without first checking whether the main pattern
11991         statement is live or relevant.
11993 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11995         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
11996         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
11998 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
12000         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
12001         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
12002         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
12003         (expand_block_compare): Change select_block_compare_mode call.
12004         (expand_strncmp_align_check): Use new functions, fix comment.
12005         (emit_final_str_compare_gpr): New function.
12006         (expand_strn_compare): Refactor and clean up code.
12007         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
12009 2018-06-19  Tony Reix  <tony.reix@atos.com>
12010             Damien Bergamini  <damien.bergamini@atos.com>
12011             David Edelsohn  <dje.gcc@gmail.com>
12013         * collect2.c (static_obj): New variable.
12014         (static_libs): New variable.
12015         (is_in_list): Uncomment declaration.
12016         (main): Track AIX libraries linked statically.
12017         (is_in_list): Uncomment definition.
12018         (scan_prog_file): Don't add AIX shared libraries initializer
12019         to constructor list if linking statically.
12021 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
12023         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
12024         constant.
12025         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
12027 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
12029         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
12030         blocks.
12032 2018-06-19  Martin Liska  <mliska@suse.cz>
12034         * config/i386/i386.c (ix86_can_inline_p): Do not use
12035         ipa_fn_summaries::get_create.
12036         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
12037         get.
12038         (devirtualization_time_bonus): Likewise.
12039         (ipcp_propagate_stage): Likewise.
12040         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12041         (edge_set_predicate): Likewise.
12042         (evaluate_conditions_for_known_args): Likewise.
12043         (evaluate_properties_for_edge): Likewise.
12044         (ipa_call_summary::reset): Tranform to ...
12045         (ipa_call_summary::~ipa_call_summary): ... this.
12046         (ipa_fn_summary::reset): Transform to ...
12047         (ipa_fn_summary::~ipa_fn_summary): ... this.
12048         (ipa_fn_summary_t::remove): Rename to ...
12049         (ipa_fn_summary_t::remove_callees): ... this.
12050         (ipa_fn_summary_t::duplicate): Use placement new
12051         instead of memory copy.
12052         (ipa_call_summary_t::duplicate): Likewise.
12053         (ipa_call_summary_t::remove): Remove.
12054         (dump_ipa_call_summary): Change get_create to get.
12055         (ipa_dump_fn_summary): Dump only when summary exists.
12056         (analyze_function_body): Use symbol_summary::get instead
12057         of get_create.
12058         (compute_fn_summary): Likewise.
12059         (estimate_edge_devirt_benefit): Likewise.
12060         (estimate_edge_size_and_time): Likewise.
12061         (inline_update_callee_summaries): Likewise.
12062         (remap_edge_change_prob): Likewise.
12063         (remap_edge_summaries): Likewise.
12064         (ipa_merge_fn_summary_after_inlining): Likewise.
12065         (write_ipa_call_summary): Likewise.
12066         (ipa_fn_summary_write): Likewise.
12067         (ipa_free_fn_summary): Likewise.
12068         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
12069         (struct ipa_call_summary): Likewise.
12070         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
12071         of get_create.
12072         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
12073         (estimate_size_after_inlining): Likewise.
12074         (estimate_growth): Likewise.
12075         (growth_likely_positive): Likewise.
12076         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12077         (inline_call): Likewise.
12078         * ipa-inline.c (caller_growth_limits): Likewise.
12079         (can_inline_edge_p): Likewise.
12080         (can_inline_edge_by_limits_p): Likewise.
12081         (compute_uninlined_call_time): Likewise.
12082         (compute_inlined_call_time): Likewise.
12083         (want_inline_small_function_p): Likewise.
12084         (edge_badness): Likewise.
12085         (update_caller_keys): Likewise.
12086         (update_callee_keys): Likewise.
12087         (inline_small_functions): Likewise.
12088         (inline_to_all_callers_1): Likewise.
12089         (dump_overall_stats): Likewise.
12090         (early_inline_small_functions): Likewise.
12091         (early_inliner): Likewise.
12092         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12093         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12094         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12095         * ipa-split.c (execute_split_functions): Likewise.
12096         * symbol-summary.h: Likewise.
12097         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12099 2018-06-19  Richard Biener  <rguenther@suse.de>
12101         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
12102         (vectorize_loops): ... here.  Fix dbgcnt handling.
12103         (try_vectorize_loop): Wrap try_vectorize_loop_1.
12105 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
12107         PR target/86197
12108         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
12109         ieee128 argument takes up only one (vector) register, not two (floating
12110         point) registers.
12112 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12114         * gimplify.c (gimplify_init_constructor): Really never clear for an
12115         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
12117 2018-06-19  Richard Biener  <rguenther@suse.de>
12119         PR tree-optimization/86179
12120         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
12121         after failed recognition.
12123 2018-06-18  Martin Sebor  <msebor@redhat.com>
12125         PR middle-end/85602
12126         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
12127         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
12128         Handle integer subtraction.
12129         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
12130         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
12132 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12134         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
12135         param from rtx to rtx_insn *.
12136         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
12137         param.
12138         (frv_ifcvt_modify_insn): Likwise.
12139         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
12140         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
12141         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
12142         as_a <rtx_insn *> cast to local "unprotected_region" once
12143         it's been established that it's not NULL or pc_rtx.
12144         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
12145         param "sethi" from rtx to rtx_insn *.
12146         (nds32_group_float_insns): Likewise for param "insn".
12147         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
12148         param.
12149         (vax_output_int_subtract): Likewise.
12150         * config/vax/vax.c (vax_output_int_add): Likewise for param
12151         "insn".
12152         (vax_output_int_subtract): Likewise.
12153         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
12154         (emit_pattern_after): Likewise for param "after".
12155         (emit_insn_after): Likewise.
12156         (emit_jump_insn_after): Likewise.
12157         (emit_call_insn_after): Likewise.
12158         (emit_debug_insn_after): Likewise.
12159         (emit_pattern_before): Likewise for param "before".
12160         (emit_insn_before): Likewise.
12161         (emit_jump_insn_before): Likewise.
12162         * final.c (get_insn_template): Likewise for param "insn", removing
12163         a cast.
12164         * output.h (get_insn_template): Likewise for 2nd param.
12165         * rtl.h (emit_insn_before): Likewise.
12166         (emit_jump_insn_before): Likewise.
12167         (emit_debug_insn_before_noloc): Likewise.
12168         (emit_insn_after): Likewise.
12169         (emit_jump_insn_after): Likewise.
12170         (emit_call_insn_after): Likewise.
12171         (emit_debug_insn_after): Likewise.
12172         (set_insn_deleted): Likewise for param.
12174 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
12176         PR target/85358
12177         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
12178         floating point modes, so that IFmode is numerically greater than
12179         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
12180         to declare the ordering.  This prevents IFmode from being
12181         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
12182         machine.  Include rs6000-modes.h to share the fractional values
12183         between genmodes* and the rest of the compiler.
12184         (IFmode): Likewise.
12185         (KFmode): Likewise.
12186         (TFmode): Likewise.
12187         * config/rs6000/rs6000-modes.h: New file.
12188         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
12189         meaning of rs6000_long_double_size so that 126..128 selects an
12190         appropriate 128-bit floating point type.
12191         (rs6000_option_override_internal): Likewise.
12192         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
12193         (TARGET_LONG_DOUBLE_128): Change the meaning of
12194         rs6000_long_double_size so that 126..128 selects an appropriate
12195         128-bit floating point type.
12196         (LONG_DOUBLE_TYPE_SIZE): Update comment.
12197         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
12198         source and destination to match the standard usage.
12199         (truncifkf2): Likewise.
12200         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
12201         ISA 2.07 to use an explicit clobber, instead of passing in a
12202         temporary.
12203         (copysign<mode>3_soft): Likewise.
12205 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12207         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12208         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
12209         (vect_slp_analyze_instance_dependence): Likewise.
12210         (vect_enhance_data_refs_alignment): Likewise.
12211         (vect_analyze_data_refs_alignment): Likewise.
12212         (vect_slp_analyze_and_verify_instance_alignment
12213         (vect_analyze_data_ref_accesses): Likewise.
12214         (vect_prune_runtime_alias_test_list): Likewise.
12215         (vect_analyze_data_refs): Likewise.
12216         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
12217         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12218         (vect_analyze_scalar_cycles_1): Likewise.
12219         (vect_get_loop_niters): Likewise.
12220         (vect_analyze_loop_form_1): Likewise.
12221         (vect_update_vf_for_slp): Likewise.
12222         (vect_analyze_loop_operations): Likewise.
12223         (vect_analyze_loop): Likewise.
12224         (vectorizable_induction): Likewise.
12225         (vect_transform_loop): Likewise.
12226         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
12227         * tree-vect-slp.c (vect_analyze_slp): Likewise.
12228         (vect_make_slp_decision): Likewise.
12229         (vect_detect_hybrid_slp): Likewise.
12230         (vect_slp_analyze_operations): Likewise.
12231         (vect_slp_bb): Likewise.
12232         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
12233         (vectorizable_bswap): Likewise.
12234         (vectorizable_call): Likewise.
12235         (vectorizable_simd_clone_call): Likewise.
12236         (vectorizable_conversion): Likewise.
12237         (vectorizable_assignment): Likewise.
12238         (vectorizable_shift): Likewise.
12239         (vectorizable_operation): Likewise.
12240         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12242 2018-06-18  Martin Sebor  <msebor@redhat.com>
12244         PR tree-optimization/81384
12245         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12246         * builtins.c (expand_builtin_strnlen): New function.
12247         (expand_builtin): Call it.
12248         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12249         * builtins.def (BUILT_IN_STRNLEN): New.
12250         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12251         Warn for bounds in excess of maximum object size.
12252         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12253         single-value ranges.  Handle strnlen.
12254         (handle_builtin_strlen): Handle strnlen.
12255         (strlen_check_and_optimize_stmt): Same.
12256         * doc/extend.texi (Other Builtins): Document strnlen.
12258 2018-06-18  Maya Rashish  <coypu@sdf.org>
12260         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12261         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12262         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12264         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12265         here to ...
12266         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12268 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12270         * tree.c (escaped_string::escape): Replace cast to char * by
12271         const_cast<char *> (unescaped).
12273 2018-06-18  Nick Clifton  <nickc@redhat.com>
12275         PR 84195
12276         * tree.c (escaped_string): New class.  Converts an unescaped
12277         string into its escaped equivalent.
12278         (warn_deprecated_use): Use the new class to convert the
12279         deprecation message, if present.
12280         (test_escaped_strings): New self test.
12281         (test_c_tests): Add test_escaped_strings.
12282         * doc/extend.texi (deprecated): Add a note that the
12283         deprecation message is affected by the -fmessage-length
12284         option, and that control characters will be escaped.
12285         (#pragma GCC error): Document this pragma.
12286         (#pragma GCC warning): Likewise.
12287         * doc/invoke.texi (-fmessage-length): Document this option's
12288         effect on the #warning and #error preprocessor directives and
12289         the deprecated attribute.
12291 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
12293         * tree.c (decl_value_expr_lookup): Revert latest change.
12294         (decl_value_expr_insert): Likewise.
12296 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
12298         * gimplify.c (nonlocal_vlas): Delete.
12299         (nonlocal_vla_vars): Likewise.
12300         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12301         referenced VLAs.
12302         (gimplify_body): Do not create and destroy nonlocal_vlas.
12303         * tree-nested.c: Include diagnostic.h.
12304         (use_pointer_in_frame): Tweak.
12305         (lookup_field_for_decl): Add assertion and declare the transformation.
12306         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12307         internal error when the reference is in a wrong context.  Do not
12308         create a debug decl by default.
12309         (note_nonlocal_block_vlas): Delete.
12310         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12311         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
12312         create a debug decl by default.
12313         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12314         call is in a wrong context.
12315         (fixup_vla_decls): New function.
12316         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
12317         debug variables were created.
12318         * tree.c (decl_value_expr_lookup): Add checking assertion.
12319         (decl_value_expr_insert): Likewise.
12321 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12323         PR middle-end/82479
12324         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12325         * tree-scalar-evolution.c (interpret_expr): Likewise.
12326         (expression_expensive_p): Likewise.
12327         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12328         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12329         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12330         (ssa_defined_by_minus_one_stmt_p): New.
12332 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12334         PR middle-end/64946
12335         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12336         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12337         * dojump.c (do_jump): Likewise.
12338         * expr.c (expand_expr_real_2): Check operand type's sign.
12339         * fold-const.c (const_unop): Handle ABSU_EXPR.
12340         (fold_abs_const): Likewise.
12341         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
12342         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12343         (strip_sign_op_1): Likesise.
12344         * match.pd: Add new pattern to generate ABSU_EXPR.
12345         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
12346         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12347         * tree-eh.c (operation_could_trap_helper_p): Likewise.
12348         * tree-inline.c (estimate_operator_cost): Likewise.
12349         * tree-pretty-print.c (dump_generic_node): Likewise.
12350         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
12351         * tree.def (ABSU_EXPR): New.
12353 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
12355         PR middle-end/86095
12356         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
12357         documented as preserved for backward compatibility only.
12358         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
12360         PR rtl-optimization/86108
12361         * bb-reorder.c (create_forwarder_block): Renamed to ...
12362         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
12363         jump from new landing pad to the second part.
12364         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
12365         Adjust callers.
12367 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
12369         PR middle-end/85878
12370         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
12371         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
12372         Only call store_expr for halves if the mode is the same.
12374         PR middle-end/86123
12375         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
12376         Fix up comment formatting.
12378 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12380         * typed-splay-tree.h (typed_splay_tree::remove): New function.
12381         (typed_splay_tree::closure,
12382         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
12383         (typed_splay_tree::typed_splay_tree,
12384         typed_splay_tree::operator =): Declared private.
12385         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
12386         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
12387         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
12388         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
12389         typed_splay_tree::splay_tree_splay,
12390         typed_splay_tree::splay_tree_foreach_helper,
12391         typed_splay_tree::splay_tree_insert,
12392         typed_splay_tree::splay_tree_remove,
12393         typed_splay_tree::splay_tree_lookup,
12394         typed_splay_tree::splay_tree_predecessor,
12395         typed_splay_tree::splay_tree_successor,
12396         typed_splay_tree::splay_tree_min,
12397         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
12398         (typed_splay_tree::root, typed_splay_tree::comp,
12399         typed_splay_tree::delete_key,
12400         typed_splay_tree::delete_value): New data members.
12401         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
12402         typed_splay_tree::remove.
12404 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
12406         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
12407         -mginv and -mno-ginv to the assembler.
12408         * config/mips/mips.opt (-mcrc): New option.
12409         (-mginv): Likewise.
12410         * doc/invoke.text (-mcrc): Document.
12411         (-mginv): Likewise.
12413 2018-06-15  Nick Clifton  <nickc@redhat.com>
12415         PR 84195
12416         * tree.c (escaped_string): New class.  Converts an unescaped
12417         string into its escaped equivalent.
12418         (warn_deprecated_use): Use the new class to convert the
12419         deprecation message, if present.
12420         (test_escaped_strings): New self test.
12421         (test_c_tests): Add test_escaped_strings.
12422         * doc/extend.texi (deprecated): Add a note that the
12423         deprecation message is affected by the -fmessage-length
12424         option, and that control characters will be escaped.
12425         (#pragma GCC error): Document this pragma.
12426         (#pragma GCC warning): Likewise.
12427         * doc/invoke.texi (-fmessage-length): Document this option's
12428         effect on the #warning and #error preprocessor directives and
12429         the deprecated attribute.
12431 2018-06-15  Richard Biener  <rguenther@suse.de>
12433         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
12434         here, also noting vector size used.
12435         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
12436         size used in MSG_OPTIMIZED_LOCATIONS dump.
12437         (pass_slp_vectorize::execute): Adjust.
12439 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
12441         PR target/85968
12442         * config/arc/arc.c (arc_return_address_register): Fix
12443         if-condition.
12445 2018-06-15  Richard Biener  <rguenther@suse.de>
12447         PR middle-end/86159
12448         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
12449         leave useless conversion stripping to force_gimple_operand_gsi.
12450         (gimplify_build2): Likewise.
12451         (gimplify_build1): Likewise.
12453 2018-06-15  Richard Biener  <rguenther@suse.de>
12455         PR middle-end/86076
12456         * tree-cfg.c (move_stmt_op): unshare invariant addresses
12457         before adjusting their block.
12459 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12461         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
12462         multilibs for *-*-rtems*.
12463         * config/riscv/t-rtems: New file.
12465 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
12467         PR middle-end/86122
12468         * match.pd ((A +- CST1) +- CST2): Punt if last resort
12469         unsigned_type_for returns NULL.
12471         PR target/85945
12472         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
12473         subregs of multi-word pseudos unless the float mode has word size.
12475 2018-06-14  Richard Biener  <rguenther@suse.de>
12477         PR middle-end/86139
12478         * tree-vect-generic.c (build_word_mode_vector_type): Remove
12479         duplicate and harmful type_hash_canon.
12480         * tree.c (type_hash_canon): Assert we didn't find ourselves.
12482 2018-06-14  Richard Biener  <rguenther@suse.de>
12484         PR ipa/86124
12485         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
12486         NULL cgraph_node.
12488 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12490         * config/rtems.h (STDINT_LONG32): Define.
12492 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
12493             Prachi Godbole  <prachi.godbole@imgtec.com>
12495         * config/mips/mips-cpus.def: Define P6600.
12496         * config/mips/mips-tables.opt: Regenerate.
12497         * config/mips/mips.c (mips_ucbranch_type): New enum.
12498         (mips_rtx_cost_data): Add support for P6600.
12499         (mips_issue_rate): Likewise.
12500         (mips_multipass_dfa_lookahead): Likewise.
12501         (mips_avoid_hazard): Likewise.
12502         (mips_reorg_process_insns): Likewise.
12503         (mips_classify_branch_p6600): New function.
12504         * config/mips/mips.h (TUNE_P6600): New define.
12505         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
12506         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
12507         * config/mips/mips.md: Include p6600.md.
12508         (processor): Add p6600.
12509         * config/mips/p6600.md: New file.
12510         * doc/invoke.texi: Add p6600 to supported architectures.
12512 2018-06-13  Martin Sebor  <msebor@redhat.com>
12514         PR tree-optimization/86114
12515         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
12516         of integer types.
12517         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
12519 2018-06-13  Richard Biener  <rguenther@suse.de>
12521         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
12522         Properly set vector type of the intermediate stmt.
12523         * tree-vect-stmts.c (vectorizable_operation): The destination
12524         var always has vectype_out type.
12526 2018-06-13  Jeff Law  <law@redhat.com>
12528         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
12529         integer 0 for argument to print_rtl_with_bb.
12530         (rl78_reorg): Likewise.
12532 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
12534         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
12535         from rtx to rtx_insn *.
12536         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
12537         "label".
12538         (add_sched_insns_for_speculation): Likewise for local "target",
12539         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12540         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
12541         from rtx_insn ** to rtx_code_label **.
12542         (reorg_emit_nops): Likewise.
12543         (c6x_reorg): Likewise for local "call_labels".
12544         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
12545         rtx to rtx_insn *.
12546         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
12547         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
12548         the loops over LABEL_REFS.
12549         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
12550         braf_label.
12551         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12552         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
12553         (split_branches): Strengthen local "olabel" from rtx to
12554         rtx_insn *, adding a safe_as_a cast.
12555         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
12556         to "rtx_insn *".
12557         (add_insn_after): Likewise for first two params.
12558         (add_insn_before): Likewise.
12559         (remove_insn): Likewise for param.
12560         (emit_pattern_before_noloc): Likewise for second and third params.
12561         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
12562         (emit_call_insn_before_noloc): Likewise.
12563         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
12564         to "rtx_insn *".
12565         (emit_barrier_before): Likewise.
12566         (emit_label_before): Strengthen "label" param from "rtx" to
12567         "rtx_code_label *".  Strengthen "before" param from "rtx" to
12568         "rtx_insn *".
12569         (emit_insn_after_1): Strengthen "after" param from "rtx" to
12570         "rtx_insn *".
12571         (emit_pattern_after_noloc): Likewise.
12572         (emit_insn_after_noloc): Likewise.
12573         (emit_jump_insn_after_noloc): Likewise.
12574         (emit_call_insn_after_noloc): Likewise.
12575         (emit_debug_insn_after_noloc): Likewise.
12576         (emit_barrier_after): Likewise.
12577         (emit_label_after): Likewise for both params.
12578         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
12579         "loc" param from "int" to "location_t".
12580         (emit_insn_after_setloc): Likewise.
12581         (emit_jump_insn_after_setloc): Likewise.
12582         (emit_call_insn_after_setloc): Likewise.
12583         (emit_debug_insn_after_setloc): Likewise.
12584         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
12585         "loc" param from "int" to "location_t".
12586         (emit_pattern_before): Convert NULL_RTX to NULL.
12587         (emit_insn_before_setloc): Convert "loc" param from "int" to
12588         "location_t".
12589         (emit_jump_insn_before_setloc): Likewise.
12590         (emit_call_insn_before_setloc): Likewise.
12591         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
12592         rtx_insn *.  Convert "loc" param from "int" to "location_t".
12593         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
12594         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
12595         Convert 3rd param from "int" to "location_t".
12596         (emit_barrier_before, emit_barrier_after, next_real_insn):
12597         Strengthen param from rtx to rtx_insn *.
12598         (emit_label_before): Strengthen 1st param from "rtx" to
12599         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
12600         "rtx_insn *".
12601         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
12602         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
12603         Strengthen 2nd param from "rtx" to "rtx_insn *".
12604         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
12605         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
12606         Likewise. Convert 3rd param from "int" to "location_t".
12607         (emit_label_after): Strengthen 1st param from "rtx" to
12608         "rtx_code_label *".
12609         (next_real_insn, remove_insn): Strengthen param from "rtx" to
12610         "rtx_insn *".
12611         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
12612         from "rtx" to "rtx_insn *".
12614 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
12616         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
12617         bodies streamed in with -Q.
12618         * dumpfile.c (dump_files): Add lto-stream-out dump file.
12619         * dumpfile.h (tree_dump_index): Add lto_stream_out.
12620         * gimple-streamer-out.c: Include gimple-pretty-print.h
12621         (output_bb): Dump stmts streamed.
12622         * lto-section-out.c: Include print-tree.h
12623         (lto_begin_section): Dump sections created.
12624         (lto_output_decl_index): Dump decl encoded.
12625         * lto-streamer-out.c: Include print-tree.h
12626         (create_output_block): Dump output block created.
12627         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
12628         (output_function): Dump function output.
12629         (output_constructor): Dump constructor streamed.
12630         (write_global_stream): Output indexes encoded.
12631         (produce_asm_for_decls): Dump streams encoded.
12632         * lto-streamer.c (streamer_dump_file): New global var.
12633         * lto-streamer.h (streamer_dump_file): Declare.
12634         * passes.c (ipa_write_summaries): Initialize streamer dump.
12635         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
12636         in.
12638 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
12640         PR target/86048
12641         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
12642         offsets for register save directives.  Emit a second batch of save
12643         directives, if need be, when the function accesses prior frames.
12645 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12647         * config/arc/fpu.md (fmasf4): Force operand to register.
12648         (fnmasf4): Likewise.
12650 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12652         * config/arc/arc-protos.h (arc_pad_return): Remove.
12653         * config/arc/arc.c (machine_function): Remove force_short_suffix
12654         and size_reason.
12655         (arc_print_operand): Adjust printing of '&'.
12656         (arc_verify_short): Remove conditional printing of short suffix.
12657         (arc_final_prescan_insn): Remove reference to size_reason.
12658         (pad_return): New function.
12659         (arc_reorg): Call pad_return.
12660         (arc_pad_return): Remove.
12661         (arc_init_machine_status): Remove reference to force_short_suffix.
12662         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
12663         (attr length): When attribute iscompact is true force to 2
12664         regardless; in the case of maybe check if we want to force the
12665         instruction to have 4 bytes length.
12666         (nopv): Change it to generate 4 byte long nop as well.
12667         (blockage): New pattern.
12668         (simple_return): Remove call to arc_pad_return.
12669         (p_return_i): Likewise.
12671 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12673         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
12675 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12677         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
12678         ARC cores.
12680 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12682         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
12683         for ARC700 and ARCv2.
12685 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
12687         PR target/86076
12688         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
12689         operands[2] instead of operands[1].
12692 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
12694         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
12695         case, check whether the outer register overlaps an unallocatable
12696         register, not just whether it fits the required class.
12698 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
12700         * poly-int.h (can_div_trunc_p): Add new overload in which all values
12701         are poly_ints.
12702         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
12703         (memrefs_conflict_p): Likewise.
12704         (init_alias_analysis): Likewise.
12705         * cfgexpand.c (expand_debug_expr): Likewise.
12706         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
12707         * cse.c (fold_rtx): Likewise.
12708         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
12709         * expr.c (emit_block_move_hints): Likewise.
12710         (clear_storage_hints, push_block, emit_push_insn): Likewise.
12711         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
12712         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
12713         (emit_group_store): Likewise.
12714         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
12715         to read the PRE/POST_MODIFY increment.
12716         * calls.c (store_one_arg): Use strip_offset.
12717         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
12718         poly_int_rtx_p.
12719         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
12720         by a VEC_SELECT.
12721         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
12722         (simplify_binary_operation_1): Extend CONST_INT handling to
12723         poly_int_rtx_p.
12724         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
12725         than a HOST_WIDE_INT.
12726         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
12727         poly_int64.
12728         (adjust_mems, add_stores): Update accodingly.
12729         (vt_canonicalize_addr): Track polynomial offsets.
12730         (emit_note_insn_var_location): Likewise.
12731         (vt_add_function_parameter): Likewise.
12732         (vt_initialize): Likewise.
12734 2018-06-12  Jeff Law  <law@redhat.com>
12736         * config.gcc (alpha*-*-freebsd*): Remove.
12737         * config/alpha/freebsd.h: Remove.
12739 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
12741         PR other/69968
12742         * spellcheck-tree.c (levenshtein_distance): Rename to...
12743         (get_edit_distance): ...this, and update for underlying renaming.
12744         * spellcheck-tree.h (levenshtein_distance): Rename to...
12745         (get_edit_distance): ...this.
12746         * spellcheck.c (levenshtein_distance): Rename to...
12747         (get_edit_distance): ...this.  Convert from Levenshtein distance
12748         to Damerau-Levenshtein distance by supporting transpositions of
12749         adjacent characters.  Rename "v1" to "v_next" and "v0" to
12750         "v_one_ago".
12751         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
12752         (selftest::test_edit_distance_unit_test_oneway): ...this, and
12753         update for underlying renaming.
12754         (selftest::levenshtein_distance_unit_test): Rename to...
12755         (selftest::test_get_edit_distance_unit): ...this, and update for
12756         underlying renaming.
12757         (selftest::test_find_closest_string): Add example from PR 69968
12758         where transposition helps
12759         (selftest::test_metric_conditions): Update for renaming.
12760         (selftest::test_metric_conditions): Likewise.
12761         (selftest::spellcheck_c_tests): Likewise.
12762         * spellcheck.h (levenshtein_distance): Rename both overloads to...
12763         (get_edit_distance): ...this.
12764         (best_match::consider): Update for renaming.
12766 2018-06-12  Martin Sebor  <msebor@redhat.com>
12768         PR tree-optimization/85259
12769         * builtins.c (compute_objsize): Handle constant offsets.
12770         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
12771         true iff a warning has been issued.
12772         * gimple.h (gimple_nonartificial_location): New function.
12773         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
12774         gimple_nonartificial_location and handle -Wno-system-headers.
12775         (handle_builtin_stxncpy): Same.
12777 2018-06-12  Martin Sebor  <msebor@redhat.com>
12779         PR c/85931
12780         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
12782 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
12784         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
12785         BUILTIN_VEC_XST entries for pointer to double and long long.
12787 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
12789         PR target/85990
12790         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12791         Update comments.
12792         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12793         Likewise.
12795 2018-06-12  Martin Liska  <mliska@suse.cz>
12797         * doc/options.texi: Document IntegerRange.
12799 2018-06-12  Martin Liska  <mliska@suse.cz>
12801         * config/i386/i386.opt: Make MPX-related options as Deprecated.
12802         * opt-functions.awk: Handle Deprecated flag.
12803         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
12804         and report error.
12805         (read_cmdline_option): Report warning for a deprecated option.
12806         * opts.h (struct cl_option): Add new field cl_deprecated.
12807         (CL_ERR_DEPRECATED): New.
12809 2018-06-12  Martin Liska  <mliska@suse.cz>
12811         * doc/options.texi: Document Deprecated option flag.
12813 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12815         * config/arc/arc-arch.h (arc_extras): New enum.
12816         (arc_cpu_t):Add field extra.
12817         (arc_cpu_types): Consider the extras.
12818         * config/arc/arc-cpus.def: Add extras info.
12819         * config/arc/arc-opts.h (processor_type): Consider extra field.
12820         * config/arc/arc.c (arc_override_options): Handle extra field.
12822 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12824         * config/arc/arc-arch.h: Update ARC_OPTX macro.
12825         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
12826         field.
12827         * config/arc/arc.c (arc_init): Update pic warning.
12828         (irq_range): Update irq range parsing warnings.
12829         (arc_override_options): Update various warning messages.
12830         (arc_handle_aux_attribute): Likewise.
12832 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
12834         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
12836 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12838         * doc/sourcebuild.texi: Document usage of line number 0 in verify
12839         compiler messages directives.
12841 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
12843         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
12844         * config/mips/mips-tables.opt: Regenerate.
12845         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
12846         mips64r6.
12847         * doc/invoke.texi: Document -march=i6500.
12849 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
12851         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
12852         (i6400_gpmul): Add cpu_unit.
12853         (i6400_gpdiv): Likewise.
12854         (i6400_msa_add_d): Update reservations.
12855         (i6400_msa_int_add) Likewise.
12856         (i6400_msa_short_logic3) Likewise.
12857         (i6400_msa_short_logic2) Likewise.
12858         (i6400_msa_short_logic) Likewise.
12859         (i6400_msa_move) Likewise.
12860         (i6400_msa_cmp) Likewise.
12861         (i6400_msa_short_float2) Likewise.
12862         (i6400_msa_div_d) Likewise.
12863         (i6400_msa_long_logic1) Likewise.
12864         (i6400_msa_long_logic2) Likewise.
12865         (i6400_msa_mult) Likewise.
12866         (i6400_msa_long_float2) Likewise.
12867         (i6400_msa_long_float4) Likewise.
12868         (i6400_msa_long_float5) Likewise.
12869         (i6400_msa_long_float8) Likewise.
12870         (i6400_fpu_fadd): Include frint type.
12871         (i6400_fpu_store): New define_insn_reservation.
12872         (i6400_fpu_load): Likewise.
12873         (i6400_fpu_move): Likewise.
12874         (i6400_fpu_fcmp): Likewise.
12875         (i6400_fpu_fmadd): Likewise.
12876         (i6400_int_mult): Include imul3nc type and update reservation.
12877         (i6400_int_div): Include idiv3 type and update reservation.
12878         (i6400_int_load): Update to check type not move_type.
12879         (i6400_int_store): Likewise.
12880         (i6400_int_prefetch): Set zero latency.
12882 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
12884         * gcc.c: Document new %@{...} sequence.
12885         (LINK_COMMAND_SPEC): Use it for the -L switches.
12886         (cpp_unique_options): Use it for the -I switches.
12887         (at_file_argbuf): New global variable.
12888         (in_at_file): Likewise.
12889         (alloc_args): Create at_file_argbuf.
12890         (clear_args): Truncate at_file_argbuf.
12891         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
12892         (open_at_file): New function.
12893         (close_at_file): Likewise.
12894         (create_at_file): Delete.
12895         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
12896         <'o'>: Likewise.
12897         <'@'>: New case.
12898         (validate_switches_from_spec): Deal with %@{...} sequence.
12899         (validate_switches): Likewise.
12900         (driver::finalize): Call clear_args.
12902 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
12904         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
12906 2018-06-11  Martin Sebor  <msebor@redhat.com>
12908         * doc/invoke.texi (-Wall): List -Wc++17-compat.
12909         (Wno-class-memaccess): Add @opindex.
12910         (Wno-templates, Wno-multiple-inheritance): Same.
12911         (Wno-virtual-inheritance, Wno-namespaces): Same.
12912         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
12913         (Wno-format-overflow, Wno-format-truncation): Same.
12914         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
12915         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
12916         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
12917         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
12918         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
12919         (Wno-misspelled-isr): Same.
12921 2018-06-11  Martin Sebor  <msebor@redhat.com>
12923         * PR tree-optimization/86083
12924         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
12926 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
12928         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
12930 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
12932         PR target/85755
12933         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
12934         on the correct operand.
12935         (*movdi_internal64): Ditto.
12937 2018-06-11  Martin Liska  <mliska@suse.cz>
12939         PR tree-optimization/86089
12940         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
12942 2018-06-11  Julia Koval  <julia.koval@intel.com>
12944         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
12945         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
12946         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
12948 2018-06-11  Olivier Hainque  <hainque@adacore.com>
12950         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
12951         for Ada with strict dwarf2.
12953 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
12955         PR target/85755
12956         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
12957         addresses.
12959 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
12961         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
12963 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
12965         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
12966         TARGET_ELF.
12968 2018-06-08  Martin Liska  <mliska@suse.cz>
12970         * tree-cfg.h (debug_function): Fix argument type to match
12971         implementation.
12973 2018-06-08  Martin Liska  <mliska@suse.cz>
12975         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
12976         Remove usage of MPX-related (and removed) fields.
12977         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
12979 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
12981         * cfg.c (debug): Use TDF_NONE rather than 0.
12982         * cfghooks.c (debug): Likewise.
12983         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
12984         (struct dump_option_value_info): Convert to...
12985         (struct kv_pair): ...this template type.
12986         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
12987         rather than 0.
12988         (optinfo_verbosity_options): Likewise.
12989         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
12990         OPTGROUP_NONE.
12991         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
12992         than int for "optgroup_flags" param.
12993         (dump_generic_expr_loc): Use dump_flags_t rather than int for
12994         "dump_kind" param.
12995         (dump_dec): Likewise.
12996         (dump_finish): Use TDF_NONE rather than 0.
12997         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
12998         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
12999         than 0.  Update for change to option_ptr.
13000         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
13001         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
13002         0.  Update for changes to optinfo_verbosity_options and
13003         optgroup_options.
13004         (opt_info_switch_p): Convert optgroup_flags from int to
13005         optgroup_flags_t.
13006         (dump_basic_block): Use dump_flags_t rather than int
13007         for "dump_kind" param.
13008         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
13009         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
13010         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
13011         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
13012         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
13013         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
13014         TDF_NONE): Convert from macros to...
13015         (enum dump_flag): ...this new enum.
13016         (dump_flags_t): Update to use enum.
13017         (operator|, operator&, operator~, operator|=, operator&=):
13018         Implement for dump_flags_t.
13019         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
13020         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
13021         Convert from macros to...
13022         (enum optgroup_flag): ...this new enum.
13023         (optgroup_flags_t): New typedef.
13024         (operator|, operator|=): Implement for optgroup_flags_t.
13025         (struct dump_file_info): Convert field "alt_flags" to
13026         dump_flags_t.  Convert field "optgroup_flags" to
13027         optgroup_flags_t.
13028         (dump_basic_block): Use dump_flags_t rather than int for param.
13029         (dump_generic_expr_loc): Likewise.
13030         (dump_dec): Likewise.
13031         (dump_register): Convert param "optgroup_flags" to
13032         optgroup_flags_t.
13033         (opt_info_enable_passes): Likewise.
13034         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
13035         than 0.
13036         * gimple-pretty-print.c (debug): Likewise.
13037         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
13038         (merged_store_group::apply_stores): Likewise.
13039         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
13040         * gimple.c (verify_gimple_pp): Likewise.
13041         * graphite-poly.c (print_pbb_body): Likewise.
13042         * passes.c (pass_manager::register_one_dump_file): Convert
13043         local "optgroup_flags" to optgroup_flags_t.
13044         * print-tree.c (print_node): Use TDF_NONE rather than 0.
13045         (debug): Likewise.
13046         (debug_body): Likewise.
13047         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
13048         to optgroup_flags_t.
13049         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
13050         than 0.
13051         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
13052         (convert_mult_to_fma): Likewise.
13053         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
13054         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
13055         * tree-vect-data-refs.c (dump_lower_bound): Convert param
13056         "dump_kind" to dump_flags_t.
13058 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13060         * config/rs6000/rs6000.c (min, max): Delete.
13062 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13064         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
13065         -mabi=spe and -mabi=no-spe.
13067 2018-06-08  Martin Liska  <mliska@suse.cz>
13069         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
13070         where we expect an existing summary.
13072 2018-06-08  Martin Liska  <mliska@suse.cz>
13074         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
13075         * ipa-inline.h (estimate_edge_growth): Likewise.
13077 2018-06-08  Martin Liska  <mliska@suse.cz>
13079         * cgraph.c (function_version_hasher::hash): Use
13080         cgraph_node::get_uid ().
13081         (function_version_hasher::equal):
13082         * cgraph.h (cgraph_node::get_uid): New method.
13083         * ipa-inline.c (update_caller_keys): Use
13084         cgraph_node::get_uid ().
13085         (update_callee_keys): Likewise.
13086         * ipa-utils.c (searchc): Likewise.
13087         (ipa_reduced_postorder): Likewise.
13088         * lto-cgraph.c (input_node): Likewise.
13089         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
13090         * symbol-summary.h (symtab_insertion): Likewise.
13091         (symtab_removal): Likewise.
13092         (symtab_duplication): Likewise.
13093         * tree-pretty-print.c (dump_function_header): Likewise.
13094         * tree-sra.c (convert_callers_for_node): Likewise.
13096 2018-06-08  Martin Liska  <mliska@suse.cz>
13098         * cgraph.c (symbol_table::create_edge): Always assign a new
13099         unique number.
13100         (symbol_table::free_edge): Do not recycle numbers.
13101         * cgraph.h (cgraph_edge::get): New method.
13102         * symbol-summary.h (symtab_removal): Use it.
13103         (symtab_duplication): Likewise.
13104         (call_summary::hashable_uid): Remove.
13106 2018-06-08  Martin Liska  <mliska@suse.cz>
13108         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
13109         (initialize_growth_caches): Remove.
13110         (free_growth_caches): Likewise.
13111         (do_estimate_edge_time): Use edge_growth_cache.
13112         (do_estimate_edge_size): Likewise.
13113         (do_estimate_edge_hints): Likewise.
13114         * ipa-inline.c (reset_edge_caches): Likewise.
13115         (recursive_inlining): Likewise.
13116         (inline_small_functions): Likewise.
13117         * ipa-inline.h (initialize_growth_caches): Remove.
13118         (estimate_edge_size): Likewise.
13119         (estimate_edge_time): Likewise.
13120         (estimate_edge_hints): Likewise.
13121         (reset_edge_growth_cache): Likewise.
13122         * symbol-summary.h (call_summary::remove): New method.
13124 2018-06-08  Martin Liska  <mliska@suse.cz>
13126         * ipa-cp.c (class edge_clone_summary): New summary.
13127         (grow_edge_clone_vectors): Remove.
13128         (ipcp_edge_duplication_hook): Remove.
13129         (class edge_clone_summary_t): New call_summary class.
13130         (ipcp_edge_removal_hook): Remove.
13131         (edge_clone_summary_t::duplicate): New function.
13132         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
13133         (create_specialized_node): Likewise.
13134         (ipcp_driver): Initialize edge_clone_summaries and do not
13135         register hooks.
13137 2018-06-08  Martin Liska  <mliska@suse.cz>
13139         * symbol-summary.h (get): New function.
13140         (call_summary::m_initialize_when_cloning): New class member.
13142 2018-06-08  Martin Liska  <mliska@suse.cz>
13144         * cgraph.c (cgraph_node::remove): Do not recycle uid.
13145         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
13146         (symbol_table::allocate_cgraph_symbol): Do not set uid.
13147         * passes.c (uid_hash_t): Record removed_nodes by their uids.
13148         (remove_cgraph_node_from_order): Use the removed_nodes set.
13149         (do_per_function_toporder): Likwise.
13150         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
13151         instead of summary_uid.
13152         (symtab_removal): Likewise.
13153         (symtab_duplication): Likewise.
13155 2018-06-08  Martin Liska  <mliska@suse.cz>
13157         * ipa-cp.c (ipcp_store_bits_results): Use
13158         ipcp_transformation_sum.
13159         (ipcp_store_vr_results): Likewise.
13160         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
13161         to ...
13162         (ipcp_transformation_initialize): ... this.
13163         (ipa_set_node_agg_value_chain):
13164         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
13165         (write_ipcp_transformation_info): Likewise.
13166         (read_ipcp_transformation_info): Likewise.
13167         (ipcp_update_bits): Likewise.
13168         (ipcp_update_vr): Likewise.
13169         (ipcp_transform_function): Likewise.
13170         * ipa-prop.h: Rename ipcp_transformation_summary to
13171         ipcp_transformation.
13172         (class ipcp_transformation_t): New function summary.
13173         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
13174         (ipa_get_agg_replacements_for_node): Likewise.
13176 2018-06-08  Martin Liska  <mliska@suse.cz>
13178         * ipa-pure-const.c (struct funct_state_d): Do it class instead
13179         of struct.
13180         (class funct_state_summary_t): New function_summary class.
13181         (has_function_state): Remove.
13182         (get_function_state): Likewise.
13183         (set_function_state): Likewise.
13184         (add_new_function): Likewise.
13185         (funct_state_summary_t::insert): New function.
13186         (duplicate_node_data): Remove.
13187         (remove_node_data): Remove.
13188         (funct_state_summary_t::duplicate): New function.
13189         (register_hooks): Create new funct_state_summaries.
13190         (pure_const_generate_summary): Use it.
13191         (pure_const_write_summary): Likewise.
13192         (pure_const_read_summary): Likewise.
13193         (propagate_pure_const): Likewise.
13194         (propagate_nothrow): Likewise.
13195         (dump_malloc_lattice): Likewise.
13196         (propagate_malloc): Likewise.
13197         (execute): Do not register hooks, just remove summary
13198         instead.
13199         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
13200         constructor.
13202 2018-06-08  Martin Liska  <mliska@suse.cz>
13204         * ipa-reference.c (remove_node_data): Remove.
13205         (duplicate_node_data): Likewise.
13206         (class ipa_ref_var_info_summary_t): New class.
13207         (class ipa_ref_opt_summary_t): Likewise.
13208         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
13209         (get_reference_optimization_summary): Use
13210         ipa_ref_opt_sum_summaries.
13211         (set_reference_vars_info): Remove.
13212         (set_reference_optimization_summary): Likewise.
13213         (ipa_init): Create summaries.
13214         (init_function_info): Use function summary.
13215         (ipa_ref_opt_summary_t::duplicate): New function.
13216         (ipa_ref_opt_summary_t::remove): New function.
13217         (get_read_write_all_from_node): Fix GNU coding style.
13218         (propagate): Use function summary.
13219         (write_node_summary_p): Fix GNU coding style.
13220         (stream_out_bitmap): Likewise.
13221         (ipa_reference_read_optimization_summary): Use function summary.
13222         (ipa_reference_c_finalize): Do not release hooks.
13224 2018-06-08  Martin Liska  <mliska@suse.cz>
13226         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
13227         (analyze_function_body): Extract multiple calls of get_create.
13228         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13229         * ipa-inline.c (recursive_inlining): Use ::get method.
13230         * ipa-inline.h (estimate_edge_growth): Likewise.
13232 2018-06-08  Martin Liska  <mliska@suse.cz>
13234         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13235         HSA_INVALID.
13236         (hsa_function_summary::hsa_function_summary): Use the new enum
13237         value.
13238         (hsa_gpu_implementation_p): Use hsa_summaries::get.
13239         * hsa-gen.c (hsa_get_host_function): Likewise.
13240         (get_brig_function_name): Likewise.
13241         * ipa-hsa.c (process_hsa_functions): Likewise.
13242         (ipa_hsa_write_summary): Likewise.
13243         * symbol-summary.h (symtab_duplication): Use ::get function/
13244         (get): New function.
13246 2018-06-08  Martin Liska  <mliska@suse.cz>
13248         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13249         of get.
13250         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13251         (hsa_register_kernel): Likewise.
13252         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13253         * hsa-gen.c (hsa_get_host_function): Likewise.
13254         (get_brig_function_name): Likewise.
13255         (generate_hsa): Likewise.
13256         (pass_gen_hsail::execute): Likewise.
13257         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13258         (devirtualization_time_bonus): Likewise.
13259         (ipcp_propagate_stage): Likewise.
13260         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13261         (edge_set_predicate): Likewise.
13262         (evaluate_conditions_for_known_args): Likewise.
13263         (evaluate_properties_for_edge): Likewise.
13264         (ipa_fn_summary::reset): Likewise.
13265         (ipa_fn_summary_t::duplicate): Likewise.
13266         (dump_ipa_call_summary): Likewise.
13267         (ipa_dump_fn_summary): Likewise.
13268         (analyze_function_body): Likewise.
13269         (compute_fn_summary): Likewise.
13270         (estimate_edge_devirt_benefit): Likewise.
13271         (estimate_edge_size_and_time): Likewise.
13272         (estimate_calls_size_and_time): Likewise.
13273         (estimate_node_size_and_time): Likewise.
13274         (inline_update_callee_summaries): Likewise.
13275         (remap_edge_change_prob): Likewise.
13276         (remap_edge_summaries): Likewise.
13277         (ipa_merge_fn_summary_after_inlining): Likewise.
13278         (ipa_update_overall_fn_summary): Likewise.
13279         (read_ipa_call_summary): Likewise.
13280         (inline_read_section): Likewise.
13281         (write_ipa_call_summary): Likewise.
13282         (ipa_fn_summary_write): Likewise.
13283         (ipa_free_fn_summary): Likewise.
13284         * ipa-hsa.c (process_hsa_functions): Likewise.
13285         (ipa_hsa_write_summary): Likewise.
13286         (ipa_hsa_read_section): Likewise.
13287         * ipa-icf.c (sem_function::merge): Likewise.
13288         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13289         (do_estimate_edge_time): Likewise.
13290         (estimate_size_after_inlining): Likewise.
13291         (estimate_growth): Likewise.
13292         (growth_likely_positive): Likewise.
13293         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13294         (inline_call): Likewise.
13295         * ipa-inline.c (caller_growth_limits): Likewise.
13296         (can_inline_edge_p): Likewise.
13297         (can_inline_edge_by_limits_p): Likewise.
13298         (compute_uninlined_call_time): Likewise.
13299         (compute_inlined_call_time): Likewise.
13300         (want_inline_small_function_p): Likewise.
13301         (edge_badness): Likewise.
13302         (update_caller_keys): Likewise.
13303         (update_callee_keys): Likewise.
13304         (recursive_inlining): Likewise.
13305         (inline_small_functions): Likewise.
13306         (inline_to_all_callers_1): Likewise.
13307         (dump_overall_stats): Likewise.
13308         (early_inline_small_functions): Likewise.
13309         (early_inliner): Likewise.
13310         * ipa-inline.h (estimate_edge_growth): Likewise.
13311         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13312         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13313         * ipa-prop.h (IPA_NODE_REF): Likewise.
13314         (IPA_EDGE_REF): Likewise.
13315         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13316         (propagate_malloc): Likewise.
13317         * ipa-split.c (execute_split_functions): Likewise.
13318         * symbol-summary.h: Rename get to get_create.
13319         (get): Likewise.
13320         (get_create): Likewise.
13321         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13323 2018-06-08  Martin Liska  <mliska@suse.cz>
13325         * symbol-summary.h (release): Move definition out of class
13326         declaration.
13327         (symtab_removal): Likewise.
13328         (symtab_duplication): Likewise.
13330 2018-06-08  Martin Liska  <mliska@suse.cz>
13332         * symbol-summary.h (function_summary): Move constructor
13333         implementation out of class declaration.
13334         (release): Likewise.
13335         (symtab_insertion): Likewise.
13336         (symtab_removal): Likewise.
13337         (symtab_duplication): Likewise.
13338         (get): Likewise.
13340 2018-06-08  Martin Liska  <mliska@suse.cz>
13342         * Makefile.in: Remove support for MPX (macros, related functions,
13343         fields in cgraph_node, ...).
13344         * builtin-types.def (BT_BND): Likewise.
13345         (BT_FN_BND_CONST_PTR): Likewise.
13346         (BT_FN_CONST_PTR_BND): Likewise.
13347         (BT_FN_VOID_PTR_BND): Likewise.
13348         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
13349         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
13350         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
13351         (expand_builtin_mempcpy_with_bounds): Likewise.
13352         (expand_builtin_memset_with_bounds): Likewise.
13353         (expand_builtin_memset_args): Likewise.
13354         (std_expand_builtin_va_start): Likewise.
13355         (expand_builtin): Likewise.
13356         (expand_builtin_with_bounds): Likewise.
13357         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
13358         (DEF_LIB_BUILTIN_CHKP): Likewise.
13359         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
13360         (DEF_CHKP_BUILTIN): Likewise.
13361         (BUILT_IN_MEMCPY): Likewise.
13362         (BUILT_IN_MEMMOVE): Likewise.
13363         (BUILT_IN_MEMPCPY): Likewise.
13364         (BUILT_IN_MEMSET): Likewise.
13365         (BUILT_IN_STPCPY): Likewise.
13366         (BUILT_IN_STRCAT): Likewise.
13367         (BUILT_IN_STRCHR): Likewise.
13368         (BUILT_IN_STRCPY): Likewise.
13369         (BUILT_IN_STRLEN): Likewise.
13370         (BUILT_IN_MEMCPY_CHK): Likewise.
13371         (BUILT_IN_MEMMOVE_CHK): Likewise.
13372         (BUILT_IN_MEMPCPY_CHK): Likewise.
13373         (BUILT_IN_MEMSET_CHK): Likewise.
13374         (BUILT_IN_STPCPY_CHK): Likewise.
13375         (BUILT_IN_STRCAT_CHK): Likewise.
13376         (BUILT_IN_STRCPY_CHK): Likewise.
13377         * calls.c (store_bounds): Likewise.
13378         (emit_call_1): Likewise.
13379         (special_function_p): Likewise.
13380         (maybe_warn_nonstring_arg): Likewise.
13381         (initialize_argument_information): Likewise.
13382         (finalize_must_preallocate): Likewise.
13383         (compute_argument_addresses): Likewise.
13384         (expand_call): Likewise.
13385         * cfgexpand.c (expand_call_stmt): Likewise.
13386         (expand_return): Likewise.
13387         (expand_gimple_stmt_1): Likewise.
13388         (pass_expand::execute): Likewise.
13389         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
13390         (cgraph_node::remove): Likewise.
13391         (cgraph_node::dump): Likewise.
13392         (cgraph_node::verify_node): Likewise.
13393         * cgraph.h (chkp_function_instrumented_p): Likewise.
13394         (symtab_node::get_alias_target): Likewise.
13395         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
13396         (cgraph_local_p): Likewise.
13397         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
13398         (cgraph_edge::rebuild_references): Likewise.
13399         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
13400         (walk_polymorphic_call_targets): Likewise.
13401         (cgraph_node::expand_thunk): Likewise.
13402         (symbol_table::output_weakrefs): Likewise.
13403         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
13404         (ix86_handle_option): Likewise.
13405         * config/i386/constraints.md: Likewise.
13406         * config/i386/i386-builtin-types.def (BND): Likewise.
13407         (VOID): Likewise.
13408         (PVOID): Likewise.
13409         (ULONG): Likewise.
13410         * config/i386/i386-builtin.def (BDESC_END): Likewise.
13411         (BDESC_FIRST): Likewise.
13412         (BDESC): Likewise.
13413         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
13414         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
13415         * config/i386/i386.c (enum reg_class): Likewise.
13416         (ix86_target_string): Likewise.
13417         (ix86_option_override_internal): Likewise.
13418         (ix86_conditional_register_usage): Likewise.
13419         (ix86_valid_target_attribute_inner_p): Likewise.
13420         (ix86_set_indirect_branch_type): Likewise.
13421         (ix86_set_current_function): Likewise.
13422         (ix86_function_arg_regno_p): Likewise.
13423         (init_cumulative_args): Likewise.
13424         (ix86_function_arg_advance): Likewise.
13425         (ix86_function_arg): Likewise.
13426         (ix86_pass_by_reference): Likewise.
13427         (ix86_function_value_regno_p): Likewise.
13428         (ix86_function_value_1): Likewise.
13429         (ix86_function_value_bounds): Likewise.
13430         (ix86_return_in_memory): Likewise.
13431         (ix86_setup_incoming_vararg_bounds): Likewise.
13432         (ix86_va_start): Likewise.
13433         (indirect_thunk_need_prefix): Likewise.
13434         (print_reg): Likewise.
13435         (ix86_print_operand): Likewise.
13436         (ix86_expand_call): Likewise.
13437         (ix86_output_function_return): Likewise.
13438         (reg_encoded_number): Likewise.
13439         (BDESC_VERIFYS): Likewise.
13440         (ix86_init_mpx_builtins): Likewise.
13441         (ix86_init_builtins): Likewise.
13442         (ix86_emit_cmove): Likewise.
13443         (ix86_emit_move_max): Likewise.
13444         (ix86_expand_builtin): Likewise.
13445         (ix86_builtin_mpx_function): Likewise.
13446         (ix86_get_arg_address_for_bt): Likewise.
13447         (ix86_load_bounds): Likewise.
13448         (ix86_store_bounds): Likewise.
13449         (ix86_load_returned_bounds): Likewise.
13450         (ix86_store_returned_bounds): Likewise.
13451         (ix86_class_likely_spilled_p): Likewise.
13452         (ix86_hard_regno_mode_ok): Likewise.
13453         (x86_order_regs_for_local_alloc): Likewise.
13454         (ix86_mitigate_rop): Likewise.
13455         (ix86_bnd_prefixed_insn_p): Likewise.
13456         (ix86_mpx_bound_mode): Likewise.
13457         (ix86_make_bounds_constant): Likewise.
13458         (ix86_initialize_bounds): Likewise.
13459         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
13460         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
13461         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
13462         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
13463         (TARGET_CHKP_BOUND_MODE): Likewise.
13464         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
13465         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
13466         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
13467         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
13468         * config/i386/i386.h (TARGET_MPX): Likewise.
13469         (TARGET_MPX_P): Likewise.
13470         (VALID_BND_REG_MODE): Likewise.
13471         (FIRST_BND_REG): Likewise.
13472         (LAST_BND_REG): Likewise.
13473         (enum reg_class): Likewise.
13474         (BND_REG_P): Likewise.
13475         (BND_REGNO_P): Likewise.
13476         (BNDmode): Likewise.
13477         (ADJUST_INSN_LENGTH): Likewise.
13478         * config/i386/i386.md: Likewise.
13479         * config/i386/i386.opt: Likewise.
13480         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
13481         (defined): Likewise.
13482         (LINK_MPX): Likewise.
13483         (MPX_SPEC): Likewise.
13484         (LIBMPX_SPEC): Likewise.
13485         (LIBMPXWRAPPERS_SPEC): Likewise.
13486         (CHKP_SPEC): Likewise.
13487         * config/i386/predicates.md: Likewise.
13488         * dbxout.c (dbxout_type): Likewise.
13489         * doc/extend.texi: Likewise.
13490         * doc/invoke.texi: Likewise.
13491         * doc/md.texi: Likewise.
13492         * doc/tm.texi: Likewise.
13493         * doc/tm.texi.in: Likewise.
13494         * dwarf2out.c (is_base_type): Likewise.
13495         (gen_formal_types_die): Likewise.
13496         (gen_subprogram_die): Likewise.
13497         (gen_type_die_with_usage): Likewise.
13498         (gen_decl_die): Likewise.
13499         (dwarf2out_late_global_decl): Likewise.
13500         * expr.c (expand_assignment): Likewise.
13501         (emit_storent_insn): Likewise.
13502         (store_expr_with_bounds): Likewise.
13503         (store_expr): Likewise.
13504         (expand_expr_real_1): Likewise.
13505         * expr.h (store_expr_with_bounds): Likewise.
13506         * function.c (use_register_for_decl): Likewise.
13507         (struct bounds_parm_data): Likewise.
13508         (assign_parms_augmented_arg_list): Likewise.
13509         (assign_parm_find_entry_rtl): Likewise.
13510         (assign_parm_is_stack_parm): Likewise.
13511         (assign_parm_load_bounds): Likewise.
13512         (assign_bounds): Likewise.
13513         (assign_parms): Likewise.
13514         (expand_function_start): Likewise.
13515         * gcc.c (CHKP_SPEC): Likewise.
13516         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
13517         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
13518         (wrestrict_dom_walker::check_call): Likewise.
13519         * gimple.c (gimple_build_call_from_tree): Likewise.
13520         * gimple.h (enum gf_mask): Likewise.
13521         (gimple_call_with_bounds_p): Likewise.
13522         (gimple_call_set_with_bounds): Likewise.
13523         * gimplify.c (gimplify_init_constructor): Likewise.
13524         * ipa-cp.c (initialize_node_lattices): Likewise.
13525         (propagate_constants_across_call): Likewise.
13526         (find_more_scalar_values_for_callers_subset): Likewise.
13527         * ipa-hsa.c (process_hsa_functions): Likewise.
13528         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
13529         * ipa-icf.c (sem_function::merge): Likewise.
13530         * ipa-inline.c (early_inliner): Likewise.
13531         * ipa-pure-const.c (warn_function_noreturn): Likewise.
13532         (warn_function_cold): Likewise.
13533         (propagate_pure_const): Likewise.
13534         * ipa-ref.h (enum GTY): Likewise.
13535         * ipa-split.c (find_retbnd): Likewise.
13536         (consider_split): Likewise.
13537         (split_function): Likewise.
13538         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
13539         * ipa.c (walk_polymorphic_call_targets): Likewise.
13540         (symbol_table::remove_unreachable_nodes): Likewise.
13541         (process_references): Likewise.
13542         (cgraph_build_static_cdtor_1): Likewise.
13543         * lto-cgraph.c (lto_output_node): Likewise.
13544         (output_refs): Likewise.
13545         (compute_ltrans_boundary): Likewise.
13546         (input_overwrite_node): Likewise.
13547         (input_node): Likewise.
13548         (input_cgraph_1): Likewise.
13549         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
13550         * passes.c (pass_manager::execute_early_local_passes): Likewise.
13551         (class pass_chkp_instrumentation_passes): Likewise.
13552         (make_pass_chkp_instrumentation_passes): Likewise.
13553         * passes.def: Likewise.
13554         * rtl.h (struct GTY): Likewise.
13555         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
13556         * stor-layout.c (layout_type): Likewise.
13557         * symtab.c: Likewise.
13558         * target.def: Likewise.
13559         * targhooks.c (default_chkp_bound_type): Likewise.
13560         (default_chkp_bound_mode): Likewise.
13561         (default_builtin_chkp_function): Likewise.
13562         (default_chkp_function_value_bounds): Likewise.
13563         (default_chkp_make_bounds_constant): Likewise.
13564         (default_chkp_initialize_bounds): Likewise.
13565         * targhooks.h (default_chkp_bound_type): Likewise.
13566         (default_chkp_bound_mode): Likewise.
13567         (default_builtin_chkp_function): Likewise.
13568         (default_chkp_function_value_bounds): Likewise.
13569         (default_chkp_make_bounds_constant): Likewise.
13570         (default_chkp_initialize_bounds): Likewise.
13571         * toplev.c (compile_file): Likewise.
13572         (process_options): Likewise.
13573         * tree-core.h (DEF_BUILTIN): Likewise.
13574         (DEF_BUILTIN_CHKP): Likewise.
13575         * tree-inline.c (declare_return_variable): Likewise.
13576         (remap_gimple_stmt): Likewise.
13577         (copy_bb): Likewise.
13578         (initialize_inlined_parameters): Likewise.
13579         (expand_call_inline): Likewise.
13580         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
13581         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
13582         (make_pass_ipa_chkp_produce_thunks): Likewise.
13583         (make_pass_chkp): Likewise.
13584         (make_pass_chkp_opt): Likewise.
13585         (make_pass_chkp_instrumentation_passes): Likewise.
13586         * tree-pretty-print.c (dump_generic_node): Likewise.
13587         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
13588         * tree-ssa-dce.c (propagate_necessity): Likewise.
13589         (eliminate_unnecessary_stmts): Likewise.
13590         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
13591         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
13592         * tree-ssa-sccvn.h: Likewise.
13593         * tree-ssa-strlen.c (get_string_length): Likewise.
13594         (valid_builtin_call): Likewise.
13595         (adjust_last_stmt): Likewise.
13596         (handle_builtin_strchr): Likewise.
13597         (handle_builtin_strcpy): Likewise.
13598         (handle_builtin_stxncpy): Likewise.
13599         (handle_builtin_memcpy): Likewise.
13600         (handle_builtin_strcat): Likewise.
13601         (strlen_check_and_optimize_stmt): Likewise.
13602         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
13603         * tree-streamer-in.c: Likewise.
13604         * tree-streamer.c (record_common_node): Likewise.
13605         * tree.c (tree_code_size): Likewise.
13606         (wide_int_to_tree_1): Likewise.
13607         (type_contains_placeholder_1): Likewise.
13608         (build_common_tree_nodes): Likewise.
13609         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
13610         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
13611         (POINTER_BOUNDS_P): Likewise.
13612         (BOUNDED_TYPE_P): Likewise.
13613         (BOUNDED_P): Likewise.
13614         (CALL_WITH_BOUNDS_P): Likewise.
13615         (pointer_bounds_type_node): Likewise.
13616         * value-prof.c (gimple_ic): Likewise.
13617         * var-tracking.c (vt_add_function_parameters): Likewise.
13618         * varasm.c (make_decl_rtl): Likewise.
13619         (assemble_start_function): Likewise.
13620         (output_constant): Likewise.
13621         (maybe_assemble_visibility): Likewise.
13622         * varpool.c (ctor_for_folding): Likewise.
13623         * chkp-builtins.def: Remove.
13624         * ipa-chkp.c: Remove.
13625         * ipa-chkp.h: Remove.
13626         * rtl-chkp.c: Remove.
13627         * rtl-chkp.h: Remove.
13628         * tree-chkp-opt.c: Remove.
13629         * tree-chkp.c: Remove.
13630         * tree-chkp.h: Remove.
13632 2018-06-07  Carl Love  <cel@us.ibm.com>
13634         * config/rs6000/vsx.md (vextract_fp_from_shorth,
13635         vextract_fp_from_shortl): Add BE support.
13637 2018-06-07  Paul Koning  <ni1d@arrl.net>
13639         * compare-elim.c (try_merge_compare): Don't merge compare if
13640         address contains a side effect.
13641         (try_eliminate_compare): Likewise.
13643 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
13645         * config.gcc: Support "tremont".
13646         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
13647         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13648         PROCESSOR_TREMONT.
13649         * config/i386/i386.c (m_TREMONT): Define.
13650         (processor_target_table): Add "tremont".
13651         (PTA_TREMONT): Define.
13652         (ix86_lea_outperforms): Add TARGET_TREMONT.
13653         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
13654         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
13655         and M_INTEL_GOLDMONT_PLUS.
13656         (fold_builtin_cpu): Add "tremont".
13657         (ix86_add_stmt_cost): Add TARGET_TREMONT.
13658         (ix86_option_override_internal): Add "tremont".
13659         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
13660         (processor_type): Add PROCESSOR_TREMONT.
13661         * config/i386/x86-tune.def: Add m_TREMONT.
13662         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
13664 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13666         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
13667         symbol defined for msp430i* devices to be lower case.
13669 2018-06-07  Richard Biener  <rguenther@suse.de>
13671         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
13672         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
13673         Properly wrap signed arithmetic if overflow wraps.
13675 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
13677         PR tree-optimization/69615
13678         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
13679         of a cast from a same precision integral SSA_NAME in a bb dominated
13680         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
13681         cast to utype if rhs2 has already a compatible type.
13683 2018-06-07  Richard Biener  <rguenther@suse.de>
13685         PR tree-optimization/85935
13686         * graphite-scop-detection.c (find_params_in_bb): Analyze
13687         condition operands with respect to the correct loop.  Assert
13688         the analysis doesn't fail.
13690 2018-06-04  Carl Love  <cel@us.ibm.com>
13692         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
13693         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
13694         as it is slightly cheaper.
13695         (first_match_or_eos_index_<mode>):
13696         Calculate index using natural element order.
13697         (first_match_index_<mode>):
13698         Calculate index using natural element order.
13699         (first_match_or_eos_index_<mode>):
13700         Calculate index using natural order.
13701         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
13702         for BE and LE modes.
13703         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
13704         P9V_BUILTIN_VCLZLSBB_V16QI.
13705         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
13706         specific.
13708 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13710         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
13711         indentation and line wrap for many prototypes.  Add missing
13712         @smallexample directives around block of prototypes for vec_xl and
13713         vec_xst.
13715 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
13717         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
13718         track if we pass or return IEEE 128-bit floating point.
13719         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
13720         C++ mangling that is compatible with GCC 8.1.
13721         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
13722         (init_cumulative_args): Note if we pass or return IEEE 128-bit
13723         floating point types.
13724         (rs6000_function_arg_advance_1): Likewise.
13725         (rs6000_mangle_type): Optionally generate mangled names that match
13726         what GCC 8.1 generated for IEEE 128-bit floating point types.
13727         (rs6000_globalize_decl_name): If we have an external function that
13728         passes or returns IEEE 128-bit types, generate a weak reference
13729         from the mangled name used in GCC 8.1 to the current mangled
13730         name.
13731         (rs6000_init_builtins): Make __ibm128 use the long double type if
13732         long double is IBM extended double.  Make __float128 use the long
13733         double type if long double is IEEE 128-bit.
13735         PR target/85657
13736         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
13737         macro for __ibm128 built-in functions.
13738         (PACK_IF): Add __ibm128 pack/unpack functions.
13739         (UNPACK_IF): Likewise.
13740         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
13741         enable long double built-in functions if long double is IEEE
13742         128-bit floating point.
13743         (rs6000_invalid_builtin): Update long double built-in function
13744         error message.
13745         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
13746         functions, adjust the built-in function to use the long double
13747         built-in function if __ibm128 and long double are the same type.
13748         * doc/extend.texi (PowerPC builtins): Update documention for
13749         __builtin_{,un}pack_longdouble.  Add documentation for
13750         __builtin_{,un}pack_ibm128.
13752 2018-06-06  Jim Wilson  <jimw@sifive.com>
13754         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
13755         (struct machine_function): New field interrupt_mode.
13756         (riscv_handle_type_attribute): New function.  Add forward declaration.
13757         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
13758         (riscv_expand_epilogue): Check interrupt_mode field.
13759         (riscv_set_current_function): Check interrupt attribute args and
13760         set interrupt_mode field.
13761         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
13762         (riscv_sret, riscv_uret): New.
13763         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
13764         new arguments to interrupt attribute.
13766 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
13768         PR target/63177
13769         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
13770         Don't handle -mcpu=power8 if -mpower9-vector is also used.
13772 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13774         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
13775         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
13776         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13777         several redundant entries.
13779 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
13781         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
13782         type from "rtx" to "rtx_insn *".
13783         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
13784         for local "call_insn", removing cast.
13785         (ix86_expand_call): Likewise, introducing a "call_insn" local.
13787 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
13789         PR tree-optimization/86066
13790         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
13791         for BIT_INSERT_EXPR stores.
13793 2018-06-06  Richard Biener  <rguenther@suse.de>
13795         PR tree-optimization/86062
13796         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13797         component refs ontop
13798         of to be offsetted base.
13800 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13802         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
13803         to be static and remove check on interrupt attribute name.
13805 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13807         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13808         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
13810 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
13812         PR target/79924
13813         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
13814         second argument.
13815         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
13816         Remove second argument, change how error is called.
13817         (aarch64_layout_arg): Remove second argument from
13818         aarch64_err_no_fpadvsimd call.
13819         (aarch64_init_cumulative_args): Ditto.
13820         (aarch64_gimplify_va_arg_expr): Ditto.
13821         * config/aarch64/aarch64.md (mov<mode>): Ditto.
13823 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
13825         * config/i386/i386.md (simple_return_indirect_internal): New expander.
13826         (*simple_return_indirect_internal<mode>): Rename from
13827         simple_return_indirect_internal.  Use W mode iterator.
13828         (rstorssp): New expander.
13829         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
13830         (clrssbsy): New expander.
13831         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
13833 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13835         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
13836         __typeof__.
13837         (cmse_check_pointed_object): Likewise.
13839 2018-06-05  Martin Liska  <mliska@suse.cz>
13841         PR gcov-profile/47618
13842         * doc/invoke.texi: Document how -fprofile-dir format
13843         is extended.
13845 2018-06-05  Richard Biener  <rguenther@suse.de>
13847         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
13848         removal pretend DOM info isn't available so we do not update
13849         it and only remove edges, not dominated blocks.  Actually free
13850         DOM info in case we removed something.  Remove unreachable blocks.
13851         (mfb_keep_latches): Work with either DOM info or marked backedges.
13852         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
13853         first.  Mark backedges if DOM info isn't available.
13854         (Re-)compute DOM info after cleanup_control_flow_pre.
13856 2018-06-05  Richard Biener  <rguenther@suse.de>
13858         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
13859         (locus_discrim_hasher::hash): Adjust.
13860         (locus_discrim_hasher::equal): Likewise.
13861         (next_discriminator_for_locus): Work on line directly.
13862         (same_line_p): Pass in expanded locus1 as well.
13863         (assign_discriminators): Avoid redundant location expansions.
13865 2018-06-05  Richard Biener  <rguenther@suse.de>
13867         PR tree-optimization/86046
13868         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
13869         if required after clearing TREE_ADDRESSABLE.
13871 2018-06-05  Richard Biener  <rguenther@suse.de>
13873         PR tree-optimization/86047
13874         * tree-ssa-loop.c (for_each_index): Glob handling of all
13875         decls and constants and really handle all of them.
13877 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13879         PR target/81497
13880         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
13881         qualifier_void_pointer and qualifier_const_void_pointer.
13882         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
13883         (arm_init_builtins): Handle the above.
13884         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
13885         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
13886         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
13887         void intrinsics.
13889 2018-06-05  Martin Liska  <mliska@suse.cz>
13891         * auto-profile.c (read_autofdo_file): Do not use
13892         gcov_ctr_summary struct.
13893         (afdo_callsite_hot_enough_for_early_inline): Likewise.
13894         * coverage.c (struct counts_entry): Likewise.
13895         (read_counts_file): Read just single summary entry.
13896         (get_coverage_counts): Use gcov_summary struct.
13897         * coverage.h (get_coverage_counts): Likewise.
13898         * gcov-dump.c (dump_working_sets): Likewise.
13899         (tag_summary): Dump just single summary.
13900         * gcov-io.c (gcov_write_summary): Write just histogram
13901         summary.
13902         (gcov_read_summary): Read just single summary.
13903         (compute_working_sets): Use gcov_summary struct.
13904         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
13905         of GCOV_COUNTERS_SUMMABLE.
13906         (GCOV_COUNTERS_SUMMABLE): Remove.
13907         (GCOV_FIRST_VALUE_COUNTER): Replace with
13908         GCOV_COUNTER_V_INTERVAL.
13909         (struct gcov_ctr_summary): Remove.
13910         (struct gcov_summary): Directly use fields of former
13911         gcov_ctr_summary.
13912         (compute_working_sets): Use gcov_summary struct.
13913         * gcov.c (read_count_file): Do not use ctrs fields.
13914         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
13915         struct.
13916         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
13917         struct.
13918         * profile.c: Likewise.
13919         * profile.h: Likewise.
13921 2018-06-05  Martin Liska  <mliska@suse.cz>
13923         PR gcov-profile/84846
13924         * gcov.c (output_lines): Print working directory only
13925         in intermediate format.
13927 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
13929         * config/s390/s390-builtin-types.def: Add void function type.
13930         * config/s390/s390-builtins.def: Use the function type for the
13931         tbeginc builtin.
13933 2018-06-04  Jim Wilson  <jimw@sifive.com>
13935         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
13936         to int.
13937         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
13938         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
13939         handle EH_RETURN_DATA_REGNO registers properly.
13940         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
13941         (riscv_expand_epilogue): Update comment.  Change argument name and
13942         type.  Update code to use new name and type.  Pass new args to
13943         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
13944         EXCEPTION_RETURN.
13945         * config/riscv/riscv.md (NORMAL_RETURN): New.
13946         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
13947         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
13948         (eh_return): Call gen_eh_return_internal and emit barrier.
13949         (eh_return_internal): Call riscv_expand_epilogue.
13951 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
13953         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
13954         bit_insertion field and declare can_be_merged_into method.
13955         (merged_store_group::can_be_merged_into): New method.
13956         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
13957         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
13958         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
13960 2018-06-04  Richard Biener  <rguenther@suse.de>
13962         PR tree-optimization/85955
13963         * builtins.c (fold_builtin_sincos): Convert pointers to
13964         destination to appropriate type before dereferencing.
13966 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
13968         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
13970 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
13972         * expr.c (expand_expr_real_1): Force the operand into memory if
13973         its TYPE_MODE is BLKmode and if there is no integer mode for
13974         the number of bits being extracted.
13976 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
13978         PR target/85832
13979         PR target/86036
13980         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13981         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
13983 2018-06-04  Richard Biener  <rguenther@suse.de>
13985         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
13986         (cleanup_tree_cfg_noloop): ... single caller.  Do
13987         start_recording_case_labels later.
13989 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
13991         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
13992         to _IMMINTRIN_H_INCLUDED.
13993         * config/i386/pconfigintrin.h: Ditto.
13994         * config/i386/waitpkgintrin.h: Ditto.
13995         * config/i386/immintrin.h: Add includes for sgxintrin.h,
13996         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
13997         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
13998         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
13999         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
14000         waitpkgintrin.h and cldemoteintrin.h.
14002 2018-06-04  Richard Biener  <rguenther@suse.de>
14004         PR tree-optimization/86038
14005         * tracer.c (find_best_successor): Check probability for
14006         being initialized, bail out if not.
14008 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
14010         PR target/86003
14011         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
14012         of bits to ignore when comparing architectures.
14014 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14016         PR tree-optimization/69615
14017         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
14018         maximum or minimum of the type, try to merge it also as if
14019         range1 is + [-, x - 1] or + [x + 1, -].
14021         PR c++/86025
14022         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
14024 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
14026         PR tree-optimization/86034
14027         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
14028         the unsigned bitfield type in a bit insertion sequence if it does not
14029         have a larger precision than the bitfield size.
14030         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
14032 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
14034         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
14036 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14038         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
14039         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
14040         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
14041         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
14043 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14045         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
14046         Disable -fdelete-null-pointer-checks for ELF toolchain.
14048 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14049             Kito Cheng  <kito.cheng@gmail.com>
14051         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
14052         (nds32le-*-*, nds32be-*-*): Integrate checking process.
14053         (nds32*-*-*): Add glibc and uclibc conditions.
14054         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
14055         (TARGET_EXCEPT_UNWIND_INFO): Define.
14056         * config/nds32/elf.h: New file.
14057         * config/nds32/linux.h: New file.
14058         * config/nds32/nds32-elf.opt: New file.
14059         * config/nds32/nds32-linux.opt: New file.
14060         * config/nds32/nds32-fp-as-gp.c
14061         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
14062         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
14063         TARGET_LINUX_ABI.
14064         (nds32_asm_file_end): Ditto.
14065         (nds32_print_operand): Ditto.
14066         (nds32_insert_attributes): Ditto.
14067         (nds32_init_libfuncs): New function.
14068         (TARGET_HAVE_TLS): Define.
14069         (TARGET_INIT_LIBFUNCS): Define.
14070         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
14071         spec content.
14072         (TARGET_ELF): Apply different mcmodel setting.
14073         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
14074         been migrated into elf.h and linux.h files.
14075         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
14076         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
14077         (mcmodel): The content has been migrated into nds32-elf.opt and
14078         nds32-linux.opt files.
14079         * config/nds32/t-elf: New file.
14080         * config/nds32/t-linux: New file.
14082 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14083             Shiva Chen  <shiva0217@gmail.com>
14085         * config/nds32/constants.md (unspec_volatile_element): Add
14086         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
14087         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
14088         optimization.
14089         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
14090         (make_pass_nds32_fp_as_gp): Declare.
14091         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
14092         optmization pass.
14093         (nds32_asm_function_end_prologue): Remove unused asm output.
14094         (nds32_asm_function_begin_epilogue): Remove unused asm output.
14095         (nds32_asm_file_start): Output necessary fp_as_gp information.
14096         (nds32_option_override): Adjust register usage.
14097         (nds32_expand_prologue): Consider fp_as_gp situation.
14098         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
14099         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
14100         (epilogue): Ditto.
14101         (return): Ditto.
14102         (simple_return): Ditto.
14103         (omit_fp_begin): Output special directive for fp_as_gp.
14104         (omit_fp_end): Output special directive for fp_as_gp.
14105         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
14106         mforbid-fp-as-gp): New options.
14108 2018-06-01  Mark Wielaard  <mark@klomp.org>
14110         * dwarf2out.c (dwarf2out_finish): Remove generation of
14111         DW_AT_loclists_base.
14113 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
14115         * gimple-ssa-store-merging.c: Include gimple-fold.h.
14116         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
14117         (struct merged_store_group): Add bit_insertion field.
14118         (dump_char_array): Use standard hexadecimal format.
14119         (merged_store_group::merged_store_group): Set bit_insertion to false.
14120         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
14121         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
14122         also print the mask in the dump file.
14123         (pass_store_merging::gate): Minor tweak.
14124         (imm_store_chain_info::coalesce_immediate): Fix wrong association
14125         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
14126         stores with INTEGER_CST stores.
14127         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
14128         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
14129         and use it throughout.  Generate bit insertion sequences if need be.
14130         (pass_store_merging::process_store): Remove redundant condition.
14131         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
14133 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
14135         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
14136         the 128-bit floating point types.  Fix function comment.
14138 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14140         * config/aarch64/aarch64-simd.md
14141         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
14142         mnemonics.
14143         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
14144         mnemonics.
14146 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14148         PR tree-optimization/85989
14149         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
14150         variable.
14151         (backprop::intersect_uses): Check it when deciding whether this
14152         is a backedge reference.
14153         (backprop::process_block): Add each phi to m_visited_phis
14154         after visiting it, then clear it at the end.
14156 2018-06-01  Richard Biener  <rguenther@suse.de>
14158         * tree-vectorizer.h (vect_dr_stmt): New function.
14159         (vect_get_load_cost): Adjust.
14160         (vect_get_store_cost): Likewise.
14161         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
14162         Use vect_dr_stmt instead of DR_SMTT.
14163         (vect_record_base_alignments): Likewise.
14164         (vect_calculate_target_alignment): Likewise.
14165         (vect_compute_data_ref_alignment): Likewise and make static.
14166         (vect_update_misalignment_for_peel): Likewise.
14167         (vect_verify_datarefs_alignment): Likewise.
14168         (vector_alignment_reachable_p): Likewise.
14169         (vect_get_data_access_cost): Likewise.  Pass down
14170         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
14171         (vect_get_peeling_costs_all_drs): Likewise.
14172         (vect_peeling_hash_get_lowest_cost): Likewise.
14173         (vect_enhance_data_refs_alignment): Likewise.
14174         (vect_find_same_alignment_drs): Likewise.
14175         (vect_analyze_data_refs_alignment): Likewise.
14176         (vect_analyze_group_access_1): Likewise.
14177         (vect_analyze_group_access): Likewise.
14178         (vect_analyze_data_ref_access): Likewise.
14179         (vect_analyze_data_ref_accesses): Likewise.
14180         (vect_vfa_segment_size): Likewise.
14181         (vect_small_gap_p): Likewise.
14182         (vectorizable_with_step_bound_p): Likewise.
14183         (vect_prune_runtime_alias_test_list): Likewise.
14184         (vect_analyze_data_refs): Likewise.
14185         (vect_supportable_dr_alignment): Likewise.
14186         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14187         (vect_gen_prolog_loop_niters): Likewise.
14188         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14189         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
14190         modify DR_STMT.
14191         (vect_recog_mask_conversion_pattern): Likewise.
14192         (vect_try_gather_scatter_pattern): Likewise.
14193         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
14194         to vect_get_store_cost.
14195         (vect_get_store_cost): Get stmt_info instead of DR.
14196         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
14197         (vect_get_load_cost): Get stmt_info instead of DR.
14199 2018-06-01  Richard Biener  <rguenther@suse.de>
14201         PR middle-end/86017
14202         * gimple-fold.c (var_decl_component_p): Also allow offsetted
14203         vars wrapped in MEM_REFs.
14205 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14207         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14208         Fix subreg tests so that we only return a choice between
14209         GENERAL_REGS and FP_REGS if the original classes included both.
14211 2018-06-01  Richard Biener  <rguenther@suse.de>
14213         PR ipa/85960
14214         * tree-ssa-structalias.c (get_function_part_constraint):
14215         Handle NULL fi->decl.
14216         (find_func_aliases_for_call): Properly handle indirect
14217         fi from direct call.
14218         (find_func_clobbers): Likewise.
14219         (ipa_pta_execute): Likewise.
14220         (create_variable_info_for): For functions that are ifunc_resolver
14221         resolve to a varinfo that contains the result of the resolver call.
14222         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
14223         aliases.
14225 2018-05-31  Michael Collison  <michael.collison@arm.com>
14227         * config/aarch64/aarch64.md:
14228         (*fix_to_zero_extenddfdi2): New pattern.
14229         * gcc.target/aarch64/fix_extend1.c: New testcase.
14231 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
14233         PR middle-end/78809
14234         PR middle-end/83026
14235         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14236         and BUILT_IN_STRNCMP_EQ.
14237         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14238         BUILT_IN_STRNCMP_EQ.
14239         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14240         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14241         (gimple_fold_builtin): Likewise.
14242         * tree-ssa-strlen.c (compute_string_length): New function.
14243         (determine_min_obsize): New function.
14244         (handle_builtin_string_cmp): New function to handle calls to
14245         string compare functions.
14246         (strlen_optimize_stmt): Add handling to builtin string compare
14247         calls.
14248         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14249         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14250         * tree.c (build_common_builtin_nodes): Add new defines of
14251         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14253 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
14255         PR target/85984
14256         * bb-reorder.c (pass_partition_blocks::gate): Return false for
14257         functions with naked attribute.
14259 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
14261         * config/i386/sse.md (avx_vec_concat<mode>):
14262         Substitute concat_tg_mode mode attribute with xtg_mode.
14263         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14264         (concat_tg_mode): Remove mode attribute.
14266 2018-05-31  Martin Sebor  <msebor@redhat.com>
14268         PR c/82063
14269         * calls.c (alloc_max_size): Correct a logic error/typo.
14270         Treat excessive arguments as infinite.  Warn for invalid arguments.
14271         * doc/invoke.texi (-Walloc-size-larger-than): Update.
14273 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
14275         PR target/85829
14276         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14277         and movx for Haswell.
14279 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
14280             Cesar Philippidis  <cesar@codesourcery.com>
14282         PR middle-end/85879
14283         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
14284         when emitting error on private/firstprivate reductions.
14285         * omp-low.c (lower_omp_target): Avoid reference-type processing
14286         on pointers for firstprivate clause.
14288 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14290         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14291         (st1x2): Likewise.
14292         (st1x3): Likewise.
14293         * config/aarch64/aarch64-simd.md
14294         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14295         (aarch64_ld1_x3_<mode>): Likewise
14296         (aarch64_st1x2<VALLDIF:mode>): Likewise
14297         (aarch64_st1_x2_<mode>): Likewise
14298         (aarch64_st1x3<VALLDIF:mode>): Likewise
14299         (aarch64_st1_x3_<mode>): Likewise
14300         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14301         (vld1_s8_x3): Likewise.
14302         (vld1_u16_x3): Likewise.
14303         (vld1_s16_x3): Likewise.
14304         (vld1_u32_x3): Likewise.
14305         (vld1_s32_x3): Likewise.
14306         (vld1_u64_x3): Likewise.
14307         (vld1_s64_x3): Likewise.
14308         (vld1_f16_x3): Likewise.
14309         (vld1_f32_x3): Likewise.
14310         (vld1_f64_x3): Likewise.
14311         (vld1_p8_x3): Likewise.
14312         (vld1_p16_x3): Likewise.
14313         (vld1_p64_x3): Likewise.
14314         (vld1q_u8_x3): Likewise.
14315         (vld1q_s8_x3): Likewise.
14316         (vld1q_u16_x3): Likewise.
14317         (vld1q_s16_x3): Likewise.
14318         (vld1q_u32_x3): Likewise.
14319         (vld1q_s32_x3): Likewise.
14320         (vld1q_u64_x3): Likewise.
14321         (vld1q_s64_x3): Likewise.
14322         (vld1q_f16_x3): Likewise.
14323         (vld1q_f32_x3): Likewise.
14324         (vld1q_f64_x3): Likewise.
14325         (vld1q_p8_x3): Likewise.
14326         (vld1q_p16_x3): Likewise.
14327         (vld1q_p64_x3): Likewise.
14328         (vst1_s64_x2): Likewise.
14329         (vst1_u64_x2): Likewise.
14330         (vst1_f64_x2): Likewise.
14331         (vst1_s8_x2): Likewise.
14332         (vst1_p8_x2): Likewise.
14333         (vst1_s16_x2): Likewise.
14334         (vst1_p16_x2): Likewise.
14335         (vst1_s32_x2): Likewise.
14336         (vst1_u8_x2): Likewise.
14337         (vst1_u16_x2): Likewise.
14338         (vst1_u32_x2): Likewise.
14339         (vst1_f16_x2): Likewise.
14340         (vst1_f32_x2): Likewise.
14341         (vst1_p64_x2): Likewise.
14342         (vst1q_s8_x2): Likewise.
14343         (vst1q_p8_x2): Likewise.
14344         (vst1q_s16_x2): Likewise.
14345         (vst1q_p16_x2): Likewise.
14346         (vst1q_s32_x2): Likewise.
14347         (vst1q_s64_x2): Likewise.
14348         (vst1q_u8_x2): Likewise.
14349         (vst1q_u16_x2): Likewise.
14350         (vst1q_u32_x2): Likewise.
14351         (vst1q_u64_x2): Likewise.
14352         (vst1q_f16_x2): Likewise.
14353         (vst1q_f32_x2): Likewise.
14354         (vst1q_f64_x2): Likewise.
14355         (vst1q_p64_x2): Likewise.
14356         (vst1_s64_x3): Likewise.
14357         (vst1_u64_x3): Likewise.
14358         (vst1_f64_x3): Likewise.
14359         (vst1_s8_x3): Likewise.
14360         (vst1_p8_x3): Likewise.
14361         (vst1_s16_x3): Likewise.
14362         (vst1_p16_x3): Likewise.
14363         (vst1_s32_x3): Likewise.
14364         (vst1_u8_x3): Likewise.
14365         (vst1_u16_x3): Likewise.
14366         (vst1_u32_x3): Likewise.
14367         (vst1_f16_x3): Likewise.
14368         (vst1_f32_x3): Likewise.
14369         (vst1_p64_x3): Likewise.
14370         (vst1q_s8_x3): Likewise.
14371         (vst1q_p8_x3): Likewise.
14372         (vst1q_s16_x3): Likewise.
14373         (vst1q_p16_x3): Likewise.
14374         (vst1q_s32_x3): Likewise.
14375         (vst1q_s64_x3): Likewise.
14376         (vst1q_u8_x3): Likewise.
14377         (vst1q_u16_x3): Likewise.
14378         (vst1q_u32_x3): Likewise.
14379         (vst1q_u64_x3): Likewise.
14380         (vst1q_f16_x3): Likewise.
14381         (vst1q_f32_x3): Likewise.
14382         (vst1q_f64_x3): Likewise.
14383         (vst1q_p64_x3): Likewise.
14385 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14387         * config/msp430/msp430.c (msp430_output_labelref): Prepend
14388         user_label_prefix to name.
14390         * tree-core.h: Update comment about the format of NAME string
14391         passed to handler in attribute_spec.
14393         * config/msp430/msp430.md: Remove erroneous subreg expression from
14394         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
14395         zero_extend{q,h}isi2.
14397 2018-05-30  Borislav Petkov  <bp@suse.de>
14399         * doc/extend.texi: Document some architecture specific
14400         constraints and sort entries.
14402 2018-05-30  Martin Sebor  <msebor@redhat.com>
14404         PR middle-end/85369
14405         * builtins.c (expand_builtin_stpcpy_1): New function.
14406         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
14407         only if the former succeeds.
14409 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14411         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
14412         in saphira.
14414 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14416         * doc/invoke.texi (-flinker-output): Document
14418 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14420         * passes.c (ipa_write_summaries): Only modify statements if body
14421         is in memory.
14422         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
14423         incrementally linking.
14424         (ipa_passes): Likewise.
14425         * lto-cgraph.c (lto_output_node): When incrementally linking do not
14426         pass down resolution info.
14427         * common.opt (flag_incremental_link): Update info.
14428         * gcc.c (plugin specs): Turn flinker-output=* to
14429         -plugin-opt=-linker-output-known
14430         * toplev.c (compile_file): Also cut compilation when doing incremental
14431         link.
14432         * flag-types. (enum lto_partition_model): Add
14433         LTO_LINKER_OUTPUT_NOLTOREL.
14434         (invoke.texi): Add -flinker-output docs.
14435         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
14436         link same way as WPA; do not stream in dead initializers.
14438         * dwarf2out.c (dwarf2out_die_ref_for_decl,
14439         darf2out_register_external_decl): Support incremental link.
14441 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14443         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
14445 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14447         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
14448         it down to simple_object_copy_lto_debug_sections.
14449         (run_gcc): Determine incremental LTO link time and configure
14450         lto1 into non-wpa mode, disable renaming of debug sections.
14452 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14454         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14455         descriptions of various incorrectly documented functions.
14457 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14459         Revert:
14460         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14461         address check not strict.
14463 2018-05-30  Richard Biener  <rguenther@suse.de>
14465         PR tree-optimization/85964
14466         * tracer.c (better_p): Drop initialized count check, we only
14467         call the function with initialized counts now.
14468         (find_best_successor): Do find a best edge if one
14469         has uninitialized count.
14470         (find_best_predecessor): Likewise.  Do BB frequency check only
14471         if count is initialized.
14473 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
14475         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14476         (aarch64_ldrstr_offset_compare): New.
14477         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
14478         load/store orderings.
14479         (aarch64_gen_adjusted_ldpstp): Likewise.
14481 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
14483         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14484         Check for subset of GENERAL_REGS and FP_REGS.
14485         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14486         r=w alternative.
14488 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
14490         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
14491         and wi::to_poly_offset.  Add the current offset and then check
14492         whether the sum fits, rather than using an unchecked addition of
14493         a checked term.  Check for a shwi rather than a uhwi.
14494         * expr.c (get_bit_range): Use tree_to_poly_uint64.
14495         (store_constructor): Use poly_int_tree_p.
14496         (expand_expr_real_1): Likewise.
14497         * function.c (assign_temp): Likewise.
14498         * fold-const.c (const_binop): Use poly_int_tree_p and
14499         wi::to_poly_offset.
14500         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
14501         division.
14502         * ipa-icf-gimple.c (func_checker::compare_operand): Use
14503         to_poly_offset for MEM offsets.
14504         * ipa-icf.c (sem_variable::equals): Likewise.
14505         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
14506         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
14507         wi::to_poly_offset for BIT_FIELD_REF offsets.
14508         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
14509         wi::to_poly_offset.
14510         * var-tracking.c (emit_note_insn_var_location): Use
14511         tree_to_poly_uint64.
14513 2018-05-29  Jim Wilson  <jimw@sifive.com>
14515         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
14517 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
14519         PR target/85950
14520         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
14521         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
14522         sequence.
14523         (sse4_1_round<mode>2): Use nonimmediate_operand
14524         for operand 1 predicate.
14526 2018-05-29  Martin Sebor  <msebor@redhat.com>
14527             Richard Biener  <rguenther@suse.de>
14529         PR testsuite/85888
14530         * calls.c (get_size_range): Call determine_value_range instead
14531         of get_value_range..
14532         * tree-vrp.h (determine_value_range): Declared new function.
14533         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
14535 2018-05-29  Richard Biener  <rguenther@suse.de>
14537         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
14538         sure to use non-pattern stmts for get_earlier_stmt arguments.
14539         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
14540         called on pattern stmts.
14541         (get_later_stmt): Likewise.
14543 2018-05-29  Martin Liska  <mliska@suse.cz>
14545         PR gcov-profile/85759
14546         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
14547         env variables.
14549 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
14551         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
14552         VEC_UNPACK_*_EXPR.
14553         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
14554         VEC_PACK_*_EXPR.
14556         PR target/85918
14557         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14558         VEC_PACK_FLOAT_EXPR): New tree codes.
14559         * tree-pretty-print.c (op_code_prio): Handle
14560         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14561         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14562         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14563         * tree-inline.c (estimate_operator_cost): Likewise.
14564         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
14565         * fold-const.c (const_binop): Likewise.
14566         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
14567         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14568         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14569         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
14570         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14571         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14572         * expr.c (expand_expr_real_2): Likewise.
14573         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
14574         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
14575         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
14576         optabs.
14577         * optabs.c (expand_widen_pattern_expr): For
14578         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
14579         sign from result type rather than operand's type.
14580         (expand_binop_directly): For vec_packu_float_optab and
14581         vec_packs_float_optab allow result type to be different from operand's
14582         type.
14583         * optabs-tree.c (optab_for_tree_code): Handle
14584         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14585         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
14586         * tree-vect-generic.c (expand_vector_operations_1):  Handle
14587         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14588         VEC_PACK_FLOAT_EXPR.
14589         * tree-vect-stmts.c (supportable_widening_operation): Handle
14590         FIX_TRUNC_EXPR.
14591         (supportable_narrowing_operation): Handle FLOAT_EXPR.
14592         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
14593         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
14594         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
14595         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
14596         mode attributes.
14597         (vec_pack<floatprefix>_float_<mode>): New expander.
14598         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
14599         attributes.
14600         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
14601         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
14602         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
14603         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
14604         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
14605         Document.
14606         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
14607         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
14608         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14609         VEC_PACK_FLOAT_EXPR): Document.
14611 2018-05-29  Richard Biener  <rguenther@suse.de>
14613         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
14614         member.
14615         (stmt_vec_info_vec): Make pointer.
14616         (init_stmt_vec_info_vec): Remove.
14617         (free_stmt_vec_info_vec): Likewise.
14618         (set_stmt_vec_info_vec): New function.
14619         (free_stmt_vec_infos): Likewise.
14620         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
14621         (set_vinfo_for_stmt): Likewise.
14622         (get_earlier_stmt): Likewise.
14623         (get_later_stmt): Likewise.
14624         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
14625         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
14626         (vec_info::~vec_info): Free stmt_vec_infos.
14627         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
14628         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
14629         (pass_slp_vectorize::execute): Likewise.
14630         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
14631         (free_stmt_vec_info_vec): Likewise.
14632         (set_stmt_vec_info_vec): New function.
14633         (free_stmt_vec_infos): Likewise.
14634         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
14635         the global stmt_vec_info_vec.
14636         * tree-parloops.c (gather_scalar_reductions): Use
14637         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
14638         vector.
14640 2018-05-29  Richard Biener  <rguenther@suse.de>
14642         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
14644 2018-05-29  Martin Liska  <mliska@suse.cz>
14645             David Malcolm  <dmalcolm@redhat.com>
14647         * vec.c (test_reverse): New.
14648         (vec_c_tests): Add new test.
14649         * vec.h (vl_ptr>::reverse): New function.
14651 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
14653         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
14655         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
14656         and later.
14658 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14660         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
14662 2018-05-28  Richard Biener  <rguenther@suse.de>
14664         PR tree-optimization/85933
14665         * tree-vect-data-refs.c (vect_record_base_alignments): Only
14666         look at stmts marked as vectorizable.
14668 2018-05-28  Richard Biener  <rguenther@suse.de>
14670         PR tree-optimization/85934
14671         * tree-vect-generic.c (expand_vector_operations_1): Hoist
14672         vector boolean check before scalar optimization.
14674 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
14676         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
14677         for armv5te.
14679 2018-05-28  Mark Wielaard  <mark@klomp.org>
14681         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
14682         if it is an expression containing a minus sign.
14684 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
14686         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
14688 2018-05-27  Paul Koning  <ni1d@arrl.net>
14690         * config/pdp11/pdp11.md (truncsihi2): Remove.
14692 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
14693             Chung-Ju Wu  <jasonwucj@gmail.com>
14695         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
14696         implementation.
14697         (unaligned_store_dw): Ditto.
14698         * config/nds32/nds32-memory-manipulation.c
14699         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
14700         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
14701         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
14702         (emit_setmem_word_loop): Rename to ...
14703         (emit_setmem_doubleword_loop): ... this.
14704         (nds32_gen_dup_4_byte_to_word_value): New function.
14705         (nds32_gen_dup_8_byte_to_double_word_value): New function.
14706         (nds32_expand_setmem_loop): Refine implementation.
14707         (nds32_expand_setmem_loop_v3m): Ditto.
14708         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
14709         pattern.
14711 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
14713         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
14715 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
14717         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
14718         (nds32_init_machine_status): Initialize machine->attr_naked_p and
14719         machine->attr_no_prologue_p.
14720         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
14721         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
14722         (nds32_expand_epilogue): Consider attr_naked_p.
14723         (nds32_expand_epilogue_v3pop): Likewise.
14724         (nds32_can_use_return_insn): Likewise.
14725         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
14726         attr_no_prologue_p fields.
14727         * config/nds32/nds32.opt (mret-in-naked-func): New option.
14729 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
14731         PR target/85918
14732         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
14733         attributes.
14734         * config/i386/sse.md
14735         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
14736         Rename to ...
14737         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
14738         ... this.
14739         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
14740         Rename to ...
14741         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
14742         ... this.
14743         (*<floatsuffix>floatv2div2sf2): Rename to ...
14744         (*float<floatunssuffix>v2div2sf2): ... this.
14745         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
14746         (float<floatunssuffix>v2div2sf2_mask): ... this.
14747         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
14748         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
14749         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
14750         to ...
14751         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
14752         ... this.
14753         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14754         Rename to ...
14755         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14756         ... this.
14757         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14758         Rename to ...
14759         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14760         ... this.
14761         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
14762         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
14763         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
14764         gen_ufix_truncv8dfv8si2.
14765         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
14766         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
14767         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
14768         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
14769         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
14770         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
14771         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
14772         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
14774 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
14776         PR target/85900
14777         PR target/85345
14778         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
14780 2018-05-25  Jim Wilson  <jimw@sifive.com>
14782         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
14783         * config/riscv/riscv.c (struct machine_function): Add
14784         interrupt_handler_p and attribute_checked_p fields.
14785         (riscv_attribute_table): Add interrupt.
14786         (riscv_interrupt_type_p): New.
14787         (riscv_save_reg_p): Save extra regs for interrupt handler.
14788         (riscv_use_save_libcall): Return false  for interrupt handler.
14789         (riscv_first_stack_step): Add forward declaration.
14790         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
14791         for interrupt handler with large frame.  Use it for saved reg list.
14792         (riscv_expand_prologue): Move flag_stack_usage_info support to
14793         eliminate duplication.
14794         (riscv_expand_epilogue): Generate mret for interrupt handler.
14795         (riscv_epilogue_uses): New.
14796         (riscv_can_use_return_insn): Return false for interrupt handler.
14797         (riscv_function_ok_for_sibcall): Likewise.
14798         (riscv_set_current_function): Add interrupt handler support.
14799         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
14800         * config/riscv/riscv.md (UNSPECV_MRET): New.
14801         (GP_REGNUM): New.
14802         (riscv_frflags, riscv_fsflags): Use tab after opcode.
14803         (riscv_mret): New.
14804         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
14806 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
14808         PR tree-optimization/85712
14809         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
14810         this candidate has already been replaced in-situ by a copy.
14812 2018-05-25  Jason Merrill  <jason@redhat.com>
14814         PR c++/80485 - inline function non-zero address.
14815         * symtab.c (nonzero_address): Check DECL_COMDAT.
14817 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
14819         PR target/83628
14820         * config/alpha/alpha.md (ashlsi3): New insn pattern.
14821         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
14822         extension of SImode operation.  Use const123_operand predicate.
14823         (*saddsi_1): Remove.
14824         (*saddl_se_1): Ditto.
14825         (*ssubsi_1): Ditto.
14826         (*ssubl_se_1): Ditto.
14827         * config/alpha/predicates.md (const123_operand): New predicate.
14828         * config/alpha/constraints.md (P): Use IN_RANGE.
14830 2018-05-25  Richard Biener  <rguenther@suse.de>
14832         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
14833         defaulted to true.
14834         (ref_maybe_used_by_stmt_p): Likewise.
14835         (stmt_may_clobber_ref_p): Likewise.
14836         (stmt_may_clobber_ref_p_1): Likewise.
14837         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
14838         and pass it along.
14839         (ref_maybe_used_by_stmt_p): Likewise.
14840         (stmt_may_clobber_ref_p): Likewise.
14841         (stmt_may_clobber_ref_p_1): Likewise.
14842         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
14843         the alias oracle to disambiguate DRs with stmts DR analysis
14844         couldn't handle.
14845         (vect_analyze_data_refs): Do not give up on not analyzable
14846         DRs for BB vectorization.  Remove code truncating the dataref
14847         vector.
14849 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
14851         PR target/85832
14852         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14853         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
14854         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
14856 2018-05-25  Richard Biener  <rguenther@suse.de>
14858         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
14859         function, combining stmt data ref gathering and fatal analysis
14860         parts.
14861         (vect_analyze_data_refs): Remove now redudnant code and simplify.
14862         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
14863         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
14864         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
14865         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
14867 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14869         PR tree-optimization/85720
14870         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
14871         SCC if all partitions are builtins.
14872         (version_loop_by_alias_check): New parameter.  Generate cancelable
14873         runtime alias check if all partitions are builtins.
14874         (distribute_loop): Update call to above function.
14876 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14878         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
14879         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
14880         (parm_default_def_partition_arg): Ditto.
14881         (set_parm_default_def_partition): Ditto.
14882         (get_parm_default_def_partitions): Ditto and make it static.
14883         (get_undefined_value_partitions): Ditto and make it static.
14884         (remove_ssa_form): Refactor call to init_var_map here.
14885         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
14886         computation for loop region.
14887         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
14888         (register_default_def): Delete.
14889         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
14890         (parm_default_def_partition_arg): Ditto.
14891         (set_parm_default_def_partition): Ditto.
14892         (get_parm_default_def_partitions): Ditto and make it static.
14893         (get_undefined_value_partitions): Ditto and make it static.
14894         (coalesce_with_default, coalesce_with_default): Update comment.
14895         (create_coalesce_list_for_region): New func factored out from
14896         create_outofssa_var_map.
14897         (populate_coalesce_list_for_outofssa): New func factored out from
14898         create_outofssa_var_map and coalesce_ssa_name.
14899         (create_outofssa_var_map): Delete.
14900         (coalesce_ssa_name): Refactor to support live range computation.
14901         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
14902         (get_parm_default_def_partitions): Delete.
14903         (get_undefined_value_partitions): Ditto.
14904         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
14905         computation for loop region.
14906         (new_tree_live_info, loe_visit_block): Ditto.
14907         (live_worklist, set_var_live_on_entry): Ditto.
14908         (calculate_live_on_exit, verify_live_on_entry): Ditto.
14909         * tree-ssa-live.h (struct _var_map): New fields.
14910         (init_var_map): Change decl.
14911         (region_contains_p): New.
14913 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14915         * tree-ssa-live.h (live_merge_and_clear): Delete.
14917 2018-05-25  Richard Biener  <rguenther@suse.de>
14919         PR c++/85912
14920         * tree-dump.c (dequeue_and_dump): Remove access to removed
14921         operand 2 of a SWITCH_EXPR.
14923 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14925         * doc/sourcebuild.texi (vect_double_cond_arith): Include
14926         multiplication and division.
14927         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
14928         (cond_udiv@var{m}, cond_umod@var{m}): Document.
14929         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
14930         (cond_udiv_optab, cond_umod_optab): New optabs.
14931         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
14932         (IFN_COND_RDIV): New internal functions.
14933         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
14934         TRUNC_MOD_EXPR and RDIV_EXPR.
14935         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
14936         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
14937         New unspecs.
14938         (SVE_INT_BINARY): Include mult.
14939         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
14940         (optab, sve_int_op): Handle mult.
14941         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
14942         UNSPEC_COND_DIV.
14943         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
14944         for SVE_INT_BINARY_SD.
14946 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14948         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
14949         (optab, sve_int_op): Handle div and udiv.
14950         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
14951         for SVE_INT_BINARY_SD.
14952         (*<optab><mode>3): New insn for the same.
14954 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14956         * tree-vect-patterns.c: Include predict.h.
14957         (vect_recog_divmod_pattern): Restrict check for division support
14958         to when optimizing for size.
14960 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14962         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
14963         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
14964         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
14965         (gimple_match_op::set_op): Likewise.
14966         (gimple_resimplify4): Declare.
14967         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
14968         (expr::gen_transform): Likewise.
14969         (decision_tree::gen): Generate a simplification routine for 4 operands.
14970         * gimple-match-head.c (gimple_simplify): Add an overload for
14971         4 operands.  In the top-level function, handle up to 4 call
14972         arguments and call gimple_resimplify4.
14973         (gimple_resimplify4): New function.
14974         (build_call_internal): Pass a fourth operand.
14975         (maybe_push_to_seq): Likewise.
14976         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
14977         Fold VEC_COND_EXPRs of an operation and a default value into
14978         an IFN_COND_* function if possible.
14979         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
14980         New unspecs.
14981         (SVE_COND_FP_BINARY): Include them.
14982         (optab, sve_fp_op): Handle them.
14983         (SVE_INT_BINARY_REV): New code iterator.
14984         (SVE_COND_FP_BINARY_REV): New int iterator.
14985         (commutative): New int attribute.
14986         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
14987         Declare.
14988         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
14989         function.
14990         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
14991         (*cond_<optab><mode>): New patterns for reversed operands.
14993 2018-05-25  Richard Biener  <rguenther@suse.de>
14995         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
14996         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
14997         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
14998         (STMT_VINFO_GROUPED_ACCESS): Adjust.
14999         * tree-vect-data-refs.c (everywhere): Adjust users.
15000         * tree-vect-loop.c (everywhere): Likewise.
15001         * tree-vect-slp.c (everywhere): Likewise.
15002         * tree-vect-stmts.c (everywhere): Likewise.
15003         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
15005 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15007         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
15008         Rename to...
15009         (gcc_cv_as_section_exclude): ... this.
15010         Try Solaris as #exclude syntax.
15011         * configure: Regenerate.
15012         * config.in: Regenerate.
15013         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
15014         SECTION_EXCLUDE.
15015         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
15016         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
15018         * varasm.c (default_elf_asm_named_section): Don't check if
15019         HAVE_GAS_SECTION_EXCLUDE is defined.
15021 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15023         * doc/md.texi: Update the documentation of the cond_* optabs
15024         to mention the new final operand.  Fix GET_MODE_NUNITS call.
15025         Describe the scalar case too.
15026         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
15027         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
15028         instead of 2.
15029         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
15030         (get_conditional_internal_fn): Update comment.
15031         * tree-vect-loop.c (vectorizable_reduction): Pass the original
15032         accumulator value as a final argument to conditional functions.
15033         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
15034         a define_expand and add an "else" operand.  Assert for now that
15035         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
15036         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
15037         (*cond_<optab><mode>): New patterns.
15038         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
15039         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15040         (UNSPEC_COND_EOR): Delete.
15041         (optab): Remove associated mappings.
15042         (SVE_INT_BINARY): New code iterator.
15043         (sve_int_op): Remove int attribute and add "minus" to the code
15044         attribute.
15045         (SVE_COND_INT_OP): Delete.
15046         (SVE_COND_FP_OP): Rename to...
15047         (SVE_COND_FP_BINARY): ...this.
15049 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15051         * optabs.c (can_reuse_operands_p): New function.
15052         (maybe_legitimize_operands): Try to reuse the results for
15053         earlier operands.
15055 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15057         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
15058         Add {q} suffix to insn mnemonic.
15060 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15062         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
15063         (msp430_warn_func_return): New.
15065 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
15067         * fold-const.c (tree_nonzero_bits): New function.
15068         * fold-const.h (tree_nonzero_bits): Likewise.
15069         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
15070         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
15072 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15074         PR target/85900
15075         PR target/85345
15076         * varasm.c (assemble_alias): Check ifunc_resolver only on
15077         FUNCTION_DECL.
15079 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15081         PR target/85903
15082         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
15083         when memory input operand is handled.
15085 2018-05-24  Luis Machado  <luis.machado@linaro.org>
15087         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
15088         global.
15089         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
15091 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15093         * match.pd: Delay FMA folds until after vectorization.
15095 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15097         PR target/83009
15098         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15099         address check not strict.
15101 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15103         * gimple-match.h (gimple_match_op): New class.
15104         (mprts_hook): Replace parameters with a gimple_match_op *.
15105         (maybe_build_generic_op): Likewise.
15106         (gimple_simplified_result_is_gimple_val): Replace parameters with
15107         a const gimple_match_op *.
15108         (gimple_simplify): Replace code_helper * and tree * parameters with
15109         a gimple_match_op * parameter.
15110         (gimple_resimplify1): Replace code_helper *, tree and tree *
15111         parameters with a gimple_match_op * parameter.
15112         (gimple_resimplify2): Likewise.
15113         (gimple_resimplify3): Likewise.
15114         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
15115         parameters with a gimple_match_op * parameter.
15116         * gimple-match-head.c (gimple_simplify): Change prototypes of
15117         auto-generated functions to take a gimple_match_op * instead of
15118         separate code_helper * and tree * parameters.  Make the same
15119         change in the top-level overload and update calls to the
15120         gimple_resimplify routines.  Update calls to the auto-generated
15121         functions and to maybe_push_res_to_seq in the publicly-facing
15122         operation-specific gimple_simplify overloads.
15123         (gimple_match_op::MAX_NUM_OPS): Define.
15124         (gimple_resimplify1): Replace rcode and ops with a single res_op
15125         parameter.  Update call to gimple_simplify.
15126         (gimple_resimplify2): Likewise.
15127         (gimple_resimplify3): Likewise.
15128         (mprts_hook): Replace parameters with a gimple_match_op *.
15129         (maybe_build_generic_op): Likewise.
15130         (build_call_internal): Replace type, nargs and ops with
15131         a gimple_match_op *.
15132         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
15133         with a single gimple_match_op *.  Update calls to mprts_hook,
15134         build_call_internal and gimple_simplified_result_is_gimple_val.
15135         Factor out code that is common to the tree_code and combined_fn cases.
15136         * genmatch.c (expr::gen_transform): Replace tem_code and
15137         tem_ops with a gimple_match_op called tem_op.  Update calls
15138         to the gimple_resimplify functions and maybe_push_res_to_seq.
15139         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
15140         res_ops.  Update call to the gimple_resimplify functions.
15141         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
15142         (decision_tree::gen): Make the functions take a gimple_match_op *
15143         called res_op instead of separate res_code and res_ops parameters.
15144         Update call accordingly.
15145         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
15146         and ops with a single res_op parameter.  Update calls to
15147         maybe_build_generic_op and maybe_push_res_to_seq.
15148         (fold_stmt_1): Update calls to gimple_simplify and
15149         replace_stmt_with_simplification.
15150         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
15151         and gimple_simplified_result_is_gimple_val.
15152         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
15153         gimple_simplify.
15154         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
15155         with a gimple_match_op *.
15156         (vn_nary_build_or_lookup): Likewise.  Update call to
15157         vn_nary_build_or_lookup_1.
15158         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
15159         gimple_match_op *.  Update calls to the gimple_resimplify routines
15160         and to gimple_simplified_result_is_gimple_val.
15161         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
15162         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
15163         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
15164         (visit_nary_op): Likewise.
15165         (visit_reference_op_load): Likewise.
15167 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15169         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
15170         modifier for printing the step amount.
15172 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
15174         PR target/78849
15175         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
15176         types.
15178 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
15180         * doc/sourcebuild.texi (Endianness): New subsubsection.
15182 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15184         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15185         <prefetch_dynamic_strides>: New const bool field.
15186         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15187         prefetch_dynamic_strides.
15188         (exynosm1_prefetch_tune): Likewise.
15189         (thunderxt88_prefetch_tune): Likewise.
15190         (thunderx_prefetch_tune): Likewise.
15191         (thunderx2t99_prefetch_tune): Likewise.
15192         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
15193         false.
15194         (aarch64_override_options_internal): Update to set
15195         PARAM_PREFETCH_DYNAMIC_STRIDES.
15196         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15197         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15198         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15199         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15200         prefetch-dynamic-strides setting.
15202 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15204         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15205         <minimum_stride>: New const int field.
15206         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15207         minimum_stride field defaulting to -1.
15208         (exynosm1_prefetch_tune): Likewise.
15209         (thunderxt88_prefetch_tune): Likewise.
15210         (thunderx_prefetch_tune): Likewise.
15211         (thunderx2t99_prefetch_tune): Likewise.
15212         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
15213         <default_opt_level>: Set to 3.
15214         (aarch64_override_options_internal): Update to set
15215         PARAM_PREFETCH_MINIMUM_STRIDE.
15216         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15217         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15218         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15219         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15220         stride is constant and is below the minimum stride threshold.
15222 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15224         * config/arm/arm-cpus.in (mode26): Delete.
15225         (armv4): Delete mode26 reference.
15226         * config/arm/arm.c (arm_configure_build_target): Delete use of
15227         isa_bit_mode26.
15229 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15231         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
15232         New insn pattern.
15233         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15234         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
15235         for non-SSE modes.
15236         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
15237         (floatunsdidf2): Ditto.
15239 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15241         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15242         (fixuns_trunc<mode>si2_avx512f): Ditto.
15243         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15244         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15245         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15247 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15249         PR rtl-optimization/79985
15250         * df-scan.c (df_insn_refs_collect): Remove special case for
15251         global registers and asm statements.
15253 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15255         * extend.texi (Global Register Variables): Rewrite the bullet list.
15256         Note that the register is available for allocation. Note that access
15257         via inline asm must use constraints. Add note about async-signal
15258         handlers. Remove paragraph about automagic register selection.
15260 2018-05-23  Richard Biener  <rguenther@suse.de>
15262         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15263         of fixed offset from memset VN.
15265 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
15267         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15268         first_interp field.
15269         (alloc_cand_and_find_basis): Initialize first_interp field.
15270         (slsr_process_mul): Modify first_interp field.
15271         (slsr_process_add): Likewise.
15272         (slsr_process_cast): Modify first_interp field for each new
15273         interpretation.
15274         (slsr_process_copy): Likewise.
15275         (dump_candidate): Dump first_interp field.
15276         (replace_mult_candidate): Process all interpretations, not just
15277         subsequent ones.
15278         (replace_rhs_if_not_dup): Likewise.
15279         (replace_one_candidate): Likewise.
15281 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
15283         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
15284         Add new boolean.
15285         (aarch64_needs_frame_chain): New function.
15286         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
15288 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
15290         PR target/84882
15291         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
15292         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15293         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15294         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15295         as true for strict-align.
15296         (aarch64_can_inline_p): Perform checks even when callee has no
15297         attributes to check for strict alignment.
15298         * doc/extend.texi (AArch64 Function Attributes): Document
15299         no-strict-align.
15300         * doc/invoke.texi: (AArch64 Options): Likewise.
15302 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
15304         PR tree-optimization/85853
15305         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15306         the handling of the root of the node to...
15307         (vect_slp_analyze_node_operations_1): ...this new function,
15308         and run the whole thing with the child nodes' def types
15309         set according to their SLP node's def type.
15311 2018-05-23  Richard Biener  <rguenther@suse.de>
15313         PR middle-end/85874
15314         * tree-data-ref.c (create_runtime_alias_checks): Defer
15315         and ignore overflow warnings.
15317 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
15319         PR tree-optimization/85822
15320         * tree-vrp.c (is_masked_range_test): Fix handling of negative
15321         constants.
15323 2018-05-23  Richard Biener  <rguenther@suse.de>
15325         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15326         memset constants via native_interpret_expr.
15328 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
15330         PR target/85345
15331         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15332         attribute.
15333         (cgraph_node::create_alias): Likewise.
15334         (cgraph_node::get_availability): Check ifunc_resolver instead
15335         of looking up ifunc attribute.
15336         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15337         * varasm.c (do_assemble_alias): Likewise.
15338         (assemble_alias): Likewise.
15339         (default_binds_local_p_3): Likewise.
15340         * cgraph.h (cgraph_node): Add ifunc_resolver.
15341         (cgraph_node::only_called_directly_or_aliased_p): Return false
15342         for IFUNC resolver.
15343         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
15344         attribute.
15345         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
15346         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
15347         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
15348         instead of looking up ifunc attribute.
15350 2018-05-22  Luis Machado  <luis.machado@linaro.org>
15352         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
15354 2018-05-22  Martin Sebor  <msebor@redhat.com>
15356         PR middle-end/85359
15357         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
15358         only when expasion succeeds.
15359         (expand_builtin_strcmp): Same.
15360         (expand_builtin_strncmp): Same.
15362 2018-05-22  Martin Sebor  <msebor@redhat.com>
15364         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
15366 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15367             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15369         * config/aarch64/aarch64-ldpstp.md: Replace uses of
15370         aarch64_mem_pair_operand with memory_operand and delete operand swapping
15371         code.
15372         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15373         Add check for legitimate_address.
15374         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
15375         (aarch64_swap_ldrstr_operands): New.
15376         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
15377         Define prototype.
15379 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15380             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15382         * config/aarch64/aarch64.md: New patterns to generate stp
15383         and ldp.
15384         (store_pair_sw, store_pair_dw): New patterns to generate stp for
15385         single words and double words.
15386         (load_pair_sw, load_pair_dw): Likewise.
15387         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
15388         Delete.
15389         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
15390         Delete.
15391         * config/aarch64/aarch64-ldpstp.md: Modify peephole
15392         for different mode ldpstp and add peephole for merged zero stores.
15393         Likewise for loads.
15394         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15395         Add size check.
15396         (aarch64_gen_store_pair): Rename calls to match new patterns.
15397         (aarch64_gen_load_pair): Rename calls to match new patterns.
15398         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
15399         (load_pair<DREG:mode><DREG2:mode>): ... This.
15400         (store_pair<mode>): Rename to...
15401         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
15402         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
15403         New mode iterators.
15404         (V_INT_EQUIV): Handle SImode.
15405         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
15406         New predicate.
15408 2018-05-22  Martin Sebor  <msebor@redhat.com>
15410         PR c/85623
15411         * calls.c (maybe_warn_nonstring_arg): Use string length to set
15412         or ajust the presumed bound on an operation to avoid unnecessary
15413         warnings.
15415 2018-05-22  Martin Sebor  <msebor@redhat.com>
15417         PR tree-optimization/85826
15418         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
15419         assuming that a DECL necesarily has a constant size.
15421 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
15423         PR middle-end/85862
15424         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
15426 2018-05-22  Richard Biener  <rguenther@suse.de>
15428         PR tree-optimization/85834
15429         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
15430         non-constant and non-zero memset arguments.
15432 2018-05-22  Martin Liska  <mliska@suse.cz>
15434         PR ipa/85607
15435         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
15437 2018-05-22  Richard Biener  <rguenther@suse.de>
15439         PR tree-optimization/85863
15440         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
15441         comparisons when vectype is specified.
15442         (vectorizable_condition): Do not specify vectype for
15443         vect_is_simple_cond when SLP vectorizing.
15445 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
15447         PR target/85657
15448         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
15449         define __ibm128 as long double.
15450         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
15451         as a distinct type when IEEE 128-bit support is enabled.
15452         (init_float128_ieee): Fix up conversions between IFmode and IEEE
15453         128-bit types to use the correct functions.
15454         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
15455         convert between 128-bit floating point types that have different
15456         modes but the same representation, instead of using gen_lowpart to
15457         makean alias.
15458         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
15459         KFmode.
15460         (IFKF_reg): New attributes to give the register constraints for
15461         IFmode and KFmode.
15462         (extend<mode>tf2_internal): New insns to mark an explicit
15463         conversion between 128-bit floating point types that have a
15464         different mode but share the same representation.
15466 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
15468         PR tree-optimization/85814
15469         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
15470         a null return from get_strinfo when unsharing the next
15471         strinfo in the chain.
15473 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
15475         PR gcc/84923
15476         * varasm.c (weak_finish): Clean up weak_decls.
15478 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15480         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
15481         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
15482         UNSPEC_UADALP values.
15483         * config/aarch64/iterators.md (ABAL): New int iterator.
15484         (ABDL2): Likewise.
15485         (ADALP): Likewise.
15486         (sur): Add mappings for the above.
15487         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
15488         New define_insn.
15489         (aarch64_<sur>abal<mode>_4): Likewise.
15490         (aarch64_<sur>adalp<mode>_3): Likewise.
15491         (<sur>sadv16qi): New define_expand.
15493 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
15495         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
15496         (*movdf_internal): Ditto.
15497         (*rcpsf2_sse): Ditto.
15498         (*rsqrtsf2_sse): Ditto.
15499         (*sqrt<mode>2_sse): Ditto.
15501 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
15503         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
15504         eor3q<mode>4.
15505         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
15506         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15507         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
15508         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
15509         vbcaxq_s64): New.
15510         * config/aarch64/arm_neon.h: Likewise.
15511         * config/aarch64/iterators.md (VQ_I): New.
15513 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
15515         * config.gcc: Add arc/t-multilib-linux to tmake_file for
15516         arc*-*-linux*.
15517         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
15518         MULTILIB_DIRNAMES
15520 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15522         * config/nds32/constraints.md (S): New constraint.
15523         * config/nds32/nds32.md (call_internal): Use constraint S.
15524         (call_value_internal): Likewise.
15525         (sibcall_internal): Likewise.
15526         (sibcall_value_internal): Likewise.
15528 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15529             Chung-Ju Wu  <jasonwucj@gmail.com>
15531         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15532         into consideration.
15534 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15535             Chung-Ju Wu  <jasonwucj@gmail.com>
15537         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
15538         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
15539         (nds32_rtx_costs_impl): Simplify.
15540         (nds32_address_cost_impl): Simplify.
15541         (nds32_init_rtx_costs): New function.
15542         (nds32_rtx_costs_speed_prefer): Likewise.
15543         (nds32_rtx_costs_size_prefer): Likewise.
15544         (nds32_address_cost_speed_prefer): Likewise.
15545         (nds32_address_cost_speed_fwprop): Likewise.
15546         (nds32_address_cost_size_prefer): Likewise.
15547         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
15548         * config/nds32/nds32.c (nds32_option_override): Use
15549         nds32_init_rtx_costs function.
15551 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15553         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
15554         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
15555         (TARGET_PIPELINE_N8): Likewise.
15556         (TARGET_PIPELINE_N10): Likewise.
15557         (TARGET_PIPELINE_N13): Likewise.
15558         (TARGET_PIPELINE_GRAYWOLF): Likewise.
15560 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
15562         * config/nds32/nds32-fpu.md: Update copyright year.
15564 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15566         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
15568 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15570         * config/nds32/nds32.c
15571         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
15572         * config/nds32/nds32.opt (minline-asm-r15): New option.
15574 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15576         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
15577         MASK_HW_ABS.
15578         * config/nds32/nds32.md (abssi2): New pattern.
15580 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
15582         * config/i386/i386.md (rex64namesuffix): New mode attribute.
15583         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
15584         Merge insn pattern from sse_cvtsi2ss<round_name> and
15585         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
15586         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
15587         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
15588         using SWI48 mode iterator.
15589         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
15590         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
15591         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
15592         pattern from sse_cvttss2si<round_saeonly_name>
15593         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
15594         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
15595         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
15596         using SWI48 mode iterator.
15597         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
15598         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
15599         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
15600         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
15601         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
15602         using SWI48 mode iterator.
15603         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
15604         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
15605         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
15606         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
15607         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
15608         SWI48 mode iterator.
15609         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
15610         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
15611         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
15612         pattern from sse_cvttsd2si<round_saeonly_name>
15613         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
15615 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15617         * config/nds32/nds32-md-auxiliary.c
15618         (nds32_valid_smw_lwm_base_p): Refine.
15619         (nds32_output_smw_single_word): Refine.
15620         (nds32_output_smw_double_word): New.
15621         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
15623 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15625         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
15626         (nds32_output_stack_pop): Refine.
15627         (nds32_expand_unaligned_load): Refine.
15628         (nds32_expand_unaligned_store): Refine.
15630 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
15631             Chung-Ju Wu  <jasonwucj@gmail.com>
15633         * config/nds32/constants.md: Add TP_REGNUM constant.
15634         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
15635         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
15636         UNSPEC_ADD32.
15637         * config/nds32/nds32-doubleword.md: Consider flag_pic.
15638         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
15639         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
15640         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
15641         and PIC code generation.
15642         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
15643         code generation.
15644         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
15645         optimization.
15646         * config/nds32/nds32.md: Support TLS and PIC.
15647         * config/nds32/nds32.c: Support TLS and PIC.
15648         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
15649         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
15650         predicate.
15652 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15654         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
15655         mode with E_ prefix.
15657 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
15658             Chung-Ju Wu  <jasonwucj@gmail.com>
15660         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
15661         * config/nds32/nds32-md-auxiliary.c
15662         (symbolic_reference_mentioned_p): New.
15663         (nds32_legitimize_ict_address): New.
15664         (nds32_expand_ict_move): New.
15665         (nds32_indirect_call_referenced_p): New.
15666         (nds32_symbol_binds_local_p): Delete.
15667         (nds32_long_call_p): Modify.
15668         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
15669         * config/nds32/nds32-protos.h
15670         (symbolic_reference_mentioned_p): Declare.
15671         (nds32_legitimize_ict_address): Declare.
15672         (nds32_expand_ict_move): Declare.
15673         (nds32_indirect_call_referenced_p): Declare.
15674         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
15675         (nds32_relax_group): Use nds32_ict_const_p as condition.
15676         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
15677         (nds32_asm_file_start): Output ict_model directive in asm code.
15678         (nds32_legitimate_address_p): Consider indirect call.
15679         (nds32_print_operand): Consider indirect call.
15680         (nds32_print_operand_address): Consider indirect call.
15681         (nds32_insert_attributes): Handle "indirect_call" attribute.
15682         (TARGET_LEGITIMATE_ADDRESS_P): Define.
15683         (TARGET_LEGITIMATE_CONSTANT_P): Define.
15684         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
15685         (TARGET_DELEGITIMIZE_ADDRESS): Define.
15686         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
15687         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
15688         (TARGET_ICT_MODEL_SMALL): Define.
15689         (TARGET_ICT_MODEL_LARGE): Define.
15690         * config/nds32/nds32.md (movsi): Consider ict model.
15691         (call, call_value): Consider ict model.
15692         (sibcall, sibcall_value): Consider ict model.
15693         * config/nds32/nds32.opt (mict-model): New option.
15694         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
15695         model.
15697 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
15698             Monk Chiang  <sh.chiang04@gmail.com>
15699             Jim Wilson <jimw@sifive.com>
15701         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
15702         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
15703         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
15704         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
15705         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
15706         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
15707         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
15708         compute save_libcall_adjustment properly.
15709         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
15710         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
15711         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
15712         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
15713         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
15714         (ABI_SPEC): Handle mabi=ilp32e.
15715         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
15716         (RVE): Add RVE mask.
15717         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
15718         <-march>: Add rv32e as an example.
15720 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
15722         PR c++/82899
15723         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
15724         (intra_create_variable_infos): Handle C++ constructors.
15726 2018-05-18  Martin Liska  <mliska@suse.cz>
15728         * passes.def: Remove a redundant pass.
15730 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
15732         PR bootstrap/85838
15733         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
15735 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15737         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
15738         (ARMv4): Update.
15739         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
15740         (ARMv6m): Update.
15741         (armv2, armv2a, armv3, armv3m): Delete architectures.
15742         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
15743         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
15744         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
15745         Delete cpus.
15746         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
15747         (*mulsidi3adddi): Likewise.
15748         (mulsidi3): Likewise.
15749         (*mulsidi3_nov6): Likewise.
15750         (umulsidi3): Likewise.
15751         (umulsidi3_nov6): Likewise.
15752         (umaddsidi4): Likewise.
15753         (*umulsidi3adddi): Likewise.
15754         (smulsi3_highpart): Likewise.
15755         (*smulsi3_highpart_nov6): Likewise.
15756         (umulsi3_highpart): Likewise.
15757         (*umulsi3_highpart_nov6): Likewise.
15758         * config/arm/arm.h (arm_arch3m): Delete.
15759         * config/arm/arm.c (arm_arch3m): Delete.
15760         (arm_option_override_internal): Update armv3-related comment.
15761         (arm_configure_build_target): Delete use of isa_bit_mode32.
15762         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
15763         (arm_rtx_costs_internal): Delete check of arm_arch3m.
15764         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
15765         (mulsa3): Likewise.
15766         (mulusa3): Likewise.
15767         * config/arm/arm-protos.h (arm_arch3m): Delete.
15768         * config/arm/arm-tables.opt: Regenerate.
15769         * config/arm/arm-tune.md: Likewise.
15770         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
15771         deleted architectures.
15773 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15775         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
15776         (armv5t, armv5te): New features.
15777         (ARMv5, ARMv5e): Delete fgroups.
15778         (ARMv5t, ARMv5te): Adjust for above changes.
15779         (ARMv6m): Likewise.
15780         (armv5, armv5e): Delete arches.
15781         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
15782         arm_arch5.
15783         (*call_reg_arm): Likewise.
15784         (*call_value_reg_armv5): Likewise.
15785         (*call_value_reg_arm): Likewise.
15786         (*call_symbol): Likewise.
15787         (*call_value_symbol): Likewise.
15788         (*sibcall_insn): Likewise.
15789         (*sibcall_value_insn): Likewise.
15790         (clzsi2): Likewise.
15791         (prefetch): Likewise.
15792         (define_split and define_peephole2 dependent on arm_arch5):
15793         Likewise.
15794         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
15795         arm_arch5e.
15796         (TARGET_ARM_QBIT): Likewise.
15797         (TARGET_DSP_MULTIPLY): Likewise.
15798         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
15799         (arm_arch5, arm_arch5e): Delete.
15800         (arm_arch5t, arm_arch5te): Declare.
15801         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
15802         (arm_arch5t): Declare.
15803         (arm_option_reconfigure_globals): Update for the above.
15804         (arm_options_perform_arch_sanity_checks): Update comment, replace
15805         use of arm_arch5 with arm_arch5t.
15806         (use_return_insn): Likewise.
15807         (arm_emit_call_insn): Likewise.
15808         (output_return_instruction): Likewise.
15809         (arm_final_prescan_insn): Likewise.
15810         (arm_coproc_builtin_available): Likewise.
15811         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
15812         arm_arch5e with arm_arch5t and arm_arch5te.
15813         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
15814         (arm_arch5t, arm_arch5te): Declare.
15815         * config/arm/arm-tables.opt: Regenerate.
15816         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
15817         * config/arm/t-multilib: Likewise.
15818         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
15819         instead of arm_arch5.
15820         (*call_reg_thumb1): Likewise.
15821         (*call_value_reg_thumb1_v5): Likewise.
15822         (*call_value_reg_thumb1): Likewise.
15823         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
15824         unreachable path.
15825         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
15827 2018-05-18  Martin Liska  <mliska@suse.cz>
15829         PR gcov-profile/84846
15830         * doc/gcov.texi: Document -t option of gcov tool.
15832 2018-05-18  Martin Liska  <mliska@suse.cz>
15834         PR gcov-profile/84846
15835         * gcov.c (print_usage): Add new -t option.
15836         (process_args): Handle the option.
15837         (generate_results): Use stdout as output when requested by
15838         the option.
15840 2018-05-18  Martin Liska  <mliska@suse.cz>
15842         PR gcov-profile/84846
15843         * coverage.c (coverage_init): Write PWD to .gcno file.
15844         * doc/gcov.texi: Document how working directory is printed.
15845         * gcov-dump.c (dump_gcov_file): Print PWD.
15846         * gcov.c (output_intermediate_file): Likewise.
15847         (read_graph_file): Read PWD string.
15848         (output_lines): Print PWD.
15850 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15852         PR middle-end/85817
15853         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
15854         for retval and return false if all args to phi are zero.
15856 2018-05-18  Richard Biener  <rguenther@suse.de>
15858         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
15859         method.
15860         (evrp_dom_walker::before_dom_children): Call it.
15862 2018-05-18  Richard Biener  <rguenther@suse.de>
15864         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
15865         results when processing array refs with variable index.
15867 2018-05-18  Toon Moene  <toon@moene.org>
15869         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
15870         directly after that of -floop-interchange. Indicate that both
15871         options are enabled by default when specifying -O3.
15873 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15875         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
15876         iterator.  Delete separate integer-mode vec_set<mode> expander.
15877         (aarch64_simd_vec_setv2di): Delete.
15878         (vec_setv2di): Delete.
15879         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
15880         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
15881         the "w, r" alternative.
15883 2018-05-18  Martin Liska  <mliska@suse.cz>
15885         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
15886         * tree-pass.h (make_pass_lower_switch_O0): New function.
15887         * tree-switch-conversion.c (node_has_low_bound): Remove.
15888         (node_has_high_bound): Likewise.
15889         (node_is_bounded): Likewise.
15890         (class pass_lower_switch): Make it a template type and create
15891         two instances.
15892         (pass_lower_switch::execute): Add template argument.
15893         (make_pass_lower_switch): New function.
15894         (make_pass_lower_switch_O0): New function.
15895         (do_jump_if_equal): Remove.
15896         (emit_case_nodes): Simplify to just handle all 3 cases and leave
15897         all the hard work to tree optimization passes.
15899 2018-05-18  Martin Liska  <mliska@suse.cz>
15901         * dbgcnt.c (limit_low): Renamed from limit.
15902         (limit_high): New variable.
15903         (dbg_cnt_is_enabled): Check for upper limit.
15904         (dbg_cnt): Adjust dumping.
15905         (dbg_cnt_set_limit_by_index): Add new argument for high
15906         value.
15907         (dbg_cnt_set_limit_by_name): Likewise.
15908         (dbg_cnt_process_single_pair): Parse new format.
15909         (dbg_cnt_process_opt): Use strtok.
15910         (dbg_cnt_list_all_counters): Remove 'value' and add
15911         'limit_high'.
15912         * doc/invoke.texi: Document changes.
15914 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
15916         * doc/sourcebuild.texi (scalar_all_fma): Document.
15917         * tree.def (FMA_EXPR): Delete.
15918         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
15919         * internal-fn.c (ternary_direct): New macro.
15920         (expand_ternary_optab_fn): Likewise.
15921         (direct_ternary_optab_supported_p): Likewise.
15922         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
15923         * builtins.c (fold_builtin_fma): Delete.
15924         (fold_builtin_3): Don't call it.
15925         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
15926         * expr.c (expand_expr_real_2): Likewise.
15927         * fold-const.c (operand_equal_p): Likewise.
15928         (fold_ternary_loc): Likewise.
15929         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
15930         * gimple.c (DEFTREECODE): Likewise.
15931         * gimplify.c (gimplify_expr): Likewise.
15932         * optabs-tree.c (optab_for_tree_code): Likewise.
15933         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
15934         * tree-eh.c (operation_could_trap_p): Likewise.
15935         (stmt_could_throw_1_p): Likewise.
15936         * tree-inline.c (estimate_operator_cost): Likewise.
15937         * tree-pretty-print.c (dump_generic_node): Likewise.
15938         (op_code_prio): Likewise.
15939         * tree-ssa-loop-im.c (stmt_cost): Likewise.
15940         * tree-ssa-operands.c (get_expr_operands): Likewise.
15941         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
15942         * fold-const-call.h (fold_fma): Delete.
15943         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
15944         CFN_FNMA and CFN_FNMS.
15945         (fold_fma): Delete.
15946         * genmatch.c (combined_fn): New enum.
15947         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
15948         (commutative_op): New function.
15949         (commutate): Use it.  Handle more than 2 operands.
15950         (dt_operand::gen_gimple_expr): Use commutative_op.
15951         (parser::parse_expr): Allow :c to be used with non-binary
15952         operators if the commutative operand is known.
15953         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
15954         CFN_FMS, CFN_FNMA and CFN_FNMS.
15955         (backprop::process_assign_use): Remove FMA_EXPR handling.
15956         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
15957         (gen_hsa_fma): New function.
15958         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
15959         IFN_FNMA and IFN_FNMS.
15960         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
15961         * gimple-fold.h (follow_all_ssa_edges): Declare.
15962         * gimple-fold.c (follow_all_ssa_edges): New function.
15963         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
15964         gimple_build interface and use follow_all_ssa_edges to fold the result.
15965         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
15966         instead of checking for optabs directly.
15967         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
15968         rather than FMA_EXPRs.
15969         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
15970         call to IFN_FMA instead of an FMA_EXPR.
15972 2018-05-17  Jim Wilson  <jimw@sifive.com>
15974         * expr.c (do_tablejump): When converting index to Pmode, if we have a
15975         sign extended promoted subreg, and the range does not have the sign bit
15976         set, then do a sign extend.
15978         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
15979         test, check for sign extended subreg and/or constant operands, and
15980         do a sign extend in that case.
15982 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
15984         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
15985         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
15986         Add untyped.
15987         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
15988         Change logics_shift_reg to logics_shift_imm.
15989         (thunderx2t99_fp_loadpair_basic): Delete.
15990         (thunderx2t99_fp_storepair_basic): Delete.
15991         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
15992         (thunderx2t99_asimd_polynomial): Delete.
15993         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
15994         and neon_fp_mul_d_scalar_q.
15995         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15996         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
15997         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
15998         (thunderx2t99_asimd_lut): Add missing tbl types.
15999         (thunderx2t99_asimd_ext): Delete.
16000         (thunderx2t99_asimd_load1_1_mult): Delete.
16001         (thunderx2t99_asimd_load1_2_mult): Delete.
16002         (thunderx2t99_asimd_load1_ldp): New.
16003         (thunderx2t99_asimd_load1): New.
16004         (thunderx2t99_asimd_load2): Add missing *load2* types.
16005         (thunderx2t99_asimd_load3): New.
16006         (thunderx2t99_asimd_load4): New.
16007         (thunderx2t99_asimd_store1_1_mult): Delete.
16008         (thunderx2t99_asimd_store1_2_mult): Delete.
16009         (thunderx2t99_asimd_store2_mult): Delete.
16010         (thunderx2t99_asimd_store2_onelane): Delete.
16011         (thunderx2t99_asimd_store_stp): New.
16012         (thunderx2t99_asimd_store1): New.
16013         (thunderx2t99_asimd_store2): New.
16014         (thunderx2t99_asimd_store3): New.
16015         (thunderx2t99_asimd_store4): New.
16017 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
16019         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
16020         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
16022 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
16023             Segher Boessenkool  <segher@kernel.crashing.org>
16025         PR target/85698
16026         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
16027         operand.
16029 2018-05-17  Richard Biener  <rguenther@suse.de>
16031         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
16032         for pruning loop and prune defs feeding only already visited PHIs.
16034 2018-05-17  Richard Biener  <rguenther@suse.de>
16036         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
16038 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
16039             Richard Biener  <rguenther@suse.de>
16041         PR tree-optimization/85793
16042         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
16043         for VMAT_ELEMENTWISE.
16045 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16047         * internal-fn.h (lookup_internal_fn): Declare
16048         * internal-fn.c (lookup_internal_fn): New function.
16049         * gimple.c (gimple_build_call_from_tree): Handle calls to
16050         internal functions.
16051         * gimple-pretty-print.c (dump_gimple_call): Print "." before
16052         internal function names.
16053         * tree-pretty-print.c (dump_generic_node): Likewise.
16054         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
16056 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16058         * gimple-fold.h (gimple_build): Make the function forms take
16059         combined_fn rather than built_in_function.
16060         (gimple_simplify): Likewise.
16061         * gimple-match-head.c (gimple_simplify): Likewise.
16062         * gimple-fold.c (gimple_build): Likewise.
16063         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
16064         rather than gimple_build_call_internal.
16065         (get_initial_defs_for_reduction): Likewise.
16066         (vect_create_epilog_for_reduction): Likewise.
16067         (vectorizable_live_operation): Likewise.
16069 2018-05-17  Martin Liska  <mliska@suse.cz>
16071         * gimple-ssa-sprintf.c (format_directive): Do not use
16072         space in between 'G_' and '('.
16074 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
16076         PR target/85323
16077         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
16078         even if the mask is not all ones.
16080         PR target/85323
16081         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
16082         vector.
16083         (ix86_gimple_fold_builtin): Likewise.
16085         PR target/85323
16086         * config/i386/i386.c: Include tree-vector-builder.h.
16087         (ix86_vector_shift_count): New function.
16088         (ix86_fold_builtin): Fold shift builtins by scalar count.
16089         (ix86_gimple_fold_builtin): Likewise.
16091         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
16092         _mm512_setzero): New intrinsics.
16094 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
16095             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16097         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
16098         code generation for cases where splatting a value is not useful.
16099         * simplify-rtx.c (simplify_ternary_operation): Simplify
16100         vec_merge across a vec_duplicate and a paradoxical subreg forming
16101         a vector mode to a vec_concat.
16103 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
16105         * config.gcc: Support "goldmont-plus".
16106         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16107         "goldmont-plus".
16108         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16109         PROCESSOR_GOLDMONT_PLUS.
16110         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
16111         (processor_target_table): Add "goldmont-plus".
16112         (PTA_GOLDMONT_PLUS): Define.
16113         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
16114         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
16115         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
16116         (fold_builtin_cpu): Add "goldmont-plus".
16117         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
16118         (ix86_option_override_internal): Add "goldmont-plus".
16119         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
16120         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
16121         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
16122         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
16124 2018-05-17  Richard Biener  <rguenther@suse.de>
16126         PR tree-optimization/85757
16127         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
16128         remove defs that only feed that PHI from further processing.
16130 2018-05-16  Jim Wilson  <jimw@sifive.com>
16132         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
16133         asterisk to name.
16134         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
16136 2018-05-16  Mark Wielaard  <mark@klomp.org>
16138         * dwarf2out.c (count_index_strings): New function.
16139         (output_indirect_strings): Call count_index_strings and generate
16140         header for dwarf_version >= 5.
16142 2018-05-16  Mark Wielaard  <mark@klomp.org>
16144         * dwarf2out.c (dwarf_FORM): New function.
16145         (set_indirect_string): Use dwarf_FORM.
16146         (reset_indirect_string): Likewise.
16147         (size_of_die): Likewise.
16148         (value_format): Likewise.
16149         (output_die): Likewise.
16150         (add_skeleton_AT_string): Likewise.
16151         (output_macinfo_op): Likewise.
16152         (index_string): Likewise.
16153         (output_index_string_offset): Likewise.
16154         (output_index_string): Likewise.
16155         (count_index_strings): Likewise.
16157 2018-05-16  Carl Love  <cel@us.ibm.com>
16159         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16160         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
16162 2018-05-16  Martin Jambor  <mjambor@suse.cz>
16164         * ipa-prop.c (ipa_free_all_edge_args): Remove.
16165         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
16167 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
16169         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
16170         (fnma<mode>4): Likewise.
16171         (fms<mode>4): Likewise.
16172         (fnms<mode>4): Likewise.
16173         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
16174         (aarch64_fnma<mode>4): Likewise.
16175         (aarch64_fms<mode>4): Likewise.
16176         (aarch64_fnms<mode>4): Likewise.
16177         (aarch64_fnmadd<mode>4): Likewise.
16179 2018-05-16  Jason Merrill  <jason@redhat.com>
16181         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
16183 2018-05-16  Richard Biener  <rguenther@suse.de>
16185         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
16186         (dump_stmt_cost): Declare.
16187         (add_stmt_cost): Dump cost we add.
16188         (add_stmt_costs): New function.
16189         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
16190         No longer exported.
16191         (vect_analyze_stmt): Adjust prototype.
16192         (vectorizable_condition): Likewise.
16193         (vectorizable_live_operation): Likewise.
16194         (vectorizable_reduction): Likewise.
16195         (vectorizable_induction): Likewise.
16196         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
16197         cost vector to pass to vectorizable_ and record afterwards.
16198         (vect_model_reduction_cost): Take cost vector argument and adjust.
16199         (vect_model_induction_cost): Likewise.
16200         (vectorizable_reduction): Likewise.
16201         (vectorizable_induction): Likewise.
16202         (vectorizable_live_operation): Likewise.
16203         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
16204         SLP_TREE_NUMBER_OF_VEC_STMTS.
16205         (vect_analyze_slp_cost_1): Remove.
16206         (vect_analyze_slp_cost): Likewise.
16207         (vect_slp_analyze_node_operations): Take visited args and
16208         a target cost vector.  Avoid processing already visited stmt sets.
16209         (vect_slp_analyze_operations): Use a local cost vector to gather
16210         costs and register those of non-discarded instances.
16211         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
16212         (vect_schedule_slp_instance): Remove copying of
16213         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
16214         zero.
16215         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
16216         adding cost.  Record cost entry location.
16217         (vect_prologue_cost_for_slp_op): Function to compute cost of
16218         a constant or invariant generated for SLP vect in the prologue,
16219         split out from vect_analyze_slp_cost_1.
16220         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
16221         (vect_model_promotion_demotion_cost): Likewise.
16222         (vect_model_store_cost): Likewise, make static.
16223         (vect_model_load_cost): Likewise.
16224         (vectorizable_bswap): Add cost vector arg and adjust.
16225         (vectorizable_call): Likewise.
16226         (vectorizable_simd_clone_call): Likewise.
16227         (vectorizable_conversion): Likewise.
16228         (vectorizable_assignment): Likewise.
16229         (vectorizable_shift): Likewise.
16230         (vectorizable_operation): Likewise.
16231         (vectorizable_store): Likewise.
16232         (vectorizable_load): Likewise.
16233         (vectorizable_condition): Likewise.
16234         (vectorizable_comparison): Likewise.
16235         (can_vectorize_live_stmts): Likewise.
16236         (vect_analyze_stmt): Likewise.
16237         (vect_transform_stmt): Adjust calls to vectorizable_*.
16238         * tree-vectorizer.c: Include gimple-pretty-print.h.
16239         (dump_stmt_cost): New function.
16241 2018-05-16  Richard Biener  <rguenther@suse.de>
16243         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16244         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16245         * tree-ssa-dse.c: Include tree-ssa-loop.h.
16246         (check_name): New callback.
16247         (dse_classify_store): Track cycles via a visited bitmap of PHI
16248         defs and simplify handling of in-loop and across loop dead stores
16249         and properly fail for loop-variant refs.  Handle byte-tracking with
16250         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16251         limiting the walk.
16253 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
16255         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16256         (vect_get_mask_type_for_stmt): Likewise.
16257         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16258         split out from...
16259         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
16260         to determine the statement's vector type and the vector type that
16261         should be used for calculating nunits.  Deal with cases in which
16262         the type has to be deferred.
16263         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16264         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16265         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16266         (vect_determine_vf_for_stmt): New functions, split out from...
16267         (vect_determine_vectorization_factor): ...here.
16268         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16269         (vect_get_mask_type_for_stmt): New functions, split out from
16270         vect_determine_vectorization_factor.
16272 2018-05-16  Richard Biener  <rguenther@suse.de>
16274         * tree-cfg.c (verify_gimple_assign_ternary): Properly
16275         verify the [VEC_]COND_EXPR embedded comparison.
16277 2018-05-15  Martin Sebor  <msebor@redhat.com>
16279         PR tree-optimization/85753
16280         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
16281         RECORD_TYPE in addition to ARRAY_TYPE.
16283 2018-05-15  Martin Sebor  <msebor@redhat.com>
16285         PR middle-end/85643
16286         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
16288 2018-05-15  Richard Biener  <rguenther@suse.de>
16290         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
16291         add by_clobber_p one.  Change algorithm to collect all defs
16292         representing uses we need to walk and try reducing them to
16293         a single one before failing.
16294         (dse_dom_walker::dse_optimize_stmt): Adjust.
16296 2018-05-13  Mark Wielaard  <mark@klomp.org>
16298         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16299         (size_of_loc_descr): Likewise.
16300         (output_loc_operands): Likewise.
16301         (output_loc_operands_raw): Likewise.
16302         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16303         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16304         (hash_loc_operands): Likewise.
16305         (compare_loc_operands): Likewise.
16307 2018-05-14  Mark Wielaard  <mark@klomp.org>
16309         * dwarf2out.c (count_index_addrs): New function.
16310         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16312 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16314         PR tree-optimization/83648
16315         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16316         return value as malloc candidate.
16318 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16320         PR ipa/85734
16321         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16322         param as true in call to suggest_attribute.
16324 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
16326         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16327         -mreadonly-in-sdata.
16329 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16331         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16332         New pattern.
16333         (aarch64_crypto_aesd_fused): Likewise.
16335 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
16337         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
16338         (movsi_aarch64): Likewise.
16339         (load_pairsi): Likewise.
16340         (load_pairdi): Likewise.
16341         (store_pairsi): Likewise.
16342         (store_pairdi): Likewise.
16343         (load_pairsf): Likewise.
16344         (load_pairdf): Likewise.
16345         (store_pairsf): Likewise.
16346         (store_pairdf): Likewise.
16347         (zero_extend): Likewise.
16348         (trunc): Swap alternatives.
16349         (fcvt_target): Add '?' to prefer w over r.
16351 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
16353         PR target/85756
16354         * config/i386/i386.md: Disallow non-commutative arithmetics in
16355         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
16356         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
16357         in the peephole2 before it.
16359 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
16361         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
16362         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
16363         (ix86_handle_option): Handle -mcldemote.
16364         * config.gcc: New header.
16365         * config/i386/cldemoteintrin.h: New file.
16366         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
16367         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16368         -mcldemote.
16369         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16370         OPTION_MASK_ISA_CLDEMOTE.
16371         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
16372         (ix86_valid_target_attribute_inner_p): Ditto.
16373         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
16374         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
16375         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
16376         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
16377         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
16378         (cldemote): New.
16379         * config/i386/i386.opt: Add -mcldemote.
16380         * config/i386/x86intrin.h: New header.
16381         * doc/invoke.texi: Add -mcldemote.
16383 2018-05-14  Richard Biener  <rguenther@suse.de>
16385         * doc/match-and-simplify.texi: Adjust :s documentation.
16387 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
16389         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
16390         intended memcpy size.
16391         (REORDER_45): Likewise.
16393 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
16395         * sort.cc: New file.
16396         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
16397         * vec.c (qsort_chk): Use gcc_qsort.
16398         * Makefile.in (OBJS-libcommon): Add sort.o.
16399         (build/sort.o): New target.  Use it...
16400         (BUILD_RTL): ... here, and...
16401         (build/gencfn-macros): ... here, and...
16402         (build/genmatch): ... here.
16404 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16405             Chung-Ju Wu  <jasonwucj@gmail.com>
16407         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
16408         * config/nds32/nds32-graywolf.md: New file.
16409         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
16410         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
16411         pipeline.
16412         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
16413         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
16414         * config/nds32/nds32.md (pipeline_model): Add graywolf.
16415         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
16416         * config/nds32/pipelines.md: Include n15 settings.
16418 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16419             Chung-Ju Wu  <jasonwucj@gmail.com>
16421         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
16422         * config/nds32/nds32-n13.md: New file.
16423         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
16424         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
16425         pipeline.
16426         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
16427         * config/nds32/nds32.md (pipeline_model): Add n13.
16428         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
16429         * config/nds32/pipelines.md: Include n13 settings.
16431 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16432             Chung-Ju Wu  <jasonwucj@gmail.com>
16434         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
16435         * config/nds32/nds32-n10.md: New file.
16436         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
16437         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
16438         pipeline.
16439         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
16440         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
16441         * config/nds32/nds32.md (pipeline_model): Add n10.
16442         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
16443         * config/nds32/pipelines.md: Include n10 settings.
16445 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
16446             Kito Cheng  <kito.cheng@gmail.com>
16447             Chung-Ju Wu  <jasonwucj@gmail.com>
16449         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
16450         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16451         Add enum values for DSP extension instructions.
16452         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
16453         New constraints.
16454         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
16455         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
16456         New code iterators.
16457         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
16458         * config/nds32/nds32-dspext.md: New file for DSP implementation.
16459         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
16460         * config/nds32/nds32-intrinsic.md: Likewise.
16461         * config/nds32/nds32_intrinsic.h: Likewise.
16462         * config/nds32/nds32-md-auxiliary.c: Likewise.
16463         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
16464         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
16465         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
16466         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
16467         * config/nds32/nds32-protos.h: New declarations for DSP extension.
16468         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
16469         TYPE_DMAC in switch statement.
16470         * config/nds32/nds32.c: New checking and implementation for DSP
16471         extension instructions.
16472         * config/nds32/nds32.h: Likewise.
16473         * config/nds32/nds32.md: Likewise.
16474         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
16475         * config/nds32/predicates.md: Implement new predicates for DSP
16476         extension.
16478 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
16480         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
16481         Reformat alternatives and attributes so it is easier to identify
16482         which constraints/attributes go with which instruction.
16483         (mov<mode>_hardfloat32, FMOVE64): Likewise.
16484         (mov<mode>_softfloat32, FMOVE64): Likewise.
16485         (mov<mode>_hardfloat64, FMOVE64): Likewise.
16486         (mov<mode>_softfloat64, FMOVE64): Likewise.
16488 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16490         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16491         subsection.
16492         (Basic PowerPC Built-in Functions): The new name of the
16493         subsection previously known as "PowerPC Built-in Functions".
16494         (Basic PowerPC Built-in Functions Available on all Configurations):
16495         New subsubsection.
16496         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
16497         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
16498         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
16499         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
16501 2018-05-11  Martin Jambor  <mjambor@suse.cz>
16503         PR ipa/85655
16504         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
16505         single const.
16507 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16509         PR target/85733
16510         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
16512 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
16514         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
16515         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
16516         (ix86_handle_option): Handle -mwaitpkg.
16517         * config.gcc: New header.
16518         * config/i386/cpuid.h (bit_WAITPKG): New bit.
16519         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
16520         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
16521         function type.
16522         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16523         OPTION_MASK_ISA_WAITPKG.
16524         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
16525         (ix86_option_override_internal): Add PTA_WAITPKG.
16526         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
16527         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
16528         IX86_BUILTIN_TPAUSE.
16529         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
16530         __builtin_ia32_umwait and __builtin_ia32_tpause.
16531         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
16532         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
16533         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
16534         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
16535         UNSPECV_TPAUSE): New.
16536         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
16537         * config/i386/i386.opt: Add -mwaitpkg.
16538         * config/i386/waitpkgintrin.h: New file.
16539         * config/i386/x86intrin.h: New header.
16540         * doc/invoke.texi: Add -mwaitpkg.
16542 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16544         PR target/85606
16545         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
16546         equivalent.
16547         (cortex-m0): Use armv6s-m isa.
16548         (cortex-m0plus): Likewise.
16549         (cortex-m1): Likewise.
16550         (cortex-m0.small-multiply): Likewise.
16551         (cortex-m0plus.small-multiply): Likewise.
16552         (cortex-m1.small-multiply): Likewise.
16554 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
16555             Jakub Jelinek  <jakub@redhat.com>
16557         PR tree-optimization/85692
16558         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
16559         source permute as well.
16561 2018-05-11  Martin Liska  <mliska@suse.cz>
16563         PR sanitizer/85556
16564         * doc/extend.texi: Document LLVM style format for no_sanitize
16565         attribute.
16567 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
16569         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
16570         mode_supports_vsx_dform_quad to mode_supports_dq_form.
16571         (mode_supports_vsx_dform_quad): Likewise.
16572         (mode_supports_vmx_dform): Move these functions to be next to the
16573         other mode_supports functions.
16574         (mode_supports_dq_form): Likewise.
16575         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
16576         mode_supports_dq_form.
16577         (reg_offset_addressing_ok_p): Likewise.
16578         (offsettable_ok_by_alignment): Likewise.
16579         (rs6000_legitimate_offset_address_p): Likewise.
16580         (legitimate_lo_sum_address_p): Likewise.
16581         (rs6000_legitimize_address): Likewise.
16582         (rs6000_legitimize_reload_address): Likewise.
16583         (rs6000_secondary_reload_inner): Likewise.
16584         (rs6000_preferred_reload_class): Likewise.
16585         (rs6000_output_move_128bit): Likewise.
16587 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
16589         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
16590         Generate SImode target register for null target.
16591         <case IX86_BUILTIN_XGETBV>: Ditto.
16592         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
16593         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
16595 2018-05-10  Carl Love  <cel@us.ibm.com>
16597         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16598         dcbtt and dcbtstt if operands[2] is 0.
16600 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
16602         PR target/85693
16603         * config/i386/sse.md (usadv64qi): New expander.
16605 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
16607         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
16608         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
16609         -maltivec=be support.
16610         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
16611         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
16612         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
16613         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
16614         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
16615         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
16616         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
16617         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
16618         altivec_vsumsws): Adjust.
16619         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
16620         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
16621         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
16622         support.
16623         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
16624         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
16625         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
16626         (altivec_lve<VI_char>x): Delete expand.
16627         (*altivec_lve<VI_char>x_internal): Rename to...
16628         (altivec_lve<VI_char>x): ... this.
16629         (altivec_lvxl_<mode>): Delete expand.
16630         (*altivec_lvxl_<mode>_internal): Rename to ...
16631         (altivec_lvxl_<mode>): ... this.
16632         (altivec_stvxl_<mode>): Delete expand.
16633         (*altivec_stvxl_<mode>_internal): Rename to ...
16634         (altivec_stvxl_<mode>): ... this.
16635         (altivec_stve<VI_char>x): Delete expand.
16636         (*altivec_stve<VI_char>x_internal): Rename to ...
16637         (altivec_stve<VI_char>x): ... this.
16638         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
16639         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
16640         reduc_plus_scal_<mode>): Adjust.
16641         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
16642         comment.
16643         (rs6000_cpu_cpp_builtins): Adjust.
16644         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
16645         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
16646         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
16647         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
16648         -maltivec=be support.
16649         (rs6000_split_vec_extract_var): Adjust.
16650         (rs6000_split_v4si_init): Adjust.
16651         (swap_selector_for_mode): Delete.
16652         (altivec_expand_lvx_be, altivec_expand_stvx_be,
16653         altivec_expand_stvex_be): Delete.
16654         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
16655         -maltivec=be support.
16656         (rs6000_gimple_fold_builtin): Ditto.
16657         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
16658         Adjust.
16659         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
16660         (TARGET_DIRECT_MOVE_64BIT): Adjust.
16661         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
16662         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
16663         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
16664         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
16665         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
16666         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
16667         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
16668         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
16669         anonymous split): Adjust.
16670         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
16671         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
16673 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
16675         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
16676         when --with-gxx-include-dir is also specified.
16677         * configure: Regenerate.
16679 2018-05-09  Jim Wilson  <jimw@sifive.com>
16681         PR target/84797
16682         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
16683         * config/riscv/t-withmultilib: New.
16684         * config/riscv/withmultilib.h: New.
16685         * doc/install.texi: Document RISC-V --with-multilib-list support.
16687 2018-05-09  Richard Biener  <rguenther@suse.de>
16689         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
16690         vector.
16691         (vect_bb_vectorization_profitable_p): Adjust.  Compute
16692         actual scalar cost using the cost vector and the add_stmt_cost
16693         machinery.
16695 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16697         PR rtl-optimization/85645
16698         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
16699         in the REG_CFA_REGISTER note for LR, don't leave it empty.
16701 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16703         PR rtl-optimization/85645
16704         * shrink-wrap.c (spread_components): Return a boolean saying if
16705         anything was changed.
16706         (try_shrink_wrapping_separate): Iterate spread_components until
16707         nothing changes anymore.
16709 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16711         PR rtl-optimization/85645
16712         * regrename.c (build_def_use): Also kill the chains that include the
16713         destination of a REG_CFA_REGISTER note.
16715 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16717         PR rtl-optimization/85645
16718         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
16719         insn that has a REG_CFA_REGISTER note.
16721 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
16723         * cfgexpand.c (expand_clobber): New function.
16724         (expand_gimple_stmt_1): Use it.
16725         * tree-vect-stmts.c (vect_clobber_variable): New function,
16726         split out from...
16727         (vectorizable_simd_clone_call): ...here.
16728         (vectorizable_store): Emit a clobber either side of an
16729         IFN_STORE_LANES sequence.
16730         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
16732 2018-05-09  Tom de Vries  <tom@codesourcery.com>
16734         PR target/85626
16735         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
16736         (define_insn "trap_if_false"): Add exit after trap.
16738 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
16740         PR rtl-optimization/85638
16741         * bb-reorder.c: Include common/common-target.h.
16742         (create_forwarder_block): New function extracted from...
16743         (fix_up_crossing_landing_pad): ...here.  Rename into...
16744         (dw2_fix_up_crossing_landing_pad): ...this.
16745         (sjlj_fix_up_crossing_landing_pad): New function.
16746         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
16747         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
16748         from both partitions and exit the loop after one iteration.
16750 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16752         Revert:
16753         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16754         subsection.
16755         (Basic PowerPC Built-in Functions): The new name of the
16756         subsection previously known as "PowerPC Built-in Functions".
16757         (Basic PowerPC Built-in Functions Available on all Configurations):
16758         New subsubsection.
16759         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16760         subsubsection.
16761         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16762         subsubsection.
16763         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16764         subsubsection.
16765         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16766         subsubsection.
16768 2018-05-08  Jim Wilson  <jimw@sifive.com>
16770         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
16771         (LD_EMUL_SUFFIX): New.
16772         (LINK_SPEC): Use it.
16774 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16776         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16777         subsection.
16778         (Basic PowerPC Built-in Functions): The new name of the
16779         subsection previously known as "PowerPC Built-in Functions".
16780         (Basic PowerPC Built-in Functions Available on all Configurations):
16781         New subsubsection.
16782         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16783         subsubsection.
16784         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16785         subsubsection.
16786         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16787         subsubsection.
16788         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16789         subsubsection.
16791 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
16793         PR target/85683
16794         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
16795         after cmpelim optimization.
16797 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
16799         * config.gcc: Support "goldmont".
16800         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
16801         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16802         PROCESSOR_GOLDMONT.
16803         * config/i386/i386.c (m_GOLDMONT): Define.
16804         (processor_target_table): Add "goldmont".
16805         (PTA_GOLDMONT): Define.
16806         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
16807         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
16808         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
16809         (fold_builtin_cpu): Add "goldmont".
16810         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
16811         (ix86_option_override_internal): Add "goldmont".
16812         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
16813         (processor_type): Add PROCESSOR_GOLDMONT.
16814         * config/i386/i386.md: Add CPU "glm".
16815         * config/i386/glm.md: New file.
16816         * config/i386/x86-tune.def: Add m_GOLDMONT.
16817         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
16819 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
16821         PR target/85572
16822         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
16823         E_V4DImode.
16824         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
16825         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
16826         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
16828         PR target/85317
16829         * config/i386/i386.c (ix86_fold_builtin): Handle
16830         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
16832         PR target/85480
16833         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
16834         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
16836 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
16838         PR target/85658
16839         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
16840         (check_arch): Likewise.
16841         (check_fpu): Return the result rather than printing it.
16842         (end arch): Fix operator precedence.
16843         (end cpu): Likewise.
16844         (END): Print the result from check_fpu.
16846 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
16847             Alan Hayward  <alan.hayward@arm.com>
16848             David Sherwood  <david.sherwood@arm.com>
16850         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
16851         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
16852         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
16853         (*fcmuo<mode>_and): New patterns.
16855 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
16857         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
16858         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
16859         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
16860         (cmp_op, sve_imm_con): New code attributes.
16861         (SVE_COND_INT_CMP, imm_con): Delete.
16862         (cmp_op): Remove above unspecs from int attribute.
16863         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
16864         to...
16865         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
16866         comparison-specific unspecs.
16867         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
16868         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
16869         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
16870         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
16871         (*vec_fcm<cmp_op><mode>): Rename to...
16872         (*fcm<cmp_op><mode>): ...this and adjust likewise.
16873         (*vec_fcmuo<mode>): Rename to...
16874         (*fcmuo<mode>): ...this and adjust likewise.
16875         (*pred_fcm<cmp_op><mode>): New pattern.
16876         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
16877         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
16878         functions.
16879         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
16880         and UNORDERED.
16881         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
16882         (aarch64_emit_sve_predicated_cond): New function.
16883         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
16884         (aarch64_emit_unspec_cond_or): Replace with...
16885         (aarch64_emit_sve_or_conds): ...this new function.  Use
16886         aarch64_emit_sve_ptrue_op for the individual comparisons and
16887         aarch64_emit_binop to OR them together.
16888         (aarch64_emit_inverted_unspec_cond): Replace with...
16889         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
16890         aarch64_emit_sve_ptrue_op for the comparison and
16891         aarch64_emit_unop to invert the result.
16892         (aarch64_expand_sve_vec_cmp_float): Update after the above
16893         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
16895 2018-05-07  Nathan Sidwell  <nathan@acm.org>
16897         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
16898         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
16899         (Backwards Compatibility): Likewise.
16901 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16903         PR bootstrap/85681
16904         Revert:
16905         2018-05-07  Luis Machado  <luis.machado@linaro.org>
16907         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16908         <prefetch_dynamic_strides>: New const bool field.
16909         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16910         prefetch_dynamic_strides.
16911         (exynosm1_prefetch_tune): Likewise.
16912         (thunderxt88_prefetch_tune): Likewise.
16913         (thunderx_prefetch_tune): Likewise.
16914         (thunderx2t99_prefetch_tune): Likewise.
16915         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16916         to false.
16917         (aarch64_override_options_internal): Update to set
16918         PARAM_PREFETCH_DYNAMIC_STRIDES.
16919         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16920         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16921         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16922         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16923         prefetch-dynamic-strides setting.
16925         2018-05-07  Luis Machado  <luis.machado@linaro.org>
16927         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16928         <minimum_stride>: New const int field.
16929         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16930         minimum_stride field.
16931         (exynosm1_prefetch_tune): Likewise.
16932         (thunderxt88_prefetch_tune): Likewise.
16933         (thunderx_prefetch_tune): Likewise.
16934         (thunderx2t99_prefetch_tune): Likewise.
16935         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16936         (aarch64_override_options_internal): Update to set
16937         PARAM_PREFETCH_MINIMUM_STRIDE.
16938         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16939         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16940         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16941         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16942         stride is constant and is below the minimum stride threshold.
16944 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16946         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
16947         to 512.
16949 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16951         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16952         <prefetch_dynamic_strides>: New const bool field.
16953         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16954         prefetch_dynamic_strides.
16955         (exynosm1_prefetch_tune): Likewise.
16956         (thunderxt88_prefetch_tune): Likewise.
16957         (thunderx_prefetch_tune): Likewise.
16958         (thunderx2t99_prefetch_tune): Likewise.
16959         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16960         to false.
16961         (aarch64_override_options_internal): Update to set
16962         PARAM_PREFETCH_DYNAMIC_STRIDES.
16963         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16964         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16965         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16966         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16967         prefetch-dynamic-strides setting.
16969 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16971         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16972         <minimum_stride>: New const int field.
16973         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16974         minimum_stride field.
16975         (exynosm1_prefetch_tune): Likewise.
16976         (thunderxt88_prefetch_tune): Likewise.
16977         (thunderx_prefetch_tune): Likewise.
16978         (thunderx2t99_prefetch_tune): Likewise.
16979         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16980         (aarch64_override_options_internal): Update to set
16981         PARAM_PREFETCH_MINIMUM_STRIDE.
16982         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16983         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16984         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16985         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16986         stride is constant and is below the minimum stride threshold.
16988 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
16990         PR c++/85659
16991         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
16992         the type is addressable.  Don't force op into register if it has
16993         BLKmode.
16995 2018-05-05  Roland McGrath  <mcgrathr@google.com>
16997         PR other/77609
16998         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
16999         any section for which we don't know a specific type it should have,
17000         regardless of name.  Previously this was done only for the exact
17001         names ".init_array", ".fini_array", and ".preinit_array".
17002         (default_elf_asm_named_section): Add comment about
17003         relationship with default_section_type_flags and SECTION_NOTYPE.
17004         (get_section): Don't consider it a type conflict if one side has
17005         SECTION_NOTYPE and the other doesn't, as long as neither has the
17006         SECTION_BSS et al used in the default_section_type_flags logic.
17008 2018-05-05  Tom de Vries  <tom@codesourcery.com>
17010         PR target/85653
17011         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
17012         (workaround_barsyncs): New function.
17013         (nvptx_reorg): Use workaround_barsyncs.
17014         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17015         (define_expand "nvptx_membar_cta"): New define_expand.
17016         (define_insn "*nvptx_membar_cta"): New insn.
17018 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
17020         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
17021         To improve optimization opportunities.
17022         * builtin-types.def: The new needed builtin types for the above.
17024 2018-05-04  Richard Biener  <rguenther@suse.de>
17026         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
17027         * gimple-ssa-store-merging.c
17028         (imm_store_chain_info::output_merged_store): Remove redundant create,
17029         release split_store vector contents on failure.
17030         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
17031         scalar stmt vector on cache hit.
17033 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
17035         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
17036         Xilinx FP support.
17037         * config.gcc (powerpc-xilinx-eabi*): Remove.
17038         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
17039         support.
17040         (fusion_addis_mem_combo_load): Ditto.
17041         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
17042         FP support.
17043         (rs6000_cpu_cpp_builtins): Ditto.
17044         * config/rs6000/rs6000-linux.c
17045         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
17046         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
17047         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
17048         support.
17049         (rs6000_setup_reg_addr_masks): Ditto.
17050         (rs6000_init_hard_regno_mode_ok): Ditto.
17051         (rs6000_option_override_internal): Ditto.
17052         (legitimate_lo_sum_address_p): Ditto.
17053         (rs6000_legitimize_address): Ditto.
17054         (rs6000_legitimize_reload_address): Ditto.
17055         (rs6000_legitimate_address_p): Ditto.
17056         (abi_v4_pass_in_fpr): Ditto.
17057         (setup_incoming_varargs): Ditto.
17058         (rs6000_gimplify_va_arg): Ditto.
17059         (rs6000_split_multireg_move): Ditto.
17060         (rs6000_savres_strategy): Ditto.
17061         (rs6000_emit_prologue_components): Ditto.
17062         (rs6000_emit_epilogue_components): Ditto.
17063         (rs6000_emit_prologue): Ditto.
17064         (rs6000_emit_epilogue): Ditto.
17065         (rs6000_elf_file_end): Ditto.
17066         (rs6000_function_value): Ditto.
17067         (rs6000_libcall_value): Ditto.
17068         * config/rs6000/rs6000.h: Ditto.
17069         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
17070         (TARGET_MINMAX): ... this.  New.
17071         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
17072         * config/rs6000/rs6000.md: Remove Xilinx FP support.
17073         (*movsi_internal1_single): Delete.
17074         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
17075         mfpu=, mxilinx-fpu): Delete.
17076         * config/rs6000/singlefp.h: Delete.
17077         * config/rs6000/sysv4.h: Remove Xilinx FP support.
17078         * config/rs6000/t-rs6000: Ditto.
17079         * config/rs6000/t-xilinx: Delete.
17080         * config/rs6000/titan.md: Adjust for fp_type removal.
17081         * config/rs6000/vsx.md: Remove Xilinx FP support.
17082         (VStype_simple): Delete.
17083         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
17084         * config/rs6000/xfpu.h: Delete.
17085         * config/rs6000/xfpu.md: Delete.
17086         * config/rs6000/xilinx.h: Delete.
17087         * config/rs6000/xilinx.opt: Delete.
17088         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
17089         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
17091 2018-05-04  Tom de Vries  <tom@codesourcery.com>
17093         PR libgomp/85639
17094         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
17095         if ignore == 0.
17097 2018-05-04  Richard Biener  <rguenther@suse.de>
17099         PR middle-end/85627
17100         * tree-complex.c (update_complex_assignment): We are always in SSA form.
17101         (expand_complex_div_wide): Likewise.
17102         (expand_complex_operations_1): Likewise.
17103         (expand_complex_libcall): Preserve EH info of the original stmt.
17104         (tree_lower_complex): Handle removed blocks.
17105         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
17106         on complex multiplication and division libcall builtins.
17108 2018-05-04  Richard Biener  <rguenther@suse.de>
17110         PR middle-end/85574
17111         * fold-const.c (negate_expr_p): Restrict negation of operand
17112         zero of a division to when we know that can happen without
17113         overflow.
17114         (fold_negate_expr_1): Likewise.
17116 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
17118         PR libstdc++/85466
17119         * real.h (real_nextafter): Declare.
17120         * real.c (real_nextafter): New function.
17121         * fold-const-call.c (fold_const_nextafter): New function.
17122         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
17123         CASE_CFN_NEXTTOWARD.
17124         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
17125         even when arg1_mode is different from arg0_mode.
17127 2018-05-03  Nathan Sidwell  <nathan@acm.org>
17129         * doc/extend.texi (Deprecated Features): Remove
17130         -ffriend-injection.
17131         (Backwards Compatibility): Likewise.
17132         * doc/invoke.texi (C++ Language Options): Likewise.
17133         (C++ Dialect Options): Likewise.
17135 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
17137         PR target/85530
17138         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
17139         _mm512_mask_mullox_epi64): New intrinsics.
17141 2018-05-03  Tom de Vries  <tom@codesourcery.com>
17143         PR testsuite/85106
17144         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17145         dump files): Add offload-tree.
17147 2018-05-03  Richard Biener  <rguenther@suse.de>
17149         PR tree-optimization/85615
17150         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17151         to loops not nested in BBs loop father to avoid creating multi-entry
17152         loops.
17154 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17156         PR tree-optimization/70291
17157         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
17158         arguments.  Change return type to tree.  Emit libcall as a new
17159         statement rather than replacing existing one when inplace_p is true.
17160         (expand_complex_multiplication_components): New function.
17161         (expand_complex_multiplication): Expand floating-point complex
17162         multiplication using the above.
17163         (expand_complex_division): Rename inner_type parameter to type.
17164         Update expand_complex_libcall call-site.
17165         (expand_complex_operations_1): Update expand_complex_multiplication
17166         and expand_complex_division call-sites.
17168 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17170         PR target/85582
17171         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17172         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17173         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
17174         the highest significant bit of the shift count mask is clear.  In
17175         check whether and[sq]i3 is needed verify that all significant bits
17176         of the shift count other than the highest are set.
17178 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17180         PR libgomp/82428
17181         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
17182         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
17183         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
17184         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
17185         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
17186         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
17187         __builtin_goacc_parlevel_size.
17189 2018-05-02  Richard Biener  <rguenther@suse.de>
17191         PR tree-optimization/85597
17192         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
17193         do not use split vect_get_vec_defs call but call vect_get_slp_defs
17194         directly.
17196 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17198         PR testsuite/85106
17199         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17200         dump files): Add ltrans-tree.
17202 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17204         PR testsuite/85106
17205         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17206         dump files): Add wpa-ipa.
17208 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
17210         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
17211         powerpc*-*-linux*paired* target.
17212         * config/rs6000/750cl.h: Delete.
17213         * config/rs6000/paired.h: Delete.
17214         * config/rs6000/paired.md: Delete.
17215         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
17216         float support.
17217         * config/rs6000/rs6000-builtin.def: Remove paired float support.
17218         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
17219         comment.  Remove paired float support.
17220         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
17221         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
17222         VECTOR_PAIRED.
17223         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
17224         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
17225         declarations.
17226         * config/rs6000/rs6000.c: Remove paired float support.
17227         (paired_expand_vector_init, paired_expand_vector_move,
17228         paired_emit_vector_compare, paired_emit_vector_cond_expr,
17229         (paired_expand_lv_builtin, paired_expand_stv_builtin,
17230         paired_expand_builtin, paired_expand_predicate_builtin,
17231         paired_init_builtins): Delete.
17232         * config/rs6000/rs6000.h: Remove paired float support.
17233         * config/rs6000/rs6000.md: Remove paired float support.
17234         (move_from_CR_ov_bit): Delete.
17235         * config/rs6000/rs6000.opt (mpaired): Delete.
17236         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17237         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17239 2018-05-02  Richard Biener  <rguenther@suse.de>
17241         PR middle-end/85567
17242         * gimplify.c (gimplify_save_expr): When in SSA form allow
17243         SAVE_EXPRs to compute to SSA vars.
17245 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17247         PR target/85582
17248         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17249         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17250         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17251         clobber operands[2], instead use a new pseudo.  Formatting fixes.
17253 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
17255         PR tree-optimization/85586
17256         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17257         exit early for statements in the same group if the accesses are
17258         not strided.
17260 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17262         PR lto/85451
17263         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17264         error message.
17266 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
17268         PR tree-optimization/85143
17269         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17271 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17273         PR lto/85451
17274         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17275         not found" error message.
17277 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17279         PR other/83786
17280         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
17281         * vec.c (test_ordered_remove_if): New function.
17282         (vec_c_tests): Call test_ordered_remove_if.
17283         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
17284         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
17285         * tree-vect-patterns.c (vect_pattern_recog_1): Use
17286         VEC_ORDERED_REMOVE_IF.
17288 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17290         PR tree-optimization/82665
17291         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
17292         pointer subtraction where arguments come from a memchr call.
17294 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
17296         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17297         --push-state --as-needed and --pop-state instead of --as-needed and
17298         --no-as-needed if ld supports it.
17299         * configure: Regenerated.
17301         PR web/85578
17302         * doc/install.texi2html: Replace _002d with - and _002a with * in
17303         generated html files using sed.
17305 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17307         PR c++/85523
17308         * gcc-rich-location.c (blank_line_before_p): New function.
17309         (use_new_line): New function.
17310         (gcc_rich_location::add_fixit_insert_formatted): New function.
17311         * gcc-rich-location.h
17312         (gcc_rich_location::add_fixit_insert_formatted): New function.
17314 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17316         * selftest.c (assert_streq): Rename "expected" and "actual" to
17317         "val1" and "val2".  Extend NULL-handling to cover both inputs
17318         symmetrically, while still requiring both to be non-NULL for a pass.
17319         * selftest.h (assert_streq): Rename "expected" and "actual" to
17320         "val1" and "val2".
17321         (ASSERT_EQ): Likewise.
17322         (ASSERT_EQ_AT): Likewise.
17323         (ASSERT_KNOWN_EQ): Likewise.
17324         (ASSERT_KNOWN_EQ_AT): Likewise.
17325         (ASSERT_NE): Likewise.
17326         (ASSERT_MAYBE_NE): Likewise.
17327         (ASSERT_MAYBE_NE_AT): Likewise.
17328         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
17329         the assertion to pass.
17330         (ASSERT_STREQ_AT): Likewise.
17332 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17334         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17335         interaction with -pie.
17337 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17339         * selftest.h: Fix alphabetization of per-source-file selftest
17340         declarations.
17342 2018-04-30  Jason Merrill  <jason@redhat.com>
17344         PR c++/61982 - dead stores to destroyed objects.
17345         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
17346         of clobber.
17348 2018-04-30  Jason Merrill  <jason@redhat.com>
17350         * tree.c (build_clobber): New.
17351         * tree.h: Declare it.
17352         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
17354 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17356         * diagnostic-show-locus.c (layout::layout): Update for
17357         location_get_source_line returning a char_span.
17358         (struct char_span): Move to input.h.
17359         (struct correction): Update for fields in char_span becoming
17360         private.
17361         (struct source_line): Update for location_get_source_line
17362         returning a char_span.
17363         (layout::print_line): Likewise.
17364         * edit-context.c (edited_file::print_content): Likewise.
17365         (edited_file::print_diff_hunk): Likewise.
17366         (edited_file::print_run_of_changed_lines): Likewise.
17367         (edited_file::get_num_lines): Likewise.
17368         (edited_line::edited_line): Likewise.
17369         * final.c (asm_show_source): Likewise.
17370         * input.c (location_get_source_line): Convert return type
17371         from const char * to char_span, losing the final "line_len"
17372         param.
17373         (dump_location_info): Update for the above.
17374         (get_substring_ranges_for_loc): Likewise.  Use a char_span
17375         when handling the literal within the line.
17376         (test_reading_source_line): Update for location_get_source_line
17377         returning a char_span.
17378         * input.h (class char_span): Move here from
17379         diagnostic-show-locus.c, converting from a struct to a class.
17380         Make data members private.
17381         (char_span::operator bool): New.
17382         (char_span::length): New.
17383         (char_span::get_buffer): New.
17384         (char_span::operator[]): New.
17385         (char_span::subspan): Make const.
17386         (char_span::xstrdup): New.
17387         (location_get_source_line): Convert return type from const char *
17388         to char_span, losing the final "line_size" param.
17390 2018-04-30  Jan Hubicka  <jh@suse.cz>
17392         * lto-wrapper.c (ltrans_priorities): New static var.
17393         (cmp_priority): New.
17394         (run_gcc): Read priorities and if doing parallel build order
17395         the Makefile by them.
17397 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17399         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
17401 2018-04-30  Richard Biener  <rguenther@suse.de>
17403         * tree-cfg.c (verify_address): Remove base argument, add
17404         flag whether to check TREE_ADDRESSABLE and do that.
17405         (verify_expr): Remove.
17406         (verify_types_in_gimple_reference): Add pieces from verify_expr.
17407         (verify_gimple_assign_single): Likewise.
17408         (verify_gimple_switch): Likewise.
17409         (verify_expr_location_1): Dereference tp once.  Add (disabled)
17410         piece from verify_expr.
17411         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
17413 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17415         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17417 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17419         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
17420         (small_data_pattern): Likewise.
17421         (arc_rewrite_small_data): Likewise.
17422         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
17423         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
17424         (get_symbol_alignment): New function.
17425         (legitimate_small_data_address_p): Likewise.
17426         (legitimate_scaled_address): Update, call
17427         legitimate_small_data_address_p.
17428         (output_sdata): New static variable.
17429         (arc_print_operand): Update how we handle small data operands.
17430         (arc_print_operand_address): Likewise.
17431         (arc_legitimate_address_p): Update, use
17432         legitimate_small_data_address_p.
17433         (arc_rewrite_small_data_p): Remove.
17434         (arc_rewrite_small_data_1): Likewise.
17435         (arc_rewrite_small_data): Likewise.
17436         (small_data_pattern): Likewise.
17437         (compact_sda_memory_operand): Update to use
17438         legitimate_small_data_address_p and get_symbol_alignment.
17439         (prepare_move_operands): Don't rewite sdata pattern.
17440         (prepare_extend_operands): Remove.
17441         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
17442         pattern.
17443         (zero_extendqisi2): Likewise.
17444         (zero_extendhisi2): Likewise.
17445         (extendqihi2): Likewise.
17446         (extendqisi2): Likewise.
17447         (extendhisi2): Likewise.
17448         (addsi3): Likewise.
17449         (subsi3): Likewise.
17450         (andsi3): Likewise.
17451         * config/arc/constraints.md (Usd): Change it to memory constraint.
17453 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17455         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
17456         as source of std instructions.
17457         * config/arc/arc.md (movsi_insn): Update pattern predicate to
17458         allow 6-bit constants as source for store instructions.
17459         (movdi_insn): Update instruction pattern to allow 6-bit constants
17460         as source for store instructions.
17462 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17464         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
17466 2018-04-30  Nathan Sidwell  <nathan@acm.org>
17467             Sandra Loosemore <sandra@codesourcery.com>
17469         * dumpfile.c (dump_open): Allow '-' for stdout.
17470         * doc/invoke.texi (Developer Options): Document dump filename
17471         determination early.  Document stdin/stdout selection.
17473 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
17475         Microblaze Target: PIC data text relative
17477         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
17478         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
17479         Add declaration.
17480         * config/microblaze/microblaze.h (microblaze_constant_address_p):
17481         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
17482         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
17483         New addressing mode for data-text relative position indepenedent code.
17484         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
17485         'ADDRESS_SYMBOLIC_TXT_REL'.
17486         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
17487         (microblaze_legitimate_pic_operand): Exclude function calls from
17488         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
17489         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
17490         addresses cases.
17491         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17492         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17493         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
17494         for 'address + offset'.
17495         (microblaze_expand_prologue): Add new function prologue call for
17496         'r20' assignation.
17497         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
17498         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
17499         table in case of TARGET_PIC_DATA_TEXT_REL.
17500         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
17501         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
17502         Add new macros 'UNSPEC_TEXT',
17503         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
17504         + exclude function calls from 'UNSPEC_PLT' in case of data text
17505         relative mode.
17506         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17507         new target hook for generating address diff vector tables in case of
17508         flag_pic.
17509         * doc/tm.texi : Regenerate.
17510         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
17511         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
17512         of addr diff vector generation.
17513         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17514         target hook definition.
17515         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
17516         Add default function for generate_pic_addr_diff_vec -> flag_pic.
17517         * doc/invoke.texi (Add new pic option): Add new microblaze pic
17518         option for data text relative.
17520 2018-04-30  Richard Biener  <rguenther@suse.de>
17522         * tree-chrec.h (evolution_function_is_constant_p): Remove
17523         redundant check.
17524         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
17526 2018-04-30  Richard Biener  <rguenther@suse.de>
17528         PR bootstrap/85571
17529         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
17531 2018-04-30  Richard Biener  <rguenther@suse.de>
17533         PR tree-optimization/28364
17534         PR tree-optimization/85275
17535         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17536         copying first exit test.
17538 2018-04-28  Mark Wielaard  <mark@klomp.org>
17540         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
17541         dwarf_version >= 5.
17542         (dwarf_AT): Handle DW_AT_addr_base.
17543         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
17545 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
17547         PR target/84431
17548         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
17549         (*ashl<dwi>3_doubleword_mask_1): Ditto.
17550         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
17551         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
17553 2018-04-28  Richard Biener  <rguenther@suse.de>
17555         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
17556         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
17557         to reflect use.  Only add interesting stmts.
17559 2018-04-27  Martin Jambor  <mjambor@suse.cz>
17561         PR ipa/85549
17562         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
17563         the jump function allows for passing through aggregate values.
17565 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
17567         * input.h (in_system_header_at): Convert from macro to inline
17568         function.
17569         (from_macro_expansion_at): Likewise.
17570         (from_macro_definition_at): Likewise.
17572 2018-04-27  Jeff Law  <law@redhat.com>
17574         * config.gcc: Mark tile* targets as deprecated/obsolete.
17576 2018-04-27  Richard Biener  <rguenther@suse.de>
17578         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
17579         fix for ILP32.
17581 2018-04-27  Richard Biener  <rguenther@suse.de>
17583         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
17585 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
17587         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
17588         with Yd constraint. Set "preferred_for_speed" attribute from
17589         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
17590         with Yd constraint.
17591         (*movdi_internal): Ditto.
17592         (movti_interunit splitters): Remove
17593         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
17594         (movdi_interunit splitters): Ditto.
17595         * config/i386/constraints.md (Ye): Remove.
17596         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
17598 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17600         PR target/85512
17601         * config/aarch64/constraints.md (Usg): Limit to 31.
17602         (Usj): Limit to 63.
17604 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
17606         PR tree-optimization/85529
17607         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
17608         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
17609         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
17610         zero extension or masking of the MSB bit.
17611         (optimize_range_tests): Add FIRST_BB argument, pass it through
17612         to optimize_range_tests_var_bound.
17613         (maybe_optimize_range_tests, reassociate_bb): Adjust
17614         optimize_range_tests callers.
17616 2018-04-26  Richard Biener  <rguenther@suse.de>
17617             Jakub Jelinek  <jakub@redhat.com>
17619         * cgraph.h (symbol_table): Just declare debug method here.
17620         * symtab.c (symbol_table::debug): Define.
17622 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
17624         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
17626 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
17628         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
17629         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
17630         (*movdi_internal): Substitute Yi and Yj constraint with x
17631         and Ym and Yn constraint with y constraint.  Update "isa"
17632         attribute and set "preferred_for_speed" attribute from
17633         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
17634         (*movsi_internal): Ditto.
17635         (*movdf_internal): Ditto.
17636         (*movsf_internal): Ditto.
17637         (*zero_extendsidi2): Ditto.
17638         * config/i386/sse.md (vec_set<mode>_0): Ditto.
17639         (sse2_loadld): Ditto.
17640         (*vec_extract<ssevecmodelower>_0): Ditto.
17641         (*vec_extractv4si_0_zext_sse4): Ditto.
17642         (vec_concatv2di): Ditto.
17643         (*vec_dup<mode>): Ditto.
17644         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
17645         * config/i386/constraints.md (Yi): Remove.
17646         (Yj): Remove.
17647         (Ym): Remove.
17648         (Yn): Remove.
17650 2018-04-26  Nathan Sidwell  <nathan@acm.org>
17652         * dumpfile.c (dump_open): New.
17653         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
17654         (dump_finish): Detect stdio/stderr by value not name.
17656 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
17658         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
17660 2018-04-26  Tom de Vries  <tom@codesourcery.com>
17662         PR target/84952
17663         * config/nvptx/nvptx.c (verify_neutering_jumps)
17664         (verify_neutering_labels): New function
17665         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
17667 2018-04-26  Tom de Vries  <tom@codesourcery.com>
17669         PR target/84025
17670         * config/nvptx/nvptx.c (needs_neutering_p): New function.
17671         (nvptx_single): Use needs_neutering_p to skip over insns that do not
17672         need neutering.
17674 2018-04-26  Richard Biener <rguenther@suse.de>
17675             Tom de Vries  <tom@codesourcery.com>
17677         PR lto/85422
17678         * lto-streamer-out.c (output_function): Fixup loops if required to match
17679         discovery done in the reader.
17681 2018-04-26  Richard Biener  <rguenther@suse.de>
17683         PR tree-optimization/85116
17684         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
17685         have a loop exit from the single latch predecessor.  Remove
17686         case of header with just condition.
17687         (ch_base::copy_headers): Exclude infinite loops from any
17688         processing.
17689         (pass_ch::execute): Record exits.
17691 2018-04-26  Richard Biener  <rguenther@suse.de>
17693         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
17694         prologue cost vector and pass it to vect_get_load_cost.
17695         (vect_get_peeling_costs_all_drs): Likewise.
17696         (vect_peeling_hash_get_lowest_cost): Likewise.
17697         (vect_enhance_data_refs_alignment): Likewise.
17699 2018-04-26  Richard Biener  <rguenther@suse.de>
17701         PR middle-end/85450
17702         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
17703         checking of integer<->pointer conversions.
17704         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
17705         sign-/zero-extending pointer types.
17706         (expand_omp_for_static_chunk): Likewise.
17708 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
17709             Jean Lee  <xiaoyur347@gmail.com>
17711         * config/mips/mips.c (mips_asan_shadow_offset): New function.
17712         (TARGET_ASAN_SHADOW_OFFSET): Define.
17713         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
17714         true for -fsanitize=address.
17716 2018-04-25  Mark Wielaard  <mark@klomp.org>
17718         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
17719         shorter ones.
17721 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
17723         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
17724         than "alu", remove explicit "memory" and "imm_disp" attributes.
17725         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
17727         PR middle-end/85414
17728         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
17729         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
17730         gen_lowpart_no_emit.
17732 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
17734         PR target/85473
17735         * config/i386/i386.c (ix86_expand_builtin): Change memory
17736         operand to XI, extend p0 to Pmode.
17737         * config/i386/i386.md: Change unspec volatile and operand
17738         1 mode to XI, change operand 0 mode to P.
17740 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17742         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
17743         GET_MODE_MASK before any checking.
17744         (nds32_can_use_bset_p): Likewise.
17745         (nds32_can_use_btgl_p): Likewise.
17747 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17749         * config/nds32/nds32-doubleword.md: New define_split pattern for
17750         illegal register number.
17752 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17754         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
17756 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17758         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
17760 2018-04-25  Richard Biener  <rguenther@suse.de>
17762         * lto-streamer.h (LTO_major_version): Bump to 8.
17764 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
17766         * BASE-VER: Set to 9.0.0.
17768 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
17770         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
17771         in __abskf2 and __powikf2.
17773 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17775         PR target/85512
17776         * config/aarch64/constraints.md (Usg, Usj): New constraints.
17777         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
17778         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
17779         Use the above on operand 2.  Reindent.
17780         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
17782 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
17784         PR target/85485
17785         * common/config/i386/i386-common.c (ix86_handle_option): Don't
17786         handle OPT_mcet.
17787         * config/i386/i386.opt (mcet): Removed.
17788         * doc/install.texi: Remove -mcet documentation.
17789         * doc/invoke.texi: Likewise.
17791 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
17793         PR target/85485
17794         * doc/install.texi: Remove -mcet from bootstrap-cet.
17796 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17798         PR target/85511
17799         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
17800         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
17801         if TARGET_64BIT.
17803         PR target/85503
17804         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
17805         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
17806         containing a CONST_VECTOR.
17808 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
17810         * doc/install.texi: Update newlib dependency for nvptx.
17812 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17814         PR target/85508
17815         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
17816         instead of INTVAL when shifting x left.
17818 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
17820         PR tree-optimization/85478
17821         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
17822         vect_grouped_store_supported for single element vectors.
17824 2018-04-24  Richard Biener  <rguenther@suse.de>
17826         PR target/85491
17827         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
17828         load cost increase to the case of non-constant step.
17830 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17832         PR target/84828
17833         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
17834         destination if any_malformed_asm.
17836 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
17838         PR middle-end/85496
17839         * expr.c (store_field): In the bitfield case, if the value comes from
17840         a function call and is returned in registers by means of a PARALLEL,
17841         do not change the mode of the temporary unless BLKmode and VOIDmode.
17843 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
17845         PR rtl-optimization/85423
17846         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
17847         dependencies to debug insns when the previous insn is non-debug.
17849 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
17851         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
17852         enums into a single definition.
17853         (fls): Fix predicates and printing.
17854         (seti): Likewise.
17856 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
17858         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
17859         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
17860         and short u6 immediate.
17861         (check_if_valid_sleep_operand): Remove.
17862         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
17864 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17866         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
17867         flag_always_save_lp condition.
17868         * config/nds32/nds32.opt (malways-save-lp): New option.
17870 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17872         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
17873         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
17874         * config/nds32/nds32.h
17875         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
17876         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
17878 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17880         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
17881         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
17883 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17884             Chung-Ju Wu  <jasonwucj@gmail.com>
17886         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
17887         Declare.
17888         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
17889         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
17891 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17893         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
17895 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17897         * config/nds32/nds32-protos.h (nds32_data_alignment,
17898         nds32_local_alignment): Declare.
17899         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
17900         nds32_local_alignment): New functions.
17901         (TARGET_CONSTANT_ALIGNMENT): Define.
17902         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
17904 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17906         * config/nds32/nds32.c
17907         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
17908         (TARGET_MODES_TIEABLE_P): Likewise.
17910 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17912         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
17913         level Ofast and Og.
17915 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
17916             Chung-Ju Wu  <jasonwucj@gmail.com>
17918         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17919         for unaligned access.
17920         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17921         unaligned access.
17922         * config/nds32/nds32-intrinsic.md: Likewise.
17923         * config/nds32/nds32_intrinsic.h: Likewise.
17924         * config/nds32/nds32.h (nds32_builtins): Likewise.
17925         * config/nds32/nds32.opt (munaligned-access): New option.
17926         * config/nds32/nds32.c (nds32_asm_file_start): Display
17927         flag_unaligned_access status.
17929 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
17931         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
17932         -mno-relax is present.
17933         * config/riscv/linux.h (LINK_SPEC): Ditto.
17935 2018-04-20  Martin Sebor  <msebor@redhat.com>
17937         PR c/85365
17938         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
17939         for null pointers.
17940         (gimple_fold_builtin_stxcpy_chk): Same.
17941         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
17943 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
17945         PR target/85456
17946         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
17947         __powikf2 when long double is IEEE 128-bit.
17949 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
17951         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
17952         step to make sure stack always aligned.
17954 2018-04-20  Carl Love  <cel@us.ibm.com>
17956         PR target/83402
17957         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
17958         size check for arg0.
17960 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
17961             Tom de Vries  <tom@codesourcery.com>
17963         PR target/85445
17964         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17965         Emit insns for calls too.
17966         (nvptx_find_par): Always look for worker-level predecessor insn.
17967         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
17968         calls.
17969         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
17970         (nvptx_process_pars): Propagate frames for calls.
17972 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
17974         PR target/85469
17975         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
17976         Removed.
17977         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
17978         (ix86_handle_option): Don't handle OPT_mibt.
17979         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
17980         __SHSTK__.
17981         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
17982         has_ibt and ibt.
17983         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
17984         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
17985         (ix86_target_macros): Define __CET__ with flag_cf_protection
17986         for -fcf-protection.
17987         * config/i386/i386.c (isa2_opts): Remove -mibt.
17988         * config/i386/i386.h (TARGET_IBT): Removed.
17989         (TARGET_IBT_P): Likewise.
17990         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
17991         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
17992         * config/i386/i386.opt (mcet): Update help message.
17993         (mshstk): Likewise.
17994         (mibt): Removed.
17995         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
17996         -mcet as an alias for -mshstk.
17998 2018-04-20  Richard Biener <rguenther@suse.de>
18000         PR middle-end/85475
18001         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
18002         complexity by forcing a single use of the multiply operand.
18004 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18006         ipa/85449
18007         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
18008         recursion dependency to only apply to non-clones.
18010 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18012         ipa/85447
18013         * ipa-cp.c (create_specialized_node): Check that clones of
18014         self-recursive edges exist during IPA-CP.
18016 2018-04-19  Toon Moene  <toon@moene.org>
18018         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
18019         by -O3.
18021 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18023         PR tree-optimization/85467
18024         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
18025         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
18026         VECTOR_CST element to type.
18028 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18030         PR target/85397
18031         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
18032         * config/i386/i386.md (builtin_setjmp_setup): Removed.
18033         (builtin_longjmp): Likewise.
18034         (save_stack_nonlocal): New pattern.
18035         (restore_stack_nonlocal): Likewise.
18037 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18039         PR target/85404
18040         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18041         Replace ASM_OUTPUT_LABEL with fprintf.
18043 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18045         PR target/85417
18046         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18047         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
18048         * config/i386/i386-c.c (ix86_target_macros_internal): Also
18049         define __IBT__ and __SHSTK__ for -fcf-protection.
18050         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
18051         TARGET_IBT.
18052         (ix86_trampoline_init): Likewise.
18053         (x86_output_mi_thunk): Likewise.
18054         (ix86_notrack_prefixed_insn_p): Likewise.
18055         (ix86_option_override_internal): Don't disallow -fcf-protection.
18056         * config/i386/i386.md (rdssp<mode>): Also enable for
18057         -fcf-protection.
18058         (incssp<mode>): Likewise.
18059         (nop_endbr): Likewise.
18060         * config/i386/i386.opt (mcet): Change help message to built-in
18061         functions only.
18062         (mibt): Likewise.
18063         (mshstk): Likewise.
18064         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
18065         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
18066         enable CET built-in functions.
18068 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
18070         * common/config/i386/i386-common.c
18071         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
18072         OPTION_MASK_ISA_MOVDIRI_UNSET,
18073         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
18074         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
18075         * config.gcc (movdirintrin.h): New header.
18076         * config/i386/cpuid.h (bit_MOVDIRI,
18077         bit_MOVDIR64B): New bits.
18078         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
18079         and -mmvodir64b.
18080         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
18081         (VOID, PVOID, PCVOID)): New function types.
18082         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
18083         __builtin_ia32_directstoreu_u64,
18084         __builtin_ia32_movdir64b): New builtins.
18085         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
18086         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
18087         and -mmovdiri.
18088         (ix86_valid_target_attribute_inner_p): Ditto.
18089         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
18090         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
18091         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
18092         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
18093         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
18094         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
18095         (movdiri<mode>, movdir64b_<mode>): New.
18096         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
18097         * config/i386/immintrin.h: Include movdirintrin.h.
18098         * config/i386/movdirintrin.h: New file.
18099         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
18101 2018-04-19  Richard Biener  <rguenther@suse.de>
18103         PR middle-end/85455
18104         * cfg.c (clear_bb_flags): When loop state says we have
18105         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
18107 2018-04-19  Richard Biener  <rguenther@suse.de>
18109         PR tree-optimization/84737
18110         * tree-vect-data-refs.c (vect_copy_ref_info): New function
18111         copying restrict info.
18112         (vect_setup_realignment): Use it.
18113         * tree-vectorizer.h (vect_copy_ref_info): Declare.
18114         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
18115         the first DR to all generated stores.
18116         (vectorizable_load): Likewise for loads.
18118 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18120         PR tree-optimization/85446
18121         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
18122         the integral and pointer types to have the same precision.
18124         * doc/install.texi: Document --disable-cet being the default and
18125         --enable-cet=auto.
18127 2018-04-18  Martin Liska  <mliska@suse.cz>
18129         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
18130         style.
18132 2018-04-18  Martin Liska  <mliska@suse.cz>
18134         Revert
18135         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18137         PR ipa/83983
18138         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18139         arguments if they are comparable.
18141 2018-04-18  Martin Liska  <mliska@suse.cz>
18143         Revert
18144         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
18146         PR lto/84805
18147         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18148         incomplete types.
18150 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
18152         PR target/85388
18153         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
18154         ENDBR after calling __morestack.
18156 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
18158         PR jit/85384
18159         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
18160         by using gcc_base_ver to generate a gcc_driver_version, and use
18161         it when generating GCC_DRIVER_NAME.
18162         * configure: Regenerate.
18164 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
18166         PR target/81084
18167         * config.gcc: Obsolete powerpc*-*-*spe*.
18169 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18171         PR debug/84637
18172         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
18173         (stabstr_D): Change type of unum from unsigned int to
18174         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
18175         type.
18177 2018-04-17  Jim Wilson  <jimw@sifive.com>
18179         PR 84856
18180         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
18181         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
18182         Set arg_pointer_offset after using pretend_args_size.
18184 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18186         PR rtl-optimization/85431
18187         * dse.c (record_store): Ignore zero width stores.
18189         PR sanitizer/85230
18190         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
18191         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
18192         __builtin_stack_restore rather than after it.
18193         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18194         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
18195         argument instead of virtual_dynamic_stack_rtx.
18197 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18199         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
18200         New prototype.
18201         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
18202         Add note to error message to explain internal mapping of overloaded
18203         built-in function name to non-overloaded built-in function name.
18204         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
18205         function.
18207 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
18209         PR target/85424
18210         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
18211         where the inputs overlap with the output.
18213 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18215         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
18216         (=v, v) alternative and explicit "memory" attribute.
18217         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
18218         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18219         attributes.
18220         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
18221         "sselog1" type instead of "sselog".
18222         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
18223         "sselog".  Remove explicit "memory" attribute.
18224         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
18225         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18226         attributes.
18227         (vec_extract_hi_v32hi): Merge all alternatives into one, use
18228         "sselog1" type instead of "sselog".  Remove explicit "memory"
18229         attribute.
18230         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
18231         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18232         attribute.
18233         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18234         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18235         attributes.
18236         (vec_extract_hi_v64qi): Merge all alternatives into one, use
18237         "sselog1" type instead of "sselog".  Remove explicit "memory"
18238         attribute.
18239         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18240         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18241         attribute.
18243         PR target/85430
18244         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18246         PR middle-end/85414
18247         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18248         on a SUBREG.
18250 2018-04-17  Martin Jambor  <mjambor@suse.cz>
18252         PR ipa/85421
18253         * ipa-cp.c (create_specialized_node): Call
18254         expand_all_artificial_thunks if necessary.
18256 2018-04-17  Martin Liska  <mliska@suse.cz>
18258         PR lto/85405
18259         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18260         in message, remote space in between '_G' and '('.
18262 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18264         PR target/85281
18265         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18266         avx512f_vmcmp<mode>3<round_saeonly_name>,
18267         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18268         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18269         avx512f_rndscale<mode><round_saeonly_name>,
18270         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18271         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18272         Use %<iptr>2 instead of %2 for -masm=intel.
18273         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18274         avx512f_vcvttss2usi<round_saeonly_name>,
18275         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18276         -masm=intel.
18277         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18278         avx512f_vcvttsd2usi<round_saeonly_name>,
18279         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18280         Use %q1 instead of %1 for -masm=intel.
18281         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
18282         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
18283         of %3 for -masm=intel.
18284         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
18285         -masm=intel.
18286         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
18287         -masm=intel.
18288         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
18289         -masm=intel.
18290         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
18291         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18292         %g1.
18293         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18294         -masm=intel.
18295         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18296         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18297         %g1 and one with %0 and %1.
18298         (avx512er_vmrcp28<mode><round_saeonly_name>,
18299         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18300         %1 for -masm=intel.
18301         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18302         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18303         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18304         of %0 and %{%4%} for -masm=intel.
18305         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18306         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18307         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18308         order of %0 and %{%5%}%{z%} for -masm=intel.
18310 2018-04-17  Jan Hubicka  <jh@suse.cz>
18312         PR lto/85405
18313         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18315 2018-04-17  Martin Liska  <mliska@suse.cz>
18317         PR ipa/85329
18318         * multiple_target.c (create_dispatcher_calls): Set apostrophes
18319         for target_clone error message.  Make default implementation
18320         clone to be a local declaration.
18321         (separate_attrs): Add new argument and check for an empty
18322         string.
18323         (expand_target_clones): Handle it.
18324         (ipa_target_clone): Make redirection just for target_clones
18325         functions.
18327 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
18328             Tom de Vries  <tom@codesourcery.com>
18330         PR middle-end/84955
18331         * omp-expand.c (expand_oacc_for): Add dummy false branch for
18332         tiled basic blocks without omp continue statements.
18334 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
18336         PR target/83660
18337         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
18338         vec_extract expression as having side effects to make sure it gets
18339         a cleanup point.
18341 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
18343         PR target/85403
18344         * config/i386/i386.c (get_builtin_code_for_version): Check
18345         error_mark_node.
18347 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
18349         PR target/84331
18350         * config.gcc: Support "skylake".
18351         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18352         PROCESSOR_SKYLAKE.
18353         * config/i386/i386.c (m_SKYLAKE): Define.
18354         (processor_target_table): Add "skylake".
18355         (ix86_option_override_internal): Add "skylake".
18356         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
18357         PROCESSOR_CANNONLAKE.
18358         (get_builtin_code_for_version): Fix priority for
18359         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
18360         PROCESSOR_SKYLAKE-AVX512.
18361         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
18362         (processor_type): Add PROCESSOR_SKYLAKE.
18364 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
18365             Jason Merrill  <jason@redhat.com>
18367         PR c++/85112
18368         * convert.c (convert_to_integer_1): Use direct recursion for
18369         enumeral types and types with a precision less than the number
18370         of bits in their mode.
18372 2018-04-16  Julia Koval  <julia.koval@intel.com>
18374         PR target/84413
18375         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
18376         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
18378 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
18380         PR target/85293
18381         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
18382         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
18383         and -mno-direct-move.
18385 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
18387         PR target/83402
18388         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
18389         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
18390         Ensure negative shifts result in {0}.
18392 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
18394         PR rtl-optimization/79916
18395         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
18396         regs (if any) to define how to gnerate SD moves when LRA is in
18397         progress.
18399 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
18401         PR rtl-optimization/85393
18402         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
18403         * except.c (expand_dw2_landing_pad_for_region): Make static.
18404         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
18405         a label and unconditional jump to old_bb, rather than
18406         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
18407         basic block.
18409         PR rtl-optimization/85376
18410         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
18411         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
18412         instead of a specific value.
18414 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18415             Bin Cheng  <bin.cheng@arm.com>
18417         PR tree-optimization/82965
18418         PR tree-optimization/83991
18419         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
18420         by_profile_only parameter.
18421         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
18422         information if the loop was predicted to iterate too many times.
18423         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
18425 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18427         PR lto/71991
18428         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
18429         always inline.
18431 2018-04-13  Martin Liska  <mliska@suse.cz>
18432             Jakub Jelinek  <jakub@redhat.com>
18434         PR middle-end/81657
18435         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
18436         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
18437         * builtins.c (expand_builtin_memory_copy_args): Use
18438         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
18439         handle dest_addr == pc_rtx.
18441 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
18443         PR target/85291
18444         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
18445         asked to not generate direct moves.
18446         (fix_trunc<mode>si2_stfiwx): Similar.
18447         (fix_trunc<mode>si2_internal): Similar.
18449 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18451         PR debug/83157
18452         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
18453         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
18454         lookup if dest in some wider mode is known to be const0_rtx and
18455         if so, record permanent equivalence for it to be ZERO_EXTEND of
18456         the narrower mode destination.
18458 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18460         * lto-streamer-out.c (output_function): Revert 259346.
18461         * omp-expand.c (expand_oacc_for): Likewise.
18463 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
18465         PR rtl-optimization/85354
18466         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
18467         * sel-sched.c (sel_global_init): ... here.
18469 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
18471         PR target/85238
18472         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
18473         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
18474         mode for PE-COFF targets.
18475         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
18476         (i386_pe_asm_lto_end): Likewise.
18477         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
18478         (TARGET_ASM_LTO_END): Likewise.
18479         * config/i386/winnt.c (saved_debug_info_level): New static variable.
18480         (i386_pe_asm_lto_start): New function.
18481         (i386_pe_asm_lto_end): Likewise.
18483 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18484             Richard Biener  <rguenther@suse.de>
18486         PR middle-end/84955
18487         * lto-streamer-out.c (output_function): Fix CFG loop state before
18488         streaming out.
18489         * omp-expand.c (expand_oacc_for): Handle calls to internal
18490         functions like regular functions.
18492 2018-04-12  Richard Biener  <rguenther@suse.de>
18494         PR lto/85371
18495         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
18496         for the early LTO debug to properly generate references to it
18497         during DIE emission.  Do not re-use that for the skeleton for
18498         split-dwarf.
18499         (dwarf2out_early_finish): Likewise.
18501 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18503         PR target/85328
18504         * config/i386/sse.md
18505         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
18506         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
18507         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
18508         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
18509         and output is a reg, avoid creating invalid lowpart subreg, but
18510         instead split into a 512-bit move.  Don't split if not AVX512VL,
18511         input is xmm16+ reg and output is a mem.
18512         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
18513         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
18514         xmm16+ reg and output is a mem.
18516 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18518         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
18519         also for flag_dwarf2_cfi_asm.
18521 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18523         PR rtl-optimization/85342
18524         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
18525         a bool scalar var inside of the loop instead.  Don't try to update
18526         recog_data.operand after failed apply_change_group.
18528 2018-04-12  Tom de Vries  <tom@codesourcery.com>
18530         PR target/85296
18531         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
18532         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
18533         array with flexible array member as array without given dimension.
18534         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
18535         argument for undefined param to true.
18537 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
18539         PR target/85321
18540         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18541         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
18542         from PowerPC section.
18543         * config/rs6000/sysv4.opt (mcall-): Improve help text.
18544         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
18545         help text that is too long.
18546         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
18547         help text that is too long.
18548         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
18549         help text that is too long.
18551 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
18553         * config/alpha/alpha.md (stack_probe_internal): Rename
18554         from "probe_stack".  Update all callers.
18556 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18558         PR rtl-optimization/84566
18559         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
18560         sched_macro_fuse_insns.
18562 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18564         PR target/84301
18565         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
18566         (compute_block_dependences): ... from here.
18568 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18570         PR tree-optimization/85331
18571         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
18572         from int to HOST_WIDE_INT.
18574 2018-04-11  Martin Jambor  <mjambor@suse.cz>
18576         PR ipa/84149
18577         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
18578         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
18579         not the same as the source val.
18580         (cgraph_edge_brings_value_p): New parameter.
18581         (gather_edges_for_value): Pass destination value to
18582         cgraph_edge_brings_value_p.
18583         (perhaps_add_new_callers): Likewise.
18584         (get_info_about_necessary_edges): Likewise and exclude values brought
18585         only by self-recursive edges.
18586         (create_specialized_node): Redirect only clones of self-calling edges.
18587         (+self_recursive_pass_through_p): New function.
18588         (find_more_scalar_values_for_callers_subset): Use it.
18589         (find_aggregate_values_for_callers_subset): Likewise.
18590         (known_aggs_to_agg_replacement_list): Removed.
18591         (decide_whether_version_node): Re-calculate known constants for all
18592         remaining context clones.
18594 2018-04-11  Richard Biener  <rguenther@suse.de>
18596         PR lto/85339
18597         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
18598         from early DWARF output.
18599         (dwarf2out_early_finish): Output line info unconditionally into
18600         early DWARF and add reference to it.
18602 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18604         PR target/85281
18605         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
18606         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
18607         other than V2DFmode using iptr mode attribute.
18608         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
18610 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18612         PR rtl-optimization/84659
18613         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
18615 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18617         PR debug/85302
18618         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
18619         SIZEP is NULL.
18620         (output_loc_list): Pass address of a dummy size variable even in the
18621         locview handling loop.
18622         (index_location_lists): Add comment on why skip_loc_list_entry can't
18623         call size_of_locs.
18625 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18627         PR target/85261
18628         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
18629         into register.
18631 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
18633         PR target/85321
18634         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18635         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
18636         and -mstring-compare-inline-limit.
18638 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18640         PR target/85287
18641         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
18642         for stack clash protection in a register whenever we need it to be in
18643         a register.
18645 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18647         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
18648         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
18650 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18652         PR target/85321
18653         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
18654         the help text.
18655         (mlong-double-): Ditto.
18656         * config/rs6000/sysv4.opt (msdata=): Ditto.
18657         (mtls-size=): Ditto.
18659 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18661         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18662         erroneous entries for
18663         "vector int vec_ldl (int, long int *)", and
18664         "vector unsigned int vec_ldl (int, unsigned long int *)".
18665         Add comments and entries for
18666         "vector bool char vec_ldl (int, bool char *)",
18667         "vector bool short vec_ldl (int, bool short *)",
18668         "vector bool int vec_ldl (int, bool int *)",
18669         "vector bool long long vec_ldl (int, bool long long *)",
18670         "vector pixel vec_ldl (int, pixel *)",
18671         "vector long long vec_ldl (int, long long *)",
18672         "vector unsigned long long vec_ldl (int, unsigned long long *)".
18673         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
18674         type tree bool_long_long_type_node and correct definition of
18675         bool_V2DI_type_node to make reference to this new type tree.
18676         (rs6000_mangle_type): Replace erroneous reference to
18677         bool_long_type_node with bool_long_long_type_node.
18678         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
18679         comments to emphasize sign distinctions for char and int types and
18680         replace RS6000_BTI_bool_long constant with
18681         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
18682         use of RS6000_BTI_pixel.
18683         (bool_long_type_node): Remove this macro definition.
18684         (bool_long_long_type_node): New macro definition
18686 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
18688         PR rtl-optimization/85300
18689         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
18690         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
18691         simplify_unary_operation fails.
18693 2018-04-10  Martin Liska  <mliska@suse.cz>
18695         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
18696         cgraph_edge and ipa_ref.
18698 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
18700         PR target/85177
18701         PR target/85255
18702         * config/i386/sse.md
18703         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
18704         computation of the VEC_MERGE selector from mask.
18705         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
18706         Fix decoding of the VEC_MERGE selector into mask.
18708 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
18710         PR tree-optimization/85286
18711         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
18713 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
18715         * final.c (final_1): Set insn_last_address as well as
18716         insn_current_address.
18718 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18720         PR target/85173
18721         * explow.c (emit_stack_probe): Call validize_mem on memory location
18722         before passing it to gen_probe_stack.  Create address operand and
18723         legitimize it for the probe_stack_address case.
18725 2018-04-09  Jan Hubicka  <jh@suse.cz>
18727         PR lto/85078
18728         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
18729         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
18730         * tree.c (free_lang_data_in_type): Fix handling of binfos;
18731         walk basetypes.
18732         (free_lang_data): Rebuild type inheritance graph.
18734 2018-04-09  Martin Sebor  <msebor@redhat.com>
18736         * invoke.texi (-finline-small-functions): Mention other optimization
18737         options.
18738         (-findirect-inlining, -fpartial-inlining): Same.
18739         (-finline-functions-called-once): Same.
18740         (-freorder-blocks-and-partition): Same.
18742 2018-04-09  Jan Hubicka  <jh@suse.cz>
18744         PR rtl/84058
18745         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
18746         jumps; choose last target that matches the criteria (i.e.
18747         no partition changes for non-crossing jumps).
18748         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
18749         support for redirecting crossing jumps to non-crossing.
18751 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
18753         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
18754         also for naked functions.
18756 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
18758         * config/arc/arc.md (add_shift): New pattern.
18759         (add_shift2): Likewise.
18760         (sub_shift): Likewise.
18761         (sub_shift_cmp0_noout): Likewise.
18762         (compare_si_ashiftsi): Likewise.
18763         (xbfu_cmp0_noout): New combine pattern.
18764         (xbfu_cmp0"): Likewise.
18765         (movsi_set_cc_insn): Place the predicable variant first.
18766         (commutative_binary_cmp0_noout): Remove clobber.
18767         (commutative_binary_cmp0): New pattern.
18768         (noncommutative_binary_cmp0): Likewise.
18769         (noncommutative_binary_cmp0_noout): Likewise.
18770         (noncommutative_binary_comparison_result_used): Removed.
18771         (rsub_cmp0): New pattern.
18772         (rsub_cmp0_noout): Likewise.
18773         (extzvsi): Changed, keep only meaningful variants.
18774         (SQH, SEZ): New iterators.
18775         (SQH_postfix): New mode attribute.
18776         (SEZ_prefix): New code attribute.
18777         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
18778         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
18779         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
18780         of numerical value.
18781         (noncommutative_operator): Check the availability of barrel
18782         shifter option.
18784 2018-04-09  Richard Biener  <rguenther@suse.de>
18786         PR tree-optimization/85284
18787         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
18788         Only use the niter constraining form of simple_iv when the exit
18789         is always executed.
18791 2018-04-09  Tom de Vries  <tom@codesourcery.com>
18793         PR target/84041
18794         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18795         (define_expand "*memory_barrier"): New define_expand.
18796         (define_insn "memory_barrier"): New insn.
18798 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18800         PR rtl-optimization/80463
18801         PR rtl-optimization/83972
18802         PR rtl-optimization/83480
18804         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
18805         correct producer for the insn.
18806         (tidy_control_flow): Fixup seqnos in case of debug insns.
18808 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18810         PR rtl-optimization/83913
18812         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
18813         different sched-times when merging exprs.
18815 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18817         PR rtl-optimization/83962
18819         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
18820         tidy_fallthru_edge and tidy_control_flow.
18822 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18824         PR rtl-optimization/83530
18826         * sel-sched.c (force_next_insn): New global variable.
18827         (remove_insn_for_debug): When force_next_insn is true, also leave only
18828         next insn in the ready list.
18829         (sel_sched_region): When the region wasn't scheduled, make another pass
18830         over it with force_next_insn set to 1.
18832 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
18834         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
18835         into tm_file.
18836         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18837         for interrupt control.
18838         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18839         functions for interrupt control.
18840         * config/nds32/nds32-intrinsic.md: Likewise.
18841         * config/nds32/nds32_intrinsic.h: Likewise.
18842         * config/nds32/nds32.h (nds32_builtins): Likewise.
18844 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
18846         * config/nds32/nds32.c (nds32_init_machine_status,
18847         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
18848         strict_aligned_p field.
18849         (nds32_expand_to_rtl_hook): New function.
18850         (TARGET_EXPAND_TO_RTL_HOOK): Define.
18851         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
18853 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18854             Chung-Ju Wu  <jasonwucj@gmail.com>
18856         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
18857         * config/nds32/nds32-n7.md: New file.
18858         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
18859         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
18860         pipeline.
18861         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
18862         * config/nds32/nds32.md (pipeline_model): Add n7.
18863         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
18864         * config/nds32/pipelines.md: Include n7 settings.
18866 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18867             Chung-Ju Wu  <jasonwucj@gmail.com>
18869         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
18870         * config/nds32/nds32-e8.md: New file.
18871         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
18872         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
18873         pipeline.
18874         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
18875         * config/nds32/nds32.md (pipeline_model): Add e8.
18876         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
18877         * config/nds32/pipelines.md: Include e8 settings.
18879 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18880             Chung-Ju Wu  <jasonwucj@gmail.com>
18882         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
18883         * config/nds32/nds32-n8.md: New file.
18884         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
18885         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
18886         pipeline.
18887         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
18888         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
18889         * config/nds32/nds32.md (pipeline_model): Add n8.
18890         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
18891         * config/nds32/pipelines.md: Include n8 settings.
18893 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18894             Chung-Ju Wu  <jasonwucj@gmail.com>
18896         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
18897         * config/nds32/nds32-n9-2r1w.md: New file.
18898         * config/nds32/nds32-n9-3r2w.md: New file.
18899         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
18900         nds32_register_ports): New or modify for cpu n9.
18901         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
18902         pipeline.
18903         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
18904         * config/nds32/nds32-utils.c: New file.
18905         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
18906         TARGET_MUL_SLOW): Define.
18907         * config/nds32/nds32.md (pipeline_model): New attribute.
18908         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
18909         New options that support cpu n9.
18910         * config/nds32/pipelines.md: Include n9 settings.
18911         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
18913 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
18915         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
18916         information if necessary.
18917         (output_cond_branch_compare_zero): Likewise.
18918         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
18919         (nds32_target_alignment): Refine for alignment.
18920         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
18921         (FUNCTION_BOUNDARY): Modify.
18922         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
18923         align case.
18924         * config/nds32/nds32.opt (malways-align, malign-functions): New.
18926 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
18928         * config/nds32/constants.md (unspec_volatile_element): Add values for
18929         TLB operation and data prefetch.
18930         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18931         functions for TLB operation and data prefetch.
18932         * config/nds32/nds32-intrinsic.md: Likewise.
18933         * config/nds32/nds32_intrinsic.h: Likewise.
18934         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
18935         (nds32_print_operand): Likewise.
18936         * config/nds32/nds32.h (nds32_builtins): Likewise.
18938 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
18939         Andrew Pinski <pinsika@gcc.gnu.org>
18941         PR middle-end/82976
18942         * match.pd: Use constant_boolean_node of correct type instead of
18943         boolean_true_node or boolean_false_node for simplifying
18944         pointer comparisons to zero.
18946 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
18948         PR tree-optimization/80021
18949         * tree.c (verify_type_variant): Make error call in verify_variant_match
18950         translatable and remove final full stop.
18952 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18954         * config/nds32/constants.md (unspec_volatile_element): Add
18955         UNSPEC_VOLATILE_EH_RETURN.
18956         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
18957         nds32_output_stack_pop): Support dwarf exception handling process.
18958         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
18959         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
18960         exception handling process.
18961         (nds32_compute_stack_frame): Likewise.
18962         (nds32_return_addr_rtx): Likewise.
18963         (nds32_initial_elimination_offset): Likewise.
18964         (nds32_expand_prologue): Likewise.
18965         (nds32_expand_epilogue): Likewise.
18966         (nds32_dynamic_chain_address): New function.
18967         * config/nds32/nds32.h (machine_function): Add fields for dwarf
18968         exception handling.
18969         (DYNAMIC_CHAIN_ADDRESS): Define.
18970         (EH_RETURN_DATA_REGNO): Define.
18971         (EH_RETURN_STACKADJ_RTX): Define.
18972         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
18973         patterns for dwarf exception handling.
18975 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18977         * config/nds32/nds32.h: Clean up obsolete macros.
18979 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18981         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18982         Add enum values for particular instructions.
18983         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18984         particular intrinsic functions.
18985         * config/nds32/nds32-intrinsic.md: Likewise.
18986         * config/nds32/nds32_intrinsic.h: Likewise.
18987         * config/nds32/nds32.h (nds32_builtins): Likewise.
18988         * config/nds32/nds32.md (type): Add pbsad and pbsada.
18989         (btst, ave): New patterns for particular instructions.
18991 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18993         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18994         Add enum values for atomic load/store and memory sync.
18995         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
18996         and memory sync.
18997         * config/nds32/nds32-intrinsic.md: Likewise.
18998         * config/nds32/nds32_intrinsic.h: Likewise.
18999         * config/nds32/nds32.h (nds32_builtins): Likewise.
19001 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19003         PR tree-optimization/85257
19004         * fold-const.c (native_encode_vector): If not all elts could fit
19005         and off is -1, return 0 rather than offset.
19006         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
19007         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
19008         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
19009         adjust buffer in native_interpret_expr call.
19011 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19013         * config/nds32/constants.md (unspec_volatile_element): Add cache
19014         control enum values.
19015         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
19016         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
19017         * config/nds32/nds32.c (nds32_cctl_names): New.
19018         (nds32_print_operand): Handle cache control register names.
19019         * config/nds32/nds32.h (nds32_builtins): New enum values.
19020         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
19021         macros.
19022         * config/nds32/nds32.md (type): Add mmu.
19023         * config/nds32/pipelines.md (simple_insn): Add mmu.
19025 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19027         * config/nds32/nds32.md (type): Remove call.
19028         * config/nds32/pipelines.md (simple_insn): Likewise.
19030 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19032         * config/nds32/constants.md (unspec_volatile_element): Add
19033         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
19034         UNSPEC_VOLATILE_FMFCFG.
19035         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
19036         description for fmfcfg and fmfcsr.
19037         (bdesc_1arg): Add fmtcsr.
19038         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
19039         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
19040         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
19041         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
19042         unspec_fmfcfg): New patterns.
19043         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
19044         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
19045         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
19046         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
19047         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
19048         __nds32__fmfcfg): Define.
19050 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19052         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
19053         intrinsic register names.
19054         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
19055         intrinsic register enum values and macros.
19057 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19059         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
19060         for load/store addressing form.
19061         (nds32_print_operand_address): Likewise.
19063 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
19065         PR target/85196
19066         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
19067         based on LABEL_REF.  Remove useless assertion.
19068         (pic_address_needs_scratch): Fix formatting.
19069         (sparc_legitimize_pic_address): Minor tweaks.
19070         (sparc_delegitimize_address): Adjust assertion accordingly.
19071         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
19072         into symbolic_operand.
19073         (movsi_high_pic_label_ref): Likewise.
19074         (movsi_lo_sum_pic_label_ref): Likewise.
19075         (movdi_pic_label_ref): Likewise.
19076         (movdi_high_pic_label_ref): Likewise.
19077         (movdi_lo_sum_pic_label_ref): Likewise.
19079 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
19081         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
19082         custom LIB_SPEC setup.
19084 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
19085             Kito Cheng  <kito.cheng@gmail.com>
19087         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
19088         * config/riscv/freebsd.h: New.
19090 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19092         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
19093         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
19094         file.
19096 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19097             Kito Cheng  <kito.cheng@gmail.com>
19099         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
19100         nds32_output_call, nds32_symbol_binds_local_p): New functions.
19101         * config/nds32/nds32-protos.h (nds32_output_call,
19102         nds32_output_return): Declare.
19103         * config/nds32/nds32.md: Refine all the call and return patterns.
19105 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19107         PR debug/85252
19108         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
19109         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
19111         PR rtl-optimization/84872
19112         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
19113         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
19114         EDGE_CROSSING edge.
19116 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
19118         * expr.c (copy_blkmode_to_reg): Revert 254862.
19119         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
19121 2018-04-06  Richard Biener  <rguenther@suse.de>
19123         PR middle-end/85244
19124         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
19125         after seeing a component reference with an adjacent field.  Treat
19126         refs to arrays at struct end of external decls similar to
19127         refs to unconstrained commons.
19129 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19131         PR sanitizer/85213
19132         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
19133         look through SAVE_EXPRs with non-side-effects argument.  Adjust
19134         recursive calls.
19135         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
19136         save_p here.
19138 2018-04-06  Richard Biener  <rguenther@suse.de>
19140         PR middle-end/85180
19141         * alias.c (find_base_term): New wrapper around find_base_term
19142         unwinding CSELIB_VAL_PTR changes.
19143         (find_base_term): Do not restore CSELIB_VAL_PTR during the
19144         recursion.
19146 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19148         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
19149         instructions.
19150         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
19151         constant definitions.
19152         ("nop"): lr 0,0 -> nopr r0
19153         ("nop_lr0", "nop_lr1"): New insn definitions.
19155 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19157         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
19158         NDS32_V3PUSH_AVAILABLE_P macro.
19160 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
19161             Chung-Ju Wu  <jasonwucj@gmail.com>
19163         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
19164         (nds32*-*-*): Add float and fpu_config into supported_defaults.
19165         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
19166         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
19167         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
19168         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
19169         * config/nds32/constraints.md: New constraints and checking for hard
19170         float configuration.
19171         * config/nds32/iterators.md: New mode iterator and attribute for hard
19172         float configuration.
19173         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
19174         patterns.
19175         * config/nds32/nds32-fpu.md: New file.
19176         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
19177         deal with hard float code generation.
19178         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
19179         ARCH_V3S.
19180         (abi_type, float_reg_number): New enum type.
19181         * config/nds32/nds32-predicates.c: New predicates for hard float.
19182         * config/nds32/nds32-protos.h: Declare functions for hard float.
19183         * config/nds32/nds32.c: Implementation for hard float configuration.
19184         * config/nds32/nds32.h: Definitions for hard float configuration.
19185         * config/nds32/nds32.md: Include hard float machine description and
19186         modify patterns for hard float configuration.
19187         * config/nds32/nds32.opt: New options for hard float configuration.
19188         * config/nds32/predicates.md: New predicates for hard float
19189         configuration.
19191 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19193         * common/config/nds32/nds32-common.c
19194         (nds32_option_optimization_table): Enable -mreleax-hint by default.
19196 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
19198         PR middle-end/85195
19199         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
19200         CONSTRUCTOR_ELT (ctor, ...)->value.
19202 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
19204         PR target/85193
19205         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
19207 2018-04-05  Tom de Vries  <tom@codesourcery.com>
19209         PR target/85204
19210         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
19211         cond jump.
19213 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19214             Kito Cheng  <kito.cheng@gmail.com>
19216         * config/nds32/constraints.md (U33): Fine-tune checking condition.
19217         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
19218         * config/nds32/nds32.h (nds32_16bit_address_type): Add
19219         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
19221 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19222             Kito Cheng  <kito.cheng@gmail.com>
19224         * config/nds32/constraints.md (Ufe): New memory constraint.
19225         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
19226         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
19227         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
19228         operands.
19229         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
19230         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
19232 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19234         * config/nds32/nds32.md: Use optimize_size in the condition for
19235         alu-shift instructions.
19237 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19239         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19241 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19243         * config/nds32/nds32.md (negsi2): Refine pattern.
19245 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19246             Chung-Ju Wu  <jasonwucj@gmail.com>
19248         * config/nds32/iterators.md (shift_rotate): New code iterator.
19249         (shift): New code attribute.
19250         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19251         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19252         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19253         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19254         bit-wise operations.
19255         (andsi3, *andsi3): Ditto.
19256         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19257         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19258         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19259         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19260         nds32_ior_operand, nds32_xor_operand): New predicates.
19262 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19264         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19265         (addsi3, subsi3): ... this.
19267 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19269         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19271 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19273         * config/nds32/nds32.md: Adjust indention.
19275 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19277         * config/nds32/nds32.md (feature): New attribute.
19279 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19281         * config/nds32/nds32.md (subtype): New attribute.
19283 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19285         PR target/85203
19286         * config/arm/arm-builtins.c (arm_expand_builtin): Change
19287         expansion to perform a bitwise AND of the argument followed by a
19288         boolean negation of the result.
19290 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
19292         PR rtl-optimization/84878
19293         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19294         the basic block.  Assert the use reference is not artificial and that
19295         it has an associated insn.
19297 2018-04-04  Michael Matz  <matz@suse.de>
19299         * builtins.c (compute_objsize): Pass correct operand
19300         to array_at_struct_end_p.
19302 2018-04-04  Richard Biener  <rguenther@suse.de>
19304         PR lto/85176
19305         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19306         from contexts for DINFO_LEVEL_TERSE and below.
19308 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19310         * config/nds32/nds32-doubleword.md (move_<mode>): Require
19311         resiter_operand condition.
19312         * config/nds32/nds32.md (*move<mode>): Ditto.
19314 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19315             Monk Chiang  <sh.chiang04@gmail.com>
19317         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19319 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19321         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19323 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19324             Kito Cheng  <kito.cheng@gmail.com>
19326         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19327         nds32_cond_code_str, output_cond_branch,
19328         output_cond_branch_compare_zero, nds32_expand_cbranch,
19329         nds32_expand_cstore, nds32_expand_movcc,
19330         nds32_output_cbranchsi4_equality_zero,
19331         nds32_output_cbranchsi4_equality_reg,
19332         nds32_output_cbranchsi4_equality_reg_or_const_int,
19333         nds32_output_cbranchsi4_greater_less_zero: New functions.
19334         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19335         nds32_expand_cstore, nds32_expand_movcc,
19336         nds32_output_cbranchsi4_equality_zero,
19337         nds32_output_cbranchsi4_equality_reg,
19338         nds32_output_cbranchsi4_equality_reg_or_const_int,
19339         nds32_output_cbranchsi4_greater_less_zero): Declare.
19340         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19341         nds32_rimm11s_operand): New predicates.
19342         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
19343         * config/nds32/nds32.md: Rewrite all the branch and conditional move
19344         patterns.
19346 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19348         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
19349         * config/nds32/nds32.md: Ditto.
19350         * config/nds32/pipelines.md: Ditto.
19352 2018-04-04  Richard Biener  <rguenther@suse.de>
19354         PR tree-optimization/85168
19355         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
19356         propagating abnormals.
19358 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19360         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
19362 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19363             Kito Cheng  <kito.cheng@gmail.com>
19365         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
19366         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
19367         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
19368         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
19369         * config/nds32/nds32.md (sibcall_internal): New.
19370         (sibcall_register): Remove.
19371         (sibcall_immediate): Remove.
19372         (sibcall_value_internal): New.
19373         (sibcall_value_register): Remove.
19374         (sibcall_value_immediate): Remove.
19375         * config/nds32/predicates.md (nds32_general_register_operand): New.
19376         (nds32_call_address_operand): New.
19378 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19380         PR rtl-optimization/85167
19381         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
19382         bb_defs if *split_p, instead preinitialize it to NULL.
19384         PR tree-optimization/85156
19385         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
19386         evaluating the argument multiple times.
19388 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
19390         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
19391         than vector.
19392         (_mm_cvtpd_ps): Likewise.
19393         (_mm_cvttpd_epi32): Likewise.
19394         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
19395         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
19396         vector, pixel, and bool following altivec.h include.
19398 2018-04-03  Martin Sebor  <msebor@redhat.com>
19400         * doc/extend.texi (Common Function Attributes): Clarify.
19401         (const attribute): Likewise.
19402         (pure attribute): Likewise.
19404 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19406         PR target/85169
19407         * config/i386/i386.c (ix86_expand_vector_set): Use
19408         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
19410 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
19412         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
19413         instructions when changing rounding bits to preserve precision bits
19414         in the x87 control word.
19416 2018-04-03  Martin Liska  <mliska@suse.cz>
19418         PR tree-optimization/82491
19419         * rtl.h (strip_offset_and_add): Replace += suboffset with
19420         poly_uint64 () + suboffset.
19422 2018-03-29  Martin Liska  <mliska@suse.cz>
19423             Martin Jambor  <mjambor@suse.cz>
19425         PR ipa/84947
19426         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
19427         param_type is not an integral or pointer type.
19429 2018-04-03  Richard Biener  <rguenther@suse.de>
19431         * sese.h (recompute_all_dominators): Remove.
19433 2018-04-02  Martin Sebor  <msebor@redhat.com>
19435         * doc/invoke.texi (-Wrestrict): Fix typos.
19437 2018-04-02  Jim Wilson  <jimw@sifive.com>
19439         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
19440         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
19441         (<optab>di3, <optab>si3_extend): Likewise.
19442         (<optab>si3_mask, <optab>si3_mask_1): New.
19443         (<optab>di3_mask, <optab>di3_mask_1): New.
19444         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
19445         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
19446         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
19448 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
19450         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
19451         example.
19453 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19455         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
19456         (nds32_canonicalize_comparison): New function.
19458 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19459             Kito Cheng  <kito.cheng@gmail.com>
19460             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19462         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
19463         * config/nds32/constants.md (unspec_volatile_element): Add
19464         UNSPEC_VOLATILE_RELAX_GROUP.
19465         * config/nds32/nds32-relax-opt.c: New file.
19466         * config/nds32/nds32-predicates.c
19467         (nds32_symbol_load_store_p): New function.
19468         * config/nds32/nds32-protos.h
19469         (nds32_symbol_load_store_p): Declare function.
19470         (make_pass_nds32_relax_opt): Declare new rtl pass function.
19471         * config/nds32/nds32.c
19472         (nds32_register_pass): New function to register pass.
19473         (nds32_register_passes): New function to register passes.
19474         * config/nds32/nds32.md (relax_group): New pattern.
19475         * config/nds32/nds32.opt (mrelax-hint): New option.
19476         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
19478 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
19480         * config/nds32/t-nds32: Modify files dependency.
19482 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19484         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
19485         (PROFILE_HOOK): Define its implementation.
19487 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19489         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
19490         type and 32-bit size.
19492 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
19494         PR middle-end/85090
19495         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
19496         (V_128_256): New mode iterator.
19497         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
19498         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
19499         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
19500         of V.
19501         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
19502         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
19504 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
19506         PR target/83315
19507         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
19508         NaN inputs correctly.
19510 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
19512         PR target/80546
19513         * config/rs6000/vsx.md (??r): New mode attribute.
19514         (*vsx_mov<mode>_64bit): Use it.
19515         (*vsx_mov<mode>_32bit): Likewise.
19517 2018-03-30  Martin Sebor  <msebor@redhat.com>
19519         PR tree-optimization/84818
19520         * builtins.c (check_access): Use warning_n.
19522 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19524         PR target/83822
19525         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
19526         condition.
19527         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
19528         condition.
19530 2018-03-30  Julia Koval  <julia.koval@intel.com>
19532         PR target/84413
19533         * x86-tune.def (movx, partial_reg_dependency): Enable for
19534         m_SKYLAKE_AVX512.
19536 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
19538         PR inline-asm/84985
19539         * lra-constraints.c (process_alt_operands): Move setting
19540         this_alternative_matches below.
19542 2018-03-29  Martin Liska  <mliska@suse.cz>
19544         PR lto/84995.
19545         * doc/invoke.texi: Document how LTO works with debug info.
19546         Describe auto-load support of binutils.  Mention 'x86-64'
19547         as valid option value of -march option.
19549 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
19551         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
19553         PR c/85094
19554         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
19555         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
19556         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
19557         checking.
19559 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
19561         PR target/84912
19562         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
19563         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
19564         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
19565         for RS6000_BTM_POWERPC64.
19566         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
19567         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
19568         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
19569         definition.
19570         (DIVDE): Use it.
19571         (DIVDEU): Likewise.
19573 2018-03-28  Carl Love  <cel@us.ibm.com>
19575         Revert
19576         2017-09-27  Carl Love  <cel@us.ibm.com>
19578         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
19579         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
19580         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
19581         fctiw instruction.
19583 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19585         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
19586         instead of __vector bool.
19587         (_mm_max_pu8): Likewise.
19588         (_mm_min_pi16): Likewise.
19590 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
19592         PR target/84912
19593         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
19594         (DIVWEUO): Likewise.
19595         (DIVDEO): Likewise.
19596         (DIVDEUO): Likewise.
19597         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
19598         DIVWEUO and DIVDEUO.
19599         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
19600         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
19601         (div_extend): Likewise.
19602         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
19603         builtin function.
19604         (__builtin_divweuo): Likewise.
19605         (__builtin_divdeo): Likewise.
19606         (__builtin_divdeuo): Likewise.
19608 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
19610         PR target/85095
19611         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
19612         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
19614         PR tree-optimization/82004
19615         * gimple-match-head.c (optimize_pow_to_exp): New function.
19616         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
19617         Don't fold to exp if optimize_pow_to_exp is false.
19619 2018-03-28  Martin Liska  <mliska@suse.cz>
19621         PR other/84819
19622         * calls.c (initialize_argument_information): Fix trailing space.
19623         * common.opt: Fix typo and provide better explanation for
19624         -fsanitize-coverage option.
19625         * config/i386/i386.opt: Fix typo.
19627 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
19628             Martin Liska  <mliska@suse.cz>
19630         PR sanitizer/85081
19631         * gimplify.c (asan_poison_variable): Don't do the check for
19632         gimplify_omp_ctxp here.
19633         (gimplify_decl_expr): Do it here.
19634         (gimplify_target_expr): Likewise.
19636 2018-03-28  Martin Liska  <mliska@suse.cz>
19638         PR target/84988
19639         * config/i386/i386.c (ix86_function_arg_advance): Do not call
19640         chkp_type_bounds_count if MPX is not enabled.
19642 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
19644         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
19646 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
19648         PR target/84914
19649         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
19650         function to create the function decl for complex long double
19651         multiply and divide for -mabi=ieeelongdouble.
19652         (init_float128_ieee): Call it.
19654 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
19656         PR target/85044
19657         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
19658         -fcf-protection=branch -mibt.
19659         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
19661 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19663         PR target/81863
19664         * config/arm/arm.c (arm_valid_symbolic_address): Handle
19665         arm_word_relocations.
19667 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
19669         PR target/85056
19670         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
19671         extern array declarations.
19673 2018-03-27  Richard Biener  <rguenther@suse.de>
19675         PR middle-end/84067
19676         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
19677         explicit single_use checks.
19679 2018-03-27  Richard Biener  <rguenther@suse.de>
19681         PR tree-optimization/85082
19682         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
19683         Valueize the VUSE.
19685 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19687         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
19688         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
19689         Turn on fasynchronous-unwind-tables and funwind-tables.
19691 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
19693         PR target/85073
19694         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
19695         (*bmi_blsr_<mode>_ccz): Ditto.
19697 2018-03-26  Tom de Vries  <tom@codesourcery.com>
19699         PR tree-optimization/85063
19700         * omp-general.c (offloading_function_p): New function.  Factor out
19701         of ...
19702         * omp-offload.c (pass_omp_target_link::gate): ... here.
19703         * omp-general.h (offloading_function_p): Declare.
19704         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
19705         with attribute omp declare target for offloading functions.
19707 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
19709         PR tree-optimization/84005
19710         * tree-data-ref.h (get_base_for_alignment): Declare.
19711         * tree-data-ref.c (get_base_for_alignment_1): New function.
19712         (get_base_for_alignment): Likewise.
19713         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
19714         get_base_for_alignment to find a suitable base object, instead
19715         of always using drb->base_address.
19717 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
19719         PR inline-asm/85022
19720         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
19721         known size by default.
19723 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
19725         PR inline-asm/85030
19726         * lra-constraints.c (process_alt_operands): Don't match BLKmode
19727         and non BLKmode operands.
19729 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19731         PR target/85026
19732         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
19733         Clean up attributes.
19735 2018-03-23  Richard Biener  <rguenther@suse.de>
19737         PR debug/85020
19738         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
19739         we are going to emit early debug for LTO.
19741 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
19743         PR inline-asm/85034
19744         * function.c (match_asm_constraints_1): Don't optimize if input
19745         doesn't satisfy general_operand predicate for output's mode.
19747         PR inline-asm/85022
19748         * alias.c (write_dependence_p): Don't require for x_canonicalized
19749         non-VOIDmode if x has VOIDmode.
19751         PR sanitizer/85029
19752         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
19753         just don't try to optimize it rather than assert it never happens.
19755 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19757         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
19758         macro expansions for definition of ST_INTERNAL_<mode> and
19759         LD_INTERNAL_<mode> builtins.
19760         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
19761         Remove prototype.
19762         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
19763         function.
19764         (altivec_expand_st_builtin): Likewise.
19765         (altivec_expand_builtin): Remove calls to deleted functions.
19766         (rs6000_address_for_altivec): Delete this function.
19767         * config/rs6000/vector.md: Remove expands for
19768         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
19770 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
19772         PR target/84826
19773         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
19774         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
19775         re-computing once computed.
19776         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
19777         (arm_init_machine_status): Initialize
19778         machine->static_chain_stack_bytes.
19780 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19782         PR target/84760
19783         * doc/extend.texi: Add four new prototypes for vec_ld.
19784         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
19785         definitions for more logical presentation.
19786         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
19787         entries for V1TI variants of __builtin_altivec_ld builtin.
19788         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
19789         handling of V1TI variant of LVX icode pattern.
19790         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
19791         (rs6000_gimple_fold_builtin): Likewise.
19792         (altivec_init_builtins): Add code to define
19793         __builtin_altivec_lvx_v1ti function.
19795 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
19797         PR inline-asm/84941
19798         * function.c (match_asm_constraints_1): Don't do the optimization
19799         if input isn't a REG, SUBREG, MEM or constant.
19801 2018-03-22  Tom de Vries  <tom@codesourcery.com>
19803         PR tree-optimization/84956
19804         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
19805         bb_has_abnormal_pred.
19807 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
19809         PR sanitizer/85018
19810         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
19811         DECL_INITIAL (decl) to decl at the end.
19812         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
19813         adjust the comment.
19815 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
19817         * doc/extend.texi (__builtin_tgmath): Document when complex
19818         integer types are treated as _Complex _Float64.
19820 2018-03-21  Tom de Vries  <tom@codesourcery.com>
19822         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
19824 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19826         PR tree-optimization/84960
19827         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
19828         if it is ENTRY block, move them into single succ of ENTRY in that case.
19830 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
19832         PR tree-optimization/84811
19833         * poly-int.h (poly_span_traits): Remove the T3 parameter and
19834         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
19835         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
19836         (known_subrange_p): Update accordingly.  Cast each value involved
19837         in the size comparison, rather than casting the result of the
19838         subtraction.
19840 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19842         PR tree-optimization/84982
19843         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
19844         by flipping the least significant bit rather than all bits from
19845         bitpos to bitpos + bitsize - 1.
19847 2018-03-21  Nathan Sidwell  <nathan@acm.org>
19849         * doc/extend.texi (Deprecated Features): Remove mention of
19850         long-deleted deprecations.
19852 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
19854         PR jit/84288
19855         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
19856         * configure: Regenerate.
19858 2018-03-21  Tom de Vries  <tom@codesourcery.com>
19860         PR tree-optimization/83126
19861         * tree-parloops.c (num_phis): New function.
19862         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
19864 2018-03-21  Nathan Sidwell  <nathan@acm.org>
19866         * doc/extend.texi (Deprecated Features): Update deprecated flags,
19867         mention anon-struct/union members and trailing attributes.
19869 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
19871         PR tree-optimization/84969
19872         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
19873         builtin memset partitions if they set different rhs values.
19875 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19877         PR rtl-optimization/84989
19878         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
19879         VEC_DUPLICATE with scalar result mode.
19881 2018-03-21  Martin Liska  <mliska@suse.cz>
19883         PR ipa/84963
19884         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
19885         not intended return statement.
19887 2018-03-21  Martin Liska  <mliska@suse.cz>
19889         PR target/84988
19890         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
19891         (chkp_find_bound_slots_1): Limit number of iterations.
19893 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
19895         PR target/84838
19896         * Minor grammar fixes for x86 options.
19898 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19900         PR debug/84875
19901         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
19902         holding REG_CFA_RESTORE notes, instead turn them into a USE.
19904 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
19906         PR target/83789
19907         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
19908         (altivec_lvx_<mode>_1op): Likewise.
19909         (altivec_stvx_<mode>_2op): Likewise.
19910         (altivec_stvx_<mode>_1op): Likewise.
19911         (altivec_lvx_<VM2:mode>): New define_expand.
19912         (altivec_stvx_<VM2:mode>): Likewise.
19913         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
19914         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19915         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
19916         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19917         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
19918         (rs6000_gen_lvx): Likewise.
19919         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
19920         (altivec_expand_stv_builtin): Likewise.
19921         (altivec_expand_builtin): Likewise.
19922         * config/rs6000/vector.md: Likewise.
19924 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19926         PR target/82518
19927         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
19928         BYTES_BIG_ENDIAN.
19930 2018-03-20  Richard Biener  <rguenther@suse.de>
19932         PR target/84986
19933         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
19934         sign-conversions as zero, fall back to standard scalar_stmt
19935         cost for the rest.
19937 2018-03-20  Martin Liska  <mliska@suse.cz>
19939         PR ipa/84825
19940         * predict.c (rebuild_frequencies): Handle case when we have
19941         PROFILE_ABSENT, but flag_guess_branch_prob is false.
19943 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19945         PR target/84990
19946         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
19947         flag_section_anchors.
19948         * varasm.c (use_blocks_for_decl_p): Remove hack for
19949         dw2_force_const_mem.
19951         PR target/84845
19952         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
19953         to ...
19954         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
19955         be created, use lowpart_subreg of operands[0] rather than operands[0]
19956         itself.
19957         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
19958         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
19959         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
19960         and n constraint instead of aarch64_shift_imm_di and Usd.
19961         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
19962         (*aarch64_<optab>_reg_minus<mode>3): ... this.
19964 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
19966         PR target/82989
19967         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
19968         to favor GPR over NEON registers.
19969         (<shift>di3_neon): Likewise.
19971 2018-03-20  Tom de Vries  <tom@codesourcery.com>
19973         PR target/84952
19974         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
19975         (nvptx_process_pars): Emit bar.sync asap and alap.
19977 2018-03-20  Tom de Vries  <tom@codesourcery.com>
19979         PR target/84954
19980         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
19981         seen_label if seen_label is already set.
19983 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19985         PR target/84945
19986         * config/i386/i386.c (fold_builtin_cpu): For features above 31
19987         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
19988         Use 1U instead of 1.  Formatting fixes.
19990         PR c/84953
19991         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
19992         instead of TREE_TYPE (s1) for the return value.
19994 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
19996         PR tree-optimization/84946
19997         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
19998         bitsize + bitsize in poly_uint64 rather than poly_int64.
20000         PR sanitizer/78651
20001         * dwarf2asm.c: Include fold-const.c.
20002         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
20003         of decl rather than decl itself.
20005         PR rtl-optimization/84643
20006         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
20008 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
20010         PR sanitizer/78651
20011         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
20012         calling assemble_variable.
20014 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
20016         PR target/81647
20017         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
20018         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
20020 2018-03-19  Jim Wilson  <jimw@sifive.com>
20022         PR bootstrap/84856
20023         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
20024         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
20025         (riscv_first_stack_step): Likewise.
20026         (riscv_option_override): Use STACK_BOUNDARY instead of
20027         MIN_STACK_BOUNDARY.
20028         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
20029         MIN_STACK_BOUNDARY.
20030         (BIGGEST_ALIGNMENT): Set to 128.
20031         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
20032         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
20033         STACK_BOUNDARY.
20035 2018-03-19  Richard Biener  <rguenther@suse.de>
20037         PR tree-optimization/84933
20038         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
20039         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
20041 2018-03-19  Richard Biener  <rguenther@suse.de>
20043         PR tree-optimization/84859
20044         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
20045         (cond_if_else_store_replacement): Perform sinking operation on
20046         single-store BBs regardless of MAX_STORES_TO_SINK setting.
20047         Generalize what a BB with a single eligible store is.
20049 2018-03-19  Richard Biener  <rguenther@suse.de>
20051         PR tree-optimization/84929
20052         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
20053         chrec_is_positive against non-chrec arg.
20055 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
20057         PR target/84711
20058         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
20060 2018-03-18  Martin Liska  <mliska@suse.cz>
20062         PR rtl-optimization/84635
20063         * regrename.c (build_def_use): Use matches_mode only when
20064         matches >= 0.
20066 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
20068         PR tree-optimization/84913
20069         * tree-vect-loop.c (vectorizable_reduction): Don't try to
20070         vectorize chains of COND_EXPRs.
20072 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20074         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
20076 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20078         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
20080 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20082         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
20084 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20085             Kito Cheng  <kito.cheng@gmail.com>
20087         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
20088         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
20089         (nds32_adjust_reg_alloc_order): New function.
20090         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
20092 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
20094         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
20095         nds32_print_operand, nds32_print_operand_address): Use
20096         HOST_WIDE_INT_PRINT_DEC instead.
20098 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20100         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
20102 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
20104         PR target/84902
20105         * config/i386/i386.c (initial_ix86_tune_features,
20106         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
20107         unsigned long long.
20108         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
20109         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
20110         rather than 1u << ix86_tune.  Formatting fix.
20111         (ix86_option_override_internal): Change ix86_arch_mask from
20112         unsigned int to unsigned HOST_WIDE_INT, initialize to
20113         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
20114         (ix86_function_specific_restore): Likewise.
20116 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20118         PR target/84899
20119         * postreload.c (reload_combine_recognize_pattern): Perform
20120         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
20121         truncate_int_for_mode the result for the destination's mode.
20123         PR c/84909
20124         * hsa-gen.c (mem_type_for_type): Fix comment typo.
20125         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
20126         Likewise.
20127         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20128         Likewise.
20130 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
20132         PR target/84876
20133         * lra-assigns.c (lra_split_hard_reg_for): Don't use
20134         regno_allocno_class_array and sorted_pseudos.
20135         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
20136         insns where regno is used.
20138 2018-03-16  Martin Liska  <mliska@suse.cz>
20140         PR ipa/84833
20141         * multiple_target.c (create_dispatcher_calls): Redirect
20142         reference in the symbol table.
20144 2018-03-16  Martin Liska  <mliska@suse.cz>
20146         PR ipa/84722
20147         * multiple_target.c (create_dispatcher_calls): Redirect also
20148         an alias.
20150 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20152         PR c++/79937
20153         PR c++/82410
20154         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
20155         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
20156         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
20158 2018-03-16  Julia Koval  <julia.koval@intel.com>
20160         * doc/invoke.texi (Skylake Server): Add CLWB.
20161         Cannonlake): Remove CLWB.
20163 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20165         PR tree-optimization/84841
20166         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
20167         1 << 3.
20168         (FLOAT_ONE_CONST_TYPE): Define.
20169         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
20170         (sort_by_operand_rank): Put entries with higher constant_type last
20171         rather than first to match comments.
20173 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
20175         * config/nios2/nios2.md (movsi_internal): Fix thinko in
20176         split predicate.
20178 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20180         PR c++/79085
20181         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
20182         check and use address of target always.
20184 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
20186         PR target/84574
20187         * config/i386/i386.c (indirect_thunk_needed): Update comments.
20188         (indirect_thunk_bnd_needed): Likewise.
20189         (indirect_thunks_used): Likewise.
20190         (indirect_thunks_bnd_used): Likewise.
20191         (indirect_return_needed): New.
20192         (indirect_return_bnd_needed): Likewise.
20193         (output_indirect_thunk_function): Add a bool argument for
20194         function return.
20195         (output_indirect_thunk_function): Don't generate alias for
20196         function return thunk.
20197         (ix86_code_end): Call output_indirect_thunk_function to generate
20198         function return thunks.
20199         (ix86_output_function_return): Set indirect_return_bnd_needed
20200         and indirect_return_needed instead of indirect_thunk_bnd_needed
20201         and indirect_thunk_needed.
20203 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
20205         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
20206         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
20207         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
20209 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
20210             Paul Hua <paul.hua.gm@gmail.com>
20212         PR c/84852
20213         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
20215 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
20217         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
20218         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
20219         resp. SFmode cases.
20221 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
20223         PR target/84711
20224         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
20225         instead of GET_MODE_SIZE when comparing Units.
20227 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
20229         PR target/68256
20230         * varasm.c (hash_section): Return an unchangeble hash value
20231         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
20232         Return !aarch64_can_use_per_function_literal_pools_p ().
20234 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20236         PR target/84860
20237         * optabs.c (emit_conditional_move): Pass address of cmode's copy
20238         rather than address of cmode as last argument to prepare_cmp_insn.
20240 2018-03-15  Julia Koval  <julia.koval@intel.com>
20242         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20243         F_AVX512VNNI, F_AVX512BITALG): New.
20245 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
20247         PR target/83451
20248         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
20249         insn for floating-point loads and stores.
20251 2018-03-14  Carl Love  <cel@us.ibm.com>
20253         * config/rs6000/rs6000-c.c: Add macro definitions for
20254         ALTIVEC_BUILTIN_VEC_PERMXOR.
20255         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20256         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20257         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20258         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20259         UNSPEC_VPERMXOR.
20260         * config/doc/extend.texi: Add prototypes for vec_permxor.
20262 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
20264         PR c/84852
20265         * diagnostic-show-locus.c (class layout_point): Convert m_line
20266         from int to linenum_type.
20267         (line_span::comparator): Use linenum "compare" function when
20268         comparing line numbers.
20269         (test_line_span): New function.
20270         (layout_range::contains_point): Convert param "row" from int to
20271         linenum_type.
20272         (layout_range::intersects_line_p): Likewise.
20273         (layout::will_show_line_p): Likewise.
20274         (layout::print_source_line): Likewise.
20275         (layout::should_print_annotation_line_p): Likewise.
20276         (layout::print_annotation_line): Likewise.
20277         (layout::print_leading_fixits): Likewise.
20278         (layout::annotation_line_showed_range_p): Likewise.
20279         (struct line_corrections): Likewise for field m_row.
20280         (line_corrections::line_corrections): Likewise for param "row".
20281         (layout::print_trailing_fixits): Likewise.
20282         (layout::get_state_at_point): Likewise.
20283         (layout::get_x_bound_for_row): Likewise.
20284         (layout::print_line): Likewise.
20285         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
20286         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
20287         * input.c (selftest::test_linenum_comparisons): New function.
20288         (selftest::input_c_tests): Call it.
20289         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
20290         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
20291         * selftest.h (ASSERT_GT): New macro.
20292         (ASSERT_GT_AT): New macro.
20293         (ASSERT_LT): New macro.
20294         (ASSERT_LT_AT): New macro.
20296 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
20298         PR rtl-optimization/84780
20299         * combine.c (distribute_links): Don't make a link based on pc_rtx.
20301 2018-03-14  Martin Liska  <mliska@suse.cz>
20303         * tree.c (record_node_allocation_statistics): Use
20304         get_stats_node_kind.
20305         (get_stats_node_kind): New function extracted from
20306         record_node_allocation_statistics.
20307         (free_node): Use get_stats_node_kind.
20309 2018-03-14  Richard Biener  <rguenther@suse.de>
20311         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20312         that the value-set of ANTIC_IN doesn't grow.
20314         Revert
20315         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20316         member.
20317         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20318         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20320 2018-03-14  Julia Koval  <julia.koval@intel.com>
20322         * config.gcc (icelake-client, icelake-server): New.
20323         (icelake): Remove.
20324         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20325         (initial_ix86_arch_features): Ditto.
20326         (PTA_SKYLAKE): Add SGX.
20327         (PTA_ICELAKE): Remove.
20328         (PTA_ICELAKE_CLIENT): New.
20329         (PTA_ICELAKE_SERVER): New.
20330         (ix86_option_override_internal): Split up icelake on icelake client and
20331         icelake server.
20332         (get_builtin_code_for_version): Ditto.
20333         (fold_builtin_cpu): Ditto.
20334         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20335         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20336         * config/i386/i386.h (processor_type): Ditto.
20337         * doc/invoke.texi: Ditto.
20339 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
20341         PR sanitizer/83392
20342         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
20343         INTEGER_CST offset, add it together with bitpos / 8 and
20344         sign extend based on POINTER_SIZE.
20346         PR target/84844
20347         Revert
20348         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
20350         PR target/78090
20351         * config/i386/constraints.md (Yc): New register constraint.
20352         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
20353         Use Yc constraint for alternative 2 of operand 0.  Remove
20354         preferred_for_speed attribute.
20356 2018-03-14  Richard Biener  <rguenther@suse.de>
20358         PR tree-optimization/84830
20359         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
20360         with the old one to avoid oscillations.
20362 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
20364         PR target/83712
20365         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
20366         pseudos.
20367         (assign_by_spills): Return a flag of reload assignment failure.
20368         Do not process the reload assignment failures.  Do not spill other
20369         reload pseudos if they has the same reg class.  Update n if
20370         necessary.
20371         (lra_assign): Add a return arg.  Set up from the result of
20372         assign_by_spills call.
20373         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20374         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20375         usage_insns if it is not NULL.
20376         (spill_hard_reg_in_range): New function.
20377         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20378         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20379         function prototypes.
20380         (lra_assign): Change prototype.
20381         * lra.c (lra): Add code to deal with fails by splitting hard reg
20382         live ranges.
20384 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
20386         * config/riscv/riscv.opt (mrelax): New option.
20387         * config/riscv/riscv.c (riscv_file_start): Emit ".option
20388         "norelax" when riscv_mrelax is disabled.
20389         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
20391 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20393         PR target/84743
20394         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
20395         reassociation for int modes.
20397 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20399         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
20400         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
20401         for big-endian.
20402         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
20403         * config/aarch64/aarch64-sve.md
20404         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
20405         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
20406         (*extend<mode><Vwide>2): Rename to...
20407         (aarch64_sve_extend<mode><Vwide>2): ...this.
20408         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
20409         renaming the old pattern to...
20410         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
20411         unsigned packs.
20412         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
20413         define_expand, renaming the old pattern to...
20414         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
20415         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
20416         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
20417         account when deciding which SVE instruction the optab should use.
20418         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
20420 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20422         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
20423         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
20424         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
20425         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
20426         (tlsdesc_small_<mode>): Turn a define_expand and use
20427         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
20428         (tlsdesc_small_advsimd_<mode>): ...this.
20429         (tlsdesc_small_sve_<mode>): New pattern.
20431 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20433         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
20434         (UNSPEC_UMUL_HIGHPART): New constants.
20435         (MUL_HIGHPART): New int iteraor.
20436         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
20437         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20438         define_expand.
20439         (*<su>mul<mode>3_highpart): New define_insn.
20441 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20443         PR lto/84805
20444         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20445         incomplete types.
20447 2018-03-13  Martin Liska  <mliska@suse.cz>
20449         PR ipa/84658.
20450         * (sem_item_optimizer::sem_item_optimizer): Initialize new
20451         vector.
20452         (sem_item_optimizer::~sem_item_optimizer): Release it.
20453         (sem_item_optimizer::merge_classes): Register variable aliases.
20454         (sem_item_optimizer::fixup_pt_set): New function.
20455         (sem_item_optimizer::fixup_points_to_sets): Likewise.
20456         * ipa-icf.h: Declare new variables and functions.
20458 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
20460         PR middle-end/84834
20461         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
20462         integer_pow2p@2 and test integer_pow2p in condition.
20463         (A < 0 ? C : 0): Similarly for @1.
20465         PR middle-end/84831
20466         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
20467         characters starting at p contain '\0' character, don't look beyond
20468         that.
20470         PR target/84827
20471         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
20472         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
20474         PR target/84828
20475         * reg-stack.c (change_stack): Change update_end var from int to
20476         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
20477         also call set_block_for_insn on the newly added insns and rescan.
20479         PR target/84786
20480         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
20481         on the last operand.
20483         PR c++/84704
20484         * tree.c (stabilize_reference_1): Return save_expr (e) for
20485         STATEMENT_LIST even if it doesn't have side-effects.
20487 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
20489         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
20491 2018-03-12  Renlin Li  <renlin.li@arm.com>
20493         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
20494         aarch64_output_scalar_simd_mov_immediate.
20496 2018-03-12  Martin Sebor  <msebor@redhat.com>
20498         PR tree-optimization/83456
20499         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
20500         for perfectly overlapping calls to memcpy.
20501         (gimple_fold_builtin_memory_chk): Same.
20502         (gimple_fold_builtin_strcpy): Handle no-warning.
20503         (gimple_fold_builtin_stxcpy_chk): Same.
20504         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
20506 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20508         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
20509         parameter.  Use it for SFmode.
20510         (rs6000_function_arg_advance_1): Adjust.
20511         (rs6000_function_arg): Adjust.
20512         (rs6000_gimplify_va_arg): Pass false for that new parameter.
20514 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20516         PR rtl-optimization/84169
20517         PR rtl-optimization/84780
20518         * combine.c (can_combine_p): Check for a 2-insn combination whether
20519         the destination register is used between the two insns, too.
20521 2018-03-12  Richard Biener  <rguenther@suse.de>
20523         PR tree-optimization/84803
20524         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20525         for refs DR analysis didn't process.
20527 2018-03-12  Richard Biener  <rguenther@suse.de>
20529         PR tree-optimization/84777
20530         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
20531         force-vectorize loops ignore whether we are optimizing for size.
20533 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
20535         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
20536         (TARGET_MD_ASM_ADJUST): Define.
20538 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
20539             Kito Cheng  <kito.cheng@gmail.com>
20540             Chung-Ju Wu  <jasonwucj@gmail.com>
20542         * config/nds32/nds32.c (nds32_compute_stack_frame,
20543         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
20544         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
20545         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
20546         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
20547         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
20548         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
20549         * config/nds32/nds32.md (prologue, epilogue): Use macro
20550         NDS32_V3PUSH_AVAILABLE_P to do checking.
20552 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
20554         PR debug/58150
20555         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
20556         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
20557         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
20558         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
20559         addition of most attributes on !orig_type_die or the attribute not
20560         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
20562 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20563             Chung-Ju Wu  <jasonwucj@gmail.com>
20565         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
20566         __NDS32_VH__ macro.
20567         * config/nds32/nds32.opt (mvh): New option.
20569 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20570             Chung-Ju Wu  <jasonwucj@gmail.com>
20572         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
20573         function.
20574         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
20575         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
20576         definition.
20578 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20579             Chung-Ju Wu  <jasonwucj@gmail.com>
20581         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
20582         function.
20583         * config/nds32/nds32-multiple.md (strlensi): New pattern.
20584         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
20586 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
20587             Kito Cheng  <kito.cheng@gmail.com>
20588             Chung-Ju Wu  <jasonwucj@gmail.com>
20590         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
20591         UNSPEC_FFMISM and UNSPEC_FLMISM.
20592         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
20593         for ffb, ffmism and flmism.
20594         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
20595         (unspec_ffmism): Ditto.
20596         (unspec_flmism): Ditto.
20597         (nds32_expand_builtin_impl): Check if string extension is available.
20598         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
20599         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
20601 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
20603         Reverting patch:
20604         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
20606         PR target/83712
20607         * lra-assigns.c (assign_by_spills): Return a flag of reload
20608         assignment failure.  Do not process the reload assignment
20609         failures.  Do not spill other reload pseudos if they has the same
20610         reg class.
20611         (lra_assign): Add a return arg.  Set up from the result of
20612         assign_by_spills call.
20613         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20614         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20615         usage_insns if it is not NULL.
20616         (spill_hard_reg_in_range): New function.
20617         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20618         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20619         function prototypes.
20620         (lra_assign): Change prototype.
20621         * lra.c (lra): Add code to deal with fails by splitting hard reg
20622         live ranges.
20624 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
20626         PR target/84807
20627         * config/i386/i386.opt: Replace Enforcment with Enforcement.
20629 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
20631         PR debug/84620
20632         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
20633         (dw_val_node): Add val_symbolic_view.
20634         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
20635         (symview_upper_bound): New.
20636         (new_line_info_table): Initialize symviews_since_reset.
20637         (dwarf2out_source_line): Count symviews_since_reset and set
20638         symview_upper_bound.
20639         (dw_val_equal_p): Handle symview.
20640         (add_AT_symview): New.
20641         (print_dw_val): Handle symview.
20642         (attr_checksum, attr_checksum_ordered): Likewise.
20643         (same_dw_val_p, size_of_die): Likewise.
20644         (value_format, output_die): Likewise.
20645         (add_high_low_attributes): Use add_AT_symview for entry_view.
20646         (dwarf2out_finish): Reset symview_upper_bound, clear
20647         zero_view_p.
20649 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
20651         PR target/83969
20652         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
20653         Add strict argument and use it.
20654         (rs6000_split_multireg_move): Update for new strict argument.
20655         (mem_operand_gpr): Disallow all non-offsettable addresses.
20656         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
20658 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
20660         PR target/84772
20661         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
20662         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
20663         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
20665         PR c++/84767
20666         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
20667         decl, use remap_type if we want to use the type.
20669 2018-03-09  Martin Sebor  <msebor@redhat.com>
20671         PR tree-optimization/84526
20672         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20673         Remove dead code.
20674         (builtin_access::generic_overlap): Be prepared to handle non-array
20675         base objects.
20677 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
20679         PR rtl-optimization/84682
20680         * lra-constraints.c (process_address_1): Check is_address flag
20681         for address constraints.
20682         (process_alt_operands): Likewise.
20683         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
20684         preprocess_constraints.
20685         * recog.h (preprocess_constraints): Add oploc parameter.
20686         Adjust callers.
20687         * recog.c (preprocess_constraints): Test address_operand for
20688         CT_ADDRESS constraints.
20690 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
20692         PR target/83712
20693         * lra-assigns.c (assign_by_spills): Return a flag of reload
20694         assignment failure.  Do not process the reload assignment
20695         failures.  Do not spill other reload pseudos if they has the same
20696         reg class.
20697         (lra_assign): Add a return arg.  Set up from the result of
20698         assign_by_spills call.
20699         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20700         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20701         usage_insns if it is not NULL.
20702         (spill_hard_reg_in_range): New function.
20703         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20704         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20705         function prototypes.
20706         (lra_assign): Change prototype.
20707         * lra.c (lra): Add code to deal with fails by splitting hard reg
20708         live ranges.
20710 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20712         PR target/83193
20713         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
20714         Accept complain bool parameter.  Only emit errors if it is true.
20715         (arm_parse_cpu_option_name): Likewise.
20716         (arm_target_thumb_only): Adjust callers of the above.
20717         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
20718         prototype to take a default true bool parameter.
20719         (arm_parse_arch_option_name): Likewise.
20721 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
20722             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
20724         PR jit/64089
20725         PR jit/84288
20726         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
20727         * configure: Regenerate.
20728         * configure.ac ("linker --version-script option"): New.
20729         ("linker soname option"): New.
20731 2018-03-09  Richard Biener  <rguenther@suse.de>
20733         PR tree-optimization/84775
20734         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
20735         immediate uses of predicate stmts and mark them modified.
20737         Revert
20738         PR tree-optimization/84178
20739         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20740         to caller.
20741         (version_loop_for_if_conversion): Delay update_ssa call.
20742         (tree_if_conversion): Delay update_ssa until after predicate
20743         insertion.
20745 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
20747         PR target/84763
20748         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
20749         when the function accesses prior frames.
20751 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
20753         PR debug/84456
20754         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
20755         gen_llsym, otherwise call maybe_gen_llsym.
20757         PR inline-asm/84742
20758         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
20759         has ',' character inside of it.
20761 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20763         PR target/84748
20764         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
20765         as clobbering CC_REGNUM.
20767 2018-03-08  Richard Biener  <rguenther@suse.de>
20769         PR middle-end/84552
20770         * tree-scalar-evolution.c: Include tree-into-ssa.h.
20771         (follow_copies_to_constant): Do not follow SSA names registered
20772         for update.
20774 2018-03-08  Richard Biener  <rguenther@suse.de>
20776         PR tree-optimization/84178
20777         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20778         to caller.
20779         (version_loop_for_if_conversion): Delay update_ssa call.
20780         (tree_if_conversion): Delay update_ssa until after predicate
20781         insertion.
20783 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
20785         PR tree-optimization/84178
20786         * tree-if-conv.c (release_bb_predicate): Remove the
20787         the assertion that the stmts have NULL use_ops.
20788         Discard the statements, asserting that they haven't
20789         yet been added to a BB.
20791 2018-03-08  Richard Biener  <rguenther@suse.de>
20793         PR tree-optimization/84746
20794         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
20795         (phi_translate): Pass in destination ANTIC_OUT set.
20796         (phi_translate_1): Likewise.  For a simplified result lookup
20797         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
20798         (phi_translate_set): Adjust.
20799         (do_pre_regular_insertion): Likewise.
20800         (do_pre_partial_partial_insertion): Likewise.
20802 2018-03-08  Martin Liska  <mliska@suse.cz>
20804         PR gcov-profile/84735
20805         * doc/gcov.texi: Document usage of profile files.
20806         * gcov-io.h: Document changes in the format.
20808 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
20810         PR debug/84404
20811         PR debug/84408
20812         * dwarf2out.c (struct dw_line_info_table): Update comments for
20813         view == -1.
20814         (FORCE_RESET_NEXT_VIEW): New.
20815         (FORCE_RESETTING_VIEW_P): New.
20816         (RESETTING_VIEW_P): Check for -1 too.
20817         (ZERO_VIEW_P): Likewise.
20818         (new_line_info_table): Force-reset next view.
20819         (dwarf2out_begin_function): Likewise.
20820         (dwarf2out_source_line): Simplify zero_view_p initialization.
20821         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
20822         view directly.  Omit view when omitting .loc at line 0.
20824 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
20826         PR tree-optimization/84740
20827         * tree-switch-conversion.c (process_switch): Call build_constructors
20828         only if info.phi_count is non-zero.
20830         PR tree-optimization/84739
20831         * tree-tailcall.c (find_tail_calls): Check call arguments against
20832         DECL_ARGUMENTS (current_function_decl) rather than
20833         DECL_ARGUMENTS (func) when checking for tail recursion.
20835 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
20837         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
20838         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
20839         Volker Reichelt's entry and add entries for people that perform
20840         GCC fuzzy testing and report numerous bugs.
20842 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
20844         PR target/82411
20845         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
20846         readonly data in sdata, if that is disabled.
20847         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
20848         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
20849         -mreadonly-in-sdata option.
20851 2018-03-07  Martin Sebor  <msebor@redhat.com>
20853         PR tree-optimization/84468
20854         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
20855         basic block when looking for nul assignment.
20857 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
20859         PR target/84277
20860         * except.h (output_function_exception_table): Adjust prototype.
20861         * except.c (output_function_exception_table): Remove FNNAME parameter
20862         and add SECTION parameter.  Ouput one part of the table at a time.
20863         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
20864         the first part of the exception table and emit unwind directives.
20865         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
20866         (i386_pe_seh_cold_init): Likewise.
20867         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
20868         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
20869         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
20870         (ix86_output_call_insn): Emit a nop in one more case for SEH.
20871         * config/i386/winnt.c: Include except.h.
20872         (struct seh_frame_state): Add reg_offset, after_prologue and
20873         in_cold_section fields.
20874         (i386_pe_seh_end_prologue): Set seh->after_prologue.
20875         (i386_pe_seh_cold_init): New function.
20876         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
20877         to seh->in_cold_section.
20878         (seh_emit_push): Record the offset of the push.
20879         (seh_emit_save): Record the offet of the save.
20880         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
20881         Test seh->after_prologue to disregard the epilogue.
20882         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
20883         (i386_pe_end_cold_function): New function.
20885 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
20887         PR fortran/84565
20888         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
20889         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
20891         PR c++/84704
20892         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
20893         on tmp_var.
20894         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
20895         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
20897         PR middle-end/84723
20898         * multiple_target.c: Include tree-inline.h and intl.h.
20899         (expand_target_clones): Diagnose and fail if node->definition and
20900         !tree_versionable_function_p (node->decl).
20902 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
20904         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
20905         sprint_ul.
20906         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
20907         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20908         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
20910 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
20912         PR target/84710
20913         * combine.c (try_combine): Use reg_or_subregno instead of handling
20914         just paradoxical SUBREGs and REGs.
20916 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
20918         * config/arc/arc.c (arc_finalize_pic): Remove function.
20919         (arc_must_save_register): We use single base PIC register, remove
20920         checks to save/restore the PIC register.
20921         (arc_expand_prologue): Likewise.
20922         * config/arc/arc-protos.h (arc_set_default_type_attributes):
20923         Remove.
20924         (arc_verify_short): Likewise.
20925         (arc_attr_type): Likewise.
20926         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
20927         (walk_stores): Likewise.
20928         (arc_address_cost): Make it static.
20929         (arc_verify_short): Likewise.
20930         (branch_dest): Likewise.
20931         (arc_attr_type): Likewise.
20932         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
20933         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
20934         (arc_final_prescan_insn): Remove inserting the nops due to
20935         hardware hazards.  It is done in reorg step.
20936         (insn_length_variant_t): Remove.
20937         (insn_length_parameters_t): Likewise.
20938         (arc_insn_length_parameters): Likewise.
20939         (arc_get_insn_variants): Likewise.
20940         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
20942 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
20944         PR inline-asm/84683
20945         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
20946         assertion failure.
20948         PR tree-optimization/84687
20949         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
20950         on new_node->decl.
20951         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
20953 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20955         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
20956         Rename to ppc_speculation_barrier.
20957         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
20958         __builtin_ppc_speculation_barrier.
20960 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
20962         PR target/84700
20963         * combine.c (combine_simplify_rtx): Don't try to simplify if
20964         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
20965         are equal to x.
20967 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
20969         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
20970         to 32 bytes when compiling for POWER9.
20972 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
20974         PR target/84564
20975         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
20976         regparm >= 3 with no arg reg available also for calls with
20977         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
20979         PR target/84524
20980         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
20981         orig,vex.
20982         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
20984 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
20986         PR target/84264
20987         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
20989 2018-03-05  Richard Biener  <rguenther@suse.de>
20991         PR tree-optimization/84486
20992         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
20993         When inserting a __builtin_assume_aligned call set the LHS
20994         SSA name alignment info accordingly.
20996 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
20998         PR tree-optimization/84114
20999         * config/aarch64/aarch64.c (aarch64_reassociation_width)
21000         Avoid reassociation of FLOAT_MODE addition.
21002 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
21004         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
21005         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
21006         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
21007         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
21008         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
21009         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
21010         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
21011         and -mwbnoinvd.
21012         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
21013         __builtin_ia32_wbinvd): New builtins.
21014         (SPECIAL_ARGS2): New.
21015         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
21016         (SPECIAL_ARGS2): New.
21017         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
21018         (ix86_valid_target_attribute_inner_p): Ditto.
21019         (ix86_init_mmx_sse_builtins): Add special_args2.
21020         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
21021         TARGET_WBNOINVD_P): New.
21022         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
21023         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
21024         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
21025         * config/i386/immintrin.h (_wbinvd): New intrinsic.
21026         * config/i386/pconfigintrin.h: New file.
21027         * config/i386/wbnoinvdintrin.h: Ditto.
21028         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
21029         wbnoinvdintrin.h.
21030         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
21032 2018-03-05  Richard Biener  <rguenther@suse.de>
21034         PR tree-optimization/84670
21035         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21036         member.
21037         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21038         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21039         (compute_antic_aux): Only assert the number of values in ANTIC_IN
21040         doesn't grow if all successors (recursively) were visited at least
21041         once.
21043 2018-03-05  Richard Biener  <rguenther@suse.de>
21045         PR tree-optimization/84650
21046         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
21047         if executed in the loop pipeline.
21049 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
21051         * doc/configfiles.texi (Configuration Files): Move info about
21052         conditionalizing $target-protos.h to...
21053         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
21054         differs from $target-protos.h.
21056 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
21057             Chung-Ju Wu  <jasonwucj@gmail.com>
21059         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
21060         * config/nds32/nds32-multiple.md (setmemsi): Define.
21061         * config/nds32/nds32-memory-manipulation.c
21062         (nds32_gen_dup_4_byte_to_word_value): New.
21063         (emit_setmem_word_loop): New.
21064         (emit_setmem_byte_loop): New.
21065         (nds32_expand_setmem_loop): New.
21066         (nds32_expand_setmem_loop_v3m): New.
21067         (nds32_expand_setmem_unroll): New.
21068         (nds32_expand_setmem): New.
21070 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21071             Chung-Ju Wu  <jasonwucj@gmail.com>
21073         * config/nds32/nds32-memory-manipulation.c
21074         (nds32_emit_load_store): New.
21075         (nds32_emit_post_inc_load_store): New.
21076         (nds32_emit_mem_move): New.
21077         (nds32_emit_mem_move_block): New.
21078         (nds32_expand_movmemsi_loop_unknown_size): New.
21079         (nds32_expand_movmemsi_loop_known_size): New.
21080         (nds32_expand_movmemsi_loop): New.
21081         (nds32_expand_movmemsi_unroll): New.
21082         (nds32_expand_movmemqi): Rename ...
21083         (nds32_expand_movmemsi): ... to this.
21084         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
21085         (movmemsi): ... to this.
21086         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
21087         (nds32_expand_movmemsi): ... to this.
21089 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21090             Monk Chiang  <sh.chiang04@gmail.com>
21091             Chung-Ju Wu  <jasonwucj@gmail.com>
21093         * config/nds32/nds32-protos.h
21094         (nds32_expand_load_multiple): New arguments.
21095         (nds32_expand_store_multiple): Ditto.
21096         (nds32_valid_multiple_load_store): Rename ...
21097         (nds32_valid_multiple_load_store_p): ... to this.
21098         * config/nds32/nds32-memory-manipulation.c
21099         (nds32_expand_load_multiple): Refine implementation.
21100         (nds32_expand_store_multiple): Ditto.
21101         * config/nds32/nds32-multiple.md
21102         (load_multiple): Update nds32_expand_load_multiple interface.
21103         (store_multiple): Update nds32_expand_store_multiple interface.
21104         * config/nds32/nds32-predicates.c
21105         (nds32_valid_multiple_load_store): Rename ...
21106         (nds32_valid_multiple_load_store_p): ... to this and refine
21107         implementation.
21108         * config/nds32/predicates.md
21109         (nds32_load_multiple_and_update_address_operation): New predicate.
21110         (nds32_store_multiple_and_update_address_operation): New predicate.
21112 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21113             Chung-Ju Wu  <jasonwucj@gmail.com>
21115         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
21116         (combo): New attribute.
21117         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
21119 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21121         * config/nds32/nds32.opt: Change -mcmodel= default value.
21123 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
21124             Monk Chiang  <sh.chiang04@gmail.com>
21125             Chung-Ju Wu  <jasonwucj@gmail.com>
21127         * config/nds32/constants.md (unspec_element): New enum.
21128         * config/nds32/constraints.md (Umw): New constraint.
21129         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
21130         * config/nds32/nds32-intrinsic.md: Likewise.
21131         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
21132         (nds32_valid_smw_lwm_base_p): New.
21133         (nds32_output_smw_single_word): New.
21134         (nds32_output_lmw_single_word): New.
21135         (nds32_expand_unaligned_load): New.
21136         (nds32_expand_unaligned_store): New.
21137         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
21138         (nds32_output_smw_single_word): Declare.
21139         (nds32_output_lmw_single_word): Declare.
21140         (nds32_expand_unaligned_load): Declare.
21141         (nds32_expand_unaligned_store): Declare.
21142         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
21143         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
21144         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
21145         NDS32_BUILTIN_UASTORE_DW.
21146         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
21147         predicate.
21149 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
21150             Kito Cheng  <kito.cheng@gmail.com>
21151             Chung-Ju Wu  <jasonwucj@gmail.com>
21153         * config/nds32/nds32-intrinsic.c
21154         (nds32_expand_builtin_null_ftype_reg): Delete.
21155         (nds32_expand_builtin_reg_ftype_imm): Ditto.
21156         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
21157         (nds32_read_argument): New.
21158         (nds32_legitimize_target): Ditto.
21159         (nds32_legitimize_argument): Ditto.
21160         (nds32_check_constant_argument): Ditto.
21161         (nds32_expand_unop_builtin): Ditto.
21162         (nds32_expand_unopimm_builtin): Ditto.
21163         (nds32_expand_binop_builtin): Ditto.
21164         (nds32_builtin_decl_impl): Ditto.
21165         (builtin_description): Ditto.
21166         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
21167         (nds32_init_builtins_impl): Ditto.
21168         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
21169         (nds32_builtin_decl): New.
21170         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
21171         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
21173 2018-03-02  Jeff Law  <law@redhat.com>
21175         * reorg.c (stop_search_p): Handle DEBUG_INSN.
21176         (redundant_insn, fill_simple_delay_slots): Likewise.
21177         (fill_slots_from_thread): Likewise.
21178         * resource.c (mark_referenced_resources): Likewise.
21179         (mark_set_resources, find_dead_or_set_registers): Likewise.
21181 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21183         * substring-locations.h (format_warning_va): Formatting fix for
21184         ATTRIBUTE_GCC_DIAG.
21185         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
21186         argument.
21187         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
21188         * substring-locations.c: Include intl.h.
21189         (format_warning_va): Turned into small wrapper around
21190         format_warning_n_va, renamed to ...
21191         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
21192         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
21193         use ngettext.
21194         (format_warning_at_substring_n): New function.
21195         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
21196         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
21197         format_warning_at_substring with just a shorter name instead of
21198         const function pointer.
21199         (fmtwarn_n): New function.
21200         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
21201         appropriate, get rid of all the fmtstr temporaries, move conditionals
21202         with G_() wrapped string literals directly into fmtwarn arguments,
21203         cast dir.len to (int), formatting fixes.
21205 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
21207         * doc/invoke.texi: Remove "Cilk Plus" references.
21209 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21210             Richard Biener  <rguenther@suse.de>
21212         PR ipa/84628
21213         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
21214         for error or warning attributes if CALL_FROM_THUNK_P is set.
21215         Formatting fixes.
21217 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21219         PR target/56540
21220         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
21221         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
21223         PR target/56540
21224         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
21225         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
21227         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
21228         instead of -1U in last predictors element's probability member.
21230 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21232         PR ipa/83983
21233         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21234         arguments if they are comparable.
21236 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21238         PR tree-optimization/84634
21239         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21240         masks and masked_loop_p with a single loop_masks, making sure it's
21241         null for bb vectorization.
21243 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21245         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21246         (vect_analyze_data_ref_access): Use loop->safe_len rather than
21247         loop->force_vectorize to check whether there is no alias.
21249 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21251         PR target/84614
21252         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21253         prototypes.
21254         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21255         comments.
21256         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21257         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21258         instead of a loop around prev_real_insn.
21259         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21260         prev_real_insn.
21262         PR inline-asm/84625
21263         * config/i386/i386.c (ix86_print_operand): Use conditional
21264         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21265         zero vector.
21267 2018-03-02  Richard Biener  <rguenther@suse.de>
21269         PR tree-optimization/84427
21270         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21271         (bitmap_set_subtract_values): Rewrite to handle multiple
21272         exprs per value.
21273         (clean): Likewise.
21274         (prune_clobbered_mems): Likewise.
21275         (phi_translate): Take edge instead of pred/phiblock.
21276         (phi_translate_1): Likewise.
21277         (phi_translate_set): Likewise.  Insert all translated
21278         exprs for a value into the set, keeping possibly multiple
21279         expressions per value.
21280         (compute_antic_aux): Adjust for phi_translate changes.
21281         When intersecting union the expressions and prune those
21282         not in the final value set, keeping possibly multiple
21283         expressions per value.  Do not use value-insertion
21284         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
21285         all expressions.  Add verification that the value-sets
21286         only shrink during iteration.
21287         (compute_partial_antic_aux): Adjust for the phi_translate changes.
21288         (do_pre_regular_insertion): Likewise.
21289         (do_pre_partial_partial_insertion): Likewise.
21291 2018-03-02  Richard Biener  <rguenther@suse.de>
21293         PR target/82005
21294         * config/darwin.c (saved_debug_info_level): New static global.
21295         (darwin_asm_lto_start): Disable debug info generation for LTO out.
21296         (darwin_asm_lto_end): Restore debug info generation settings.
21298 2018-03-01  Martin Liska  <mliska@suse.cz>
21300         PR sanitizer/82484
21301         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21302         volatile arguments.
21304 2018-03-01  Richard Biener  <rguenther@suse.de>
21306         PR debug/84645
21307         * dwarf2out.c (gen_variable_die): Properly handle late VLA
21308         type annotation with LTO when debug was disabled at compile-time.
21310 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
21312         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21313         XINT with INTVAL.
21314         (mips_final_postscan_insn): Likewise.
21316 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
21318         PR rtl-optimization/84528
21319         * alias.c (init_alias_target): Add commentary.
21320         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21321         a unique base value if the frame pointer is not eliminated
21322         to the stack pointer.
21324 2018-03-01  Tom de Vries  <tom@codesourcery.com>
21326         PR rtl-optimization/83327
21327         * lra-int.h (hard_regs_spilled_into): Declare.
21328         * lra.c (hard_regs_spilled_into): Define.
21329         (init_reg_info): Init hard_regs_spilled_into.
21330         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21331         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21332         (process_bb_lives): Handle hard_regs_spilled_into.
21333         (lra_create_live_ranges_1): Before doing liveness propagation, clear
21334         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21336 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
21338         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
21339         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
21340         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
21341         * config/rs6000/aix72.h: New file.
21343 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
21345         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
21346         instead of warning_at with conditional singular and plural messages
21347         where possible.
21349         PR target/52991
21350         * stor-layout.c (update_alignment_for_field): For
21351         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
21352         && !DECL_PACKED (field), do the alignment update, just use
21353         only desired_align instead of MAX (type_align, desired_align)
21354         as the alignment.
21355         (place_field): Don't do known_align < desired_align handling
21356         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
21357         is non-NULL, instead do it after rli->prev_field handling and
21358         only if not within a bitfield word.  For DECL_PACKED (field)
21359         use type_align of BITS_PER_UNIT.
21361 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
21363         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
21364         superfluous parentheses and trailing spaces.
21366 2018-02-28  Richard Biener  <rguenther@suse.de>
21368         PR tree-optimization/84584
21369         * graphite-scop-detection.c (scop_detection::add_scop): Discard
21370         SCoPs with fake exit edge.
21372 2018-02-28  Martin Liska  <mliska@suse.cz>
21374         PR testsuite/84597
21375         * timevar.c (timer::print): Fix format to properly print 100%
21376         values.
21378 2018-02-28  Richard Biener  <rguenther@suse.de>
21380         PR middle-end/84607
21381         * genmatch.c (capture_info::walk_match): Do not mark
21382         captured expressions without operands as expr_p given
21383         they act more like predicates and should be subject to
21384         "lost tail" side-effect preserving.
21386 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
21388         PR rtl-optimization/81611
21389         * auto-inc-dec.c (attempt_change): Move dead note from
21390         mem_insn if it's the next use of regno
21391         (find_address): Take address use of reg holding
21392         non-incremented value.  Add parm to limit search to the named
21393         reg only.
21394         (merge_in_block): Attempt to use a mem insn that is the next
21395         use of the original regno.
21397 2018-02-27  Martin Sebor  <msebor@redhat.com>
21399         PR c++/83871
21400         * doc/invoke.texi (-Wmissing-attributes): New option.
21401         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
21403 2018-02-27  Martin Sebor  <msebor@redhat.com>
21405         PR translation/84207
21406         * diagnostic-core.h (warning_n, error_n, inform_n): Change
21407         n argument to unsigned HOST_WIDE_INT.
21408         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
21409         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
21410         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
21411         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
21413 2018-02-27  Richard Biener  <rguenther@suse.de>
21415         PR tree-optimization/84512
21416         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21417         Do not use the estimate returned from record_stmt_cost for
21418         the scalar iteration cost but sum properly using add_stmt_cost.
21420 2018-02-27  Richard Biener  <rguenther@suse.de>
21422         PR tree-optimization/84466
21423         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21424         Adjust last change to less strictly validate use operands.
21426 2018-02-27  Martin Liska  <mliska@suse.cz>
21428         PR gcov-profile/84548
21429         * gcov.c (process_file): Allow partial overlap and consider it
21430         also as group functions.
21431         (output_lines): Properly calculate range of lines for a group.
21433 2018-02-27  Martin Liska  <mliska@suse.cz>
21435         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
21436         'ggc' suffixes.  Change first column width.
21437         (timer::print): Fix formatting of the column.
21439 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
21441         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
21442         preserve inline entry blocks for the sake of debug inline
21443         entry point markers alone.
21444         (remove_unused_locals): Suggest in comments a better place to
21445         force the preservation of inline entry blocks that are
21446         otherwise unused, but do not preserve them.
21448 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21450         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
21452 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21454         PR target/84039
21455         * config/i386/constraints.md (Bs): Replace
21456         ix86_indirect_branch_register with
21457         TARGET_INDIRECT_BRANCH_REGISTER.
21458         (Bw): Likewise.
21459         * config/i386/i386.md (indirect_jump): Likewise.
21460         (tablejump): Likewise.
21461         (*sibcall_memory): Likewise.
21462         (*sibcall_value_memory): Likewise.
21463         Peepholes of indirect call and jump via memory: Likewise.
21464         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
21465         (*sibcall_value_GOT_32): Likewise.
21466         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21467         (GOT_memory_operand): Likewise.
21468         (call_insn_operand): Likewise.
21469         (sibcall_insn_operand): Likewise.
21470         (GOT32_symbol_operand): Likewise.
21471         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
21473 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21475         PR rtl-optimization/83496
21476         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
21477         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
21478         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
21479         redundant insn, if any.
21480         (relax_delay_slots): Likewise.
21481         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
21483 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
21485         PR tree-optimization/83965
21486         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
21487         that grouped statements are part of a reduction chain.  Return
21488         true if the statement is not marked as a reduction itself but
21489         is part of a group.
21490         (vect_recog_dot_prod_pattern): Don't check whether the statement
21491         is part of a group here.
21492         (vect_recog_sad_pattern): Likewise.
21493         (vect_recog_widen_sum_pattern): Likewise.
21495 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21497         PR debug/84545
21498         * final.c (rest_of_clean_state): Also look for calls inside sequences.
21500 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21502         PR target/84530
21503         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
21504         the bool argument.
21505         (ix86_output_indirect_function_return): New prototype.
21506         (ix86_split_simple_return_pop_internal): Likewise.
21507         * config/i386/i386.c (indirect_return_via_cx): New.
21508         (indirect_return_via_cx_bnd): Likewise.
21509         (indirect_thunk_name): Handle return va CX_REG.
21510         (output_indirect_thunk_function): Create alias for
21511         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
21512         (ix86_output_indirect_jmp): Remove the bool argument.
21513         (ix86_output_indirect_function_return): New function.
21514         (ix86_split_simple_return_pop_internal): Likewise.
21515         * config/i386/i386.md (*indirect_jump): Don't pass false
21516         to ix86_output_indirect_jmp.
21517         (*tablejump_1): Likewise.
21518         (simple_return_pop_internal): Change it to define_insn_and_split.
21519         Call ix86_split_simple_return_pop_internal to split it for
21520         -mfunction-return=.
21521         (simple_return_indirect_internal): Call
21522         ix86_output_indirect_function_return instead of
21523         ix86_output_indirect_jmp.
21525 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
21527         PR bootstrap/84405
21528         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
21529         memset and value initialization afterwards.
21531 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
21533         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
21535 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
21537         PR target/84521
21538         * common/config/aarch64/aarch64-common.c
21539         (aarch_option_optimization_table[]): Switch
21540         off fomit-frame-pointer
21542 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
21543             Chung-Ju Wu  <jasonwucj@gmail.com>
21545         * config/nds32/nds32-multiple.md (load_multiple): Disallow
21546         volatile memory.
21547         (store_multiple): Ditto.
21549 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
21551         * config.gcc: Add --with-cpu support for nds32 target.
21552         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
21553         * config/nds32/nds32.opt: Add -mcpu= option.
21555 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
21557         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
21558         isel=yes): Warn for these deprecated options.
21560 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
21562         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
21563         ISA_2_5_MASKS_EMBEDDED.
21565 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
21567         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
21568         p->max as pointers rather than using iterative_hash_expr.
21570 2018-02-23  Carl Love  <cel@us.ibm.com>
21572         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
21573         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
21574         BU_P8V_OVERLOAD_2.
21575         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
21576         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
21577         P8V_BUILTIN_VEC_VUNSIGNED2.
21579 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
21581         PR target/81572
21582         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
21583         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
21584         LRA_UNKNOWN_ALT.
21585         * lra-constraints.c (curr_insn_transform): Set up
21586         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
21587         LRA_UNKNOWN_ALT.
21588         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
21589         * lra-eliminations.c (spill_pseudos): Ditto.
21590         (process_insn_for_elimination): Ditto.
21591         * lra-lives.c (reg_early_clobber_p): Use the new macros.
21592         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
21593         LRA_NON_CLOBBERED_ALT.
21595 2018-02-22  Martin Sebor  <msebor@redhat.com>
21597         PR tree-optimization/84480
21598         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
21599         to maybe_diag_stxncpy_trunc.  Call it.
21600         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
21601         from gimple_fold_builtin_strcpy.  Print inlining stack.
21602         (handle_builtin_stxncpy): Print inlining stack.
21603         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
21605 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
21607         PR target/84176
21608         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
21609         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
21610         and -fcheck-pointer-bounds are used together.
21611         (indirect_thunk_prefix): New enum.
21612         (indirect_thunk_need_prefix): New function.
21613         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
21614         "_nt" instead of "_bnd" for NOTRACK prefix.
21615         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
21616         (output_indirect_thunk_function): Likewise.
21617         (): Likewise.
21618         (ix86_code_end): Update output_indirect_thunk_function calls.
21619         (ix86_output_indirect_branch_via_reg): Replace
21620         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
21621         (ix86_output_indirect_branch_via_push): Likewise.
21622         (ix86_output_function_return): Likewise.
21623         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
21624         incompatible with -fcf-protection=branch and
21625         -fcheck-pointer-bounds.
21627 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
21629         PR target/83335
21630         * config/aarch64/aarch64.c (aarch64_print_address_internal):
21631         Change gcc_assert call to output_operand_lossage.
21633 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
21635         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
21637 2018-02-22  DJ Delorie  <dj@redhat.com>
21638             Sebastian Perta  <sebastian.perta@renesas.com>
21639             Oleg Endo  <olegendo@gcc.gnu.org>
21641         * config/rx/rx.c (rx_rtx_costs): New function.
21642         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
21644 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21646         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
21648 2018-02-22  Martin Liska  <mliska@suse.cz>
21650         PR driver/83193
21651         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
21652         Add "native" as a possible value.
21654 2018-02-22  Martin Liska  <mliska@suse.cz>
21656         PR driver/83193
21657         * config/i386/i386.c (ix86_option_override_internal):
21658         Add "native" as a possible value for -march and -mtune.
21660 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
21662         PR target/84502
21663         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
21664         to all type variants.
21666         PR tree-optimization/84503
21667         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
21668         width as info->bitpos + info->bitsize - start.
21669         (merged_store_group::merge_overlapping): Simplify width computation.
21670         (check_no_overlap): New function.
21671         (imm_store_chain_info::try_coalesce_bswap): Compute expected
21672         start + width and last_order of the group, fail if check_no_overlap
21673         fails.
21674         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
21675         to group if check_no_overlap fails.
21677 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
21679         * config/rs6000/altivec.md: Delete contraint arguments to
21680         define_expand, define_split, and define_peephole2, and in
21681         define_insn_and_split if always unused.
21682         * config/rs6000/darwin.md: Ditto.
21683         * config/rs6000/dfp.md: Ditto.
21684         * config/rs6000/rs6000.md: Ditto.
21685         * config/rs6000/sync.md: Ditto.
21686         * config/rs6000/vector.md: Ditto.
21687         * config/rs6000/vsx.md: Ditto.
21689 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
21691         * config/rs6000/altivec.md: Write output control strings as braced
21692         blocks instead of double-quoted strings.
21693         * config/rs6000/darwin.md: Ditto.
21694         * config/rs6000/rs6000.md: Ditto.
21695         * config/rs6000/vector.md: Ditto.
21696         * config/rs6000/vsx.md: Ditto.
21698 2018-02-21  Jason Merrill  <jason@redhat.com>
21700         PR c++/84314 - ICE with templates and fastcall attribute.
21701         * attribs.c (build_type_attribute_qual_variant): Remove assert.
21703 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
21705         * ipa-cp.c (determine_versionability): Fix comment typos.
21707 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
21709         PR c/84229
21710         * ipa-cp.c (determine_versionability): Do not version functions caling
21711         va_arg_pack.
21713 2018-02-21  Martin Liska  <mliska@suse.cz>
21715         PR driver/83193
21716         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
21717         Add "native" as a possible value.
21718         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
21719         the macro when native cpu detection is available.
21721 2018-02-21  Martin Liska  <mliska@suse.cz>
21723         PR driver/83193
21724         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
21725         Add "native" as a possible value.
21726         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
21727         when native cpu detection is available.
21729 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
21730             Martin Sebor  <msebor@redhat.com>
21732         PR tree-optimization/84478
21733         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
21734         false.
21735         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
21736         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
21737         support which is conservatively correct, for 2 only stay conservative
21738         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
21739         argument to the 2 argument get_range_strlen, adjust 6 arg
21740         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
21741         false.
21742         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
21743         (gimple_fold_builtin_strlen): Pass true as last argument to
21744         get_range_strlen.
21746 2018-02-20  Martin Sebor  <msebor@redhat.com>
21748         PR middle-end/84095
21749         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
21750         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
21751         (builtin_memref::builtin_memref): Factor out parts into
21752         set_base_and_offset and call it.
21754 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
21756         PR middle-end/84406
21757         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
21758         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
21759         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
21760         search at the associated MODE_INT.
21762 2018-02-20  Jeff Law  <law@redhat.com>
21764         PR middle-end/82123
21765         PR tree-optimization/81592
21766         PR middle-end/79257
21767         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
21768         for range data rather than using global data.
21769         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
21770         range data rather than using global data.
21771         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
21772         pass it to children as needed.
21773         (struct directive::fmtresult): Similarly.
21774         (struct directive::set_width): Similarly.
21775         (struct directive::set_precision): Similarly.
21776         (format_integer, format_directive, parse_directive): Similarly.
21777         (format_none): Accept unnamed vr_values parameter.
21778         (format_percent, format_floating, format_character): Similarly.
21779         (format_string, format_plain): Similarly.
21780         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
21781         the EVRP range analyzer for range data rather than using global data.
21782         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
21783         gimple-ssa-evrp-analyze.h
21784         (class sprintf_dom_walker): Add after_dom_children member function.
21785         Add evrp_range_analyzer member.
21786         (sprintf_dom_walker::before_dom_children): Call into the EVRP
21787         range analyzer as needed.
21788         (sprintf_dom_walker::after_dom_children): New member function.
21789         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
21790         if not optimizing.
21791         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
21792         (evrp_range_analyzer::pop_to_marker): Likewise.
21794 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
21796         PR tree-optimization/84419
21797         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
21798         with the required type if its current type is compatible but
21799         different.
21801 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
21803         PR middle-end/82004
21804         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
21805         after vectorization.
21807 2018-02-20  Martin Liska  <mliska@suse.cz>
21809         PR driver/83193
21810         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
21811         possible values if we don't have a hint.
21813 2018-02-20  Martin Liska  <mliska@suse.cz>
21815         PR c/84310
21816         PR target/79747
21817         * final.c (shorten_branches): Build align_tab array with one
21818         more element.
21819         * opts.c (finish_options): Add alignment option limit check.
21820         (MAX_CODE_ALIGN): Likewise.
21821         (MAX_CODE_ALIGN_VALUE): Likewise.
21822         * doc/invoke.texi: Document maximum allowed option value for
21823         all -falign-* options.
21825 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
21827         PR target/84146
21828         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
21829         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
21830         * var-tracking.c (emit_note_insn_var_location): Remove all references
21831         to NOTE_INSN_CALL_ARG_LOCATION.
21832         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
21833         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
21834         Use copy_rtx_if_shared.
21835         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
21836         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
21837         (dwarf2out_var_location): Remove handling of
21838         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
21839         on call_insn.
21840         * final.c (final_scan_insn): Remove all references to
21841         NOTE_INSN_CALL_ARG_LOCATION.
21842         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
21843         before dumping final insns.
21844         * except.c (emit_note_eh_region_end): Remove all references to
21845         NOTE_INSN_CALL_ARG_LOCATION.
21846         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
21847         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
21848         * config/arc/arc.c (hwloop_optimize): Likewise.
21849         * config/arm/arm.c (create_fix_barrier): Likewise.
21850         * config/s390/s390.c (s390_chunkify_start): Likewise.
21851         * config/sh/sh.c (find_barrier): Likewise.
21852         * config/i386/i386.c (rest_of_insert_endbranch,
21853         ix86_seh_fixup_eh_fallthru): Likewise.
21854         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
21855         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
21856         * config/frv/frv.c (frv_function_prologue): Likewise.
21857         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
21858         reg note.
21859         (note_outside_basic_block_p): Remove all references to
21860         NOTE_INSN_CALL_ARG_LOCATION.
21861         * gengtype.c (adjust_field_rtx_def): Likewise.
21862         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
21863         Likewise.
21864         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
21865         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
21867         PR c++/84444
21868         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
21869         is ADDR_EXPR.
21871         PR tree-optimization/84452
21872         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
21873         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
21874         is NULL.
21876 2018-02-19  Martin Liska  <mliska@suse.cz>
21878         PR sanitizer/82183
21879         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
21881 2018-02-19  Martin Liska  <mliska@suse.cz>
21882             Richard Sandiford  <richard.sandiford@linaro.org>
21884         PR tree-optimization/82491
21885         * gimple-fold.c (get_base_constructor): Make earlier bail out
21886         to prevent ubsan.
21888 2018-02-19  Carl Love  <cel@us.ibm.com>
21890         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
21891         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
21892         BU_P8V_OVERLOAD_1.
21893         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
21894         P8V_BUILTIN_VEC_NEG.
21896 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
21898         * config/rl78/rl78.md (movdf): New define expand.
21900 2018-02-19  Martin Liska  <mliska@suse.cz>
21902         PR other/80589
21903         * doc/invoke.texi: Fix typo.
21904         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
21906 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
21908         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
21909         handle rs6000_single_float and rs6000_double_float specially for
21910         e500 family CPUs.
21912 2018-02-16  Jeff Law  <law@redhat.com>
21914         * config/rx/rx.c (add_pop_cfi_notes): New function.;
21915         (pop_regs): Use it.
21917 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
21919         PR ipa/84425
21920         * ipa-inline.c (inline_small_functions): Fix a typo.
21922 2018-02-16  Nathan Sidwell  <nathan@acm.org>
21924         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
21926 2018-02-16  Carl Love  <cel@us.ibm.com>
21928         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
21929         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
21930         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
21931         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
21932         expansion to P8V_BUILTIN_VEC_FLOAT2.
21934 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
21936         PR rtl-optimization/70023
21937         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
21938         src_regno into account.
21940 2018-02-16  Carl Love  <cel@us.ibm.com>
21942         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
21943         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
21944         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
21945         * config/rs6000/rs6000.c: Remove case statements for
21946         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
21947         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
21948         and P9V_BUILTIN_VEC_VINSERT4B.
21949         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
21950         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
21951         * config/rs6000/vsx.md:
21952         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
21953         vec_insert4b.
21955 2018-02-16  Carl Love  <cel@us.ibm.com>
21957         * config/rs6000/altivec.h: Add builtin names vec_extract4b
21958         vec_insert4b.
21959         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
21960         definitions.
21961         * config/rs6000/rs6000-c.c: Add the definitions for
21962         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
21963         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
21964         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
21965         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
21966         definition for insert4b and define insn *insert3b_internal.
21967         * doc/extend.texi: Add documentation for vec_extract4b.
21969 2018-02-16  Nathan Sidwell  <nathan@acm.org>
21971         * doc/extend.texi (Backwards Compatibility): Mention friend
21972         injection.  Note for-scope is deprecated.
21973         * doc/invoke.texi (-ffriend-injection): Deprecate.
21975 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
21977         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
21978         that moved to I2, also allow destinations that are a paradoxical
21979         subreg (instead of a normal reg).
21981 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
21983         PR target/83831
21984         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
21985         to QImode.
21987 2018-02-16  Richard Biener  <rguenther@suse.de>
21989         PR tree-optimization/84037
21990         PR tree-optimization/84016
21991         PR target/82862
21992         * config/i386/i386.c (ix86_builtin_vectorization_cost):
21993         Adjust vec_construct for the fact we need additional higher latency
21994         128bit inserts for AVX256 and AVX512 vector builds.
21995         (ix86_add_stmt_cost): Scale vector construction cost for
21996         elementwise loads.
21998 2018-02-16  Richard Biener  <rguenther@suse.de>
22000         PR tree-optimization/84417
22001         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
22002         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
22003         (non_rewritable_lvalue_p): Likewise, use poly-ints.
22005 2018-02-16  Martin Liska  <mliska@suse.cz>
22007         PR sanitizer/84307
22008         * internal-fn.def (ASAN_CHECK): Set proper flags.
22009         (ASAN_MARK): Likewise.
22011 2018-02-16  Julia Koval  <julia.koval@intel.com>
22013         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
22014         from PTA_CANNONLAKE.
22016 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22018         PR target/84272
22019         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
22020         Use ++iter rather than iter++ for std::list iterators.
22021         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
22022         defer deleting them until all nodes in the forest are processed.  Do
22023         free even leaf nodes.  Change to_process into auto_vec.
22025         PR bootstrap/84405
22026         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
22027         * vec.h (vec_default_construct): Use memset instead of placement new
22028         if BROKEN_VALUE_INITIALIZATION is defined.
22029         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
22030         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
22031         is defined.
22033         PR rtl-optimization/83723
22034         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
22035         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
22036         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
22037         recursive calls.
22038         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
22039         callers.
22040         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
22042 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
22044         PR rtl-optimization/81443
22045         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
22046         from inner REGs to paradoxical SUBREGs.
22048 2018-02-16  Richard Biener  <rguenther@suse.de>
22050         PR tree-optimization/84399
22051         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22052         For operands we can analyze at their definition make sure we can
22053         analyze them at each use as well.
22055 2018-02-16  Richard Biener  <rguenther@suse.de>
22057         PR tree-optimization/84190
22058         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
22059         volatile accesses if the decl isn't volatile.
22061 2018-02-15  Jason Merrill  <jason@redhat.com>
22063         PR c++/84314 - ICE with templates and fastcall attribute.
22064         * attribs.c (build_type_attribute_qual_variant): Don't clobber
22065         TYPE_CANONICAL on an existing type.
22067 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
22069         PR tree-optimization/84383
22070         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
22071         dstoff nor call operand_equal_p if dstbase is NULL.
22073         PR tree-optimization/84334
22074         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
22075         also a CONSTANT_CLASS_P, punt.
22077 2018-02-14  Jim Wilson  <jimw@sifive.com>
22079         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
22080         first SMALL_OPERAND check.  New local min_second_step.  Move assert
22081         to where locals are set.  Add TARGET_RVC support.
22082         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
22084 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
22086         * doc/invoke.texi: Correct -Wformat-overflow code sample.
22088 2018-02-14  Martin Sebor  <msebor@redhat.com>
22090         PR tree-optimization/83698
22091         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
22092         arrays constrain the offset range to their bounds.
22093         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
22094         (builtin_access::overlap): Avoid setting the size of overlap if it's
22095         already been set.
22096         (maybe_diag_overlap): Also consider arrays when deciding what values
22097         of offsets to include in diagnostics.
22099 2018-02-14  Martin Sebor  <msebor@redhat.com>
22101         PR c/84108
22102         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
22103         that correspond to the kind of a declaration.
22105 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
22107         PR target/83984
22108         * config/pa/pa.md: Load address of PIC label using the linkage table
22109         if the label is nonlocal.
22111 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22113         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
22114         warning message if user requests -maltivec=be.
22115         * doc/invoke.texi: Document deprecation of -maltivec=be.
22117 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
22119         PR target/84220
22120         * config/rs6000/rs6000-c.c: Update definitions for
22121         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
22122         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
22124 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22126         PR target/84239
22127         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
22128         add _get_ssp intrinsics. Remove argument from
22129         __builtin_ia32_rdssp[d|q].
22130         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
22131         * config/i386/i386-builtin.def: Remove argument from
22132         __builtin_ia32_rdssp[d|q].
22133         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
22134         ix86_expand_special_args_builtin for _rdssp[d|q].
22135         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
22136         Clear register before usage.
22137         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
22138         Add documentation for new _get_ssp and _inc_ssp intrinsics.
22140 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
22142         PR tree-optimization/84357
22143         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
22144         operand 1 of an ARRAY_REF too.
22146 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
22148         PR target/83831
22149         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
22150         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22151         declarations.
22152         (set_of_reg): New struct.
22153         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
22154         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
22155         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22156         functions.
22157         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
22158         Split into bitclr, bitset, bitinvert patterns if appropriate.
22159         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
22160         use rx_fuse_in_memory_bitop.
22161         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
22162         to named insn, correct maximum insn length.
22164 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
22166         PR target/79242
22167         * machmode.def: Define a complex mode for PARTIAL_INT.
22168         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
22169         MODE_PARTIAL_INT.
22170         * doc/rtl.texi: Document CSPImode.
22171         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
22172         handling.
22173         (msp430_hard_regno_nregs_with_padding): Likewise.
22175 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
22177         PR target/84279
22178         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
22180 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
22182         PR rtl-optimization/84169
22183         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
22184         we generated a parallel as new i3 and we split that to new i2 and i3
22185         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
22186         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
22187         those to i2, not i1.  Partially rewrite this scan code.
22189 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22191         PR c/82210
22192         * stor-layout.c (place_field): For variable length fields, adjust
22193         offset_align afterwards not just based on the field's alignment,
22194         but also on the size.
22196         PR middle-end/84309
22197         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
22198         of exps and logs in the use_exp2 case.
22200 2018-02-13  Jeff Law  <law@redhat.com>
22202         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
22203         entry for "vector".
22205         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
22206         ARGS as unused.
22208 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
22210         PR debug/84342
22211         PR debug/84319
22212         * common.opt (gas-loc-support, gas-locview-support): New.
22213         (ginline-points, ginternal-reset-location-views): New.
22214         * doc/invoke.texi: Document them.  Use @itemx where intended.
22215         (gvariable-location-views): Adjust.
22216         * target.def (reset_location_view): New.
22217         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
22218         (TARGET_RESET_LOCATION_VIEW): New.
22219         * doc/tm.texi: Rebuilt.
22220         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
22221         (dwarf2out_default_as_locview_support): New.
22222         (output_asm_line_debug_info): Use option variables.
22223         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
22224         (output_loc_list): Likewise.
22225         (add_high_low_attributes): Check option variables.
22226         Don't output entry view attribute in strict mode.
22227         (gen_inlined_subroutine_die): Check option variables.
22228         (dwarf2out_inline_entry): Likewise.
22229         (init_sections_and_labels): Likewise.
22230         (dwarf2out_early_finish): Likewise.
22231         (maybe_reset_location_view): New, from...
22232         (dwarf2out_var_location): ... here.  Call it.
22233         * debug.h (dwarf2out_default_as_loc_support): Declare.
22234         (dwarf2out_default_as_locview_support): Declare.
22235         * hooks.c (hook_int_rtx_insn_0): New.
22236         * hooks.h (hook_int_rtx_insn_0): Declare.
22237         * toplev.c (process_options): Take -gas-loc-support and
22238         -gas-locview-support from dwarf2out.  Enable
22239         -gvariable-location-views by default only with locview
22240         assembler support.  Enable -ginternal-reset-location-views by
22241         default only if the target defines the corresponding hook.
22242         Enable -ginline-points by default if location views are
22243         enabled; force it disabled if statement frontiers are
22244         disabled.
22245         * tree-inline.c (expand_call_inline): Check option variables.
22246         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22248 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
22250         PR tree-optimization/84321
22251         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22252         handling.  Also check whether the anti-range contains any values
22253         that satisfy the mask; switch to a VR_RANGE if not.
22255 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
22257         PR sanitizer/84340
22258         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22260 2018-02-13  Martin Jambor  <mjambor@suse.cz>
22262         PR c++/83990
22263         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22264         of call statements, also set location of a load to a temporary.
22266 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
22268         * config/rl78/rl78.c (add_vector_labels): New function.
22269         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22270         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22271         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22272         which checks that no arguments are passed.
22273         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22274         * doc/extend.texi: Documentation for the new attribute.
22276 2018-02-13  Andreas Schwab  <schwab@suse.de>
22278         * config/riscv/linux.h (CPP_SPEC): Define.
22280 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22282         PR target/84335
22283         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
22284         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
22285         OPTION_MASK_ISA_AES as first argument to def_builtin_const
22286         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
22287         instead of OPTION_MASK_ISA_PCLMUL as first argument to
22288         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
22289         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
22290         temporarily for AES and PCLMUL builtins.
22292         PR tree-optimization/84339
22293         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22294         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22295         Formatting fixes.
22297         PR middle-end/84309
22298         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22299         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22300         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22301         inline function.
22302         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22303         inline function.
22304         * omp-simd-clone.h: New file.
22305         * omp-simd-clone.c: Include omp-simd-clone.h.
22306         (expand_simd_clones): No longer static.
22307         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22308         cgraph.h and omp-simd-clone.h.
22309         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22310         (vect_recog_widen_shift_pattern): Formatting fix.
22311         (vect_pattern_recog_1): Don't check optab for calls.
22313         PR target/84336
22314         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22315         operands[2] into a REG before using gen_lowpart on it.
22317 2018-02-12  Jeff Law  <law@redhat.com>
22319         PR target/83760
22320         * config/sh/sh.c (find_barrier): Consider a sibling call
22321         a barrier as well.
22323         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22324         successfully back substituting a reg.
22326 2018-02-12  Richard Biener  <rguenther@suse.de>
22328         PR tree-optimization/84037
22329         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22330         parameter, move visited init to caller.
22331         (vect_slp_analyze_operations): Separate cost from validity
22332         check, initialize visited once for all instances.
22333         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22334         for all instances.
22335         * tree-vect-stmts.c (vect_model_simple_cost): Make early
22336         out an assert.
22337         (vect_model_promotion_demotion_cost): Likewise.
22338         (vectorizable_bswap): Guard cost modeling with !slp_node
22339         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
22340         SLP stmts.
22341         (vectorizable_call): Likewise.
22342         (vectorizable_conversion): Likewise.
22343         (vectorizable_assignment): Likewise.
22344         (vectorizable_shift): Likewise.
22345         (vectorizable_operation): Likewise.
22346         (vectorizable_store): Likewise.
22347         (vectorizable_load): Likewise.
22348         (vectorizable_condition): Likewise.
22349         (vectorizable_comparison): Likewise.
22351 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
22353         PR sanitizer/84307
22354         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
22355         (ASAN_MARK): Fix fnspec to account for return value, change pointer
22356         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
22358 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22360         PR middle-end/83665
22361         * params.def (inline-min-speedup): Increase from 8 to 15.
22362         (max-inline-insns-auto): Decrease from 40 to 30.
22363         * ipa-split.c (consider_split): Add some buffer for function to
22364         be considered inlining candidate.
22365         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
22366         default values.
22368 2018-02-12  Richard Biener  <rguenther@suse.de>
22370         PR tree-optimization/84037
22371         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
22372         matched stmts if we cannot swap the non-matched ones.
22374 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22376         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
22377         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
22378         _mm_maskz_scalef_round_ss): New intrinsics.
22379         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
22380         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
22381         __builtin_ia32_scalefss_round): Remove.
22382         (__builtin_ia32_scalefsd_mask_round,
22383         __builtin_ia32_scalefss_mask_round): New intrinsics.
22384         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
22385         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
22386         ((match_operand:VF_128 2 "<round_nimm_predicate>"
22387         "<round_constraint>")): Changed to ...
22388         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
22389         "<round_scalar_constraint>")): ... this.
22390         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
22391         %0, %1, %2<round_op3>}"): Changed to ...
22392         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
22393         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
22394         %2<round_scalar_mask_op3>}"): ... this.
22395         * config/i386/subst.md (round_scalar_nimm_predicate): New.
22397 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22399         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
22400         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
22401         (_mm_maskz_sqrt_round_ss): New intrinsics.
22402         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
22403         (__builtin_ia32_sqrtsd_mask_round)
22404         (__builtin_ia32_sqrtss_mask_round): New builtins.
22405         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
22406         (__builtin_ia32_sqrtss_round): Remove.
22407         (__builtin_ia32_sqrtsd_mask_round)
22408         (__builtin_ia32_sqrtss_mask_round): New builtins.
22409         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
22410         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
22411         ((match_operand:VF_128 1 "vector_operand"
22412         "xBm,<round_constraint>")): Changed to ...
22413         ((match_operand:VF_128 1 "vector_operand"
22414         "xBm,<round_scalar_constraint>")): ... this.
22415         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
22416         %0, %2, %<iptr>1<round_op3>}): Changed to ...
22417         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
22418         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
22419         %<iptr>1<round_scalar_mask_op3>}): ... this.
22420         ((set_attr "prefix" "<round_prefix>")): Changed to ...
22421         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
22423 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
22425         PR target/84266
22426         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
22427         Cast vec_cmpeq result to correct type.
22428         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
22429         Cast vec_cmpgt result to correct type.
22431 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
22433         * final.c (final_scan_insn_1): Renamed from...
22434         (final_scan_insn): ... this.  New wrapper, to recover
22435         seen from the outermost call in recursive ones.
22436         * config/sparc/sparc.c (output_return): Drop seen from call.
22437         (output_sibcall): Likewise.
22438         * config/visium/visium.c (output_branch): Likewise.
22440 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
22442         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
22443         function label.
22445 2018-02-10  Alan Modra  <amodra@gmail.com>
22447         PR target/84300
22448         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
22449         Specify LR as an input.
22451 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
22453         PR sanitizer/83987
22454         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
22455         remove_member_access_dummy_vars): New functions.
22456         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
22457         lower_omp_1, execute_lower_omp): Use them.
22459         PR rtl-optimization/84308
22460         * shrink-wrap.c (spread_components): Release todo vector.
22462 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
22464         PR rtl-optimization/57193
22465         * ira-color.c (struct allocno_color_data): Add member
22466         conflict_allocno_hard_prefs.
22467         (update_conflict_allocno_hard_prefs): New.
22468         (bucket_allocno_compare_func): Add a preference based on
22469         conflict_allocno_hard_prefs.
22470         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
22471         (color_allocnos): Remove a dead code.  Initiate
22472         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
22474 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22476         PR target/84226
22477         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
22478         constraint from =wa to wa.  Avoid a subreg on the output operand,
22479         instead use a pseudo and subreg it in a move.
22480         (p9_xxbrd_<mode>): Changed to ...
22481         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
22482         (p9_xxbrd_v2df): New expander.
22483         (p9_xxbrw_<mode>): Changed to ...
22484         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
22485         (p9_xxbrw_v4sf): New expander.
22487 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22489         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
22491 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
22493         PR target/83926
22494         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
22495         multiply in 32-bit mode.
22496         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
22497         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
22498         mode.
22500 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22502         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
22503         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
22504         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
22505         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
22507 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
22509         PR lto/84213
22510         * dwarf2out.c (is_trivial_indirect_ref): New function.
22511         (dwarf2out_late_global_decl): Do not generate a location
22512         attribute for variables that have a non-trivial DECL_VALUE_EXPR
22513         and that are not defined in the current unit.
22515 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22517         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
22518         instead of a libcall for UNORDERED.
22520 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
22522         PR target/82641
22523         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
22524         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
22526 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22528         PR target/PR84295
22529         * config/s390/s390.c (s390_set_current_function): Invoke
22530         s390_indirect_branch_settings also if fndecl didn't change.
22532 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22534         * config/rs6000/rs6000.md (blockage): Set length to zero.
22536 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22538         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
22540 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22542         PR sanitizer/84285
22543         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
22544         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
22545         -static-lib*san.
22547         PR debug/84252
22548         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
22549         PARALLEL incoming that failed vt_get_decl_and_offset check.
22551         PR middle-end/84237
22552         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
22553         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
22554         TREE_READONLY bit.
22555         (get_variable_section): For decls in named .bss* sections pass true as
22556         second argument to bss_initializer_p.
22558 2018-02-09  Marek Polacek  <polacek@redhat.com>
22559             Jakub Jelinek  <jakub@redhat.com>
22561         PR c++/83659
22562         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
22563         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
22564         Sync some changes from cxx_fold_indirect_ref.
22566 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22568         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
22569         markers.
22570         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
22571         (BLOCK_INLINE_ENTRY_LABEL): New.
22572         (dwarf2out_var_location): Disregard inline entry markers.
22573         (inline_entry_data): New struct.
22574         (inline_entry_data_hasher): New hashtable type.
22575         (inline_entry_data_hasher::hash): New.
22576         (inline_entry_data_hasher::equal): New.
22577         (inline_entry_data_table): New variable.
22578         (add_high_low_attributes): Add DW_AT_entry_pc and
22579         DW_AT_GNU_entry_view attributes if a pending entry is found
22580         in inline_entry_data_table.  Add old entry_pc attribute only
22581         if debug nonbinding markers are disabled.
22582         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
22583         markers are enabled.
22584         (block_within_block_p, dwarf2out_inline_entry): New.
22585         (dwarf2out_finish): Check that no entries remained in
22586         inline_entry_data_table.
22587         * final.c (reemit_insn_block_notes): Handle inline entry notes.
22588         (final_scan_insn, notice_source_line): Likewise.
22589         (rest_of_clean_state): Skip inline entry markers.
22590         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
22591         markers.
22592         * gimple.c (gimple_build_debug_inline_entry): New.
22593         * gimple.h (enum gimple_debug_subcode): Add
22594         GIMPLE_DEBUG_INLINE_ENTRY.
22595         (gimple_build_debug_inline_entry): Declare.
22596         (gimple_debug_inline_entry_p): New.
22597         (gimple_debug_nonbind_marker_p): Adjust.
22598         * insn-notes.def (INLINE_ENTRY): New.
22599         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
22600         inline entry marker notes.
22601         (print_insn): Likewise.
22602         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
22603         (INSN_DEBUG_MARKER_KIND): Likewise.
22604         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
22605         * tree-inline.c (expand_call_inline): Build and insert
22606         debug_inline_entry stmt.
22607         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
22608         inline entry blocks early, if nonbind markers are enabled.
22609         (dump_scope_block): Dump fragment info.
22610         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
22611         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
22612         (gimple_build_debug_inline_entry): New.
22613         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
22614         Enable/disable inline entry points too.
22615         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
22616         (DEBUG_INSN): Describe inline entry markers.
22618         * common.opt (gvariable-location-views): New.
22619         (gvariable-location-views=incompat5): New.
22620         * config.in: Rebuilt.
22621         * configure: Rebuilt.
22622         * configure.ac: Test assembler for view support.
22623         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
22624         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
22625         * dwarf2out.c (var_loc_view): New typedef.
22626         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
22627         (dwarf2out_locviews_in_attribute): New.
22628         (dwarf2out_locviews_in_loclist): New.
22629         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
22630         (enum dw_line_info_opcode): Add LI_adv_address.
22631         (struct dw_line_info_table): Add view.
22632         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
22633         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
22634         (zero_view_p): New variable.
22635         (ZERO_VIEW_P): New macro.
22636         (output_asm_line_debug_info): New.
22637         (struct var_loc_node): Add view.
22638         (add_AT_view_list, AT_loc_list): New.
22639         (add_var_loc_to_decl): Add view param.  Test it against last.
22640         (new_loc_list): Add view params.  Record them.
22641         (AT_loc_list_ptr): Handle loc and view lists.
22642         (view_list_to_loc_list_val_node): New.
22643         (print_dw_val): Handle dw_val_class_view_list.
22644         (size_of_die): Likewise.
22645         (value_format): Likewise.
22646         (loc_list_has_views): New.
22647         (gen_llsym): Set vl_symbol too.
22648         (maybe_gen_llsym, skip_loc_list_entry): New.
22649         (dwarf2out_maybe_output_loclist_view_pair): New.
22650         (output_loc_list): Output view list or entries too.
22651         (output_view_list_offset): New.
22652         (output_die): Handle dw_val_class_view_list.
22653         (output_dwarf_version): New.
22654         (output_compilation_unit_header): Use it.
22655         (output_skeleton_debug_sections): Likewise.
22656         (output_rnglists, output_line_info): Likewise.
22657         (output_pubnames, output_aranges): Update version comments.
22658         (output_one_line_info_table): Output view numbers in asm comments.
22659         (dw_loc_list): Determine current endview, pass it to new_loc_list.
22660         Call maybe_gen_llsym.
22661         (loc_list_from_tree_1): Adjust.
22662         (add_AT_location_description): Create view list attribute if
22663         needed, check it's absent otherwise.
22664         (convert_cfa_to_fb_loc_list): Adjust.
22665         (maybe_emit_file): Call output_asm_line_debug_info for test.
22666         (dwarf2out_var_location): Reset views as needed.  Precompute
22667         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
22668         attribute.  Set view.
22669         (new_line_info_table): Reset next view.
22670         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
22671         (dwarf2out_source_line): Likewise.  Output view resets and labels to
22672         the assembler, or select appropriate line info opcodes.
22673         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
22674         (optimize_string_length): Catch it.  Adjust.
22675         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
22676         dw_val_class_view_list, and remove it if no longer needed.
22677         (hash_loc_list): Hash view numbers.
22678         (loc_list_hasher::equal): Compare them.
22679         (optimize_location_lists): Check whether a view list symbol is
22680         needed, and whether the locview attribute is present, and
22681         whether they match.  Remove the locview attribute if no longer
22682         needed.
22683         (index_location_lists): Call skip_loc_list_entry for test.
22684         (dwarf2out_finish): Call output_asm_line_debug_info for test.
22685         Use output_dwarf_version.
22686         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
22687         (struct dw_val_node): Add val_view_list.
22688         * final.c (SEEN_NEXT_VIEW): New.
22689         (set_next_view_needed): New.
22690         (clear_next_view_needed): New.
22691         (maybe_output_next_view): New.
22692         (final_start_function): Rename to...
22693         (final_start_function_1): ... this.  Take pointer to FIRST,
22694         add SEEN parameter.  Emit param bindings in the initial view.
22695         (final_start_function): Reintroduce SEEN-less interface.
22696         (final): Rename to...
22697         (final_1): ... this.  Take SEEN parameter.  Output final pending
22698         next view at the end.
22699         (final): Reintroduce seen-less interface.
22700         (final_scan_insn): Output pending next view before switching
22701         sections or ending a block.  Mark the next view as needed when
22702         outputting variable locations.  Notify debug backend of section
22703         changes, and of location view changes.
22704         (rest_of_handle_final): Adjust.
22705         * toplev.c (process_options): Autodetect value for debug variable
22706         location views option.  Warn on incompat5 without -gdwarf-5.
22707         * doc/invoke.texi (gvariable-location-views): New.
22708         (gvariable-location-views=incompat5): New.
22709         (gno-variable-location-views): New.
22711 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
22713         PR tree-optimization/84136
22714         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
22715         that the result of find_edge is non-NULL.
22717 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
22719         PR target/83008
22720         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
22721         storing integer register in SImode.  Fix cost of 256 and 512
22722         byte aligned SSE register store.
22724 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
22726         * config/i386/i386.c (ix86_multiplication_cost): Fix
22727         multiplication cost for TARGET_AVX512DQ.
22729 2018-02-08  Marek Polacek  <polacek@redhat.com>
22731         PR tree-optimization/84238
22732         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
22733         get_range_strlen.
22735 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22737         PR tree-optimization/84265
22738         * tree-vect-stmts.c (vectorizable_store): Don't treat
22739         VMAT_CONTIGUOUS accesses as grouped.
22740         (vectorizable_load): Likewise.
22742 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22744         PR tree-optimization/81635
22745         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
22746         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
22747         (test_round_for_mask): New functions.
22748         (wide_int_cc_tests): Call test_round_for_mask.
22749         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
22750         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
22751         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
22752         range returned by get_range_info.
22754 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22756         PR ipa/81360
22757         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
22758         * symtab.c: Include builtins.h
22759         (symtab_node::output_to_lto_symbol_table_p): Move here
22760         from lto-streamer-out.c:output_symbol_p.
22761         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
22762         (output_symbol_p): Move all logic to symtab.c
22763         (produce_symtab): Update.
22765 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22767         * config/s390/s390-opts.h (enum indirect_branch): Define.
22768         * config/s390/s390-protos.h (s390_return_addr_from_memory)
22769         (s390_indirect_branch_via_thunk)
22770         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
22771         (enum s390_indirect_branch_type): Define.
22772         * config/s390/s390.c (struct s390_frame_layout, struct
22773         machine_function): Remove.
22774         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
22775         (indirect_branch_table_label_no, indirect_branch_table_name):
22776         Define variables.
22777         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
22778         (enum s390_indirect_branch_option): Define.
22779         (s390_return_addr_from_memory): New function.
22780         (s390_handle_string_attribute): New function.
22781         (s390_attribute_table): Add new attribute handler.
22782         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
22783         (s390_indirect_branch_via_thunk): New function.
22784         (s390_indirect_branch_via_inline_thunk): New function.
22785         (s390_function_ok_for_sibcall): When jumping via thunk disallow
22786         sibling call optimization for non z10 compiles.
22787         (s390_emit_call): Force indirect branch target to be a single
22788         register.  Add r1 clobber for non-z10 compiles.
22789         (s390_emit_epilogue): Emit return jump via return_use expander.
22790         (s390_reorg): Handle JUMP_INSNs as execute targets.
22791         (s390_option_override_internal): Perform validity checks for the
22792         new command line options.
22793         (s390_indirect_branch_attrvalue): New function.
22794         (s390_indirect_branch_settings): New function.
22795         (s390_set_current_function): Invoke s390_indirect_branch_settings.
22796         (s390_output_indirect_thunk_function):  New function.
22797         (s390_code_end): Implement target hook.
22798         (s390_case_values_threshold): Implement target hook.
22799         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
22800         macros.
22801         * config/s390/s390.h (struct s390_frame_layout)
22802         (struct machine_function): Move here from s390.c.
22803         (TARGET_INDIRECT_BRANCH_NOBP_RET)
22804         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
22805         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
22806         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
22807         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
22808         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
22809         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
22810         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
22811         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
22812         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
22813         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
22814         (mnemonic attribute): Add values which aren't recognized
22815         automatically.
22816         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
22817         pattern for branch conversion.  Fix mnemonic attribute.
22818         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
22819         indirect branch via thunk if requested.
22820         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
22821         ("*indirect_jump"): Disable for branch conversion using out of
22822         line thunks.
22823         ("indirect_jump_via_thunk<mode>_z10")
22824         ("indirect_jump_via_thunk<mode>")
22825         ("indirect_jump_via_inlinethunk<mode>_z10")
22826         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
22827         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
22828         ("casesi_jump_via_inlinethunk<mode>_z10")
22829         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
22830         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
22831         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
22832         ("*indirect2_jump"): Disable for branch conversion.
22833         ("casesi_jump"): Turn into expander and expand patterns for branch
22834         conversion.
22835         ("return_use"): New expander.
22836         ("*return"): Emit return via thunk and rename it to ...
22837         ("*return<mode>"): ... this one.
22838         * config/s390/s390.opt: Add new options and and enum for the
22839         option values.
22841 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22843         * lra-constraints.c (match_reload): Unconditionally use
22844         gen_lowpart_SUBREG, rather than selecting between that
22845         and equivalent gen_rtx_SUBREG code.
22847 2018-02-08  Richard Biener  <rguenther@suse.de>
22849         PR tree-optimization/84233
22850         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
22851         changed flag instead of boguously re-using phi_inserted.
22853 2018-02-08  Martin Jambor  <mjambor@suse.cz>
22855         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
22856         static local variables.
22858 2018-02-08  Richard Biener  <rguenther@suse.de>
22860         PR tree-optimization/84278
22861         * tree-vect-stmts.c (vectorizable_store): When looking for
22862         smaller vector types to perform grouped strided loads/stores
22863         make sure the mode is supported by the target.
22864         (vectorizable_load): Likewise.
22866 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
22868         * config/aarch64/aarch64.c (aarch64_components_for_bb):
22869         Increase LDP/STP opportunities by adding adjacent callee-saves.
22871 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
22873         PR rtl-optimization/84068
22874         PR rtl-optimization/83459
22875         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
22877 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
22879         PR tree-optimization/84224
22880         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
22881         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
22882         non-zero arguments.
22884 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
22886         PR target/84113
22887         * config/rs6000/altivec.md (*restore_world): Remove LR use.
22888         * config/rs6000/predicates.md (restore_world_operation): Adjust op
22889         count, remove one USE.
22891 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22893         * doc/install.texi (Configuration): Document the
22894         --with-long-double-format={ibm,ieee} PowerPC configuration
22895         options.
22897         PR target/84154
22898         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
22899         Convert from define_expand to be define_insn_and_split.  Rework
22900         float/double/_Float128 conversions to QI/HI/SImode to work with
22901         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
22902         conversions to QI/HImode types did a store and then a load to
22903         truncate the value.  For conversions to VSX registers, don't split
22904         the insn, instead emit the code directly.  Use the code iterator
22905         any_fix to combine signed and unsigned conversions.
22906         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
22907         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
22908         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22909         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
22910         (fix_<mode>di2_hw): Likewise.
22911         (fixuns_<mode>di2_hw): Likewise.
22912         (fix_<mode>si2_hw): Likewise.
22913         (fixuns_<mode>si2_hw): Likewise.
22914         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
22915         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22916         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
22917         fix<uns>_trunc<SFDF:mode>si2_p8.
22918         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
22919         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
22920         (fix<uns>_<mode>_mem): Likewise.
22921         (fctiw<u>z_<mode>_mem): Likewise.
22922         (fix<uns>_<mode>_mem): Likewise.
22923         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
22924         the register allocator from doing a direct move to the GPRs to do
22925         a store, and instead use the ISA 3.0 store byte/half-word from
22926         vector register instruction.  For IEEE 128-bit floating point,
22927         also optimize stores of 32-bit ints.
22928         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
22930 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
22932         * genextract.c (push_pathstr_operand): New function to support
22933         [a-zA-Z].
22934         (walk_rtx): Call push_pathstr_operand.
22935         (print_path): Support [a-zA-Z].
22937 2018-02-07  Richard Biener  <rguenther@suse.de>
22939         PR tree-optimization/84037
22940         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
22941         (cse_and_gimplify_to_preheader): Declare.
22942         (vect_get_place_in_interleaving_chain): Likewise.
22943         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22944         ivexpr_map.
22945         (_loop_vec_info::~_loop_vec_info): Delete it.
22946         (cse_and_gimplify_to_preheader): New function.
22947         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
22948         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
22949         (vectorizable_load): Likewise.  For grouped stores always base
22950         the IV on the first element.
22951         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
22952         condition before gimplifying.
22954 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
22956         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
22957         *DIV_EXPR and *MOD_EXPR.
22959 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
22961         PR target/84248
22962         * config/i386/i386.c (ix86_option_override_internal): Mask out
22963         the CF_SET bit when checking -fcf-protection.
22965 2018-02-07  Tom de Vries  <tom@codesourcery.com>
22967         PR libgomp/84217
22968         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
22969         enough.
22971 2018-02-07  Richard Biener  <rguenther@suse.de>
22973         PR tree-optimization/84204
22974         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
22975         this place.
22977         PR tree-optimization/84205
22978         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
22979         special-case isl_ast_op_zdiv_r.
22981         PR tree-optimization/84223
22982         * graphite-scop-detection.c (gather_bbs::before_dom_children):
22983         Only add conditions from within the region.
22984         (gather_bbs::after_dom_children): Adjust.
22986 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
22988         PR target/84209
22989         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
22990         * config/avr/avr.md: Only post-reload split REG-REG moves if
22991         either register is GENERAL_REG_P.
22993 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
22995         PR tree-optimization/84235
22996         * tree-ssa-scopedtables.c
22997         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
22998         if the subtraction is performed in floating point type where NaNs are
22999         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
23000         build 1.  Formatting fix.
23002 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
23004         PR target/84146
23005         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
23006         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
23007         and skip it regardless of bb boundaries.  Use CALL_P macro,
23008         don't test INSN_P (insn) together with CALL_P or JUMP_P check
23009         unnecessarily, formatting fix.
23011 2018-02-06  Michael Collison  <michael.collison@arm.com>
23013         * config/arm/thumb2.md:
23014         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
23015         (*thumb_mov_notscc): Ditto.
23017 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
23019         PR target/84154
23020         * config/rs6000/rs6000.md (su code attribute): Use "u" for
23021         unsigned_fix, not "s".
23023 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23025         * configure.ac (gcc_fn_eh_frame_ro): New function.
23026         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
23027         correct .eh_frame permissions.
23028         * configure: Regenerate.
23030 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
23032         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
23033         irrelevant options.
23035 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23037         * config/rs6000/rs6000.c (rs6000_option_override_internal):
23038         Display warning message for -mno-speculate-indirect-jumps.
23040 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
23042         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
23043         Undocumented.
23044         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
23046 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
23048         PR tree-optimization/84225
23049         * tree-eh.c (find_trapping_overflow): Only call
23050         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
23052 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23054         PR target/84145
23055         * config/i386/i386.c: Reimplement the check of possible options
23056         -mibt/-mshstk conbination. Change error messages.
23057         * doc/invoke.texi: Fix a typo: remove extra '='.
23059 2018-02-06  Marek Polacek  <polacek@redhat.com>
23061         PR tree-optimization/84228
23062         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
23064 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
23066         PR target/82641
23067         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
23068         emitted arch directives.
23069         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
23070         __ARM_FEATURE_COPROC before changing architectures.
23072 2018-02-06  Richard Biener  <rguenther@suse.de>
23074         * config/i386/i386.c (print_reg): Fix typo.
23075         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
23077 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
23079         * configure: Regenerate.
23081 2018-02-05  Martin Sebor  <msebor@redhat.com>
23083         PR tree-optimization/83369
23084         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
23085         inlining context.
23087 2018-02-05  Martin Liska  <mliska@suse.cz>
23089         * doc/invoke.texi: Cherry-pick upstream r323995.
23091 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
23093         * ira.c (ira_init_register_move_cost): Adjust comment.
23095 2018-02-05  Martin Liska  <mliska@suse.cz>
23097         PR gcov-profile/84137
23098         * doc/gcov.texi: Fix typo in documentation.
23100 2018-02-05  Martin Liska  <mliska@suse.cz>
23102         PR gcov-profile/83879
23103         * doc/gcov.texi: Document necessity of --dynamic-list-data when
23104         using dlopen functionality.
23106 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
23108         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
23109         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
23110         _mm_maskz_range_ss, _mm_mask_range_round_ss,
23111         _mm_maskz_range_round_ss): New intrinsics.
23112         (__builtin_ia32_rangesd128_round)
23113         (__builtin_ia32_rangess128_round): Remove.
23114         (__builtin_ia32_rangesd128_mask_round,
23115         __builtin_ia32_rangess128_mask_round): New builtins.
23116         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
23117         __builtin_ia32_rangess128_round): Remove.
23118         (__builtin_ia32_rangesd128_mask_round,
23119         __builtin_ia32_rangess128_mask_round): New builtins.
23120         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
23121         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
23122         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
23123         "<round_saeonly_constraint>")): Changed to ...
23124         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
23125         "<round_saeonly_scalar_constraint>")): ... this.
23126         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
23127         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
23128         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
23129         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
23130         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
23132 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
23134         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
23135         options.
23136         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
23137         Remove all values except native, 8540 and 8548.
23139 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
23141         * config/i386/i386.c (ix86_output_function_return): Pass
23142         INVALID_REGNUM, instead of -1, as invalid register number to
23143         indirect_thunk_name and output_indirect_thunk.
23145 2018-02-02  Julia Koval  <julia.koval@intel.com>
23147         * config.gcc: Add -march=icelake.
23148         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
23149         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
23150         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
23151         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
23152         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
23153         (processor_target_table): Add icelake.
23154         (ix86_option_override_internal): Handle new PTAs.
23155         (get_builtin_code_for_version): Handle icelake.
23156         (M_INTEL_COREI7_ICELAKE): New.
23157         (fold_builtin_cpu): Handle icelake.
23158         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
23159         * doc/invoke.texi: Add -march=icelake.
23161 2018-02-02  Julia Koval  <julia.koval@intel.com>
23163         * config/i386/i386.c (ix86_option_override_internal): Change flags type
23164         to wide_int_bitmask.
23165         * wide-int-bitmask.h: New.
23167 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23169         PR target/84066
23170         * config/i386/i386.md: Replace Pmode with word_mode in
23171         builtin_setjmp_setup and builtin_longjmp to support x32.
23173 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
23175         PR target/56010
23176         PR target/83743
23177         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
23178         #include "opts.h".
23179         (rs6000_supported_cpu_names): New static variable.
23180         (linux_cpu_translation_table): Likewise.
23181         (elf_platform) <cpu>: Define new static variable and use it.
23182         Translate kernel AT_PLATFORM name to canonical name if needed.
23183         Error if platform name is unknown.
23185 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
23187         PR target/84089
23188         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
23190 2018-02-01  Jeff Law  <law@redhat.com>
23192         PR target/84128
23193         * config/i386/i386.c (release_scratch_register_on_entry): Add new
23194         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
23195         the scratch if RELEASE_VIA_POP is false.
23196         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
23197         If we have to save a temporary register, decrement SIZE appropriately.
23198         Pass new arguments to release_scratch_register_on_entry.
23199         (ix86_adjust_stack_and_probe): Likewise.
23200         (ix86_emit_probe_stack_range): Pass new arguments to
23201         release_scratch_register_on_entry.
23203 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
23205         PR rtl-optimization/84157
23206         * combine.c (change_zero_ext): Use REG_P predicate in
23207         front of HARD_REGISTER_P predicate.
23209 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
23211         * config/avr/avr.c (avr_option_override): Move disabling of
23212         -fdelete-null-pointer-checks to...
23213         * common/config/avr/avr-common.c (avr_option_optimization_table):
23214         ...here.
23216 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23218         PR tree-optimization/81635
23219         * tree-data-ref.c (split_constant_offset_1): For types that
23220         wrap on overflow, try to use range info to prove that wrapping
23221         cannot occur.
23223 2018-02-01  Renlin Li  <renlin.li@arm.com>
23225         PR target/83370
23226         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
23227         TAILCALL_ADDR_REGS.
23228         (aarch64_register_move_cost): Likewise.
23229         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
23230         TAILCALL_ADDR_REGS.
23231         (REG_CLASS_NAMES): Likewise.
23232         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
23233         TAILCALL_ADDR_REGS. Remove IP registers.
23234         * config/aarch64/aarch64.md (Ucs): Update register constraint.
23236 2018-02-01  Richard Biener  <rguenther@suse.de>
23238         * domwalk.h (dom_walker::dom_walker): Add additional constructor
23239         for specifying RPO order and allow NULL for that.
23240         * domwalk.c (dom_walker::dom_walker): Likewise.
23241         (dom_walker::walk): Handle NULL RPO order.
23242         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23243         in RPO order.
23244         (rewrite_update_dom_walker): Likewise.
23245         (mark_def_dom_walker): Likewise.
23247 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23249         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23250         (aarch64_maybe_expand_sve_subreg_move): Declare.
23251         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23252         * config/aarch64/predicates.md (aarch64_any_register_operand): New
23253         predicate.
23254         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23255         that are semantically a reverse operation.
23256         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23257         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23258         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23259         functions.
23260         (aarch64_can_change_mode_class): For big-endian, forbid changes
23261         between two SVE modes if they have different element sizes.
23263 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23265         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23266         the TImode handling for big-endian targets.
23268 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23270         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23271         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
23272         not just bytes.
23273         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23274         Remove BSWAP handing for big-endian targets and use the form of
23275         LD1RQ appropariate for the mode.
23277 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23279         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23280         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
23281         duplicated element.
23283 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23285         PR tearget/83845
23286         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
23287         check for operands that need to go through aarch64_sve_reload_be.
23289 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
23291         PR tree-optimization/81661
23292         PR tree-optimization/84117
23293         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23294         * tree-eh.c: Include gimplify.h.
23295         (find_trapping_overflow, replace_trapping_overflow,
23296         rewrite_to_non_trapping_overflow): New functions.
23297         * tree-vect-loop.c: Include tree-eh.h.
23298         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23299         * tree-data-ref.c: Include tree-eh.h.
23300         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23302 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
23304         PR rtl-optimization/84123
23305         * combine.c (change_zero_ext): Check if hard register satisfies
23306         can_change_dest_mode before calling gen_lowpart_SUBREG.
23308 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
23310         PR target/82444
23311         * ira.c (ira_init_register_move_cost): Remove assert.
23313 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23315         PR rtl-optimization/84071
23316         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23317         * doc/tm.texi: Regenerate.
23319 2018-01-31  Richard Biener  <rguenther@suse.de>
23321         PR tree-optimization/84132
23322         * tree-data-ref.c (analyze_miv_subscript): Properly
23323         check whether evolution_function_is_affine_multivariate_p
23324         before calling gcd_of_steps_may_divide_p.
23326 2018-01-31  Julia Koval  <julia.koval@intel.com>
23328         PR target/83618
23329         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23330         * config/i386/i386.md (rdpid_rex64) New.
23331         (rdpid): Make 32bit only.
23333 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
23335         PR lto/84105
23336         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23337         an IDENTIFIER_NODE for FUNCTION_TYPE's.
23339 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23341         Revert
23342         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23344         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23346 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23348         PR rtl-optimization/84071
23349         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
23350         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
23352 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23354         * config/arc/arc.c (arc_handle_aux_attribute): New function.
23355         (arc_attribute_table): Add 'aux' attribute.
23356         (arc_in_small_data_p): Consider aux like variables.
23357         (arc_is_aux_reg_p): New function.
23358         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
23359         (arc_get_aux_arg): New function.
23360         (prepare_move_operands): Handle aux-register access.
23361         (arc_handle_aux_attribute): New function.
23362         * doc/extend.texi (ARC Variable attributes): Add subsection.
23364 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23366         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
23367         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
23368         (arc_attribute_table): Add 'uncached' attribute.
23369         (arc_print_operand): Print '.di' flag for uncached memory
23370         accesses.
23371         (arc_in_small_data_p): Do not consider for small data the uncached
23372         types.
23373         (arc_is_uncached_mem_p): New function.
23374         * config/arc/predicates.md (compact_store_memory_operand): Check
23375         for uncached memory accesses.
23376         (nonvol_nonimm_operand): Likewise.
23377         * doc/extend.texi (ARC Type Attribute): New subsection.
23379 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
23381         PR c/84100
23382         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
23383         falign-loops=): Add Optimization flag.
23385 2018-01-30  Jeff Law  <law@redhat.com>
23387         PR target/84064
23388         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
23389         INT_REGISTERS_SAVED.  Check it prior to calling
23390         get_scratch_register_on_entry.
23391         (ix86_adjust_stack_and_probe): Similarly.
23392         (ix86_emit_probe_stack_range): Similarly.
23393         (ix86_expand_prologue): Corresponding changes.
23395 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23397         PR target/40411
23398         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
23399         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
23401 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
23403         PR target/84112
23404         * lra-constraints.c (curr_insn_transform): Process AND in the
23405         address.
23407 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
23409         PR rtl-optimization/83986
23410         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
23411         dependence against last_pending_memory_flush in addition to
23412         pending_jump_insns.
23414 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
23416         PR tree-optimization/81611
23417         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
23418         copies.
23420 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23422         PR target/83758
23423         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
23424         a reg rtx.
23426 2018-01-30  Richard Biener  <rguenther@suse.de>
23427             Jakub Jelinek  <jakub@redhat.com>
23429         PR tree-optimization/84111
23430         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
23431         inner loops added during recursion, as they don't have up-to-date
23432         SSA form.
23434 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23436         PR ipa/81360
23437         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
23438         (can_inline_edge_by_limits_p): ... here.
23439         (can_early_inline_edge_p, check_callers,
23440         update_caller_keys, update_callee_keys, recursive_inlining,
23441         add_new_edges_to_heap, speculation_useful_p,
23442         inline_small_functions,
23443         inline_small_functions, flatten_function,
23444         inline_to_all_callers_1): Update.
23446 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23448         * profile-count.c (profile_count::combine_with_ipa_count): Handle
23449         zeros correctly.
23451 2018-01-30  Richard Biener  <rguenther@suse.de>
23453         PR tree-optimization/83008
23454         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
23455         invariant and constant vector uses in stmts when they need
23456         more than one stmt.
23458 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23460         PR bootstrap/84017
23461         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
23462         * configure: Regenerate.
23464 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23466         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
23467         pattern.
23468         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
23469         Use gen_rtx_REG rather than gen_lowpart.
23471 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23473         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
23474         rather than 0 when creating partial subregs.
23476 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23478         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
23479         of usage.
23481 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
23483         PR target/81550
23484         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
23485         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
23486         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
23487         flags.  This restores the settings used before the 2017-07-24.
23488         Turning off pre increment/decrement/modify allows IVOPTS to
23489         optimize DF/SF loops where the index is an int.
23491 2018-01-29  Richard Biener  <rguenther@suse.de>
23492             Kelvin Nilsen  <kelvin@gcc.gnu.org>
23494         PR bootstrap/80867
23495         * tree-vect-stmts.c (vectorizable_call): Don't call
23496         targetm.vectorize_builtin_md_vectorized_function if callee is
23497         NULL.
23499 2018-01-22  Carl Love  <cel@us.ibm.com>
23501         * doc/extend.tex: Fix typo in second arg in
23502         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
23504 2018-01-29  Richard Biener  <rguenther@suse.de>
23506         PR tree-optimization/84086
23507         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
23508         (flush_ssaname_freelist): When SSA names were released reset
23509         the SCEV hash table.
23511 2018-01-29  Richard Biener  <rguenther@suse.de>
23513         PR tree-optimization/84057
23514         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
23515         removed paths when removing edges.
23517 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
23519         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
23520         -mfunction-return=@var{choice}.
23522 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
23524         PR diagnostic/84034
23525         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
23526         Handle CR like TAB.
23527         (layout::print_source_line): Likewise.
23528         (test_get_line_width_without_trailing_whitespace): Add test cases.
23530 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
23532         PR middle-end/84040
23533         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
23534         debug insns.
23536 2018-01-26  Jim Wilson  <jimw@sifive.com>
23538         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
23540         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
23541         specified.
23543 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23545         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
23546         and CMP + SUB-immediate -> SUBS.
23548 2018-01-26  Martin Sebor  <msebor@redhat.com>
23550         PR tree-optimization/83896
23551         * tree-ssa-strlen.c (get_string_len): Rename...
23552         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
23553         Avoid assuming length is constant.
23554         (handle_char_store): Use HOST_WIDE_INT for string length.
23556 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
23558         PR target/81763
23559         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
23560         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
23562 2018-01-26  Richard Biener  <rguenther@suse.de>
23564         PR rtl-optimization/84003
23565         * dse.c (record_store): Only record redundant stores when
23566         the earlier store aliases at least all accesses the later one does.
23568 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
23570         PR rtl-optimization/83985
23571         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
23572         REG_CFA_RESTORE insns.
23573         (delete_unmarked_insns): Don't ignore separate shrink wrapping
23574         REG_CFA_RESTORE insns here.
23576         PR c/83989
23577         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
23578         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
23580 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23582         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
23583         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
23584         (arc_init): Likewise.
23585         (arc_override_options): Likewise.
23586         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
23587         value.
23588         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
23589         support.
23590         * config/arc/arc.h (TARGET_DBNZ): Define.
23591         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
23592         properly set the tune attribute.
23593         (dbnz): Use TARGET_DBNZ guard.
23594         * config/arc/arc.opt (mtune): Add core3 option.
23596 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23598         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
23599         recognize new pic like addresses.
23600         (arc_delegitimize_address): Clean up.
23602 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23604         * config/arc/arc-arches.def: Option mrf16 valid for all
23605         architectures.
23606         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
23607         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
23608         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
23609         * config/arc/arc-tables.opt: Regenerate.
23610         * config/arc/arc.c (arc_conditional_register_usage): Handle
23611         reduced register file case.
23612         (arc_file_start): Set must have build attributes.
23613         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
23614         mrf16 option value.
23615         * config/arc/arc.opt (mrf16): Add new option.
23616         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
23617         * config/arc/genmultilib.awk: Handle new mrf16 option.
23618         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
23619         * config/arc/t-multilib: Regenerate.
23620         * doc/invoke.texi (ARC Options): Document mrf16 option.
23622 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23624         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
23625         * config/arc/arc.c (arc_handle_secure_attribute): New function.
23626         (arc_attribute_table): Add 'secure_call' attribute.
23627         (arc_print_operand): Print secure call operand.
23628         (arc_function_ok_for_sibcall): Don't optimize tail calls when
23629         secure.
23630         (arc_is_secure_call_p): New function.  * config/arc/arc.md
23631         (call_i): Add support for sjli instruction.
23632         (call_value_i): Likewise.
23633         * config/arc/constraints.md (Csc): New constraint.
23635 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23636             John Eric Martin  <John.Martin@emmicro-us.com>
23638         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
23639         * config/arc/arc.c (_arc_jli_section): New struct.
23640         (arc_jli_section): New type.
23641         (rc_jli_sections): New static variable.
23642         (arc_handle_jli_attribute): New function.
23643         (arc_attribute_table): Add jli_always and jli_fixed attribute.
23644         (arc_file_end): New function.
23645         (TARGET_ASM_FILE_END): Define.
23646         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
23647         (arc_add_jli_section): New function.
23648         (jli_call_scan): Likewise.
23649         (arc_reorg): Call jli_call_scan.
23650         (arc_output_addsi): Remove 'S' from printing asm operand.
23651         (arc_is_jli_call_p): New function.
23652         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
23653         operand.
23654         (movhi_insn): Likewise.
23655         (movsi_insn): Likewise.
23656         (movsi_set_cc_insn): Likewise.
23657         (loadqi_update): Likewise.
23658         (load_zeroextendqisi_update): Likewise.
23659         (load_signextendqisi_update): Likewise.
23660         (loadhi_update): Likewise.
23661         (load_zeroextendhisi_update): Likewise.
23662         (load_signextendhisi_update): Likewise.
23663         (loadsi_update): Likewise.
23664         (loadsf_update): Likewise.
23665         (movsicc_insn): Likewise.
23666         (bset_insn): Likewise.
23667         (bxor_insn): Likewise.
23668         (bclr_insn): Likewise.
23669         (bmsk_insn): Likewise.
23670         (bicsi3_insn): Likewise.
23671         (cmpsi_cc_c_insn): Likewise.
23672         (movsi_ne): Likewise.
23673         (movsi_cond_exec): Likewise.
23674         (clrsbsi2): Likewise.
23675         (norm_f): Likewise.
23676         (normw): Likewise.
23677         (swap): Likewise.
23678         (divaw): Likewise.
23679         (flag): Likewise.
23680         (sr): Likewise.
23681         (kflag): Likewise.
23682         (ffs): Likewise.
23683         (ffs_f): Likewise.
23684         (fls): Likewise.
23685         (call_i): Remove 'S' asm letter, add jli instruction.
23686         (call_value_i): Likewise.
23687         * config/arc/arc.op (mjli-always): New option.
23688         * config/arc/constraints.md (Cji): New constraint.
23689         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
23690         operand.
23691         (subsf3_fpx): Likewise.
23692         (mulsf3_fpx): Likewise.
23693         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
23694         asm operand.
23695         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
23696         function attrbutes.
23697         * doc/invoke.texi (ARC): Document mjli-always option.
23699 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
23701         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
23702         avoid addition with 0 and use incw and decw where possible.
23704 2018-01-26  Richard Biener  <rguenther@suse.de>
23706         PR tree-optimization/81082
23707         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
23708         association if it requires casting to unsigned.
23709         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
23710         from fold_plusminus_mult_expr to catch important cases late when
23711         range info is available.
23713 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23715         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
23716         * configure.ac (hidden_linkonce): New test.
23717         * configure: Regenerate.
23718         * config.in: Regenerate.
23720 2018-01-26  Julia Koval  <julia.koval@intel.com>
23722         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
23723         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
23724         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
23725         _mm_mask_bitshuffle_epi64_mask): Fix type.
23726         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
23727         USI_FTYPE_V4DI_V4DI_USI): Remove.
23728         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
23729         __builtin_ia32_vpshufbitqmb256_mask,
23730         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
23731         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
23732         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
23734 2018-01-26  Alan Modra  <amodra@gmail.com>
23736         PR target/84033
23737         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
23738         UNSPEC_VBPERMQ.  Sort other unspecs.
23740 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
23742         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
23744 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
23746         PR middle-end/83055
23747         * predict.c (drop_profile): Do not push/pop cfun; update also
23748         node->count.
23749         (handle_missing_profiles): Fix logic looking for zero profiles.
23751 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
23753         PR middle-end/83977
23754         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
23755         on functions with #pragma omp declare simd or functions with simd
23756         attribute.
23757         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
23758         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
23759         Remove trailing \n from warning_at calls.
23761 2018-01-25  Tom de Vries  <tom@codesourcery.com>
23763         PR target/84028
23764         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23765         for neutered workers.
23767 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
23769         PR target/68467
23770         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
23771         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
23773 2018-01-24  Jeff Law  <law@redhat.com>
23775         PR target/83994
23776         * i386.c (get_probe_interval): Move to earlier point.
23777         (ix86_compute_frame_layout): If -fstack-clash-protection and
23778         the frame is larger than the probe interval, then use pushes
23779         to save registers rather than reg->mem moves.
23780         (ix86_expand_prologue): Remove conditional for int_registers_saved
23781         assertion.
23783 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
23785         PR target/84014
23786         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
23787         min/max for never referenced object.
23789 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
23791         PR middle-end/83977
23792         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
23793         here.
23794         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
23795         attributes from DECL_ATTRIBUTES (decl) without affecting
23796         DECL_ATTRIBUTES (current_function_decl).
23797         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
23798         functions with non-NULL DECL_ABSTRACT_ORIGIN.
23800 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
23802         PR tree-optimization/83979
23803         * fold-const.c (fold_comparison): Use constant_boolean_node
23804         instead of boolean_{true,false}_node.
23806 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
23808         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
23809         with zero counts.
23811 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23813         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23814         Simplify the clause that sets the length attribute.
23815         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23816         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
23817         clause that sets the length attribute.
23818         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23820 2018-01-24  Tom de Vries  <tom@codesourcery.com>
23822         PR target/83589
23823         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
23824         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
23825         Add strict parameter.
23826         (prevent_branch_around_nothing): Insert dummy insn between branch to
23827         label and label with no ptx insn inbetween.
23828         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
23830 2018-01-24  Tom de Vries  <tom@codesourcery.com>
23832         PR target/81352
23833         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23834         for neutered threads in warp.
23835         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
23837 2018-01-24  Richard Biener  <rguenther@suse.de>
23839         PR tree-optimization/83176
23840         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
23841         operands.
23843 2018-01-24  Richard Biener  <rguenther@suse.de>
23845         PR tree-optimization/82819
23846         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
23847         code generating pluses that are no-ops in the target precision.
23849 2018-01-24  Richard Biener  <rguenther@suse.de>
23851         PR middle-end/84000
23852         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
23854 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23856         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
23857         to merge probabilities.
23858         * predict.c (probably_never_executed): Also mark as cold functions
23859         with global 0 profile and guessed local profile.
23860         * profile-count.c (profile_probability::combine_with_count): New
23861         member function.
23862         * profile-count.h (profile_probability::operator*,
23863         profile_probability::operator*=, profile_probability::operator/,
23864         profile_probability::operator/=): Reduce precision to adjusted
23865         and set value to guessed on contradictory divisions.
23866         (profile_probability::combine_with_freq): Remove.
23867         (profile_probability::combine_wiht_count): Declare.
23868         (profile_count::force_nonzero):: Set to adjusted.
23869         (profile_count::probability_in):: Set quality to adjusted.
23870         * tree-ssa-tail-merge.c (replace_block_by): Use
23871         combine_with_count.
23873 2018-01-23  Andrew Waterman  <andrew@sifive.com>
23874             Jim Wilson  <jimw@sifive.com>
23876         * config/riscv/riscv.c (riscv_stack_boundary): New.
23877         (riscv_option_override): Set riscv_stack_boundary.  Handle
23878         riscv_preferred_stack_boundary_arg.
23879         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
23880         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
23881         (STACK_BOUNDARY): Set to riscv_stack_boundary.
23882         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
23883         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
23884         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
23886 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
23888         PR target/83905
23889         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
23890         of struct ix86_frame.
23891         (ix86_expand_epilogue): Likewise.  Add a local variable for
23892         the reg_save_offset field in struct ix86_frame.
23894 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
23896         PR tree-optimization/82604
23897         * tree-loop-distribution.c (enum partition_kind): New enum item
23898         PKIND_PARTIAL_MEMSET.
23899         (partition_builtin_p): Support above new enum item.
23900         (generate_code_for_partition): Ditto.
23901         (compute_access_range): Differentiate cases that equality can be
23902         proven at all loops, the innermost loops or no loops.
23903         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
23904         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
23905         (finalize_partitions, distribute_loop): Don't fuse partition of
23906         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
23907         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
23908         parloop is enabled.
23910 2018-01-23  Martin Liska  <mliska@suse.cz>
23912         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
23913         order to ignore the predictor.
23914         (PRED_POLYMORPHIC_CALL): Likewise.
23915         (PRED_RECURSIVE_CALL): Likewise.
23917 2018-01-23  Martin Liska  <mliska@suse.cz>
23919         * tree-profile.c (tree_profiling): Print function header to
23920         aware reader which function we are working on.
23921         * value-prof.c (gimple_find_values_to_profile): Do not print
23922         not interesting value histograms.
23924 2018-01-23  Martin Liska  <mliska@suse.cz>
23926         * profile-count.h (enum profile_quality): Add
23927         profile_uninitialized as the first value. Do not number values
23928         as they are zero based.
23929         (profile_count::verify): Update sanity check.
23930         (profile_probability::verify): Likewise.
23932 2018-01-23  Nathan Sidwell  <nathan@acm.org>
23934         * doc/invoke.texi (ffor-scope): Deprecate.
23936 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23938         PR tree-optimization/83510
23939         * domwalk.c (set_all_edges_as_executable): New function.
23940         (dom_walker::dom_walker): Convert bool param
23941         "skip_unreachable_blocks" to enum reachability.  Move setup of
23942         edge flags to set_all_edges_as_executable and only do it when
23943         reachability is REACHABLE_BLOCKS.
23944         * domwalk.h (enum dom_walker::reachability): New enum.
23945         (dom_walker::dom_walker): Convert bool param
23946         "skip_unreachable_blocks" to enum reachability.
23947         (set_all_edges_as_executable): New decl.
23948         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
23949         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
23950         "reachability".
23951         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
23952         but converting true to REACHABLE_BLOCKS.
23953         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
23954         * tree-vrp.c
23955         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
23956         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
23957         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
23958         REACHABLE_BLOCKS.
23959         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
23960         if check_all_array_refs will be called.
23962 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23964         * tree.c (selftest::test_location_wrappers): Add more test
23965         coverage.
23967 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23969         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
23970         (selftest::test_bit_in_range): Likewise.
23972 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
23974         PR testsuite/83888
23975         * doc/sourcebuild.texi (vect_float): Say that the selector
23976         only describes the situation when -funsafe-math-optimizations is on.
23977         (vect_float_strict): Document.
23979 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
23981         PR tree-optimization/83965
23982         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
23983         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
23984         instead of checking only for a reduction.
23985         (vect_recog_widen_sum_pattern): Likewise.
23987 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23989         * predict.c (probably_never_executed): Only use precise profile info.
23990         (compute_function_frequency): Skip after inlining hack since we now
23991         have quality checking.
23993 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23995         * profile-count.h (profile_probability::very_unlikely,
23996         profile_probability::unlikely, profile_probability::even): Set
23997         precision to guessed.
23999 2018-01-23  Richard Biener  <rguenther@suse.de>
24001         PR tree-optimization/83963
24002         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
24003         Properly terminate dominator walk when crossing the exit edge not
24004         when visiting its source block.
24006 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
24008         PR c++/83918
24009         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
24010         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
24012 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
24014         PR tree-optimization/83957
24015         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
24016         semicolon after for body surrounded by braces.
24018         PR tree-optimization/83081
24019         * profile-count.h (profile_probability::split): New method.
24020         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
24021         Use profile_probability::split.
24022         (do_compare_rtx_and_jump): Fix adjustment of probabilities
24023         when splitting a single conditional jump into 2.
24025 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
24027         PR tree-optimization/69452
24028         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
24029         decl.
24031 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24033         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
24034         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
24035         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
24037 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24039         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
24040         declaration.
24041         * config/rl78/rl78.md (movdi): New define_expand.
24042         * config/rl78/rl78.c (rl78_split_movdi): New function.
24044 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
24046         PR target/83862
24047         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
24048         no longer used.
24049         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
24050         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
24051         128-bit to produce an UNSPEC move to get the double word with the
24052         signbit and then a shift directly to do signbit.
24053         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
24054         implementation with a new version that just does either a direct
24055         move or a regular move.  Move memory interface to separate insns.
24056         Move insns so they are next to the expander.
24057         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
24058         with signbit move.  Split big and little endian case.
24059         (signbit<mode>2_dm_mem_le): Likewise.
24060         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
24061         (signbit<mode>2_dm2): Likewise.
24063 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24065         * config/rl78/rl78.md (anddi3): New define_expand.
24067 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24069         * config/rl78/rl78.md (umindi3): New define_expand.
24071 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24073         * config/rl78/rl78.md (smindi3): New define_expand.
24075 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24077         * config/rl78/rl78.md (smaxdi3): New define_expand.
24079 2018-01-22  Carl Love  <cel@us.ibm.com>
24081         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
24082         LVX_V1TI): Add macro expansion.
24083         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
24084         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
24085         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
24086         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
24087         Change check to determine if the instruction is a byte reversing
24088         entry.  Fix typo in comment.
24089         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
24090         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
24091         Add def_builtin calls for new builtins.
24092         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
24093         Add define_insn expansion.
24095 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24097         * config/rl78/rl78.md (umaxdi3): New define_expand.
24099 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24101         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
24102         for non-QImode registers.
24104 2018-01-22  Richard Biener  <rguenther@suse.de>
24106         PR tree-optimization/83963
24107         * graphite-scop-detection.c (scop_detection::get_sese): Delay
24108         including the loop exit block.
24109         (scop_detection::merge_sese): Likewise.
24110         (scop_detection::add_scop): Do it here instead.
24112 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24114         * doc/sourcebuild.texi (arm_softfloat): Document.
24116 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
24118         PR gcc/77734
24119         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
24120         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
24121         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
24123 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24124             David Edelsohn  <dje.gcc@gmail.com>
24126         PR target/83946
24127         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24128         Change "crset eq" to "crset 2".
24129         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24130         (*call_indirect_aix<mode>_nospec): Likewise.
24131         (*call_value_indirect_aix<mode>_nospec): Likewise.
24132         (*call_indirect_elfv2<mode>_nospec): Likewise.
24133         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
24134         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
24135         change assembly output from . to $.
24136         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24137         (indirect_jump<mode>_nospec): Change assembly output from . to $.
24138         (*tablejump<mode>_internal1_nospec): Likewise.
24140 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
24142         PR target/80870
24143         * config/sh/sh_optimize_sett_clrt.cc:
24144         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
24146 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
24148         PR tree-optimization/83940
24149         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
24150         offset_dt to vect_constant_def rather than vect_unknown_def_type.
24151         (vect_check_load_store_mask): Add a mask_dt_out parameter and
24152         use it to pass back the definition type.
24153         (vect_check_store_rhs): Likewise rhs_dt_out.
24154         (vect_build_gather_load_calls): Add a mask_dt argument and use
24155         it instead of a call to vect_is_simple_use.
24156         (vectorizable_store): Update calls to vect_check_load_store_mask
24157         and vect_check_store_rhs.  Use the dt returned by the latter instead
24158         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
24159         instead of calls to vect_is_simple_use.  Pass the scalar rather
24160         than the vector operand to vect_is_simple_use when handling
24161         second and subsequent copies of an rhs value.
24162         (vectorizable_load): Update calls to vect_check_load_store_mask
24163         and vect_build_gather_load_calls.  Use the cached mask_dt and
24164         gs_info.offset_dt instead of calls to vect_is_simple_use.
24166 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
24168         PR middle-end/83945
24169         * tree-emutls.c: Include gimplify.h.
24170         (lower_emutls_2): New function.
24171         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
24172         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
24173         it before further processing.
24175         PR target/83930
24176         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
24177         UINTVAL (trueop1) instead of INTVAL (op1).
24179 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
24181         PR debug/81570
24182         PR debug/83728
24183         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
24184         INCOMING_FRAME_SP_OFFSET if not defined.
24185         (scan_trace): Add ENTRY argument.  If true and
24186         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
24187         emit a note to adjust the CFA offset.
24188         (create_cfi_notes): Adjust scan_trace callers.
24189         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
24190         INCOMING_FRAME_SP_OFFSET in the CIE.
24191         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
24192         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
24193         Likewise.
24194         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
24195         * doc/tm.texi: Regenerated.
24197 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24199         PR rtl-optimization/83147
24200         * lra-constraints.c (remove_inheritance_pseudos): Use
24201         lra_substitute_pseudo_within_insn.
24203 2018-01-19  Tom de Vries  <tom@codesourcery.com>
24204             Cesar Philippidis  <cesar@codesourcery.com>
24206         PR target/83920
24207         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
24209 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
24211         PR target/83790
24212         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
24213         spaces for function labels.
24215 2018-01-19  Martin Liska  <mliska@suse.cz>
24217         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
24218         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
24219         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
24220         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
24221         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
24222         (PRED_CONST_RETURN): Change from 69 to 65.
24223         (PRED_NULL_RETURN): Change from 91 to 71.
24224         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
24225         (PRED_LOOP_GUARD): Change from 66 to 73.
24227 2018-01-19  Martin Liska  <mliska@suse.cz>
24229         * predict.c (predict_insn_def): Add new assert.
24230         (struct branch_predictor): Change type to signed integer.
24231         (test_prediction_value_range): Amend test to cover
24232         PROB_UNINITIALIZED.
24233         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24234         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24235         (PRED_LOOP_ITERATIONS_MAX): Likewise.
24236         (PRED_LOOP_IV_COMPARE): Likewise.
24237         * predict.h (PROB_UNINITIALIZED): Define new constant.
24239 2018-01-19  Martin Liska  <mliska@suse.cz>
24241         * predict.c (dump_prediction): Add new format for
24242         analyze_brprob.py script which is enabled with -details
24243         suboption.
24244         * profile-count.h (precise_p): New function.
24246 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24248         PR tree-optimization/83922
24249         * tree-vect-loop.c (vect_verify_full_masking): Return false if
24250         there are no statements that need masking.
24251         (vect_active_double_reduction_p): New function.
24252         (vect_analyze_loop_operations): Use it when handling phis that
24253         are not in the loop header.
24255 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24257         PR tree-optimization/83914
24258         * tree-vect-loop.c (vectorizable_induction): Don't convert
24259         init_expr or apply the peeling adjustment for inductions
24260         that are nested within the vectorized loop.
24262 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24264         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24265         instead of NEG.
24267 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24269         PR sanitizer/81715
24270         PR testsuite/83882
24271         * function.h (gimplify_parameters): Add gimple_seq * argument.
24272         * function.c: Include gimple.h and options.h.
24273         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24274         for the added local temporaries if needed.
24275         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24276         if there are any parameter cleanups, wrap whole body into a
24277         try/finally with the cleanups.
24279 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
24281         PR target/82964
24282         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
24283         Use GET_MODE_CLASS for scalar floating point.
24285 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24287         PR ipa/82256
24288         patch by PaX Team
24289         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
24290         Fix call of call_cgraph_insertion_hooks.
24292 2018-01-18  Martin Sebor  <msebor@redhat.com>
24294         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24296 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24298         PR ipa/83619
24299         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24300         frequencies.
24302 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
24304         PR other/70268
24305         * common.opt: (-ffile-prefix-map): New option.
24306         * opts.c (common_handle_option): Defer it.
24307         * opts-global.c (handle_common_deferred_options): Handle it.
24308         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24309         * file-prefix-map.h: New file.
24310         (remap_debug_filename, add_debug_prefix_map): ...here.
24311         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24312         * final.c (debug_prefix_map, add_debug_prefix_map
24313         remap_debug_filename): Move to...
24314         * file-prefix-map.c: New file.
24315         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24316         generalize, get rid of alloca(), use strrchr() instead of strchr().
24317         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24318         Implement in terms of add_prefix_map().
24319         (remap_macro_filename, remap_debug_filename): Implement in term of
24320         remap_filename().
24321         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24322         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24323         * dbxout.c: Include file-prefix-map.h.
24324         * varasm.c: Likewise.
24325         * vmsdbgout.c: Likewise.
24326         * xcoffout.c: Likewise.
24327         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24328         * doc/cppopts.texi (-fmacro-prefix-map): Document.
24329         * doc/invoke.texi (-ffile-prefix-map): Document.
24330         (-fdebug-prefix-map): Update description.
24332 2018-01-18  Martin Liska  <mliska@suse.cz>
24334         * config/i386/i386.c (indirect_thunk_name): Document that also
24335         lfence is emitted.
24336         (output_indirect_thunk): Document why both instructions
24337         (pause and lfence) are generated.
24339 2018-01-18  Richard Biener  <rguenther@suse.de>
24341         PR tree-optimization/83887
24342         * graphite-scop-detection.c
24343         (scop_detection::get_nearest_dom_with_single_entry): Remove.
24344         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
24345         (scop_detection::merge_sese): Re-implement with a flood-fill
24346         algorithm that properly finds a SESE region if it exists.
24348 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24350         PR c/61240
24351         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
24352         pointer_diff optimizations use view_convert instead of convert.
24354 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24356         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24357         Generate different code for -mno-speculate-indirect-jumps.
24358         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24359         (*call_indirect_aix<mode>): Disable for
24360         -mno-speculate-indirect-jumps.
24361         (*call_indirect_aix<mode>_nospec): New define_insn.
24362         (*call_value_indirect_aix<mode>): Disable for
24363         -mno-speculate-indirect-jumps.
24364         (*call_value_indirect_aix<mode>_nospec): New define_insn.
24365         (*sibcall_nonlocal_sysv<mode>): Generate different code for
24366         -mno-speculate-indirect-jumps.
24367         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24369 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
24371         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
24372         long double type, set the flags for noting the default long double
24373         type, even if we don't pass or return a long double type.
24375 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
24377         PR ipa/83051
24378         * ipa-inline.c (flatten_function): Do not overwrite final inlining
24379         failure.
24381 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
24383         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
24384         support for merge[hl].
24385         (fold_mergehl_helper): New helper function.
24386         (tree-vector-builder.h): New #include for tree_vector_builder usage.
24387         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
24388         (altivec_vmrglw_direct): Add xxmrglw insn.
24390 2018-01-17  Andrew Waterman  <andrew@sifive.com>
24392         * config/riscv/riscv.c (riscv_conditional_register_usage): If
24393         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
24395 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
24397         PR lto/83121
24398         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
24399         call the lto_location_cache before reading the
24400         DECL_SOURCE_LOCATION of the types.
24402 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
24403             Richard Sandiford  <richard.sandiford@linaro.org>
24405         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
24406         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
24407         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
24408         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
24409         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
24410         Add declaration.
24411         * config/aarch64/constraints.md (aarch64_movti_operand):
24412         Limit immediates.
24413         * config/aarch64/predicates.md (Uti): Add new constraint.
24415 2018-01-17  Carl Love  <cel@us.ibm.com>
24417         * config/rs6000/vsx.md (define_expand xl_len_r,
24418         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
24419         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
24420         lxvll.
24421         (define_expand, define_insn): Move the shift left from  the
24422         define_insn to the define_expand for lxvl and stxvl instructions.
24423         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
24424         and XL_LEN_R definitions to PURE.
24426 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
24428         * config/i386/i386.c (indirect_thunk_name): Declare regno
24429         as unsigned int.  Compare regno with INVALID_REGNUM.
24430         (output_indirect_thunk): Ditto.
24431         (output_indirect_thunk_function): Ditto.
24432         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
24433         in the call to output_indirect_thunk_function.
24435 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
24437         PR middle-end/83884
24438         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
24439         rather than the size of inner_type to determine the stack slot size
24440         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
24442 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
24444         PR target/83546
24445         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
24446         to PTA_SILVERMONT.
24448 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
24450         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
24451         endian Linux systems to optionally enable multilibs for selecting
24452         the long double type if the user configured an explicit type.
24453         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
24454         have no long double multilibs if not defined.
24455         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
24456         warn if the user used -mabi={ieee,ibm}longdouble and we built
24457         multilibs for long double.
24458         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
24459         appropriate multilib option.
24460         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
24461         multilib options.
24462         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
24463         for building long double multilibs.
24464         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
24466 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
24468         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
24469         copies.
24471         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
24472         64 bits.
24473         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
24474         128 bits.
24476         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
24477         variables.
24479         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
24480         return value.
24482 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24484         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
24485         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
24487 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24489         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
24490         different rtl trees depending on TARGET_64BIT.
24491         (rs6000_gen_lvx): Likewise.
24493 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24495         * config/visium/visium.md (nop): Tweak comment.
24496         (hazard_nop): Likewise.
24498 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24500         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
24501         -mspeculate-indirect-jumps.
24502         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
24503         for -mno-speculate-indirect-jumps.
24504         (*call_indirect_elfv2<mode>_nospec): New define_insn.
24505         (*call_value_indirect_elfv2<mode>): Disable for
24506         -mno-speculate-indirect-jumps.
24507         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
24508         (indirect_jump): Emit different RTL for
24509         -mno-speculate-indirect-jumps.
24510         (*indirect_jump<mode>): Disable for
24511         -mno-speculate-indirect-jumps.
24512         (*indirect_jump<mode>_nospec): New define_insn.
24513         (tablejump): Emit different RTL for
24514         -mno-speculate-indirect-jumps.
24515         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
24516         (tablejumpsi_nospec): New define_expand.
24517         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
24518         (tablejumpdi_nospec): New define_expand.
24519         (*tablejump<mode>_internal1): Disable for
24520         -mno-speculate-indirect-jumps.
24521         (*tablejump<mode>_internal1_nospec): New define_insn.
24522         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
24523         option.
24525 2018-01-16  Artyom Skrobov tyomitch@gmail.com
24527         * caller-save.c (insert_save): Drop unnecessary parameter.  All
24528         callers updated.
24530 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24531             Richard Biener  <rguenth@suse.de>
24533         PR libgomp/83590
24534         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
24535         return early, inline manually is_gimple_sizepos.  Make sure if we
24536         call gimplify_expr we don't end up with a gimple constant.
24537         * tree.c (variably_modified_type_p): Don't return true for
24538         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
24539         * gimplify.h (is_gimple_sizepos): Remove.
24541 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24543         PR tree-optimization/83857
24544         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
24545         vectorizable_live_operation for pure SLP statements.
24546         (vectorizable_live_operation): Handle PHIs.
24548 2018-01-16  Richard Biener  <rguenther@suse.de>
24550         PR tree-optimization/83867
24551         * tree-vect-stmts.c (vect_transform_stmt): Precompute
24552         nested_in_vect_loop_p since the scalar stmt may get invalidated.
24554 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24556         PR c/83844
24557         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
24558         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
24559         If off is not INTEGER_CST, issue a may not be aligned warning
24560         rather than isn't aligned.  Use isn%'t rather than isn't.
24561         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
24562         into MULT_EXPR.
24563         <case MULT_EXPR>: Improve the case when bottom and one of the
24564         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
24565         operand, in that case check if the other operand is multiple of
24566         bottom divided by the INTEGER_CST operand.
24568 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24570         PR target/83858
24571         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
24572         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
24573         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
24574         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
24575         * config/pa/pa.c (pa_function_arg_advance): Likewise.
24576         (pa_function_arg, pa_arg_partial_bytes): Likewise.
24577         (pa_function_arg_size): New function.
24579 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24581         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
24582         in a separate statement.
24584 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24586         PR tree-optimization/83847
24587         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
24588         group gathers and scatters.
24590 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24592         PR rtl-optimization/86620
24593         * params.def (max-sched-ready-insns): Bump minimum value to 1.
24595         PR rtl-optimization/83213
24596         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
24597         to last if both are JUMP_INSNs.
24599         PR tree-optimization/83843
24600         * gimple-ssa-store-merging.c
24601         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
24602         store_immediate_info for bswap/nop orig_stores.
24604 2018-01-15  Andrew Waterman  <andrew@sifive.com>
24606         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
24607         !TARGET_MUL.
24608         <UDIV>: Increase cost if !TARGET_DIV.
24610 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
24612         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
24613         (define_attr "cr_logical_3op"): New.
24614         (cceq_ior_compare): Adjust.
24615         (cceq_ior_compare_complement): Adjust.
24616         (*cceq_rev_compare): Adjust.
24617         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
24618         (is_cracked_insn): Adjust.
24619         (insn_must_be_first_in_group): Adjust.
24620         * config/rs6000/40x.md: Adjust.
24621         * config/rs6000/440.md: Adjust.
24622         * config/rs6000/476.md: Adjust.
24623         * config/rs6000/601.md: Adjust.
24624         * config/rs6000/603.md: Adjust.
24625         * config/rs6000/6xx.md: Adjust.
24626         * config/rs6000/7450.md: Adjust.
24627         * config/rs6000/7xx.md: Adjust.
24628         * config/rs6000/8540.md: Adjust.
24629         * config/rs6000/cell.md: Adjust.
24630         * config/rs6000/e300c2c3.md: Adjust.
24631         * config/rs6000/e500mc.md: Adjust.
24632         * config/rs6000/e500mc64.md: Adjust.
24633         * config/rs6000/e5500.md: Adjust.
24634         * config/rs6000/e6500.md: Adjust.
24635         * config/rs6000/mpc.md: Adjust.
24636         * config/rs6000/power4.md: Adjust.
24637         * config/rs6000/power5.md: Adjust.
24638         * config/rs6000/power6.md: Adjust.
24639         * config/rs6000/power7.md: Adjust.
24640         * config/rs6000/power8.md: Adjust.
24641         * config/rs6000/power9.md: Adjust.
24642         * config/rs6000/rs64.md: Adjust.
24643         * config/rs6000/titan.md: Adjust.
24645 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24647         * config/i386/predicates.md (indirect_branch_operand): Rewrite
24648         ix86_indirect_branch_register logic.
24650 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24652         * config/i386/constraints.md (Bs): Update
24653         ix86_indirect_branch_register check.  Don't check
24654         ix86_indirect_branch_register with GOT_memory_operand.
24655         (Bw): Likewise.
24656         * config/i386/predicates.md (GOT_memory_operand): Don't check
24657         ix86_indirect_branch_register here.
24658         (GOT32_symbol_operand): Likewise.
24660 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24662         * config/i386/predicates.md (constant_call_address_operand):
24663         Rewrite ix86_indirect_branch_register logic.
24664         (sibcall_insn_operand): Likewise.
24666 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24668         * config/i386/constraints.md (Bs): Replace
24669         ix86_indirect_branch_thunk_register with
24670         ix86_indirect_branch_register.
24671         (Bw): Likewise.
24672         * config/i386/i386.md (indirect_jump): Likewise.
24673         (tablejump): Likewise.
24674         (*sibcall_memory): Likewise.
24675         (*sibcall_value_memory): Likewise.
24676         Peepholes of indirect call and jump via memory: Likewise.
24677         * config/i386/i386.opt: Likewise.
24678         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24679         (GOT_memory_operand): Likewise.
24680         (call_insn_operand): Likewise.
24681         (sibcall_insn_operand): Likewise.
24682         (GOT32_symbol_operand): Likewise.
24684 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
24686         PR middle-end/83837
24687         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
24688         type rather than type addr's type points to.
24689         (expand_omp_atomic_mutex): Likewise.
24690         (expand_omp_atomic): Likewise.
24692 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24694         PR target/83839
24695         * config/i386/i386.c (output_indirect_thunk_function): Use
24696         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
24697         for  __x86_return_thunk.
24699 2018-01-15  Richard Biener  <rguenther@suse.de>
24701         PR middle-end/83850
24702         * expmed.c (extract_bit_field_1): Fix typo.
24704 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24706         PR target/83687
24707         * config/arm/iterators.md (VF): New mode iterator.
24708         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
24709         Remove integer-related logic from pattern.
24710         (neon_vabd<mode>_3): Likewise.
24712 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
24714         PR middle-end/82694
24715         * common.opt (fstrict-overflow): No longer an alias.
24716         (fwrapv-pointer): New option.
24717         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
24718         also for pointer types based on flag_wrapv_pointer.
24719         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
24720         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
24721         opts->x_flag_wrapv got set.
24722         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
24723         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
24724         POINTER_TYPE_OVERFLOW_UNDEFINED.
24725         * match.pd: Likewise in address comparison pattern.
24726         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
24728 2018-01-15  Richard Biener  <rguenther@suse.de>
24730         PR lto/83804
24731         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
24732         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
24733         Reset type names to their identifier if their TYPE_DECL doesn't
24734         have linkage (and thus is used for ODR and devirt).
24735         (save_debug_info_for_decl): Remove.
24736         (save_debug_info_for_type): Likewise.
24737         (add_tree_to_fld_list): Adjust.
24738         * tree-pretty-print.c (dump_generic_node): Make dumping of
24739         type names more robust.
24741 2018-01-15  Richard Biener  <rguenther@suse.de>
24743         * BASE-VER: Bump to 8.0.1.
24745 2018-01-14  Martin Sebor  <msebor@redhat.com>
24747         PR other/83508
24748         * builtins.c (check_access): Avoid warning when the no-warning bit
24749         is set.
24751 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
24753         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
24754         * ira-color (allocno_hard_regs_compare): Likewise.
24756 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
24758         PR target/83013
24759         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
24760         Use .pushsection/.popsection.
24762 2018-01-14  Martin Sebor  <msebor@redhat.com>
24764         PR c++/81327
24765         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
24767 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
24769         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
24770         entry from extra_headers.
24771         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
24772         extra_headers, make the list bitwise identical to the i?86-*-* one.
24774 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24776         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
24777         -mcmodel=large with -mindirect-branch=thunk,
24778         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
24779         -mfunction-return=thunk-extern.
24780         * doc/invoke.texi: Document -mcmodel=large is incompatible with
24781         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
24782         -mfunction-return=thunk and -mfunction-return=thunk-extern.
24784 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24786         * config/i386/i386.c (print_reg): Print the name of the full
24787         integer register without '%'.
24788         (ix86_print_operand): Handle 'V'.
24789         * doc/extend.texi: Document 'V' modifier.
24791 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24793         * config/i386/constraints.md (Bs): Disallow memory operand for
24794         -mindirect-branch-register.
24795         (Bw): Likewise.
24796         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24797         (GOT_memory_operand): Likewise.
24798         (call_insn_operand): Likewise.
24799         (sibcall_insn_operand): Likewise.
24800         (GOT32_symbol_operand): Likewise.
24801         * config/i386/i386.md (indirect_jump): Call convert_memory_address
24802         for -mindirect-branch-register.
24803         (tablejump): Likewise.
24804         (*sibcall_memory): Likewise.
24805         (*sibcall_value_memory): Likewise.
24806         Disallow peepholes of indirect call and jump via memory for
24807         -mindirect-branch-register.
24808         (*call_pop): Replace m with Bw.
24809         (*call_value_pop): Likewise.
24810         (*sibcall_pop_memory): Replace m with Bs.
24811         * config/i386/i386.opt (mindirect-branch-register): New option.
24812         * doc/invoke.texi: Document -mindirect-branch-register option.
24814 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24816         * config/i386/i386-protos.h (ix86_output_function_return): New.
24817         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
24818         set function_return_type.
24819         (indirect_thunk_name): Add ret_p to indicate thunk for function
24820         return.
24821         (output_indirect_thunk_function): Pass false to
24822         indirect_thunk_name.
24823         (ix86_output_indirect_branch_via_reg): Likewise.
24824         (ix86_output_indirect_branch_via_push): Likewise.
24825         (output_indirect_thunk_function): Create alias for function
24826         return thunk if regno < 0.
24827         (ix86_output_function_return): New function.
24828         (ix86_handle_fndecl_attribute): Handle function_return.
24829         (ix86_attribute_table): Add function_return.
24830         * config/i386/i386.h (machine_function): Add
24831         function_return_type.
24832         * config/i386/i386.md (simple_return_internal): Use
24833         ix86_output_function_return.
24834         (simple_return_internal_long): Likewise.
24835         * config/i386/i386.opt (mfunction-return=): New option.
24836         (indirect_branch): Mention -mfunction-return=.
24837         * doc/extend.texi: Document function_return function attribute.
24838         * doc/invoke.texi: Document -mfunction-return= option.
24840 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24842         * config/i386/i386-opts.h (indirect_branch): New.
24843         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
24844         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
24845         with local indirect jump when converting indirect call and jump.
24846         (ix86_set_indirect_branch_type): New.
24847         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
24848         (indirectlabelno): New.
24849         (indirect_thunk_needed): Likewise.
24850         (indirect_thunk_bnd_needed): Likewise.
24851         (indirect_thunks_used): Likewise.
24852         (indirect_thunks_bnd_used): Likewise.
24853         (INDIRECT_LABEL): Likewise.
24854         (indirect_thunk_name): Likewise.
24855         (output_indirect_thunk): Likewise.
24856         (output_indirect_thunk_function): Likewise.
24857         (ix86_output_indirect_branch_via_reg): Likewise.
24858         (ix86_output_indirect_branch_via_push): Likewise.
24859         (ix86_output_indirect_branch): Likewise.
24860         (ix86_output_indirect_jmp): Likewise.
24861         (ix86_code_end): Call output_indirect_thunk_function if needed.
24862         (ix86_output_call_insn): Call ix86_output_indirect_branch if
24863         needed.
24864         (ix86_handle_fndecl_attribute): Handle indirect_branch.
24865         (ix86_attribute_table): Add indirect_branch.
24866         * config/i386/i386.h (machine_function): Add indirect_branch_type
24867         and has_local_indirect_jump.
24868         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
24869         to true.
24870         (tablejump): Likewise.
24871         (*indirect_jump): Use ix86_output_indirect_jmp.
24872         (*tablejump_1): Likewise.
24873         (simple_return_indirect_internal): Likewise.
24874         * config/i386/i386.opt (mindirect-branch=): New option.
24875         (indirect_branch): New.
24876         (keep): Likewise.
24877         (thunk): Likewise.
24878         (thunk-inline): Likewise.
24879         (thunk-extern): Likewise.
24880         * doc/extend.texi: Document indirect_branch function attribute.
24881         * doc/invoke.texi: Document -mindirect-branch= option.
24883 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
24885         PR ipa/83051
24886         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
24888 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
24890         * ipa-inline.c (want_inline_small_function_p): Return false if
24891         inlining has already failed with CIF_FINAL_ERROR.
24892         (update_caller_keys): Call want_inline_small_function_p before
24893         can_inline_edge_p.
24894         (update_callee_keys): Likewise.
24896 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24898         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
24899         New function.
24900         (rs6000_quadword_masked_address_p): Likewise.
24901         (quad_aligned_load_p): Likewise.
24902         (quad_aligned_store_p): Likewise.
24903         (const_load_sequence_p): Add comment to describe the outer-most loop.
24904         (mimic_memory_attributes_and_flags): New function.
24905         (rs6000_gen_stvx): Likewise.
24906         (replace_swapped_aligned_store): Likewise.
24907         (rs6000_gen_lvx): Likewise.
24908         (replace_swapped_aligned_load): Likewise.
24909         (replace_swapped_load_constant): Capitalize argument name in
24910         comment describing this function.
24911         (rs6000_analyze_swaps): Add a third pass to search for vector loads
24912         and stores that access quad-word aligned addresses and replace
24913         with stvx or lvx instructions when appropriate.
24914         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
24915         New function prototype.
24916         (rs6000_quadword_masked_address_p): Likewise.
24917         (rs6000_gen_lvx): Likewise.
24918         (rs6000_gen_stvx): Likewise.
24919         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
24920         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
24921         when memory address is aligned.
24922         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
24923         this split to select lvx instruction when memory address is aligned.
24924         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
24925         instruction when memory address is aligned.
24926         (*vsx_le_perm_load_v16qi): Likewise.
24927         (four unnamed splitters): Modify to select the stvx instruction
24928         when memory is aligned.
24930 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
24932         * predict.c (determine_unlikely_bbs): Handle correctly BBs
24933         which appears in the queue multiple times.
24935 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24936             Alan Hayward  <alan.hayward@arm.com>
24937             David Sherwood  <david.sherwood@arm.com>
24939         * tree-vectorizer.h (vec_lower_bound): New structure.
24940         (_loop_vec_info): Add check_nonzero and lower_bounds.
24941         (LOOP_VINFO_CHECK_NONZERO): New macro.
24942         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
24943         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
24944         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
24945         fields.  Make seg_len the distance travelled, not including the
24946         access size.
24947         (dr_direction_indicator): Declare.
24948         (dr_zero_step_indicator): Likewise.
24949         (dr_known_forward_stride_p): Likewise.
24950         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
24951         tree-ssanames.h.
24952         (runtime_alias_check_p): Allow runtime alias checks with
24953         variable strides.
24954         (operator ==): Compare access_size and align.
24955         (prune_runtime_alias_test_list): Rework for new distinction between
24956         the access_size and seg_len.
24957         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
24958         segment lengths.
24959         (get_segment_min_max): New function.
24960         (create_intersect_range_checks): Use it.
24961         (dr_step_indicator): New function.
24962         (dr_direction_indicator): Likewise.
24963         (dr_zero_step_indicator): Likewise.
24964         (dr_known_forward_stride_p): Likewise.
24965         * tree-loop-distribution.c (data_ref_segment_size): Return
24966         DR_STEP * (niters - 1).
24967         (compute_alias_check_pairs): Update call to the dr_with_seg_len
24968         constructor.
24969         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
24970         (vect_preserves_scalar_order_p): New function, split out from...
24971         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
24972         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
24973         (vect_vfa_access_size): New function.
24974         (vect_vfa_align): Likewise.
24975         (vect_compile_time_alias): Take access_size_a and access_b arguments.
24976         (dump_lower_bound): New function.
24977         (vect_check_lower_bound): Likewise.
24978         (vect_small_gap_p): Likewise.
24979         (vectorizable_with_step_bound_p): Likewise.
24980         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
24981         depencies if the vectorization factor is 1.  Convert the checks
24982         for nonzero steps into checks on the bounds of DR_STEP.  Try using
24983         a bunds check for variable steps if the minimum required step is
24984         relatively small. Update calls to the dr_with_seg_len
24985         constructor and to vect_compile_time_alias.
24986         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
24987         function.
24988         (vect_loop_versioning): Call it.
24989         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
24990         when retrying.
24991         (vect_estimate_min_profitable_iters): Account for any bounds checks.
24993 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24994             Alan Hayward  <alan.hayward@arm.com>
24995             David Sherwood  <david.sherwood@arm.com>
24997         * doc/sourcebuild.texi (vect_scatter_store): Document.
24998         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
24999         optabs.
25000         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
25001         Document.
25002         * genopinit.c (main): Add supports_vec_scatter_store and
25003         supports_vec_scatter_store_cached to target_optabs.
25004         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
25005         IFN_MASK_SCATTER_STORE.
25006         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
25007         functions.
25008         * internal-fn.h (internal_store_fn_p): Declare.
25009         (internal_fn_stored_value_index): Likewise.
25010         * internal-fn.c (scatter_store_direct): New macro.
25011         (expand_scatter_store_optab_fn): New function.
25012         (direct_scatter_store_optab_supported_p): New macro.
25013         (internal_store_fn_p): New function.
25014         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
25015         IFN_MASK_SCATTER_STORE.
25016         (internal_fn_mask_index): Likewise.
25017         (internal_fn_stored_value_index): New function.
25018         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
25019         for scatter stores.
25020         * optabs-query.h (supports_vec_scatter_store_p): Declare.
25021         * optabs-query.c (supports_vec_scatter_store_p): New function.
25022         * tree-vectorizer.h (vect_get_store_rhs): Declare.
25023         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
25024         true for scatter stores.
25025         (vect_gather_scatter_fn_p): Handle scatter stores too.
25026         (vect_check_gather_scatter): Consider using scatter stores if
25027         supports_vec_scatter_store_p.
25028         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
25029         scatter stores too.
25030         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25031         internal_fn_stored_value_index.
25032         (check_load_store_masking): Handle scatter stores too.
25033         (vect_get_store_rhs): Make public.
25034         (vectorizable_call): Use internal_store_fn_p.
25035         (vectorizable_store): Handle scatter store internal functions.
25036         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
25037         when deciding whether the end of the group has been reached.
25038         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
25039         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
25040         (mask_scatter_store<mode>): New insns.
25042 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25043             Alan Hayward  <alan.hayward@arm.com>
25044             David Sherwood  <david.sherwood@arm.com>
25046         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
25047         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
25048         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
25049         function.
25050         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
25051         Use vect_truncate_gather_scatter_offset if we can't treat the
25052         operation as a normal gather load or scatter store.
25053         (get_group_load_store_type): Take the gather_scatter_info
25054         as argument.  Try using a gather load or scatter store for
25055         single-element groups.
25056         (get_load_store_type): Update calls to get_group_load_store_type
25057         and vect_use_strided_gather_scatters_p.
25059 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25060             Alan Hayward  <alan.hayward@arm.com>
25061             David Sherwood  <david.sherwood@arm.com>
25063         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
25064         optional tree argument.
25065         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
25066         null target hooks.
25067         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
25068         but continue to use the current value as a fallback.
25069         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
25070         to compare the updates.
25071         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
25072         (get_load_store_type): Use it when handling a strided access.
25073         (vect_get_strided_load_store_ops): New function.
25074         (vect_get_data_ptr_increment): Likewise.
25075         (vectorizable_load): Handle strided gather loads.  Always pass
25076         a step to vect_create_data_ref_ptr and bump_vector_ptr.
25078 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25079             Alan Hayward  <alan.hayward@arm.com>
25080             David Sherwood  <david.sherwood@arm.com>
25082         * doc/md.texi (gather_load@var{m}): Document.
25083         (mask_gather_load@var{m}): Likewise.
25084         * genopinit.c (main): Add supports_vec_gather_load and
25085         supports_vec_gather_load_cached to target_optabs.
25086         * optabs-tree.c (init_tree_optimization_optabs): Use
25087         ggc_cleared_alloc to allocate target_optabs.
25088         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
25089         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
25090         functions.
25091         * internal-fn.h (internal_load_fn_p): Declare.
25092         (internal_gather_scatter_fn_p): Likewise.
25093         (internal_fn_mask_index): Likewise.
25094         (internal_gather_scatter_fn_supported_p): Likewise.
25095         * internal-fn.c (gather_load_direct): New macro.
25096         (expand_gather_load_optab_fn): New function.
25097         (direct_gather_load_optab_supported_p): New macro.
25098         (direct_internal_fn_optab): New function.
25099         (internal_load_fn_p): Likewise.
25100         (internal_gather_scatter_fn_p): Likewise.
25101         (internal_fn_mask_index): Likewise.
25102         (internal_gather_scatter_fn_supported_p): Likewise.
25103         * optabs-query.c (supports_at_least_one_mode_p): New function.
25104         (supports_vec_gather_load_p): Likewise.
25105         * optabs-query.h (supports_vec_gather_load_p): Declare.
25106         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
25107         and memory_type field.
25108         (NUM_PATTERNS): Bump to 15.
25109         * tree-vect-data-refs.c: Include internal-fn.h.
25110         (vect_gather_scatter_fn_p): New function.
25111         (vect_describe_gather_scatter_call): Likewise.
25112         (vect_check_gather_scatter): Try using internal functions for
25113         gather loads.  Recognize existing calls to a gather load function.
25114         (vect_analyze_data_refs): Consider using gather loads if
25115         supports_vec_gather_load_p.
25116         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
25117         (vect_get_gather_scatter_offset_type): Likewise.
25118         (vect_convert_mask_for_vectype): Likewise.
25119         (vect_add_conversion_to_patterm): Likewise.
25120         (vect_try_gather_scatter_pattern): Likewise.
25121         (vect_recog_gather_scatter_pattern): New pattern recognizer.
25122         (vect_vect_recog_func_ptrs): Add it.
25123         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25124         internal_fn_mask_index and internal_gather_scatter_fn_p.
25125         (check_load_store_masking): Take the gather_scatter_info as an
25126         argument and handle gather loads.
25127         (vect_get_gather_scatter_ops): New function.
25128         (vectorizable_call): Check internal_load_fn_p.
25129         (vectorizable_load): Likewise.  Handle gather load internal
25130         functions.
25131         (vectorizable_store): Update call to check_load_store_masking.
25132         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
25133         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
25134         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
25135         (aarch64_gather_scale_operand_d): New predicates.
25136         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
25137         (mask_gather_load<mode>): New insns.
25139 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25140             Alan Hayward  <alan.hayward@arm.com>
25141             David Sherwood  <david.sherwood@arm.com>
25143         * optabs.def (fold_left_plus_optab): New optab.
25144         * doc/md.texi (fold_left_plus_@var{m}): Document.
25145         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
25146         * internal-fn.c (fold_left_direct): Define.
25147         (expand_fold_left_optab_fn): Likewise.
25148         (direct_fold_left_optab_supported_p): Likewise.
25149         * fold-const-call.c (fold_const_fold_left): New function.
25150         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
25151         * tree-parloops.c (valid_reduction_p): New function.
25152         (gather_scalar_reductions): Use it.
25153         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
25154         (vect_finish_replace_stmt): Declare.
25155         * tree-vect-loop.c (fold_left_reduction_fn): New function.
25156         (needs_fold_left_reduction_p): New function, split out from...
25157         (vect_is_simple_reduction): ...here.  Accept reductions that
25158         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
25159         (vect_force_simple_reduction): Also store the reduction type in
25160         the assignment's STMT_VINFO_REDUC_TYPE.
25161         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
25162         (merge_with_identity): New function.
25163         (vect_expand_fold_left): Likewise.
25164         (vectorize_fold_left_reduction): Likewise.
25165         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
25166         scalar phi in place for it.  Check for target support and reject
25167         cases that would reassociate the operation.  Defer the transform
25168         phase to vectorize_fold_left_reduction.
25169         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
25170         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
25171         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
25173 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25175         * tree-if-conv.c (predicate_mem_writes): Remove redundant
25176         call to ifc_temp_var.
25178 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25179             Alan Hayward  <alan.hayward@arm.com>
25180             David Sherwood  <david.sherwood@arm.com>
25182         * target.def (legitimize_address_displacement): Take the original
25183         offset as a poly_int.
25184         * targhooks.h (default_legitimize_address_displacement): Update
25185         accordingly.
25186         * targhooks.c (default_legitimize_address_displacement): Likewise.
25187         * doc/tm.texi: Regenerate.
25188         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
25189         as an argument, moving assert of ad->disp == ad->disp_term to...
25190         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
25191         Try calling targetm.legitimize_address_displacement before expanding
25192         the address rather than afterwards, and adjust for the new interface.
25193         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
25194         Match the new hook interface.  Handle SVE addresses.
25195         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
25196         new hook interface.
25198 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25200         * Makefile.in (OBJS): Add early-remat.o.
25201         * target.def (select_early_remat_modes): New hook.
25202         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
25203         * doc/tm.texi: Regenerate.
25204         * targhooks.h (default_select_early_remat_modes): Declare.
25205         * targhooks.c (default_select_early_remat_modes): New function.
25206         * timevar.def (TV_EARLY_REMAT): New timevar.
25207         * passes.def (pass_early_remat): New pass.
25208         * tree-pass.h (make_pass_early_remat): Declare.
25209         * early-remat.c: New file.
25210         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
25211         function.
25212         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
25214 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25215             Alan Hayward  <alan.hayward@arm.com>
25216             David Sherwood  <david.sherwood@arm.com>
25218         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
25219         vfm1 with a bound_epilog parameter.
25220         (vect_do_peeling): Update calls accordingly, and move the prologue
25221         call earlier in the function.  Treat the base bound_epilog as 0 for
25222         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
25223         this base when peeling for gaps.
25224         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
25225         with fully-masked loops.
25226         (vect_estimate_min_profitable_iters): Handle the single peeled
25227         iteration in that case.
25229 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25230             Alan Hayward  <alan.hayward@arm.com>
25231             David Sherwood  <david.sherwood@arm.com>
25233         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25234         single-element interleaving even if the size is not a power of 2.
25235         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25236         accesses for single-element interleaving if the group size is
25237         not a power of 2.
25239 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25240             Alan Hayward  <alan.hayward@arm.com>
25241             David Sherwood  <david.sherwood@arm.com>
25243         * doc/md.texi (fold_extract_last_@var{m}): Document.
25244         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25245         * optabs.def (fold_extract_last_optab): New optab.
25246         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25247         * internal-fn.c (fold_extract_direct): New macro.
25248         (expand_fold_extract_optab_fn): Likewise.
25249         (direct_fold_extract_optab_supported_p): Likewise.
25250         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25251         * tree-vect-loop.c (vect_model_reduction_cost): Handle
25252         EXTRACT_LAST_REDUCTION.
25253         (get_initial_def_for_reduction): Do not create an initial vector
25254         for EXTRACT_LAST_REDUCTION reductions.
25255         (vectorizable_reduction): Leave the scalar phi in place for
25256         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
25257         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
25258         epilogue code for EXTRACT_LAST_REDUCTION and defer the
25259         transform phase to vectorizable_condition.
25260         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25261         split out from...
25262         (vect_finish_stmt_generation): ...here.
25263         (vect_finish_replace_stmt): New function.
25264         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25265         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25266         pattern.
25267         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25269 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25270             Alan Hayward  <alan.hayward@arm.com>
25271             David Sherwood  <david.sherwood@arm.com>
25273         * doc/md.texi (extract_last_@var{m}): Document.
25274         * optabs.def (extract_last_optab): New optab.
25275         * internal-fn.def (EXTRACT_LAST): New internal function.
25276         * internal-fn.c (cond_unary_direct): New macro.
25277         (expand_cond_unary_optab_fn): Likewise.
25278         (direct_cond_unary_optab_supported_p): Likewise.
25279         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25280         loops using EXTRACT_LAST.
25281         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
25282         (extract_last_<mode>): ...this optab.
25283         (vec_extract<mode><Vel>): Update accordingly.
25285 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25286             Alan Hayward  <alan.hayward@arm.com>
25287             David Sherwood  <david.sherwood@arm.com>
25289         * target.def (empty_mask_is_expensive): New hook.
25290         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
25291         * doc/tm.texi: Regenerate.
25292         * targhooks.h (default_empty_mask_is_expensive): Declare.
25293         * targhooks.c (default_empty_mask_is_expensive): New function.
25294         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25295         if the target says that empty masks are expensive.
25296         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25297         New function.
25298         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25300 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25301             Alan Hayward  <alan.hayward@arm.com>
25302             David Sherwood  <david.sherwood@arm.com>
25304         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25305         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25306         (vect_use_loop_mask_for_alignment_p): New function.
25307         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25308         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25309         niters_skip argument.  Make sure that the first niters_skip elements
25310         of the first iteration are inactive.
25311         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25312         Update call to vect_set_loop_masks_directly.
25313         (get_misalign_in_elems): New function, split out from...
25314         (vect_gen_prolog_loop_niters): ...here.
25315         (vect_update_init_of_dr): Take a code argument that specifies whether
25316         the adjustment should be added or subtracted.
25317         (vect_update_init_of_drs): Likewise.
25318         (vect_prepare_for_masked_peels): New function.
25319         (vect_do_peeling): Skip prologue peeling if we're using a mask
25320         instead.  Update call to vect_update_inits_of_drs.
25321         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25322         mask_skip_niters.
25323         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25324         alignment.  Do not include the number of peeled iterations in
25325         the minimum threshold in that case.
25326         (vectorizable_induction): Adjust the start value down by
25327         LOOP_VINFO_MASK_SKIP_NITERS iterations.
25328         (vect_transform_loop): Call vect_prepare_for_masked_peels.
25329         Take the number of skipped iterations into account when calculating
25330         the loop bounds.
25331         * tree-vect-stmts.c (vect_gen_while_not): New function.
25333 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25334             Alan Hayward  <alan.hayward@arm.com>
25335             David Sherwood  <david.sherwood@arm.com>
25337         * doc/sourcebuild.texi (vect_fully_masked): Document.
25338         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
25339         default value to 0.
25340         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
25341         split out from...
25342         (vect_analyze_loop_2): ...here. Don't check the vectorization
25343         factor against the number of loop iterations if the loop is
25344         fully-masked.
25346 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25347             Alan Hayward  <alan.hayward@arm.com>
25348             David Sherwood  <david.sherwood@arm.com>
25350         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
25351         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
25352         (dump_groups): Update accordingly.
25353         (iv_use::mem_type): New member variable.
25354         (address_p): New function.
25355         (record_use): Add a mem_type argument and initialize the new
25356         mem_type field.
25357         (record_group_use): Add a mem_type argument.  Use address_p.
25358         Remove obsolete null checks of base_object.  Update call to record_use.
25359         (find_interesting_uses_op): Update call to record_group_use.
25360         (find_interesting_uses_cond): Likewise.
25361         (find_interesting_uses_address): Likewise.
25362         (get_mem_type_for_internal_fn): New function.
25363         (find_address_like_use): Likewise.
25364         (find_interesting_uses_stmt): Try find_address_like_use before
25365         calling find_interesting_uses_op.
25366         (addr_offset_valid_p): Use the iv mem_type field as the type
25367         of the addressed memory.
25368         (add_autoinc_candidates): Likewise.
25369         (get_address_cost): Likewise.
25370         (split_small_address_groups_p): Use address_p.
25371         (split_address_groups): Likewise.
25372         (add_iv_candidate_for_use): Likewise.
25373         (autoinc_possible_for_pair): Likewise.
25374         (rewrite_groups): Likewise.
25375         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
25376         (determine_group_iv_cost): Update after split of USE_ADDRESS.
25377         (get_alias_ptr_type_for_ptr_address): New function.
25378         (rewrite_use_address): Rewrite address uses in calls that were
25379         identified by find_address_like_use.
25381 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25382             Alan Hayward  <alan.hayward@arm.com>
25383             David Sherwood  <david.sherwood@arm.com>
25385         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
25386         TARGET_MEM_REFs.
25387         * gimple-expr.h (is_gimple_addressable: Likewise.
25388         * gimple-expr.c (is_gimple_address): Likewise.
25389         * internal-fn.c (expand_call_mem_ref): New function.
25390         (expand_mask_load_optab_fn): Use it.
25391         (expand_mask_store_optab_fn): Likewise.
25393 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25394             Alan Hayward  <alan.hayward@arm.com>
25395             David Sherwood  <david.sherwood@arm.com>
25397         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
25398         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
25399         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
25400         (cond_umax@var{mode}): Document.
25401         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
25402         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
25403         (cond_umin_optab, cond_umax_optab): New optabs.
25404         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
25405         (COND_IOR, COND_XOR): New internal functions.
25406         * internal-fn.h (get_conditional_internal_fn): Declare.
25407         * internal-fn.c (cond_binary_direct): New macro.
25408         (expand_cond_binary_optab_fn): Likewise.
25409         (direct_cond_binary_optab_supported_p): Likewise.
25410         (get_conditional_internal_fn): New function.
25411         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
25412         Cope with reduction statements that are vectorized as calls rather
25413         than assignments.
25414         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
25415         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
25416         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
25417         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
25418         (UNSPEC_COND_EOR): New unspecs.
25419         (optab): Add mappings for them.
25420         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
25421         (sve_int_op, sve_fp_op): New int attributes.
25423 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25424             Alan Hayward  <alan.hayward@arm.com>
25425             David Sherwood  <david.sherwood@arm.com>
25427         * optabs.def (while_ult_optab): New optab.
25428         * doc/md.texi (while_ult@var{m}@var{n}): Document.
25429         * internal-fn.def (WHILE_ULT): New internal function.
25430         * internal-fn.h (direct_internal_fn_supported_p): New override
25431         that takes two types as argument.
25432         * internal-fn.c (while_direct): New macro.
25433         (expand_while_optab_fn): New function.
25434         (convert_optab_supported_p): Likewise.
25435         (direct_while_optab_supported_p): New macro.
25436         * wide-int.h (wi::udiv_ceil): New function.
25437         * tree-vectorizer.h (rgroup_masks): New structure.
25438         (vec_loop_masks): New typedef.
25439         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
25440         and fully_masked_p.
25441         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
25442         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
25443         (vect_max_vf): New function.
25444         (slpeel_make_loop_iterate_ntimes): Delete.
25445         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
25446         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
25447         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
25448         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
25449         internal-fn.h, stor-layout.h and optabs-query.h.
25450         (vect_set_loop_mask): New function.
25451         (add_preheader_seq): Likewise.
25452         (add_header_seq): Likewise.
25453         (interleave_supported_p): Likewise.
25454         (vect_maybe_permute_loop_masks): Likewise.
25455         (vect_set_loop_masks_directly): Likewise.
25456         (vect_set_loop_condition_masked): Likewise.
25457         (vect_set_loop_condition_unmasked): New function, split out from
25458         slpeel_make_loop_iterate_ntimes.
25459         (slpeel_make_loop_iterate_ntimes): Rename to..
25460         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
25461         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
25462         (vect_do_peeling): Update call accordingly.
25463         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
25464         loops.
25465         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25466         mask_compare_type, can_fully_mask_p and fully_masked_p.
25467         (release_vec_loop_masks): New function.
25468         (_loop_vec_info): Use it to free the loop masks.
25469         (can_produce_all_loop_masks_p): New function.
25470         (vect_get_max_nscalars_per_iter): Likewise.
25471         (vect_verify_full_masking): Likewise.
25472         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
25473         retries, and free the mask rgroups before retrying.  Check loop-wide
25474         reasons for disallowing fully-masked loops.  Make the final decision
25475         about whether use a fully-masked loop or not.
25476         (vect_estimate_min_profitable_iters): Do not assume that peeling
25477         for the number of iterations will be needed for fully-masked loops.
25478         (vectorizable_reduction): Disable fully-masked loops.
25479         (vectorizable_live_operation): Likewise.
25480         (vect_halve_mask_nunits): New function.
25481         (vect_double_mask_nunits): Likewise.
25482         (vect_record_loop_mask): Likewise.
25483         (vect_get_loop_mask): Likewise.
25484         (vect_transform_loop): Handle the case in which the final loop
25485         iteration might handle a partial vector.  Call vect_set_loop_condition
25486         instead of slpeel_make_loop_iterate_ntimes.
25487         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
25488         (check_load_store_masking): New function.
25489         (prepare_load_store_mask): Likewise.
25490         (vectorizable_store): Handle fully-masked loops.
25491         (vectorizable_load): Likewise.
25492         (supportable_widening_operation): Use vect_halve_mask_nunits for
25493         booleans.
25494         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
25495         (vect_gen_while): New function.
25496         * config/aarch64/aarch64.md (umax<mode>3): New expander.
25497         (aarch64_uqdec<mode>): New insn.
25499 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25500             Alan Hayward  <alan.hayward@arm.com>
25501             David Sherwood  <david.sherwood@arm.com>
25503         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
25504         (reduc_xor_scal_optab): New optabs.
25505         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
25506         (reduc_xor_scal_@var{m}): Document.
25507         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
25508         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
25509         internal functions.
25510         * fold-const-call.c (fold_const_call): Handle them.
25511         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
25512         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
25513         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
25514         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
25515         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
25516         (UNSPEC_XORV): New unspecs.
25517         (optab): Add entries for them.
25518         (BITWISEV): New int iterator.
25519         (bit_reduc_op): New int attributes.
25521 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25522             Alan Hayward  <alan.hayward@arm.com>
25523             David Sherwood  <david.sherwood@arm.com>
25525         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
25526         * internal-fn.def (VEC_SHL_INSERT): New internal function.
25527         * optabs.def (vec_shl_insert_optab): New optab.
25528         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
25529         (duplicate_and_interleave): Likewise.
25530         * tree-vect-loop.c: Include internal-fn.h.
25531         (neutral_op_for_slp_reduction): New function, split out from
25532         get_initial_defs_for_reduction.
25533         (get_initial_def_for_reduction): Handle option 2 for variable-length
25534         vectors by loading the neutral value into a vector and then shifting
25535         the initial value into element 0.
25536         (get_initial_defs_for_reduction): Replace the code argument with
25537         the neutral value calculated by neutral_op_for_slp_reduction.
25538         Use gimple_build_vector for constant-length vectors.
25539         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
25540         but the first group_size elements have a neutral value.
25541         Use duplicate_and_interleave otherwise.
25542         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
25543         Update call to get_initial_defs_for_reduction.  Handle SLP
25544         reductions for variable-length vectors by creating one vector
25545         result for each scalar result, with the elements associated
25546         with other scalar results stubbed out with the neutral value.
25547         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
25548         Require IFN_VEC_SHL_INSERT for double reductions on
25549         variable-length vectors, or SLP reductions that have
25550         a neutral value.  Require can_duplicate_and_interleave_p
25551         support for variable-length unchained SLP reductions if there
25552         is no neutral value, such as for MIN/MAX reductions.  Also require
25553         the number of vector elements to be a multiple of the number of
25554         SLP statements when doing variable-length unchained SLP reductions.
25555         Update call to vect_create_epilog_for_reduction.
25556         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
25557         and remove initial values.
25558         (duplicate_and_interleave): Make public.
25559         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
25560         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
25562 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25563             Alan Hayward  <alan.hayward@arm.com>
25564             David Sherwood  <david.sherwood@arm.com>
25566         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
25567         (can_duplicate_and_interleave_p): New function.
25568         (vect_get_and_check_slp_defs): Take the vector of statements
25569         rather than just the current one.  Remove excess parentheses.
25570         Restriction rejectinon of vect_constant_def and vect_external_def
25571         for variable-length vectors to boolean types, or types for which
25572         can_duplicate_and_interleave_p is false.
25573         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
25574         (duplicate_and_interleave): New function.
25575         (vect_get_constant_vectors): Use gimple_build_vector for
25576         constant-length vectors and suitable variable-length constant
25577         vectors.  Use duplicate_and_interleave for other variable-length
25578         vectors.  Don't defer the update when inserting new statements.
25580 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25581             Alan Hayward  <alan.hayward@arm.com>
25582             David Sherwood  <david.sherwood@arm.com>
25584         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
25585         min_profitable_iters doesn't go negative.
25587 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25588             Alan Hayward  <alan.hayward@arm.com>
25589             David Sherwood  <david.sherwood@arm.com>
25591         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
25592         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
25593         * optabs.def (vec_mask_load_lanes_optab): New optab.
25594         (vec_mask_store_lanes_optab): Likewise.
25595         * internal-fn.def (MASK_LOAD_LANES): New internal function.
25596         (MASK_STORE_LANES): Likewise.
25597         * internal-fn.c (mask_load_lanes_direct): New macro.
25598         (mask_store_lanes_direct): Likewise.
25599         (expand_mask_load_optab_fn): Handle masked operations.
25600         (expand_mask_load_lanes_optab_fn): New macro.
25601         (expand_mask_store_optab_fn): Handle masked operations.
25602         (expand_mask_store_lanes_optab_fn): New macro.
25603         (direct_mask_load_lanes_optab_supported_p): Likewise.
25604         (direct_mask_store_lanes_optab_supported_p): Likewise.
25605         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
25606         parameter.
25607         (vect_load_lanes_supported): Likewise.
25608         * tree-vect-data-refs.c (strip_conversion): New function.
25609         (can_group_stmts_p): Likewise.
25610         (vect_analyze_data_ref_accesses): Use it instead of checking
25611         for a pair of assignments.
25612         (vect_store_lanes_supported): Take a masked_p parameter.
25613         (vect_load_lanes_supported): Likewise.
25614         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
25615         vect_store_lanes_supported and vect_load_lanes_supported.
25616         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
25617         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
25618         parameter.  Don't allow gaps for masked accesses.
25619         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
25620         and vect_load_lanes_supported.
25621         (get_load_store_type): Take a masked_p parameter and update
25622         call to get_group_load_store_type.
25623         (vectorizable_store): Update call to get_load_store_type.
25624         Handle IFN_MASK_STORE_LANES.
25625         (vectorizable_load): Update call to get_load_store_type.
25626         Handle IFN_MASK_LOAD_LANES.
25628 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25629             Alan Hayward  <alan.hayward@arm.com>
25630             David Sherwood  <david.sherwood@arm.com>
25632         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
25633         modes for SVE.
25634         * config/aarch64/aarch64-protos.h
25635         (aarch64_sve_struct_memory_operand_p): Declare.
25636         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
25637         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
25638         (VPRED, vpred): Handle SVE structure modes.
25639         * config/aarch64/constraints.md (Utx): New constraint.
25640         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
25641         (aarch64_sve_struct_nonimmediate_operand): New predicates.
25642         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
25643         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
25644         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
25645         structure modes.  Split into pieces after RA.
25646         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
25647         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
25648         New patterns.
25649         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
25650         SVE structure modes.
25651         (aarch64_classify_address): Likewise.
25652         (sizetochar): Move earlier in file.
25653         (aarch64_print_operand): Handle SVE register lists.
25654         (aarch64_array_mode): New function.
25655         (aarch64_sve_struct_memory_operand_p): Likewise.
25656         (TARGET_ARRAY_MODE): Redefine.
25658 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25659             Alan Hayward  <alan.hayward@arm.com>
25660             David Sherwood  <david.sherwood@arm.com>
25662         * target.def (array_mode): New target hook.
25663         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
25664         * doc/tm.texi: Regenerate.
25665         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
25666         * hooks.c (hook_optmode_mode_uhwi_none): New function.
25667         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
25668         targetm.array_mode.
25669         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
25670         type sizes.
25672 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25673             Alan Hayward  <alan.hayward@arm.com>
25674             David Sherwood  <david.sherwood@arm.com>
25676         * fold-const.c (fold_binary_loc): Check the argument types
25677         rather than the result type when testing for a vector operation.
25679 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25681         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
25682         * doc/tm.texi: Regenerate.
25684 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25685             Alan Hayward  <alan.hayward@arm.com>
25686             David Sherwood  <david.sherwood@arm.com>
25688         * doc/invoke.texi (-msve-vector-bits=): Document new option.
25689         (sve): Document new AArch64 extension.
25690         * doc/md.texi (w): Extend the description of the AArch64
25691         constraint to include SVE vectors.
25692         (Upl, Upa): Document new AArch64 predicate constraints.
25693         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
25694         enum.
25695         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
25696         (msve-vector-bits=): New option.
25697         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
25698         SVE when these are disabled.
25699         (sve): New extension.
25700         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
25701         modes.  Adjust their number of units based on aarch64_sve_vg.
25702         (MAX_BITSIZE_MODE_ANY_MODE): Define.
25703         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
25704         aarch64_addr_query_type.
25705         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
25706         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
25707         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
25708         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
25709         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
25710         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
25711         (aarch64_simd_imm_zero_p): Delete.
25712         (aarch64_check_zero_based_sve_index_immediate): Declare.
25713         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25714         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25715         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25716         (aarch64_sve_float_mul_immediate_p): Likewise.
25717         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25718         rather than an rtx.
25719         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
25720         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
25721         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
25722         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
25723         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
25724         (aarch64_regmode_natural_size): Likewise.
25725         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
25726         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
25727         left one place.
25728         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
25729         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
25730         for VG and the SVE predicate registers.
25731         (V_ALIASES): Add a "z"-prefixed alias.
25732         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
25733         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
25734         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
25735         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
25736         (REG_CLASS_NAMES): Add entries for them.
25737         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
25738         and the predicate registers.
25739         (aarch64_sve_vg): Declare.
25740         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
25741         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
25742         (REGMODE_NATURAL_SIZE): Define.
25743         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
25744         SVE macros.
25745         * config/aarch64/aarch64.c: Include cfgrtl.h.
25746         (simd_immediate_info): Add a constructor for series vectors,
25747         and an associated step field.
25748         (aarch64_sve_vg): New variable.
25749         (aarch64_dbx_register_number): Handle VG and the predicate registers.
25750         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
25751         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
25752         (VEC_ANY_DATA, VEC_STRUCT): New constants.
25753         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
25754         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
25755         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
25756         (aarch64_get_mask_mode): New functions.
25757         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
25758         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
25759         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
25760         predicate modes and predicate registers.  Explicitly restrict
25761         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
25762         to store a vector mode if it is recognized by
25763         aarch64_classify_vector_mode.
25764         (aarch64_regmode_natural_size): New function.
25765         (aarch64_hard_regno_caller_save_mode): Return the original mode
25766         for predicates.
25767         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
25768         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
25769         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
25770         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
25771         functions.
25772         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
25773         does not overlap dest if the function is frame-related.  Handle
25774         SVE constants.
25775         (aarch64_split_add_offset): New function.
25776         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
25777         them aarch64_add_offset.
25778         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
25779         and update call to aarch64_sub_sp.
25780         (aarch64_add_cfa_expression): New function.
25781         (aarch64_expand_prologue): Pass extra temporary registers to the
25782         functions above.  Handle the case in which we need to emit new
25783         DW_CFA_expressions for registers that were originally saved
25784         relative to the stack pointer, but now have to be expressed
25785         relative to the frame pointer.
25786         (aarch64_output_mi_thunk): Pass extra temporary registers to the
25787         functions above.
25788         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
25789         IP0 and IP1 values for SVE frames.
25790         (aarch64_expand_vec_series): New function.
25791         (aarch64_expand_sve_widened_duplicate): Likewise.
25792         (aarch64_expand_sve_const_vector): Likewise.
25793         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
25794         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
25795         into the register, rather than emitting a SET directly.
25796         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
25797         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
25798         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
25799         (offset_9bit_signed_scaled_p): New functions.
25800         (aarch64_replicate_bitmask_imm): New function.
25801         (aarch64_bitmask_imm): Use it.
25802         (aarch64_cannot_force_const_mem): Reject expressions involving
25803         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
25804         (aarch64_classify_index): Handle SVE indices, by requiring
25805         a plain register index with a scale that matches the element size.
25806         (aarch64_classify_address): Handle SVE addresses.  Assert that
25807         the mode of the address is VOIDmode or an integer mode.
25808         Update call to aarch64_classify_symbol.
25809         (aarch64_classify_symbolic_expression): Update call to
25810         aarch64_classify_symbol.
25811         (aarch64_const_vec_all_in_range_p): New function.
25812         (aarch64_print_vector_float_operand): Likewise.
25813         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
25814         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
25815         and the FP immediates 1.0 and 0.5.
25816         (aarch64_print_address_internal): Handle SVE addresses.
25817         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
25818         (aarch64_regno_regclass): Handle predicate registers.
25819         (aarch64_secondary_reload): Handle big-endian reloads of SVE
25820         data modes.
25821         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
25822         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
25823         (aarch64_convert_sve_vector_bits): New function.
25824         (aarch64_override_options): Use it to handle -msve-vector-bits=.
25825         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25826         rather than an rtx.
25827         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
25828         Handle SVE vector and predicate modes.  Accept VL-based constants
25829         that need only one temporary register, and VL offsets that require
25830         no temporary registers.
25831         (aarch64_conditional_register_usage): Mark the predicate registers
25832         as fixed if SVE isn't available.
25833         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
25834         Return true for SVE vector and predicate modes.
25835         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
25836         rather than an unsigned int.  Handle SVE modes.
25837         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
25838         SVE modes.
25839         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
25840         if SVE is enabled.
25841         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25842         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25843         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25844         (aarch64_sve_float_mul_immediate_p): New functions.
25845         (aarch64_sve_valid_immediate): New function.
25846         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
25847         Explicitly reject structure modes.  Check for INDEX constants.
25848         Handle PTRUE and PFALSE constants.
25849         (aarch64_check_zero_based_sve_index_immediate): New function.
25850         (aarch64_simd_imm_zero_p): Delete.
25851         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
25852         vector modes.  Accept constants in the range of CNT[BHWD].
25853         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
25854         ask for an Advanced SIMD mode.
25855         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
25856         (aarch64_simd_vector_alignment): Handle SVE predicates.
25857         (aarch64_vectorize_preferred_vector_alignment): New function.
25858         (aarch64_simd_vector_alignment_reachable): Use it instead of
25859         the vector size.
25860         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
25861         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
25862         functions.
25863         (MAX_VECT_LEN): Delete.
25864         (expand_vec_perm_d): Add a vec_flags field.
25865         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
25866         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
25867         (aarch64_evpc_ext): Don't apply a big-endian lane correction
25868         for SVE modes.
25869         (aarch64_evpc_rev): Rename to...
25870         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
25871         (aarch64_evpc_rev_global): New function.
25872         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
25873         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
25874         MAX_VECT_LEN.
25875         (aarch64_evpc_sve_tbl): New function.
25876         (aarch64_expand_vec_perm_const_1): Update after rename of
25877         aarch64_evpc_rev.  Handle SVE permutes too, trying
25878         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
25879         than aarch64_evpc_tbl.
25880         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
25881         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
25882         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
25883         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
25884         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
25885         (aarch64_expand_sve_vcond): New functions.
25886         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
25887         of aarch64_vector_mode_p.
25888         (aarch64_dwarf_poly_indeterminate_value): New function.
25889         (aarch64_compute_pressure_classes): Likewise.
25890         (aarch64_can_change_mode_class): Likewise.
25891         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
25892         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
25893         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
25894         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
25895         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
25896         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
25897         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
25898         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
25899         constraints.
25900         (Dn, Dl, Dr): Accept const as well as const_vector.
25901         (Dz): Likewise.  Compare against CONST0_RTX.
25902         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
25903         of "vector" where appropriate.
25904         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
25905         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
25906         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
25907         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
25908         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
25909         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
25910         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
25911         (v_int_equiv): Extend to SVE modes.
25912         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
25913         mode attributes.
25914         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
25915         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
25916         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
25917         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
25918         (SVE_COND_FP_CMP): New int iterators.
25919         (perm_hilo): Handle the new unpack unspecs.
25920         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
25921         attributes.
25922         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
25923         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
25924         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
25925         (aarch64_equality_operator, aarch64_constant_vector_operand)
25926         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
25927         (aarch64_sve_nonimmediate_operand): Likewise.
25928         (aarch64_sve_general_operand): Likewise.
25929         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
25930         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
25931         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
25932         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
25933         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
25934         (aarch64_sve_float_arith_immediate): Likewise.
25935         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
25936         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
25937         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
25938         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
25939         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
25940         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
25941         (aarch64_sve_float_arith_operand): Likewise.
25942         (aarch64_sve_float_arith_with_sub_operand): Likewise.
25943         (aarch64_sve_float_mul_operand): Likewise.
25944         (aarch64_sve_vec_perm_operand): Likewise.
25945         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
25946         (aarch64_mov_operand): Accept const_poly_int and const_vector.
25947         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
25948         as well as const_vector.
25949         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
25950         in file.  Use CONST0_RTX and CONSTM1_RTX.
25951         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
25952         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
25953         Use aarch64_simd_imm_zero.
25954         * config/aarch64/aarch64-sve.md: New file.
25955         * config/aarch64/aarch64.md: Include it.
25956         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
25957         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
25958         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
25959         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
25960         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
25961         (sve): New attribute.
25962         (enabled): Disable instructions with the sve attribute unless
25963         TARGET_SVE.
25964         (movqi, movhi): Pass CONST_POLY_INT operaneds through
25965         aarch64_expand_mov_immediate.
25966         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
25967         CNT[BHSD] immediates.
25968         (movti): Split CONST_POLY_INT moves into two halves.
25969         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
25970         Split additions that need a temporary here if the destination
25971         is the stack pointer.
25972         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
25973         (*add<mode>3_poly_1): New instruction.
25974         (set_clobber_cc): New expander.
25976 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25978         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
25979         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
25980         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
25981         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
25982         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
25983         Change innermode from fixed_mode_size to machine_mode.
25984         (simplify_subreg): Update call accordingly.  Handle a constant-sized
25985         subreg of a variable-length CONST_VECTOR.
25987 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25988             Alan Hayward  <alan.hayward@arm.com>
25989             David Sherwood  <david.sherwood@arm.com>
25991         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
25992         (add_offset_to_base): New function, split out from...
25993         (create_mem_ref): ...here.  When handling a scale other than 1,
25994         check first whether the address is valid without the offset.
25995         Add it into the base if so, leaving the index and scale as-is.
25997 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
25999         PR c++/83778
26000         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
26001         fold_for_warn before checking if arg2 is INTEGER_CST.
26003 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
26005         * config/rs6000/predicates.md (load_multiple_operation): Delete.
26006         (store_multiple_operation): Delete.
26007         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
26008         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
26009         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
26010         guarded by TARGET_STRING.
26011         (rs6000_output_load_multiple): Delete.
26012         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
26013         OPTION_MASK_STRING / TARGET_STRING handling.
26014         (print_operand) <'N', 'O'>: Add comment that these are unused now.
26015         (const rs6000_opt_masks) <"string">: Change mask to 0.
26016         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
26017         (MASK_STRING): Delete.
26018         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
26019         parts.  Simplify.
26020         (load_multiple): Delete.
26021         (*ldmsi8): Delete.
26022         (*ldmsi7): Delete.
26023         (*ldmsi6): Delete.
26024         (*ldmsi5): Delete.
26025         (*ldmsi4): Delete.
26026         (*ldmsi3): Delete.
26027         (store_multiple): Delete.
26028         (*stmsi8): Delete.
26029         (*stmsi7): Delete.
26030         (*stmsi6): Delete.
26031         (*stmsi5): Delete.
26032         (*stmsi4): Delete.
26033         (*stmsi3): Delete.
26034         (movmemsi_8reg): Delete.
26035         (corresponding unnamed define_insn): Delete.
26036         (movmemsi_6reg): Delete.
26037         (corresponding unnamed define_insn): Delete.
26038         (movmemsi_4reg): Delete.
26039         (corresponding unnamed define_insn): Delete.
26040         (movmemsi_2reg): Delete.
26041         (corresponding unnamed define_insn): Delete.
26042         (movmemsi_1reg): Delete.
26043         (corresponding unnamed define_insn): Delete.
26044         * config/rs6000/rs6000.opt (mno-string): New.
26045         (mstring): Replace by deprecation warning stub.
26046         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
26048 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26050         * regrename.c (regrename_do_replace): If replacing the same
26051         reg multiple times, try to reuse last created gen_raw_REG.
26053         PR debug/81155
26054         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
26055         main to workaround a bug in GDB.
26057 2018-01-12  Tom de Vries  <tom@codesourcery.com>
26059         PR target/83737
26060         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
26062 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
26064         PR rtl-optimization/80481
26065         * ira-color.c (get_cap_member): New function.
26066         (allocnos_conflict_by_live_ranges_p): Use it.
26067         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
26068         (setup_slot_coalesced_allocno_live_ranges): Ditto.
26070 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
26072         PR target/83628
26073         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
26074         (*saddl_se_1): Ditto.
26075         (*ssubsi_1): Ditto.
26076         (*ssubl_se_1): Ditto.
26078 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26080         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
26081         rather than wi::to_widest for DR_INITs.
26082         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
26083         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
26084         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
26085         INTEGER_CSTs.
26086         (vect_analyze_group_access_1): Note that here.
26088 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26090         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
26091         polynomial type sizes.
26093 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26095         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
26096         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
26097         (gimple_add_tmp_var): Likewise.
26099 2018-01-12  Martin Liska  <mliska@suse.cz>
26101         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
26102         (gimple_alloc_sizes): Likewise.
26103         (dump_gimple_statistics): Use PRIu64 in printf format.
26104         * gimple.h: Change uint64_t to int.
26106 2018-01-12  Martin Liska  <mliska@suse.cz>
26108         * tree-core.h: Use uint64_t instead of int.
26109         * tree.c (tree_node_counts): Likewise.
26110         (tree_node_sizes): Likewise.
26111         (dump_tree_statistics): Use PRIu64 in printf format.
26113 2018-01-12  Martin Liska  <mliska@suse.cz>
26115         * Makefile.in: As qsort_chk is implemented in vec.c, add
26116         vec.o to linkage of gencfn-macros.
26117         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
26118         passing the info to record_node_allocation_statistics.
26119         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
26120         and pass the info.
26121         * ggc-common.c (struct ggc_usage): Add operator== and use
26122         it in operator< and compare function.
26123         * mem-stats.h (struct mem_usage): Likewise.
26124         * vec.c (struct vec_usage): Remove operator< and compare
26125         function. Can be simply inherited.
26127 2018-01-12  Martin Jambor  <mjambor@suse.cz>
26129         PR target/81616
26130         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
26131         * tree-ssa-math-opts.c: Include domwalk.h.
26132         (convert_mult_to_fma_1): New function.
26133         (fma_transformation_info): New type.
26134         (fma_deferring_state): Likewise.
26135         (cancel_fma_deferring): New function.
26136         (result_of_phi): Likewise.
26137         (last_fma_candidate_feeds_initial_phi): Likewise.
26138         (convert_mult_to_fma): Added deferring logic, split actual
26139         transformation to convert_mult_to_fma_1.
26140         (math_opts_dom_walker): New type.
26141         (math_opts_dom_walker::after_dom_children): New method, body moved
26142         here from pass_optimize_widening_mul::execute, added deferring logic
26143         bits.
26144         (pass_optimize_widening_mul::execute): Moved most of code to
26145         math_opts_dom_walker::after_dom_children.
26146         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
26147         * config/i386/i386.c (ix86_option_override_internal): Added
26148         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
26150 2018-01-12  Richard Biener  <rguenther@suse.de>
26152         PR debug/83157
26153         * dwarf2out.c (gen_variable_die): Do not reset old_die for
26154         inline instance vars.
26156 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
26158         PR target/81819
26159         * config/rx/rx.c (rx_is_restricted_memory_address):
26160         Handle SUBREG case.
26162 2018-01-12  Richard Biener  <rguenther@suse.de>
26164         PR tree-optimization/80846
26165         * target.def (split_reduction): New target hook.
26166         * targhooks.c (default_split_reduction): New function.
26167         * targhooks.h (default_split_reduction): Declare.
26168         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
26169         target requests first reduce vectors by combining low and high
26170         parts.
26171         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
26172         (get_vectype_for_scalar_type_and_size): Export.
26173         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
26174         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
26175         * doc/tm.texi: Regenerate.
26176         * config/i386/i386.c (ix86_split_reduction): Implement
26177         TARGET_VECTORIZE_SPLIT_REDUCTION.
26179 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26181         PR target/83368
26182         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
26183         in PIC mode except for TARGET_VXWORKS_RTP.
26184         * config/sparc/sparc.c: Include cfgrtl.h.
26185         (TARGET_INIT_PIC_REG): Define.
26186         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
26187         (sparc_pic_register_p): New predicate.
26188         (sparc_legitimate_address_p): Use it.
26189         (sparc_legitimize_pic_address): Likewise.
26190         (sparc_delegitimize_address): Likewise.
26191         (sparc_mode_dependent_address_p): Likewise.
26192         (gen_load_pcrel_sym): Remove 4th parameter.
26193         (load_got_register): Adjust call to above.  Remove obsolete stuff.
26194         (sparc_expand_prologue): Do not call load_got_register here.
26195         (sparc_flat_expand_prologue): Likewise.
26196         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
26197         (sparc_use_pseudo_pic_reg): New function.
26198         (sparc_init_pic_reg): Likewise.
26199         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26200         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
26202 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
26204         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
26205         Add item for branch_cost.
26207 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26209         PR rtl-optimization/83565
26210         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
26211         not extend the result to a larger mode for rotate operations.
26212         (num_sign_bit_copies1): Likewise.
26214 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26216         PR target/40411
26217         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
26218         -symbolic.
26219         Use values-Xc.o for -pedantic.
26220         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
26222 2018-01-12  Martin Liska  <mliska@suse.cz>
26224         PR ipa/83054
26225         * ipa-devirt.c (final_warning_record::grow_type_warnings):
26226         New function.
26227         (possible_polymorphic_call_targets): Use it.
26228         (ipa_devirt): Likewise.
26230 2018-01-12  Martin Liska  <mliska@suse.cz>
26232         * profile-count.h (enum profile_quality): Use 0 as invalid
26233         enum value of profile_quality.
26235 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26237         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26238         -mext-string options.
26240 2018-01-12  Richard Biener  <rguenther@suse.de>
26242         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26243         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26244         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26245         Likewise.
26246         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26248 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
26250         * configure.ac (--with-long-double-format): Add support for the
26251         configuration option to change the default long double format on
26252         PowerPC systems.
26253         * config.gcc (powerpc*-linux*-*): Likewise.
26254         * configure: Regenerate.
26255         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26256         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26257         used without modification.
26259 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26261         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26262         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26263         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26264         MISC_BUILTIN_SPEC_BARRIER.
26265         (rs6000_init_builtins): Likewise.
26266         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26267         enum value.
26268         (speculation_barrier): New define_insn.
26269         * doc/extend.texi: Document __builtin_speculation_barrier.
26271 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26273         PR target/83203
26274         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26275         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26276         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26277         iterators.
26278         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
26279         integral modes instead of "ss" and "sd".
26280         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
26281         vectors with 32-bit and 64-bit elements.
26282         (vecdupssescalarmodesuffix): New mode attribute.
26283         (vec_dup<mode>): Use it.
26285 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
26287         PR target/83330
26288         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
26289         frame if argument is passed on stack.
26291 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26293         PR target/82682
26294         * ree.c (combine_reaching_defs): Optimize also
26295         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26296         reg2=any_extend(exp); reg1=reg2;, formatting fix.
26298 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26300         PR middle-end/83189
26301         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26303 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26305         PR middle-end/83718
26306         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26307         after they are computed.
26309 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
26311         PR tree-optimization/83695
26312         * gimple-loop-linterchange.cc
26313         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26314         reset cached scev information after interchange.
26315         (pass_linterchange::execute): Remove call to scev_reset_htab.
26317 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26319         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26320         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26321         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26322         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26323         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26324         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26325         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26326         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26327         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26328         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26329         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26330         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26331         (V_lane_reg): Likewise.
26332         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26333         New define_expand.
26334         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26335         (vfmal_lane_low<mode>_intrinsic,
26336         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26337         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
26338         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
26339         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
26340         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
26341         vfmsl_lane_high<mode>_intrinsic): New define_insns.
26343 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26345         * config/arm/arm-cpus.in (fp16fml): New feature.
26346         (ALL_SIMD): Add fp16fml.
26347         (armv8.2-a): Add fp16fml as an option.
26348         (armv8.3-a): Likewise.
26349         (armv8.4-a): Add fp16fml as part of fp16.
26350         * config/arm/arm.h (TARGET_FP16FML): Define.
26351         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
26352         when appropriate.
26353         * config/arm/arm-modes.def (V2HF): Define.
26354         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
26355         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
26356         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
26357         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
26358         vfmsl_low, vfmsl_high): New set of builtins.
26359         * config/arm/iterators.md (PLUSMINUS): New code iterator.
26360         (vfml_op): New code attribute.
26361         (VFMLHALVES): New int iterator.
26362         (VFML, VFMLSEL): New mode attributes.
26363         (V_reg): Define mapping for V2HF.
26364         (V_hi, V_lo): New mode attributes.
26365         (VF_constraint): Likewise.
26366         (vfml_half, vfml_half_selector): New int attributes.
26367         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
26368         define_expand.
26369         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
26370         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
26371         New define_insn.
26372         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
26373         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
26374         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
26375         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
26376         documentation.
26377         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
26378         Document new effective target and option set.
26380 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26382         * config/arm/arm-cpus.in (armv8_4): New feature.
26383         (ARMv8_4a): New fgroup.
26384         (armv8.4-a): New arch.
26385         * config/arm/arm-tables.opt: Regenerate.
26386         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
26387         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
26388         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
26389         Add matching rules for -march=armv8.4-a and extensions.
26390         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
26392 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
26394         PR target/81821
26395         * config/rx/rx.md (BW): New mode attribute.
26396         (sync_lock_test_and_setsi): Add mode suffix to insn output.
26398 2018-01-11  Richard Biener  <rguenther@suse.de>
26400         PR tree-optimization/83435
26401         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
26402         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
26403         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
26405 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26406             Alan Hayward  <alan.hayward@arm.com>
26407             David Sherwood  <david.sherwood@arm.com>
26409         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
26410         field.
26411         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
26412         (aarch64_print_address_internal): Use it to check for a zero offset.
26414 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26415             Alan Hayward  <alan.hayward@arm.com>
26416             David Sherwood  <david.sherwood@arm.com>
26418         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
26419         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
26420         Return a poly_int64 rather than a HOST_WIDE_INT.
26421         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
26422         rather than a HOST_WIDE_INT.
26423         * config/aarch64/aarch64.h (aarch64_frame): Protect with
26424         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
26425         hard_fp_offset, frame_size, initial_adjust, callee_offset and
26426         final_offset from HOST_WIDE_INT to poly_int64.
26427         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
26428         to_constant when getting the number of units in an Advanced SIMD
26429         mode.
26430         (aarch64_builtin_vectorized_function): Check for a constant number
26431         of units.
26432         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
26433         GET_MODE_SIZE.
26434         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
26435         attribute instead of GET_MODE_NUNITS.
26436         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
26437         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
26438         GET_MODE_SIZE for fixed-size registers.
26439         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
26440         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
26441         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
26442         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
26443         (aarch64_print_operand, aarch64_print_address_internal)
26444         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
26445         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
26446         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
26447         Handle polynomial GET_MODE_SIZE.
26448         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
26449         wider than SImode without modification.
26450         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
26451         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
26452         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
26453         passing and returning SVE modes.
26454         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
26455         rather than GEN_INT.
26456         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
26457         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
26458         (aarch64_allocate_and_probe_stack_space): Likewise.
26459         (aarch64_layout_frame): Cope with polynomial offsets.
26460         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
26461         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
26462         polynomial offsets.
26463         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
26464         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
26465         poly_int64 rather than a HOST_WIDE_INT.
26466         (aarch64_get_separate_components, aarch64_process_components)
26467         (aarch64_expand_prologue, aarch64_expand_epilogue)
26468         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
26469         (aarch64_anchor_offset): New function, split out from...
26470         (aarch64_legitimize_address): ...here.
26471         (aarch64_builtin_vectorization_cost): Handle polynomial
26472         TYPE_VECTOR_SUBPARTS.
26473         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
26474         GET_MODE_NUNITS.
26475         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
26476         number of elements from the PARALLEL rather than the mode.
26477         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
26478         rather than GET_MODE_BITSIZE.
26479         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
26480         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
26481         (aarch64_expand_vec_perm_const_1): Handle polynomial
26482         d->perm.length () and d->perm elements.
26483         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
26484         Apply to_constant to d->perm elements.
26485         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
26486         polynomial CONST_VECTOR_NUNITS.
26487         (aarch64_move_pointer): Take amount as a poly_int64 rather
26488         than an int.
26489         (aarch64_progress_pointer): Avoid temporary variable.
26490         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
26491         the mode attribute instead of GET_MODE.
26493 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26494             Alan Hayward  <alan.hayward@arm.com>
26495             David Sherwood  <david.sherwood@arm.com>
26497         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
26498         x exists before using it.
26499         (aarch64_add_constant_internal): Rename to...
26500         (aarch64_add_offset_1): ...this.  Replace regnum with separate
26501         src and dest rtxes.  Handle the case in which they're different,
26502         including when the offset is zero.  Replace scratchreg with an rtx.
26503         Use 2 additions if there is no spare register into which we can
26504         move a 16-bit constant.
26505         (aarch64_add_constant): Delete.
26506         (aarch64_add_offset): Replace reg with separate src and dest
26507         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
26508         Use aarch64_add_offset_1.
26509         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
26510         an rtx rather than an int.  Take the delta as a poly_int64
26511         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
26512         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
26513         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
26514         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
26515         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
26516         and aarch64_add_sp.
26517         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
26518         aarch64_add_constant.
26520 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26522         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
26523         Use scalar_float_mode.
26525 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26527         * config/aarch64/aarch64-simd.md
26528         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
26529         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
26530         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
26531         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
26532         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
26533         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
26534         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
26535         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
26536         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
26537         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
26539 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26541         PR target/83514
26542         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
26543         targ_options->x_arm_arch_string is non NULL.
26545 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
26547         * config/aarch64/aarch64.h
26548         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
26550 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
26552         PR target/82096
26553         * expmed.c (emit_store_flag_force): Swap if const op0
26554         and change VOIDmode to mode of op0.
26556 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26558         PR rtl-optimization/83761
26559         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
26560         than bytes to mode_for_size.
26562 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26564         PR middle-end/83189
26565         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
26566         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
26567         profile.
26569 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26571         PR middle-end/83575
26572         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
26573         when in layout mode.
26574         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
26575         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
26576         partition fixup.
26578 2018-01-10  Michael Collison  <michael.collison@arm.com>
26580         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
26581         * config/aarch64/aarch64-option-extension.def: Add
26582         AARCH64_OPT_EXTENSION of 'fp16fml'.
26583         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26584         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
26585         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
26586         * config/aarch64/constraints.md (Ui7): New constraint.
26587         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
26588         (VFMLA_SEL_W): Ditto.
26589         (f16quad): Ditto.
26590         (f16mac1): Ditto.
26591         (VFMLA16_LOW): New int iterator.
26592         (VFMLA16_HIGH): Ditto.
26593         (UNSPEC_FMLAL): New unspec.
26594         (UNSPEC_FMLSL): Ditto.
26595         (UNSPEC_FMLAL2): Ditto.
26596         (UNSPEC_FMLSL2): Ditto.
26597         (f16mac): New code attribute.
26598         * config/aarch64/aarch64-simd-builtins.def
26599         (aarch64_fmlal_lowv2sf): Ditto.
26600         (aarch64_fmlsl_lowv2sf): Ditto.
26601         (aarch64_fmlalq_lowv4sf): Ditto.
26602         (aarch64_fmlslq_lowv4sf): Ditto.
26603         (aarch64_fmlal_highv2sf): Ditto.
26604         (aarch64_fmlsl_highv2sf): Ditto.
26605         (aarch64_fmlalq_highv4sf): Ditto.
26606         (aarch64_fmlslq_highv4sf): Ditto.
26607         (aarch64_fmlal_lane_lowv2sf): Ditto.
26608         (aarch64_fmlsl_lane_lowv2sf): Ditto.
26609         (aarch64_fmlal_laneq_lowv2sf): Ditto.
26610         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
26611         (aarch64_fmlalq_lane_lowv4sf): Ditto.
26612         (aarch64_fmlsl_lane_lowv4sf): Ditto.
26613         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
26614         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
26615         (aarch64_fmlal_lane_highv2sf): Ditto.
26616         (aarch64_fmlsl_lane_highv2sf): Ditto.
26617         (aarch64_fmlal_laneq_highv2sf): Ditto.
26618         (aarch64_fmlsl_laneq_highv2sf): Ditto.
26619         (aarch64_fmlalq_lane_highv4sf): Ditto.
26620         (aarch64_fmlsl_lane_highv4sf): Ditto.
26621         (aarch64_fmlalq_laneq_highv4sf): Ditto.
26622         (aarch64_fmlsl_laneq_highv4sf): Ditto.
26623         * config/aarch64/aarch64-simd.md:
26624         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
26625         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26626         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
26627         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26628         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
26629         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
26630         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
26631         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
26632         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
26633         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
26634         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
26635         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
26636         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
26637         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
26638         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
26639         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
26640         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
26641         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
26642         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
26643         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
26644         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
26645         (vfmlsl_low_u32): Ditto.
26646         (vfmlalq_low_u32): Ditto.
26647         (vfmlslq_low_u32): Ditto.
26648         (vfmlal_high_u32): Ditto.
26649         (vfmlsl_high_u32): Ditto.
26650         (vfmlalq_high_u32): Ditto.
26651         (vfmlslq_high_u32): Ditto.
26652         (vfmlal_lane_low_u32): Ditto.
26653         (vfmlsl_lane_low_u32): Ditto.
26654         (vfmlal_laneq_low_u32): Ditto.
26655         (vfmlsl_laneq_low_u32): Ditto.
26656         (vfmlalq_lane_low_u32): Ditto.
26657         (vfmlslq_lane_low_u32): Ditto.
26658         (vfmlalq_laneq_low_u32): Ditto.
26659         (vfmlslq_laneq_low_u32): Ditto.
26660         (vfmlal_lane_high_u32): Ditto.
26661         (vfmlsl_lane_high_u32): Ditto.
26662         (vfmlal_laneq_high_u32): Ditto.
26663         (vfmlsl_laneq_high_u32): Ditto.
26664         (vfmlalq_lane_high_u32): Ditto.
26665         (vfmlslq_lane_high_u32): Ditto.
26666         (vfmlalq_laneq_high_u32): Ditto.
26667         (vfmlslq_laneq_high_u32): Ditto.
26668         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
26669         (AARCH64_FL_FOR_ARCH8_4): New.
26670         (AARCH64_ISA_F16FML): New ISA flag.
26671         (TARGET_F16FML): New feature flag for fp16fml.
26672         (doc/invoke.texi): Document new fp16fml option.
26674 2018-01-10  Michael Collison  <michael.collison@arm.com>
26676         * config/aarch64/aarch64-builtins.c:
26677         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
26678         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26679         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
26680         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
26681         (AARCH64_ISA_SHA3): New ISA flag.
26682         (TARGET_SHA3): New feature flag for sha3.
26683         * config/aarch64/iterators.md (sha512_op): New int attribute.
26684         (CRYPTO_SHA512): New int iterator.
26685         (UNSPEC_SHA512H): New unspec.
26686         (UNSPEC_SHA512H2): Ditto.
26687         (UNSPEC_SHA512SU0): Ditto.
26688         (UNSPEC_SHA512SU1): Ditto.
26689         * config/aarch64/aarch64-simd-builtins.def
26690         (aarch64_crypto_sha512hqv2di): New builtin.
26691         (aarch64_crypto_sha512h2qv2di): Ditto.
26692         (aarch64_crypto_sha512su0qv2di): Ditto.
26693         (aarch64_crypto_sha512su1qv2di): Ditto.
26694         (aarch64_eor3qv8hi): Ditto.
26695         (aarch64_rax1qv2di): Ditto.
26696         (aarch64_xarqv2di): Ditto.
26697         (aarch64_bcaxqv8hi): Ditto.
26698         * config/aarch64/aarch64-simd.md:
26699         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
26700         (aarch64_crypto_sha512su0qv2di): Ditto.
26701         (aarch64_crypto_sha512su1qv2di): Ditto.
26702         (aarch64_eor3qv8hi): Ditto.
26703         (aarch64_rax1qv2di): Ditto.
26704         (aarch64_xarqv2di): Ditto.
26705         (aarch64_bcaxqv8hi): Ditto.
26706         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
26707         (vsha512h2q_u64): Ditto.
26708         (vsha512su0q_u64): Ditto.
26709         (vsha512su1q_u64): Ditto.
26710         (veor3q_u16): Ditto.
26711         (vrax1q_u64): Ditto.
26712         (vxarq_u64): Ditto.
26713         (vbcaxq_u16): Ditto.
26714         * config/arm/types.md (crypto_sha512): New type attribute.
26715         (crypto_sha3): Ditto.
26716         (doc/invoke.texi): Document new sha3 option.
26718 2018-01-10  Michael Collison  <michael.collison@arm.com>
26720         * config/aarch64/aarch64-builtins.c:
26721         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
26722         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26723         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
26724         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
26725         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
26726         (AARCH64_ISA_SM4): New ISA flag.
26727         (TARGET_SM4): New feature flag for sm4.
26728         * config/aarch64/aarch64-simd-builtins.def
26729         (aarch64_sm3ss1qv4si): Ditto.
26730         (aarch64_sm3tt1aq4si): Ditto.
26731         (aarch64_sm3tt1bq4si): Ditto.
26732         (aarch64_sm3tt2aq4si): Ditto.
26733         (aarch64_sm3tt2bq4si): Ditto.
26734         (aarch64_sm3partw1qv4si): Ditto.
26735         (aarch64_sm3partw2qv4si): Ditto.
26736         (aarch64_sm4eqv4si): Ditto.
26737         (aarch64_sm4ekeyqv4si): Ditto.
26738         * config/aarch64/aarch64-simd.md:
26739         (aarch64_sm3ss1qv4si): Ditto.
26740         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
26741         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
26742         (aarch64_sm4eqv4si): Ditto.
26743         (aarch64_sm4ekeyqv4si): Ditto.
26744         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
26745         (sm3part_op): Ditto.
26746         (CRYPTO_SM3TT): Ditto.
26747         (CRYPTO_SM3PART): Ditto.
26748         (UNSPEC_SM3SS1): New unspec.
26749         (UNSPEC_SM3TT1A): Ditto.
26750         (UNSPEC_SM3TT1B): Ditto.
26751         (UNSPEC_SM3TT2A): Ditto.
26752         (UNSPEC_SM3TT2B): Ditto.
26753         (UNSPEC_SM3PARTW1): Ditto.
26754         (UNSPEC_SM3PARTW2): Ditto.
26755         (UNSPEC_SM4E): Ditto.
26756         (UNSPEC_SM4EKEY): Ditto.
26757         * config/aarch64/constraints.md (Ui2): New constraint.
26758         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
26759         * config/arm/types.md (crypto_sm3): New type attribute.
26760         (crypto_sm4): Ditto.
26761         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
26762         (vsm3tt1aq_u32): Ditto.
26763         (vsm3tt1bq_u32): Ditto.
26764         (vsm3tt2aq_u32): Ditto.
26765         (vsm3tt2bq_u32): Ditto.
26766         (vsm3partw1q_u32): Ditto.
26767         (vsm3partw2q_u32): Ditto.
26768         (vsm4eq_u32): Ditto.
26769         (vsm4ekeyq_u32): Ditto.
26770         (doc/invoke.texi): Document new sm4 option.
26772 2018-01-10  Michael Collison  <michael.collison@arm.com>
26774         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
26775         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
26776         (AARCH64_FL_FOR_ARCH8_4): New.
26777         (AARCH64_FL_V8_4): New flag.
26778         (doc/invoke.texi): Document new armv8.4-a option.
26780 2018-01-10  Michael Collison  <michael.collison@arm.com>
26782         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26783         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
26784         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
26785         * config/aarch64/aarch64-option-extension.def: Add
26786         AARCH64_OPT_EXTENSION of 'sha2'.
26787         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
26788         (crypto): Disable sha2 and aes if crypto disabled.
26789         (crypto): Enable aes and sha2 if enabled.
26790         (simd): Disable sha2 and aes if simd disabled.
26791         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
26792         New flags.
26793         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
26794         (TARGET_SHA2): New feature flag for sha2.
26795         (TARGET_AES): New feature flag for aes.
26796         * config/aarch64/aarch64-simd.md:
26797         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
26798         conditional on TARGET_AES.
26799         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
26800         (aarch64_crypto_sha1hsi): Make pattern conditional
26801         on TARGET_SHA2.
26802         (aarch64_crypto_sha1hv4si): Ditto.
26803         (aarch64_be_crypto_sha1hv4si): Ditto.
26804         (aarch64_crypto_sha1su1v4si): Ditto.
26805         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
26806         (aarch64_crypto_sha1su0v4si): Ditto.
26807         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
26808         (aarch64_crypto_sha256su0v4si): Ditto.
26809         (aarch64_crypto_sha256su1v4si): Ditto.
26810         (doc/invoke.texi): Document new aes and sha2 options.
26812 2018-01-10  Martin Sebor  <msebor@redhat.com>
26814         PR tree-optimization/83781
26815         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
26816         as string arrays.
26818 2018-01-11  Martin Sebor  <msebor@gmail.com>
26819             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26821         PR tree-optimization/83501
26822         PR tree-optimization/81703
26824         * tree-ssa-strlen.c (get_string_cst): Rename...
26825         (get_string_len): ...to this.  Handle global constants.
26826         (handle_char_store): Adjust.
26828 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
26829             Jim Wilson  <jimw@sifive.com>
26831         * config/riscv/riscv-protos.h (riscv_output_return): New.
26832         * config/riscv/riscv.c (struct machine_function): New naked_p field.
26833         (riscv_attribute_table, riscv_output_return),
26834         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
26835         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
26836         (riscv_compute_frame_info): Only compute frame->mask if not a naked
26837         function.
26838         (riscv_expand_prologue): Add early return for naked function.
26839         (riscv_expand_epilogue): Likewise.
26840         (riscv_function_ok_for_sibcall): Return false for naked function.
26841         (riscv_set_current_function): New.
26842         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
26843         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
26844         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
26845         * doc/extend.texi (RISC-V Function Attributes): New.
26847 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
26849         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
26850         check for 128-bit long double before checking TCmode.
26851         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
26852         128-bit long doubles before checking TFmode or TCmode.
26853         (FLOAT128_IBM_P): Likewise.
26855 2018-01-10  Martin Sebor  <msebor@redhat.com>
26857         PR tree-optimization/83671
26858         * builtins.c (c_strlen): Unconditionally return zero for the empty
26859         string.
26860         Use -Warray-bounds for warnings.
26861         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
26862         for non-constant array indices with COMPONENT_REF, arrays of
26863         arrays, and pointers to arrays.
26864         (gimple_fold_builtin_strlen): Determine and set length range for
26865         non-constant character arrays.
26867 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
26869         PR middle-end/81897
26870         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
26871         empty blocks.
26873 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
26875         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
26877 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
26879         PR target/83399
26880         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
26881         VECTOR_MEM_ALTIVEC_OR_VSX_P.
26882         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
26883         indexed_or_indirect_operand predicate.
26884         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
26885         (*vsx_le_perm_load_v8hi): Likewise.
26886         (*vsx_le_perm_load_v16qi): Likewise.
26887         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
26888         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
26889         (*vsx_le_perm_store_v8hi): Likewise.
26890         (*vsx_le_perm_store_v16qi): Likewise.
26891         (eight unnamed splitters): Likewise.
26893 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
26895         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
26896         * config/rs6000/emmintrin.h: Likewise.
26897         * config/rs6000/mmintrin.h: Likewise.
26898         * config/rs6000/xmmintrin.h: Likewise.
26900 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
26902         PR c++/43486
26903         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
26904         "public_flag".
26905         * tree.c (tree_nop_conversion): Return true for location wrapper
26906         nodes.
26907         (maybe_wrap_with_location): New function.
26908         (selftest::check_strip_nops): New function.
26909         (selftest::test_location_wrappers): New function.
26910         (selftest::tree_c_tests): Call it.
26911         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
26912         (maybe_wrap_with_location): New decl.
26913         (EXPR_LOCATION_WRAPPER_P): New macro.
26914         (location_wrapper_p): New inline function.
26915         (tree_strip_any_location_wrapper): New inline function.
26917 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
26919         PR target/83735
26920         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
26921         stack_realign_offset for the largest alignment of stack slot
26922         actually used.
26923         (ix86_find_max_used_stack_alignment): New function.
26924         (ix86_finalize_stack_frame_flags): Use it.  Set
26925         max_used_stack_alignment if we don't realign stack.
26926         * config/i386/i386.h (machine_function): Add
26927         max_used_stack_alignment.
26929 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
26931         * config/arm/arm.opt (-mbranch-cost): New option.
26932         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
26933         account.
26935 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
26937         PR target/83629
26938         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
26939         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
26941 2018-01-10  Richard Biener  <rguenther@suse.de>
26943         PR debug/83765
26944         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
26945         early out so it also covers the case where we have a non-NULL
26946         origin.
26948 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
26950         PR tree-optimization/83753
26951         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
26952         for non-strided grouped accesses if the number of elements is 1.
26954 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26956         PR target/81616
26957         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
26958         * i386.h (TARGET_USE_GATHER): Define.
26959         * x86-tune.def (X86_TUNE_USE_GATHER): New.
26961 2018-01-10  Martin Liska  <mliska@suse.cz>
26963         PR bootstrap/82831
26964         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
26965         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
26966         partitioning.
26967         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
26968         CLEANUP_NO_PARTITIONING is not set.
26970 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
26972         * doc/rtl.texi: Remove documentation of (const ...) wrappers
26973         for vectors, as a partial revert of r254296.
26974         * rtl.h (const_vec_p): Delete.
26975         (const_vec_duplicate_p): Don't test for vector CONSTs.
26976         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
26977         * expmed.c (make_tree): Likewise.
26979         Revert:
26980         * common.md (E, F): Use CONSTANT_P instead of checking for
26981         CONST_VECTOR.
26982         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
26983         checking for CONST_VECTOR.
26985 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26987         PR middle-end/83575
26988         * predict.c (force_edge_cold): Handle in more sane way edges
26989         with no prediction.
26991 2018-01-09  Carl Love  <cel@us.ibm.com>
26993         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
26994         V4SI, V4SF types.
26995         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
26996         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
26997         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
26998         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
26999         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
27000         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
27001         * config/rs6000/rs6000-protos.h: Add extern defition for
27002         rs6000_generate_float2_double_code.
27003         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
27004         function.
27005         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
27006         (float2_v2df): Add define_expand.
27008 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
27010         PR target/83628
27011         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
27012         op_mode in the force_to_mode call.
27014 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27016         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
27017         instead of checking each element individually.
27018         (aarch64_evpc_uzp): Likewise.
27019         (aarch64_evpc_zip): Likewise.
27020         (aarch64_evpc_ext): Likewise.
27021         (aarch64_evpc_rev): Likewise.
27022         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
27023         instead of checking each element individually.  Return true without
27024         generating rtl if
27025         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
27026         whether all selected elements come from the same input, instead of
27027         checking each element individually.  Remove calls to gen_rtx_REG,
27028         start_sequence and end_sequence and instead assert that no rtl is
27029         generated.
27031 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27033         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
27034         order of HIGH and CONST checks.
27036 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27038         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
27039         if the destination isn't an SSA_NAME.
27041 2018-01-09  Richard Biener  <rguenther@suse.de>
27043         PR tree-optimization/83668
27044         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
27045         move prologue...
27046         (canonicalize_loop_form): ... here, renamed from ...
27047         (canonicalize_loop_closed_ssa_form): ... this and amended to
27048         swap successor edges for loop exit blocks to make us use
27049         the RPO order we need for initial schedule generation.
27051 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
27053         PR tree-optimization/64811
27054         * match.pd: When optimizing comparisons with Inf, avoid
27055         introducing or losing exceptions from comparisons with NaN.
27057 2018-01-09  Martin Liska  <mliska@suse.cz>
27059         PR sanitizer/82517
27060         * asan.c (shadow_mem_size): Add gcc_assert.
27062 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
27064         Don't save registers in main().
27066         PR target/83738
27067         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
27068         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
27069         * config/avr/avr.c (avr_set_current_function): Don't error if
27070         naked, OS_task or OS_main are specified at the same time.
27071         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
27072         OS_main.
27073         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
27074         attribute.
27075         * common/config/avr/avr-common.c (avr_option_optimization_table):
27076         Switch on -mmain-is-OS_task for optimizing compilations.
27078 2018-01-09  Richard Biener  <rguenther@suse.de>
27080         PR tree-optimization/83572
27081         * graphite.c: Include cfganal.h.
27082         (graphite_transform_loops): Connect infinite loops to exit
27083         and remove fake edges at the end.
27085 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27087         * ipa-inline.c (edge_badness): Revert accidental checkin.
27089 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27091         PR ipa/80763
27092         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
27093         symbols; not inline clones.
27095 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
27097         PR target/83507
27098         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
27099         hard registers.  Formatting fixes.
27101         PR preprocessor/83722
27102         * gcc.c (try_generate_repro): Pass
27103         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
27104         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
27105         do_report_bug.
27107 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
27108             Kito Cheng  <kito.cheng@gmail.com>
27110         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
27111         (riscv_leaf_function_p): Delete.
27112         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
27114 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27116         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
27117         function.
27118         (do_ifelse): New function.
27119         (do_isel): New function.
27120         (do_sub3): New function.
27121         (do_add3): New function.
27122         (do_load_mask_compare): New function.
27123         (do_overlap_load_compare): New function.
27124         (expand_compare_loop): New function.
27125         (expand_block_compare): Call expand_compare_loop() when appropriate.
27126         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
27127         option description.
27128         (-mblock-compare-inline-loop-limit): New option.
27130 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27132         PR target/83677
27133         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
27134         Reverse order of second and third operands in first alternative.
27135         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
27136         of first and second elements in UNSPEC_VPERMR vector.
27137         (altivec_expand_vec_perm_le): Likewise.
27139 2018-01-08  Jeff Law  <law@redhat.com>
27141         PR rtl-optimizatin/81308
27142         * tree-switch-conversion.c (cfg_altered): New file scoped static.
27143         (process_switch): If group_case_labels makes a change, then set
27144         cfg_altered.
27145         (pass_convert_switch::execute): If a switch is converted, then
27146         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
27148         PR rtl-optimization/81308
27149         * recog.c (split_all_insns): Conditionally cleanup the CFG after
27150         splitting insns.
27152 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
27154         PR target/83663 - Revert r255946
27155         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
27156         generation for cases where splatting a value is not useful.
27157         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
27158         across a vec_duplicate and a paradoxical subreg forming a vector
27159         mode to a vec_concat.
27161 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27163         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
27164         -march=armv8.3-a variants.
27165         * config/arm/t-multilib: Likewise.
27166         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
27168 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27170         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
27171         to generate rtl.
27172         (cceq_ior_compare_complement): Give it a name so I can use it, and
27173         change boolean_or_operator predicate to boolean_operator so it can
27174         be used to generate a crand.
27175         (eqne): New code iterator.
27176         (bd/bd_neg): New code_attrs.
27177         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
27178         a single define_insn.
27179         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
27180         decrement (bdnzt/bdnzf/bdzt/bdzf).
27181         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
27182         with the new names of the branch decrement patterns, and added the
27183         names of the branch decrement conditional patterns.
27185 2018-01-08  Richard Biener  <rguenther@suse.de>
27187         PR tree-optimization/83563
27188         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
27189         cache.
27191 2018-01-08  Richard Biener  <rguenther@suse.de>
27193         PR middle-end/83713
27194         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
27196 2018-01-08  Richard Biener  <rguenther@suse.de>
27198         PR tree-optimization/83685
27199         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
27200         references to abnormals.
27202 2018-01-08  Richard Biener  <rguenther@suse.de>
27204         PR lto/83719
27205         * dwarf2out.c (output_indirect_strings): Handle empty
27206         skeleton_debug_str_hash.
27207         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
27209 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27211         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
27212         (emit_store_direct): Likewise.
27213         (arc_trampoline_adjust_address): Likewise.
27214         (arc_asm_trampoline_template): New function.
27215         (arc_initialize_trampoline): Use asm_trampoline_template.
27216         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
27217         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
27218         * config/arc/arc.md (flush_icache): Delete pattern.
27220 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27222         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
27223         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
27224         munaligned-access.
27226 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27228         PR target/83681
27229         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
27230         by not USED_FOR_TARGET.
27231         (make_pass_resolve_sw_modes): Likewise.
27233 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27235         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27236         USED_FOR_TARGET.
27238 2018-01-08  Richard Biener  <rguenther@suse.de>
27240         PR middle-end/83580
27241         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27243 2018-01-08  Richard Biener  <rguenther@suse.de>
27245         PR middle-end/83517
27246         * match.pd ((t * 2) / 2) -> t): Add missing :c.
27248 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
27250         PR middle-end/81897
27251         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27252         basic blocks with a small number of successors.
27253         (convert_control_dep_chain_into_preds): Improve handling of
27254         forwarder blocks.
27255         (dump_predicates): Split apart into...
27256         (dump_pred_chain): ...here...
27257         (dump_pred_info): ...and here.
27258         (can_one_predicate_be_invalidated_p): Add debugging printfs.
27259         (can_chain_union_be_invalidated_p): Improve check for invalidation
27260         of paths.
27261         (uninit_uses_cannot_happen): Avoid unnecessary if
27262         convert_control_dep_chain_into_preds yielded nothing.
27264 2018-01-06  Martin Sebor  <msebor@redhat.com>
27266         PR tree-optimization/83640
27267         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27268         subtracting negative offset from size.
27269         (builtin_access::overlap): Adjust offset bounds of the access to fall
27270         within the size of the object if possible.
27272 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
27274         PR rtl-optimization/83699
27275         * expmed.c (extract_bit_field_1): Restrict the vector usage of
27276         extract_bit_field_as_subreg to cases in which the extracted
27277         value is also a vector.
27279         * lra-constraints.c (process_alt_operands): Test for the equivalence
27280         substitutions when detecting a possible reload cycle.
27282 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
27284         PR debug/83480
27285         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
27286         by default if flag_selective_schedling{,2}.  Formatting fixes.
27288         PR rtl-optimization/83682
27289         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
27290         if it has non-VECTOR_MODE element mode.
27291         (vec_duplicate_p): Likewise.
27293         PR middle-end/83694
27294         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27295         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27297 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27299         PR target/83604
27300         * config/i386/i386-builtin.def
27301         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27302         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27303         Require also OPTION_MASK_ISA_AVX512F in addition to
27304         OPTION_MASK_ISA_GFNI.
27305         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27306         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27307         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27308         to OPTION_MASK_ISA_GFNI.
27309         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27310         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27311         OPTION_MASK_ISA_AVX512BW.
27312         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27313         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27314         addition to OPTION_MASK_ISA_GFNI.
27315         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27316         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27317         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27318         to OPTION_MASK_ISA_GFNI.
27319         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27320         a requirement for all ISAs rather than any of them with a few
27321         exceptions.
27322         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27323         processing.
27324         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27325         bitmasks to be enabled with 3 exceptions, instead of requiring any
27326         enabled ISA with lots of exceptions.
27327         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27328         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27329         Change avx512bw in isa attribute to avx512f.
27330         * config/i386/sgxintrin.h: Add license boilerplate.
27331         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
27332         to __AVX512F__ and __AVX512VL to __AVX512VL__.
27333         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27334         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27335         defined.
27336         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27337         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
27338         temporarily sse2 rather than sse if not enabled already.
27340         PR target/83604
27341         * config/i386/sse.md (VI248_VLBW): Rename to ...
27342         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
27343         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
27344         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
27345         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
27346         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
27347         mode iterator instead of VI248_VLBW.
27349 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
27351         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
27352         (record_modified): Skip clobbers; add debug output.
27353         (param_change_prob): Use sreal frequencies.
27355 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27357         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
27358         punt for user-aligned variables.
27360 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27362         * tree-chrec.c (chrec_contains_symbols): Return true for
27363         POLY_INT_CST.
27365 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
27367         PR target/82439
27368         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
27369         of (x|y) == x for BICS pattern.
27371 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27373         PR tree-optimization/83605
27374         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
27375         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
27376         can throw.
27378 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27380         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
27381         * config/epiphany/rtems.h: New file.
27383 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27384             Uros Bizjak  <ubizjak@gmail.com>
27386         PR target/83554
27387         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
27388         QIreg_operand instead of register_operand predicate.
27389         * config/i386/i386.c (ix86_rop_should_change_byte_p,
27390         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
27391         comments instead of -fmitigate[-_]rop.
27393 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27395         PR bootstrap/81926
27396         * cgraphunit.c (symbol_table::compile): Switch to text_section
27397         before calling assembly_start debug hook.
27398         * run-rtl-passes.c (run_rtl_passes): Likewise.
27399         Include output.h.
27401 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27403         * tree-vrp.c (extract_range_from_binary_expr_1): Check
27404         range_int_cst_p rather than !symbolic_range_p before calling
27405         extract_range_from_multiplicative_op_1.
27407 2018-01-04  Jeff Law  <law@redhat.com>
27409         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
27410         redundant test in assertion.
27412 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27414         * doc/rtl.texi: Document machine_mode wrapper classes.
27416 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27418         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
27419         using tree_to_uhwi.
27421 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27423         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
27424         the VEC_PERM_EXPR fold to fail.
27426 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27428         PR debug/83585
27429         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
27430         to switched_sections.
27432 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27434         PR target/83680
27435         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
27436         test for d.testing.
27438 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
27440         PR target/83387
27441         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
27442         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
27444 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27446         PR debug/83666
27447         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
27448         is BLKmode and bitpos not zero or mode change is needed.
27450 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27452         PR target/83675
27453         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
27454         TARGET_VIS2.
27456 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
27458         PR target/83628
27459         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
27460         instead of MULT rtx.  Update all corresponding splitters.
27461         (*saddl_se): Ditto.
27462         (*ssub<modesuffix>): Ditto.
27463         (*ssubl_se): Ditto.
27464         (*cmp_sadd_di): Update split patterns.
27465         (*cmp_sadd_si): Ditto.
27466         (*cmp_sadd_sidi): Ditto.
27467         (*cmp_ssub_di): Ditto.
27468         (*cmp_ssub_si): Ditto.
27469         (*cmp_ssub_sidi): Ditto.
27470         * config/alpha/predicates.md (const23_operand): New predicate.
27471         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
27472         Look for ASHIFT, not MULT inner operand.
27473         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
27475 2018-01-04  Martin Liska  <mliska@suse.cz>
27477         PR gcov-profile/83669
27478         * gcov.c (output_intermediate_file): Add version to intermediate
27479         gcov file.
27480         * doc/gcov.texi: Document new field 'version' in intermediate
27481         file format. Fix location of '-k' option of gcov command.
27483 2018-01-04  Martin Liska  <mliska@suse.cz>
27485         PR ipa/82352
27486         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
27488 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27490         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
27492 2018-01-03  Martin Sebor  <msebor@redhat.com>
27494         PR tree-optimization/83655
27495         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
27496         checking calls with invalid arguments.
27498 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27500         * tree-vect-stmts.c (vect_get_store_rhs): New function.
27501         (vectorizable_mask_load_store): Delete.
27502         (vectorizable_call): Return false for masked loads and stores.
27503         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
27504         instead of gimple_assign_rhs1.
27505         (vectorizable_load): Handle IFN_MASK_LOAD.
27506         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
27508 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27510         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
27511         split out from..,
27512         (vectorizable_mask_load_store): ...here.
27513         (vectorizable_load): ...and here.
27515 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27517         * tree-vect-stmts.c (vect_build_all_ones_mask)
27518         (vect_build_zero_merge_argument): New functions, split out from...
27519         (vectorizable_load): ...here.
27521 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27523         * tree-vect-stmts.c (vect_check_store_rhs): New function,
27524         split out from...
27525         (vectorizable_mask_load_store): ...here.
27526         (vectorizable_store): ...and here.
27528 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27530         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
27531         split out from...
27532         (vectorizable_mask_load_store): ...here.
27534 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27536         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
27537         (vect_model_store_cost): Take a vec_load_store_type instead of a
27538         vect_def_type.
27539         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
27540         (vect_model_store_cost): Take a vec_load_store_type instead of a
27541         vect_def_type.
27542         (vectorizable_mask_load_store): Update accordingly.
27543         (vectorizable_store): Likewise.
27544         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
27546 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27548         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
27549         IFN_MASK_LOAD calls here rather than...
27550         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
27552 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27553             Alan Hayward  <alan.hayward@arm.com>
27554             David Sherwood  <david.sherwood@arm.com>
27556         * expmed.c (extract_bit_field_1): For vector extracts,
27557         fall back to extract_bit_field_as_subreg if vec_extract
27558         isn't available.
27560 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27561             Alan Hayward  <alan.hayward@arm.com>
27562             David Sherwood  <david.sherwood@arm.com>
27564         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
27565         they are variable or constant sized.
27566         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
27567         slots for constant-sized data.
27569 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27570             Alan Hayward  <alan.hayward@arm.com>
27571             David Sherwood  <david.sherwood@arm.com>
27573         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
27574         handling COND_EXPRs with boolean comparisons, try to find a better
27575         basis for the mask type than the boolean itself.
27577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27579         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
27580         is calculated and how it can be overridden.
27581         * genmodes.c (max_bitsize_mode_any_mode): New variable.
27582         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
27583         if defined.
27584         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
27585         if nonzero.
27587 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27588             Alan Hayward  <alan.hayward@arm.com>
27589             David Sherwood  <david.sherwood@arm.com>
27591         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
27592         Remove the mode argument.
27593         (aarch64_simd_valid_immediate): Remove the mode and inverse
27594         arguments.
27595         * config/aarch64/iterators.md (bitsize): New iterator.
27596         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
27597         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
27598         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
27599         aarch64_simd_valid_immediate.
27600         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
27601         (aarch64_reg_or_bic_imm): Likewise.
27602         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
27603         with an insn_type enum and msl with a modifier_type enum.
27604         Replace element_width with a scalar_mode.  Change the shift
27605         to unsigned int.  Add constructors for scalar_float_mode and
27606         scalar_int_mode elements.
27607         (aarch64_vect_float_const_representable_p): Delete.
27608         (aarch64_can_const_movi_rtx_p)
27609         (aarch64_simd_scalar_immediate_valid_for_move)
27610         (aarch64_simd_make_constant): Update call to
27611         aarch64_simd_valid_immediate.
27612         (aarch64_advsimd_valid_immediate_hs): New function.
27613         (aarch64_advsimd_valid_immediate): Likewise.
27614         (aarch64_simd_valid_immediate): Remove mode and inverse
27615         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
27616         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
27617         and aarch64_float_const_representable_p on the result.
27618         (aarch64_output_simd_mov_immediate): Remove mode argument.
27619         Update call to aarch64_simd_valid_immediate and use of
27620         simd_immediate_info.
27621         (aarch64_output_scalar_simd_mov_immediate): Update call
27622         accordingly.
27624 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27625             Alan Hayward  <alan.hayward@arm.com>
27626             David Sherwood  <david.sherwood@arm.com>
27628         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
27629         (mode_nunits): Likewise CONST_MODE_NUNITS.
27630         * machmode.def (ADJUST_NUNITS): Document.
27631         * genmodes.c (mode_data::need_nunits_adj): New field.
27632         (blank_mode): Update accordingly.
27633         (adj_nunits): New variable.
27634         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
27635         parameter.
27636         (emit_mode_size_inline): Set need_bytesize_adj for all modes
27637         listed in adj_nunits.
27638         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
27639         listed in adj_nunits.  Don't emit case statements for such modes.
27640         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
27641         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
27642         nothing if adj_nunits is nonnull.
27643         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
27644         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
27645         (emit_mode_fbit): Update use of print_maybe_const_decl.
27646         (emit_move_size): Likewise.  Treat the array as non-const
27647         if adj_nunits.
27648         (emit_mode_adjustments): Handle adj_nunits.
27650 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27652         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
27653         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
27654         (VECTOR_MODES): Use it.
27655         (make_vector_modes): Take the prefix as an argument.
27657 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27658             Alan Hayward  <alan.hayward@arm.com>
27659             David Sherwood  <david.sherwood@arm.com>
27661         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
27662         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
27663         for MODE_VECTOR_BOOL.
27664         * machmode.def (VECTOR_BOOL_MODE): Document.
27665         * genmodes.c (VECTOR_BOOL_MODE): New macro.
27666         (make_vector_bool_mode): New function.
27667         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
27668         MODE_VECTOR_BOOL.
27669         * lto-streamer-in.c (lto_input_mode_table): Likewise.
27670         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
27671         Likewise.
27672         * stor-layout.c (int_mode_for_mode): Likewise.
27673         * tree.c (build_vector_type_for_mode): Likewise.
27674         * varasm.c (output_constant_pool_2): Likewise.
27675         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
27676         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
27677         for MODE_VECTOR_BOOL.
27678         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
27679         of mode class checks.
27680         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
27681         instead of a list of mode class checks.
27682         (expand_vector_scalar_condition): Likewise.
27683         (type_for_widest_vector_mode): Handle BImode as an inner mode.
27685 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27686             Alan Hayward  <alan.hayward@arm.com>
27687             David Sherwood  <david.sherwood@arm.com>
27689         * machmode.h (mode_size): Change from unsigned short to
27690         poly_uint16_pod.
27691         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
27692         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27693         or if measurement_type is not polynomial.
27694         (fixed_size_mode::includes_p): Check for constant-sized modes.
27695         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
27696         return a poly_uint16 rather than an unsigned short.
27697         (emit_mode_size): Change the type of mode_size from unsigned short
27698         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
27699         (emit_mode_adjustments): Cope with polynomial vector sizes.
27700         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27701         for GET_MODE_SIZE.
27702         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27703         for GET_MODE_SIZE.
27704         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
27705         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
27706         * caller-save.c (setup_save_areas): Likewise.
27707         (replace_reg_with_saved_mem): Likewise.
27708         * calls.c (emit_library_call_value_1): Likewise.
27709         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
27710         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
27711         (gen_lowpart_for_combine): Likewise.
27712         * convert.c (convert_to_integer_1): Likewise.
27713         * cse.c (equiv_constant, cse_insn): Likewise.
27714         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
27715         (cselib_subst_to_values): Likewise.
27716         * dce.c (word_dce_process_block): Likewise.
27717         * df-problems.c (df_word_lr_mark_ref): Likewise.
27718         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
27719         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
27720         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
27721         (rtl_for_decl_location): Likewise.
27722         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
27723         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
27724         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
27725         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
27726         (expand_expr_real_1): Likewise.
27727         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
27728         (pad_below): Likewise.
27729         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27730         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
27731         * ira.c (get_subreg_tracking_sizes): Likewise.
27732         * ira-build.c (ira_create_allocno_objects): Likewise.
27733         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
27734         (ira_sort_regnos_for_alter_reg): Likewise.
27735         * ira-costs.c (record_operand_costs): Likewise.
27736         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
27737         (resolve_simple_move): Likewise.
27738         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
27739         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
27740         (lra_constraints): Likewise.
27741         (CONST_POOL_OK_P): Reject variable-sized modes.
27742         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
27743         (add_pseudo_to_slot, lra_spill): Likewise.
27744         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27745         * optabs-query.c (get_best_extraction_insn): Likewise.
27746         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27747         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
27748         (expand_mult_highpart, valid_multiword_target_p): Likewise.
27749         * recog.c (offsettable_address_addr_space_p): Likewise.
27750         * regcprop.c (maybe_mode_change): Likewise.
27751         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
27752         * regrename.c (build_def_use): Likewise.
27753         * regstat.c (dump_reg_info): Likewise.
27754         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
27755         (find_reloads, find_reloads_subreg_address): Likewise.
27756         * reload1.c (eliminate_regs_1): Likewise.
27757         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
27758         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
27759         (simplify_binary_operation_1, simplify_subreg): Likewise.
27760         * targhooks.c (default_function_arg_padding): Likewise.
27761         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
27762         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
27763         (verify_gimple_assign_ternary): Likewise.
27764         * tree-inline.c (estimate_move_cost): Likewise.
27765         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27766         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
27767         (get_address_cost_ainc): Likewise.
27768         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
27769         (vect_supportable_dr_alignment): Likewise.
27770         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27771         (vectorizable_reduction): Likewise.
27772         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
27773         (vectorizable_operation, vectorizable_load): Likewise.
27774         * tree.c (build_same_sized_truth_vector_type): Likewise.
27775         * valtrack.c (cleanup_auto_inc_dec): Likewise.
27776         * var-tracking.c (emit_note_insn_var_location): Likewise.
27777         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
27778         (ADDR_VEC_ALIGN): Likewise.
27780 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27781             Alan Hayward  <alan.hayward@arm.com>
27782             David Sherwood  <david.sherwood@arm.com>
27784         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
27785         unsigned short.
27786         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27787         or if measurement_type is polynomial.
27788         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
27789         * combine.c (make_extraction): Likewise.
27790         * dse.c (find_shift_sequence): Likewise.
27791         * dwarf2out.c (mem_loc_descriptor): Likewise.
27792         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
27793         (extract_bit_field, extract_low_bits): Likewise.
27794         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
27795         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
27796         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
27797         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
27798         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27799         * reload.c (find_reloads): Likewise.
27800         * reload1.c (alter_reg): Likewise.
27801         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
27802         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
27803         * tree-if-conv.c (predicate_mem_writes): Likewise.
27804         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
27805         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
27806         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
27807         * valtrack.c (dead_debug_insert_temp): Likewise.
27808         * varasm.c (mergeable_constant_section): Likewise.
27809         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
27811 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27812             Alan Hayward  <alan.hayward@arm.com>
27813             David Sherwood  <david.sherwood@arm.com>
27815         * expr.c (expand_assignment): Cope with polynomial mode sizes
27816         when assigning to a CONCAT.
27818 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27819             Alan Hayward  <alan.hayward@arm.com>
27820             David Sherwood  <david.sherwood@arm.com>
27822         * machmode.h (mode_precision): Change from unsigned short to
27823         poly_uint16_pod.
27824         (mode_to_precision): Return a poly_uint16 rather than an unsigned
27825         short.
27826         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
27827         or if measurement_type is not polynomial.
27828         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
27829         in which the mode is already known to be a scalar_int_mode.
27830         * genmodes.c (emit_mode_precision): Change the type of mode_precision
27831         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
27832         initializer.
27833         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27834         for GET_MODE_PRECISION.
27835         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27836         for GET_MODE_PRECISION.
27837         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
27838         as polynomial.
27839         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
27840         (expand_field_assignment, make_extraction): Likewise.
27841         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
27842         (get_last_value): Likewise.
27843         * convert.c (convert_to_integer_1): Likewise.
27844         * cse.c (cse_insn): Likewise.
27845         * expr.c (expand_expr_real_1): Likewise.
27846         * lra-constraints.c (simplify_operand_subreg): Likewise.
27847         * optabs-query.c (can_atomic_load_p): Likewise.
27848         * optabs.c (expand_atomic_load): Likewise.
27849         (expand_atomic_store): Likewise.
27850         * ree.c (combine_reaching_defs): Likewise.
27851         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
27852         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
27853         * tree.h (type_has_mode_precision_p): Likewise.
27854         * ubsan.c (instrument_si_overflow): Likewise.
27856 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27857             Alan Hayward  <alan.hayward@arm.com>
27858             David Sherwood  <david.sherwood@arm.com>
27860         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
27861         polynomial numbers of units.
27862         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
27863         (valid_vector_subparts_p): New function.
27864         (build_vector_type): Remove temporary shim and take the number
27865         of units as a poly_uint64 rather than an int.
27866         (build_opaque_vector_type): Take the number of units as a
27867         poly_uint64 rather than an int.
27868         * tree.c (build_vector_from_ctor): Handle polynomial
27869         TYPE_VECTOR_SUBPARTS.
27870         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
27871         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
27872         (build_vector_from_val): If the number of units is variable,
27873         use build_vec_duplicate_cst for constant operands and
27874         VEC_DUPLICATE_EXPR otherwise.
27875         (make_vector_type): Remove temporary is_constant ().
27876         (build_vector_type, build_opaque_vector_type): Take the number of
27877         units as a poly_uint64 rather than an int.
27878         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
27879         VECTOR_CST_NELTS.
27880         * cfgexpand.c (expand_debug_expr): Likewise.
27881         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
27882         (store_constructor, expand_expr_real_1): Likewise.
27883         (const_scalar_mask_from_tree): Likewise.
27884         * fold-const-call.c (fold_const_reduction): Likewise.
27885         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
27886         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
27887         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
27888         (fold_relational_const): Likewise.
27889         (native_interpret_vector): Likewise.  Change the size from an
27890         int to an unsigned int.
27891         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
27892         TYPE_VECTOR_SUBPARTS.
27893         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
27894         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
27895         duplicating a non-constant operand into a variable-length vector.
27896         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
27897         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
27898         * ipa-icf.c (sem_variable::equals): Likewise.
27899         * match.pd: Likewise.
27900         * omp-simd-clone.c (simd_clone_subparts): Likewise.
27901         * print-tree.c (print_node): Likewise.
27902         * stor-layout.c (layout_type): Likewise.
27903         * targhooks.c (default_builtin_vectorization_cost): Likewise.
27904         * tree-cfg.c (verify_gimple_comparison): Likewise.
27905         (verify_gimple_assign_binary): Likewise.
27906         (verify_gimple_assign_ternary): Likewise.
27907         (verify_gimple_assign_single): Likewise.
27908         * tree-pretty-print.c (dump_generic_node): Likewise.
27909         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27910         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
27911         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
27912         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
27913         (vect_shift_permute_load_chain): Likewise.
27914         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
27915         (expand_vector_condition, optimize_vector_constructor): Likewise.
27916         (lower_vec_perm, get_compute_type): Likewise.
27917         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27918         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
27919         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
27920         (vect_recog_mask_conversion_pattern): Likewise.
27921         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
27922         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
27923         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27924         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
27925         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
27926         (vectorizable_shift, vectorizable_operation, vectorizable_store)
27927         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
27928         (supportable_widening_operation): Likewise.
27929         (supportable_narrowing_operation): Likewise.
27930         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
27931         Likewise.
27932         * varasm.c (output_constant): Likewise.
27934 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27935             Alan Hayward  <alan.hayward@arm.com>
27936             David Sherwood  <david.sherwood@arm.com>
27938         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
27939         so that both the length == 3 and length != 3 cases set up their
27940         own permute vectors.  Add comments explaining why we know the
27941         number of elements is constant.
27942         (vect_permute_load_chain): Likewise.
27944 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27945             Alan Hayward  <alan.hayward@arm.com>
27946             David Sherwood  <david.sherwood@arm.com>
27948         * machmode.h (mode_nunits): Change from unsigned char to
27949         poly_uint16_pod.
27950         (ONLY_FIXED_SIZE_MODES): New macro.
27951         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
27952         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
27953         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
27954         New typedefs.
27955         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
27956         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
27957         or if measurement_type is not polynomial.
27958         * genmodes.c (ZERO_COEFFS): New macro.
27959         (emit_mode_nunits_inline): Make mode_nunits_inline return a
27960         poly_uint16.
27961         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
27962         Use ZERO_COEFFS when emitting initializers.
27963         * data-streamer.h (bp_pack_poly_value): New function.
27964         (bp_unpack_poly_value): Likewise.
27965         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27966         for GET_MODE_NUNITS.
27967         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27968         for GET_MODE_NUNITS.
27969         * tree.c (make_vector_type): Remove temporary shim and make
27970         the real function take the number of units as a poly_uint64
27971         rather than an int.
27972         (build_vector_type_for_mode): Handle polynomial nunits.
27973         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
27974         * emit-rtl.c (const_vec_series_p_1): Likewise.
27975         (gen_rtx_CONST_VECTOR): Likewise.
27976         * fold-const.c (test_vec_duplicate_folding): Likewise.
27977         * genrecog.c (validate_pattern): Likewise.
27978         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
27979         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27980         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
27981         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
27982         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
27983         * rtlanal.c (subreg_get_info): Likewise.
27984         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27985         (vect_grouped_load_supported): Likewise.
27986         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
27987         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
27988         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
27989         (simplify_const_unary_operation, simplify_binary_operation_1)
27990         (simplify_const_binary_operation, simplify_ternary_operation)
27991         (test_vector_ops_duplicate, test_vector_ops): Likewise.
27992         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
27993         instead of CONST_VECTOR_NUNITS.
27994         * varasm.c (output_constant_pool_2): Likewise.
27995         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
27996         explicit-encoded elements in the XVEC for variable-length vectors.
27998 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28000         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
28002 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28003             Alan Hayward  <alan.hayward@arm.com>
28004             David Sherwood  <david.sherwood@arm.com>
28006         * coretypes.h (fixed_size_mode): Declare.
28007         (fixed_size_mode_pod): New typedef.
28008         * builtins.h (target_builtins::x_apply_args_mode)
28009         (target_builtins::x_apply_result_mode): Change type to
28010         fixed_size_mode_pod.
28011         * builtins.c (apply_args_size, apply_result_size, result_vector)
28012         (expand_builtin_apply_args_1, expand_builtin_apply)
28013         (expand_builtin_return): Update accordingly.
28015 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28017         * cse.c (hash_rtx_cb): Hash only the encoded elements.
28018         * cselib.c (cselib_hash_rtx): Likewise.
28019         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
28020         CONST_VECTOR encoding.
28022 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28023             Jeff Law  <law@redhat.com>
28025         PR target/83641
28026         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
28027         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
28028         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
28029         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
28031         PR target/83641
28032         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
28033         explicitly probe *sp in a noreturn function if there were any callee
28034         register saves or frame pointer is needed.
28036 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28038         PR debug/83621
28039         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
28040         BLKmode for ternary, binary or unary expressions.
28042         PR debug/83645
28043         * var-tracking.c (delete_vta_debug_insn): New inline function.
28044         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
28045         insns from get_insns () to NULL instead of each bb separately.
28046         Use delete_vta_debug_insn.  No longer static.
28047         (vt_debug_insns_local, variable_tracking_main_1): Adjust
28048         delete_vta_debug_insns callers.
28049         * rtl.h (delete_vta_debug_insns): Declare.
28050         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
28051         instead of variable_tracking_main.
28053 2018-01-03  Martin Sebor  <msebor@redhat.com>
28055         PR tree-optimization/83603
28056         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
28057         arguments past the endof the argument list in functions declared
28058         without a prototype.
28059         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
28060         Avoid checking when arguments are null.
28062 2018-01-03  Martin Sebor  <msebor@redhat.com>
28064         PR c/83559
28065         * doc/extend.texi (attribute const): Fix a typo.
28066         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
28067         issuing -Wsuggest-attribute for void functions.
28069 2018-01-03  Martin Sebor  <msebor@redhat.com>
28071         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
28072         offset_int::from instead of wide_int::to_shwi.
28073         (maybe_diag_overlap): Remove assertion.
28074         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
28075         * gimple-ssa-sprintf.c (format_directive): Same.
28076         (parse_directive): Same.
28077         (sprintf_dom_walker::compute_format_length): Same.
28078         (try_substitute_return_value): Same.
28080 2018-01-03  Jeff Law  <law@redhat.com>
28082         PR middle-end/83654
28083         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
28084         non-constant residual for zero at runtime and avoid probing in
28085         that case.  Reorganize code for trailing problem to mirror handling
28086         of the residual.
28088 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28090         PR tree-optimization/83501
28091         * tree-ssa-strlen.c (get_string_cst): New.
28092         (handle_char_store): Call get_string_cst.
28094 2018-01-03  Martin Liska  <mliska@suse.cz>
28096         PR tree-optimization/83593
28097         * tree-ssa-strlen.c: Include tree-cfg.h.
28098         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
28099         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
28100         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
28101         to false.
28102         (strlen_dom_walker::before_dom_children): Call
28103         gimple_purge_dead_eh_edges. Dump tranformation with details
28104         dump flags.
28105         (strlen_dom_walker::before_dom_children): Update call by adding
28106         new argument cleanup_eh.
28107         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
28109 2018-01-03  Martin Liska  <mliska@suse.cz>
28111         PR ipa/83549
28112         * cif-code.def (VARIADIC_THUNK): New enum value.
28113         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
28114         thunks.
28116 2018-01-03  Jan Beulich  <jbeulich@suse.com>
28118         * sse.md (mov<mode>_internal): Tighten condition for when to use
28119         vmovdqu<ssescalarsize> for TI and OI modes.
28121 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28123         Update copyright years.
28125 2018-01-03  Martin Liska  <mliska@suse.cz>
28127         PR ipa/83594
28128         * ipa-visibility.c (function_and_variable_visibility): Skip
28129         functions with noipa attribure.
28131 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28133         * gcc.c (process_command): Update copyright notice dates.
28134         * gcov-dump.c (print_version): Ditto.
28135         * gcov.c (print_version): Ditto.
28136         * gcov-tool.c (print_version): Ditto.
28137         * gengtype.c (create_file): Ditto.
28138         * doc/cpp.texi: Bump @copying's copyright year.
28139         * doc/cppinternals.texi: Ditto.
28140         * doc/gcc.texi: Ditto.
28141         * doc/gccint.texi: Ditto.
28142         * doc/gcov.texi: Ditto.
28143         * doc/install.texi: Ditto.
28144         * doc/invoke.texi: Ditto.
28146 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28148         * vector-builder.h (vector_builder::m_full_nelts): Change from
28149         unsigned int to poly_uint64.
28150         (vector_builder::full_nelts): Update prototype accordingly.
28151         (vector_builder::new_vector): Likewise.
28152         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
28153         (vector_builder::operator ==): Likewise.
28154         (vector_builder::finalize): Likewise.
28155         * int-vector-builder.h (int_vector_builder::int_vector_builder):
28156         Take the number of elements as a poly_uint64 rather than an
28157         unsigned int.
28158         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
28159         from unsigned int to poly_uint64.
28160         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
28161         (vec_perm_indices::new_vector): Likewise.
28162         (vec_perm_indices::length): Likewise.
28163         (vec_perm_indices::nelts_per_input): Likewise.
28164         (vec_perm_indices::input_nelts): Likewise.
28165         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
28166         number of elements per input as a poly_uint64 rather than an
28167         unsigned int.  Use the original encoding for variable-length
28168         vectors, rather than clamping each individual element.
28169         For the second and subsequent elements in each pattern,
28170         clamp the step and base before clamping their sum.
28171         (vec_perm_indices::series_p): Handle polynomial element counts.
28172         (vec_perm_indices::all_in_range_p): Likewise.
28173         (vec_perm_indices_to_tree): Likewise.
28174         (vec_perm_indices_to_rtx): Likewise.
28175         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
28176         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
28177         (tree_vector_builder::new_binary_operation): Handle polynomial
28178         element counts.  Return false if we need to know the number
28179         of elements at compile time.
28180         * fold-const.c (fold_vec_perm): Punt if the number of elements
28181         isn't known at compile time.
28183 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28185         * vec-perm-indices.h (vec_perm_builder): Change element type
28186         from HOST_WIDE_INT to poly_int64.
28187         (vec_perm_indices::element_type): Update accordingly.
28188         (vec_perm_indices::clamp): Handle polynomial element_types.
28189         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28190         (vec_perm_indices::all_in_range_p): Likewise.
28191         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
28192         than shwi trees.
28193         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
28194         polynomial vec_perm_indices element types.
28195         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
28196         * fold-const.c (fold_vec_perm): Likewise.
28197         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
28198         * tree-vect-generic.c (lower_vec_perm): Likewise.
28199         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28200         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
28201         element type to HOST_WIDE_INT.
28203 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28204             Alan Hayward  <alan.hayward@arm.com>
28205             David Sherwood  <david.sherwood@arm.com>
28207         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
28208         rather than an int.  Use plus_constant.
28209         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
28210         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
28212 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28213             Alan Hayward  <alan.hayward@arm.com>
28214             David Sherwood  <david.sherwood@arm.com>
28216         * calls.c (emit_call_1, expand_call): Change struct_value_size from
28217         a HOST_WIDE_INT to a poly_int64.
28219 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28220             Alan Hayward  <alan.hayward@arm.com>
28221             David Sherwood  <david.sherwood@arm.com>
28223         * calls.c (load_register_parameters): Cope with polynomial
28224         mode sizes.  Require a constant size for BLKmode parameters
28225         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
28226         forces a parameter to be padded at the lsb end in order to
28227         fill a complete number of words, require the parameter size
28228         to be ordered wrt UNITS_PER_WORD.
28230 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28231             Alan Hayward  <alan.hayward@arm.com>
28232             David Sherwood  <david.sherwood@arm.com>
28234         * reload1.c (spill_stack_slot_width): Change element type
28235         from unsigned int to poly_uint64_pod.
28236         (alter_reg): Treat mode sizes as polynomial.
28238 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28239             Alan Hayward  <alan.hayward@arm.com>
28240             David Sherwood  <david.sherwood@arm.com>
28242         * reload.c (complex_word_subreg_p): New function.
28243         (reload_inner_reg_of_subreg, push_reload): Use it.
28245 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28246             Alan Hayward  <alan.hayward@arm.com>
28247             David Sherwood  <david.sherwood@arm.com>
28249         * lra-constraints.c (process_alt_operands): Reject matched
28250         operands whose sizes aren't ordered.
28251         (match_reload): Refer to this check here.
28253 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28254             Alan Hayward  <alan.hayward@arm.com>
28255             David Sherwood  <david.sherwood@arm.com>
28257         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28258         that the mode size is in the set {1, 2, 4, 8, 16}.
28260 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28261             Alan Hayward  <alan.hayward@arm.com>
28262             David Sherwood  <david.sherwood@arm.com>
28264         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28265         Use plus_constant instead of gen_rtx_PLUS.
28267 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28268             Alan Hayward  <alan.hayward@arm.com>
28269             David Sherwood  <david.sherwood@arm.com>
28271         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28272         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28273         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28274         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28275         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28276         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28277         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28278         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28279         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28280         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
28281         a poly_int64.
28282         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
28283         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
28284         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
28285         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
28286         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
28287         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
28288         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
28289         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
28290         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
28291         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28292         function.
28293         * expr.c (emit_move_resolve_push): Treat the input and result
28294         of PUSH_ROUNDING as a poly_int64.
28295         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28296         (emit_push_insn): Likewise.
28297         * lra-eliminations.c (mark_not_eliminable): Likewise.
28298         * recog.c (push_operand): Likewise.
28299         * reload1.c (elimination_effects): Likewise.
28300         * rtlanal.c (nonzero_bits1): Likewise.
28301         * calls.c (store_one_arg): Likewise.  Require the padding to be
28302         known at compile time.
28304 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28305             Alan Hayward  <alan.hayward@arm.com>
28306             David Sherwood  <david.sherwood@arm.com>
28308         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28309         Use plus_constant instead of gen_rtx_PLUS.
28311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28312             Alan Hayward  <alan.hayward@arm.com>
28313             David Sherwood  <david.sherwood@arm.com>
28315         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28316         rather than an int.
28318 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28319             Alan Hayward  <alan.hayward@arm.com>
28320             David Sherwood  <david.sherwood@arm.com>
28322         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28323         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28324         via stack temporaries.  Treat the mode size as polynomial too.
28326 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28327             Alan Hayward  <alan.hayward@arm.com>
28328             David Sherwood  <david.sherwood@arm.com>
28330         * expr.c (expand_expr_real_2): When handling conversions involving
28331         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28332         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
28333         as a poly_uint64 too.
28335 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28336             Alan Hayward  <alan.hayward@arm.com>
28337             David Sherwood  <david.sherwood@arm.com>
28339         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
28341 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28342             Alan Hayward  <alan.hayward@arm.com>
28343             David Sherwood  <david.sherwood@arm.com>
28345         * combine.c (can_change_dest_mode): Handle polynomial
28346         REGMODE_NATURAL_SIZE.
28347         * expmed.c (store_bit_field_1): Likewise.
28348         * expr.c (store_constructor): Likewise.
28349         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
28350         and polynomial REGMODE_NATURAL_SIZE.
28351         (gen_lowpart_common): Likewise.
28352         * reginfo.c (record_subregs_of_mode): Likewise.
28353         * rtlanal.c (read_modify_subreg_p): Likewise.
28355 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28356             Alan Hayward  <alan.hayward@arm.com>
28357             David Sherwood  <david.sherwood@arm.com>
28359         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
28360         numbers of elements.
28362 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28363             Alan Hayward  <alan.hayward@arm.com>
28364             David Sherwood  <david.sherwood@arm.com>
28366         * match.pd: Cope with polynomial numbers of vector elements.
28368 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28369             Alan Hayward  <alan.hayward@arm.com>
28370             David Sherwood  <david.sherwood@arm.com>
28372         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
28373         in a POINTER_PLUS_EXPR.
28375 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28376             Alan Hayward  <alan.hayward@arm.com>
28377             David Sherwood  <david.sherwood@arm.com>
28379         * omp-simd-clone.c (simd_clone_subparts): New function.
28380         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
28381         (ipa_simd_modify_function_body): Likewise.
28383 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28384             Alan Hayward  <alan.hayward@arm.com>
28385             David Sherwood  <david.sherwood@arm.com>
28387         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
28388         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
28389         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
28390         (expand_vector_condition, vector_element): Likewise.
28391         (subparts_gt): New function.
28392         (get_compute_type): Use subparts_gt.
28393         (count_type_subparts): Delete.
28394         (expand_vector_operations_1): Use subparts_gt instead of
28395         count_type_subparts.
28397 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28398             Alan Hayward  <alan.hayward@arm.com>
28399             David Sherwood  <david.sherwood@arm.com>
28401         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
28402         (vect_compile_time_alias): ...this new function.  Do the calculation
28403         on poly_ints rather than trees.
28404         (vect_prune_runtime_alias_test_list): Update call accordingly.
28406 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28407             Alan Hayward  <alan.hayward@arm.com>
28408             David Sherwood  <david.sherwood@arm.com>
28410         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
28411         numbers of units.
28412         (vect_schedule_slp_instance): Likewise.
28414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28415             Alan Hayward  <alan.hayward@arm.com>
28416             David Sherwood  <david.sherwood@arm.com>
28418         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
28419         constant and extern definitions for variable-length vectors.
28420         (vect_get_constant_vectors): Note that the number of units
28421         is known to be constant.
28423 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28424             Alan Hayward  <alan.hayward@arm.com>
28425             David Sherwood  <david.sherwood@arm.com>
28427         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
28428         of units as polynomial.  Choose between WIDE and NARROW based
28429         on multiple_p.
28431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28432             Alan Hayward  <alan.hayward@arm.com>
28433             David Sherwood  <david.sherwood@arm.com>
28435         * tree-vect-stmts.c (simd_clone_subparts): New function.
28436         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
28438 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28439             Alan Hayward  <alan.hayward@arm.com>
28440             David Sherwood  <david.sherwood@arm.com>
28442         * tree-vect-stmts.c (vectorizable_call): Treat the number of
28443         vectors as polynomial.  Use build_index_vector for
28444         IFN_GOMP_SIMD_LANE.
28446 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28447             Alan Hayward  <alan.hayward@arm.com>
28448             David Sherwood  <david.sherwood@arm.com>
28450         * tree-vect-stmts.c (get_load_store_type): Treat the number of
28451         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
28452         for variable-length vectors.
28453         (vectorizable_mask_load_store): Treat the number of units as
28454         polynomial, asserting that it is constant if the condition has
28455         already been enforced.
28456         (vectorizable_store, vectorizable_load): Likewise.
28458 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28459             Alan Hayward  <alan.hayward@arm.com>
28460             David Sherwood  <david.sherwood@arm.com>
28462         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
28463         of units as polynomial.  Punt if we can't tell at compile time
28464         which vector contains the final result.
28466 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28467             Alan Hayward  <alan.hayward@arm.com>
28468             David Sherwood  <david.sherwood@arm.com>
28470         * tree-vect-loop.c (vectorizable_induction): Treat the number
28471         of units as polynomial.  Punt on SLP inductions.  Use an integer
28472         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
28473         cast of such a series for variable-length floating-point
28474         reductions.
28476 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28477             Alan Hayward  <alan.hayward@arm.com>
28478             David Sherwood  <david.sherwood@arm.com>
28480         * tree.h (build_index_vector): Declare.
28481         * tree.c (build_index_vector): New function.
28482         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
28483         of units as polynomial, forcibly converting it to a constant if
28484         vectorizable_reduction has already enforced the condition.
28485         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
28486         to create a {1,2,3,...} vector.
28487         (vectorizable_reduction): Treat the number of units as polynomial.
28488         Choose vectype_in based on the largest scalar element size rather
28489         than the smallest number of units.  Enforce the restrictions
28490         relied on above.
28492 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28493             Alan Hayward  <alan.hayward@arm.com>
28494             David Sherwood  <david.sherwood@arm.com>
28496         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
28497         number of units as polynomial.
28499 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28500             Alan Hayward  <alan.hayward@arm.com>
28501             David Sherwood  <david.sherwood@arm.com>
28503         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
28504         * target.def (autovectorize_vector_sizes): Return the vector sizes
28505         by pointer, using vector_sizes rather than a bitmask.
28506         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
28507         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
28508         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
28509         Likewise.
28510         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
28511         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
28512         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
28513         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
28514         * omp-general.c (omp_max_vf): Likewise.
28515         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28516         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
28517         * tree-vect-loop.c (vect_analyze_loop): Likewise.
28518         * tree-vect-slp.c (vect_slp_bb): Likewise.
28519         * doc/tm.texi: Regenerate.
28520         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
28521         to a poly_uint64.
28522         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
28523         the vector size as a poly_uint64 rather than an unsigned int.
28524         (current_vector_size): Change from an unsigned int to a poly_uint64.
28525         (get_vectype_for_scalar_type): Update accordingly.
28526         * tree.h (build_truth_vector_type): Take the size and number of
28527         units as a poly_uint64 rather than an unsigned int.
28528         (build_vector_type): Add a temporary overload that takes
28529         the number of units as a poly_uint64 rather than an unsigned int.
28530         * tree.c (make_vector_type): Likewise.
28531         (build_truth_vector_type): Take the number of units as a poly_uint64
28532         rather than an unsigned int.
28534 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28535             Alan Hayward  <alan.hayward@arm.com>
28536             David Sherwood  <david.sherwood@arm.com>
28538         * target.def (get_mask_mode): Take the number of units and length
28539         as poly_uint64s rather than unsigned ints.
28540         * targhooks.h (default_get_mask_mode): Update accordingly.
28541         * targhooks.c (default_get_mask_mode): Likewise.
28542         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
28543         * doc/tm.texi: Regenerate.
28545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28546             Alan Hayward  <alan.hayward@arm.com>
28547             David Sherwood  <david.sherwood@arm.com>
28549         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
28550         * omp-general.c (omp_max_vf): Likewise.
28551         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
28552         (expand_omp_simd): Handle polynomial safelen.
28553         * omp-low.c (omplow_simd_context): Add a default constructor.
28554         (omplow_simd_context::max_vf): Change from int to poly_uint64.
28555         (lower_rec_simd_input_clauses): Update accordingly.
28556         (lower_rec_input_clauses): Likewise.
28558 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28559             Alan Hayward  <alan.hayward@arm.com>
28560             David Sherwood  <david.sherwood@arm.com>
28562         * tree-vectorizer.h (vect_nunits_for_cost): New function.
28563         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
28564         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
28565         (vect_analyze_slp_cost): Likewise.
28566         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
28567         (vect_model_load_cost): Likewise.
28569 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28570             Alan Hayward  <alan.hayward@arm.com>
28571             David Sherwood  <david.sherwood@arm.com>
28573         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
28574         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
28575         from an unsigned int * to a poly_uint64_pod *.
28576         (calculate_unrolling_factor): New function.
28577         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
28579 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28580             Alan Hayward  <alan.hayward@arm.com>
28581             David Sherwood  <david.sherwood@arm.com>
28583         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
28584         from an unsigned int to a poly_uint64.
28585         (_loop_vec_info::slp_unrolling_factor): Likewise.
28586         (_loop_vec_info::vectorization_factor): Change from an int
28587         to a poly_uint64.
28588         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
28589         (vect_get_num_vectors): New function.
28590         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
28591         (vect_get_num_copies): Use vect_get_num_vectors.
28592         (vect_analyze_data_ref_dependences): Change max_vf from an int *
28593         to an unsigned int *.
28594         (vect_analyze_data_refs): Change min_vf from an int * to a
28595         poly_uint64 *.
28596         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28597         than an unsigned HOST_WIDE_INT.
28598         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
28599         (vect_analyze_data_ref_dependence): Change max_vf from an int *
28600         to an unsigned int *.
28601         (vect_analyze_data_ref_dependences): Likewise.
28602         (vect_compute_data_ref_alignment): Handle polynomial vf.
28603         (vect_enhance_data_refs_alignment): Likewise.
28604         (vect_prune_runtime_alias_test_list): Likewise.
28605         (vect_shift_permute_load_chain): Likewise.
28606         (vect_supportable_dr_alignment): Likewise.
28607         (dependence_distance_ge_vf): Take the vectorization factor as a
28608         poly_uint64 rather than an unsigned HOST_WIDE_INT.
28609         (vect_analyze_data_refs): Change min_vf from an int * to a
28610         poly_uint64 *.
28611         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
28612         vfm1 as a poly_uint64 rather than an int.  Make the same change
28613         for the returned bound_scalar.
28614         (vect_gen_vector_loop_niters): Handle polynomial vf.
28615         (vect_do_peeling): Likewise.  Update call to
28616         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
28617         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
28618         be constant.
28619         * tree-vect-loop.c (vect_determine_vectorization_factor)
28620         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
28621         (vect_get_known_peeling_cost): Likewise.
28622         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
28623         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
28624         (vect_transform_loop): Likewise.  Use the lowest possible VF when
28625         updating the upper bounds of the loop.
28626         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
28627         rather than an int.
28628         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
28629         polynomial unroll factors.
28630         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
28631         (vect_make_slp_decision): Likewise.
28632         (vect_supported_load_permutation_p): Likewise, and polynomial
28633         vf too.
28634         (vect_analyze_slp_cost): Handle polynomial vf.
28635         (vect_slp_analyze_node_operations): Likewise.
28636         (vect_slp_analyze_bb_1): Likewise.
28637         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28638         than an unsigned HOST_WIDE_INT.
28639         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
28640         (vectorizable_load): Handle polynomial vf.
28641         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
28642         a poly_uint64.
28643         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
28645 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28646             Alan Hayward  <alan.hayward@arm.com>
28647             David Sherwood  <david.sherwood@arm.com>
28649         * match.pd: Handle bit operations involving three constants
28650         and try to fold one pair.
28652 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28654         * tree-vect-loop-manip.c: Include gimple-fold.h.
28655         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
28656         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
28657         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
28658         Add a path that uses a step of VF instead of 1, but disable it
28659         for now.
28660         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
28661         and niters_no_overflow parameters.  Update calls to
28662         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
28663         Create a new SSA name if the latter choses to use a ste other
28664         than zero, and return it via niters_vector_mult_vf_var.
28665         * tree-vect-loop.c (vect_transform_loop): Update calls to
28666         vect_do_peeling, vect_gen_vector_loop_niters and
28667         slpeel_make_loop_iterate_ntimes.
28668         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
28669         (vect_gen_vector_loop_niters): Update declarations after above changes.
28671 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
28673         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
28674         128-bit round to integer instructions.
28675         (ceil<mode>2): Likewise.
28676         (btrunc<mode>2): Likewise.
28677         (round<mode>2): Likewise.
28679 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28681         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
28682         unaligned VSX load/store on P8/P9.
28683         (expand_block_clear): Allow the use of unaligned VSX
28684         load/store on P8/P9.
28686 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28688         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
28689         New function.
28690         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
28691         swap associated with both a load and a store.
28693 2018-01-02  Andrew Waterman  <andrew@sifive.com>
28695         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
28696         * config/riscv/riscv.md (clear_cache): Use it.
28698 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
28700         * web.c: Remove out-of-date comment.
28702 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28704         * expr.c (fixup_args_size_notes): Check that any existing
28705         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
28706         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
28707         (emit_single_push_insn): ...here.
28709 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28711         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
28712         (const_vector_encoded_nelts): New function.
28713         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
28714         (const_vector_int_elt, const_vector_elt): Declare.
28715         * emit-rtl.c (const_vector_int_elt_1): New function.
28716         (const_vector_elt): Likewise.
28717         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
28718         of CONST_VECTOR_ELT.
28720 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28722         * expr.c: Include rtx-vector-builder.h.
28723         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
28724         directly on the tree encoding.
28725         (const_vector_from_tree): Likewise.
28726         * optabs.c: Include rtx-vector-builder.h.
28727         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
28728         sequence of "u" values.
28729         * vec-perm-indices.c: Include rtx-vector-builder.h.
28730         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
28731         directly on the vec_perm_indices encoding.
28733 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28735         * doc/rtl.texi (const_vector): Describe new encoding scheme.
28736         * Makefile.in (OBJS): Add rtx-vector-builder.o.
28737         * rtx-vector-builder.h: New file.
28738         * rtx-vector-builder.c: Likewise.
28739         * rtl.h (rtx_def::u2): Add a const_vector field.
28740         (CONST_VECTOR_NPATTERNS): New macro.
28741         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
28742         (CONST_VECTOR_DUPLICATE_P): Likewise.
28743         (CONST_VECTOR_STEPPED_P): Likewise.
28744         (CONST_VECTOR_ENCODED_ELT): Likewise.
28745         (const_vec_duplicate_p): Check for a duplicated vector encoding.
28746         (unwrap_const_vec_duplicate): Likewise.
28747         (const_vec_series_p): Check for a non-duplicated vector encoding.
28748         Say that the function only returns true for integer vectors.
28749         * emit-rtl.c: Include rtx-vector-builder.h.
28750         (gen_const_vec_duplicate_1): Delete.
28751         (gen_const_vector): Call gen_const_vec_duplicate instead of
28752         gen_const_vec_duplicate_1.
28753         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
28754         (gen_const_vec_duplicate): Use rtx_vector_builder.
28755         (gen_const_vec_series): Likewise.
28756         (gen_rtx_CONST_VECTOR): Likewise.
28757         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
28758         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28759         Build a new vector rather than modifying a CONST_VECTOR in-place.
28760         (handle_special_swappables): Update call accordingly.
28761         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
28762         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28763         Build a new vector rather than modifying a CONST_VECTOR in-place.
28764         (handle_special_swappables): Update call accordingly.
28766 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28768         * simplify-rtx.c (simplify_const_binary_operation): Use
28769         CONST_VECTOR_ELT instead of XVECEXP.
28771 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28773         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
28774         the selector elements to be different from the data elements
28775         if the selector is a VECTOR_CST.
28776         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
28777         ssizetype for the selector.
28779 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28781         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
28782         before testing each element individually.
28783         * tree-vect-generic.c (lower_vec_perm): Likewise.
28785 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28787         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
28788         * selftest-run-tests.c (selftest::run_tests): Call it.
28789         * vector-builder.h (vector_builder::operator ==): New function.
28790         (vector_builder::operator !=): Likewise.
28791         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
28792         (vec_perm_indices::all_from_input_p): New function.
28793         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28794         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
28795         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
28796         instead of reading the VECTOR_CST directly.  Detect whether both
28797         vector inputs are the same before constructing the vec_perm_indices,
28798         and update the number of inputs argument accordingly.  Use the
28799         utility functions added above.  Only construct sel2 if we need to.
28801 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28803         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
28804         the broadcast of the low byte.
28805         (expand_mult_highpart): Use an explicit encoding for the permutes.
28806         * optabs-query.c (can_mult_highpart_p): Likewise.
28807         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
28808         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28809         (vectorizable_bswap): Likewise.
28810         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
28811         explicit encoding for the power-of-2 permutes.
28812         (vect_permute_store_chain): Likewise.
28813         (vect_grouped_load_supported): Likewise.
28814         (vect_permute_load_chain): Likewise.
28816 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28818         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
28819         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
28820         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
28821         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28822         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
28823         (vect_gen_perm_mask_any): Likewise.
28825 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28827         * int-vector-builder.h: New file.
28828         * vec-perm-indices.h: Include int-vector-builder.h.
28829         (vec_perm_indices): Redefine as an int_vector_builder.
28830         (auto_vec_perm_indices): Delete.
28831         (vec_perm_builder): Redefine as a stand-alone class.
28832         (vec_perm_indices::vec_perm_indices): New function.
28833         (vec_perm_indices::clamp): Likewise.
28834         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
28835         (vec_perm_indices::new_vector): New function.
28836         (vec_perm_indices::new_expanded_vector): Update for new
28837         vec_perm_indices class.
28838         (vec_perm_indices::rotate_inputs): New function.
28839         (vec_perm_indices::all_in_range_p): Operate directly on the
28840         encoded form, without computing elided elements.
28841         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
28842         encoding.  Update for new vec_perm_indices class.
28843         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
28844         the given vec_perm_builder.
28845         (expand_vec_perm_var): Update vec_perm_builder constructor.
28846         (expand_mult_highpart): Use vec_perm_builder instead of
28847         auto_vec_perm_indices.
28848         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
28849         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
28850         or double series encoding as appropriate.
28851         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
28852         vec_perm_indices instead of auto_vec_perm_indices.
28853         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28854         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28855         (vect_permute_store_chain): Likewise.
28856         (vect_grouped_load_supported): Likewise.
28857         (vect_permute_load_chain): Likewise.
28858         (vect_shift_permute_load_chain): Likewise.
28859         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28860         (vect_transform_slp_perm_load): Likewise.
28861         (vect_schedule_slp_instance): Likewise.
28862         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28863         (vectorizable_mask_load_store): Likewise.
28864         (vectorizable_bswap): Likewise.
28865         (vectorizable_store): Likewise.
28866         (vectorizable_load): Likewise.
28867         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
28868         vec_perm_indices instead of auto_vec_perm_indices.  Use
28869         tree_to_vec_perm_builder to read the vector from a tree.
28870         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
28871         vec_perm_builder instead of a vec_perm_indices.
28872         (have_whole_vector_shift): Use vec_perm_builder and
28873         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
28874         truncation to calc_vec_perm_mask_for_shift.
28875         (vect_create_epilog_for_reduction): Likewise.
28876         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
28877         from auto_vec_perm_indices to vec_perm_indices.
28878         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28879         instead of changing individual elements.
28880         (aarch64_vectorize_vec_perm_const): Use new_vector to install
28881         the vector in d.perm.
28882         * config/arm/arm.c (expand_vec_perm_d::perm): Change
28883         from auto_vec_perm_indices to vec_perm_indices.
28884         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28885         instead of changing individual elements.
28886         (arm_vectorize_vec_perm_const): Use new_vector to install
28887         the vector in d.perm.
28888         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
28889         Update vec_perm_builder constructor.
28890         (rs6000_expand_interleave): Likewise.
28891         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
28892         (rs6000_expand_interleave): Likewise.
28894 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28896         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
28897         to qimode could truncate the indices.
28898         * optabs.c (expand_vec_perm_var): Likewise.
28900 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28902         * Makefile.in (OBJS): Add vec-perm-indices.o.
28903         * vec-perm-indices.h: New file.
28904         * vec-perm-indices.c: Likewise.
28905         * target.h (vec_perm_indices): Replace with a forward class
28906         declaration.
28907         (auto_vec_perm_indices): Move to vec-perm-indices.h.
28908         * optabs.h: Include vec-perm-indices.h.
28909         (expand_vec_perm): Delete.
28910         (selector_fits_mode_p, expand_vec_perm_var): Declare.
28911         (expand_vec_perm_const): Declare.
28912         * target.def (vec_perm_const_ok): Replace with...
28913         (vec_perm_const): ...this new hook.
28914         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
28915         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
28916         * doc/tm.texi: Regenerate.
28917         * optabs.def (vec_perm_const): Delete.
28918         * doc/md.texi (vec_perm_const): Likewise.
28919         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
28920         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
28921         expand_vec_perm for constant permutation vectors.  Assert that
28922         the mode of variable permutation vectors is the integer equivalent
28923         of the mode that is being permuted.
28924         * optabs-query.h (selector_fits_mode_p): Declare.
28925         * optabs-query.c: Include vec-perm-indices.h.
28926         (selector_fits_mode_p): New function.
28927         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
28928         is defined, instead of checking whether the vec_perm_const_optab
28929         exists.  Use targetm.vectorize.vec_perm_const instead of
28930         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
28931         fit in the vector mode before using a variable permute.
28932         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
28933         vec_perm_indices instead of an rtx.
28934         (expand_vec_perm): Replace with...
28935         (expand_vec_perm_const): ...this new function.  Take the selector
28936         as a vec_perm_indices rather than an rtx.  Also take the mode of
28937         the selector.  Update call to shift_amt_for_vec_perm_mask.
28938         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
28939         Use vec_perm_indices::new_expanded_vector to expand the original
28940         selector into bytes.  Check whether the indices fit in the vector
28941         mode before using a variable permute.
28942         (expand_vec_perm_var): Make global.
28943         (expand_mult_highpart): Use expand_vec_perm_const.
28944         * fold-const.c: Includes vec-perm-indices.h.
28945         * tree-ssa-forwprop.c: Likewise.
28946         * tree-vect-data-refs.c: Likewise.
28947         * tree-vect-generic.c: Likewise.
28948         * tree-vect-loop.c: Likewise.
28949         * tree-vect-slp.c: Likewise.
28950         * tree-vect-stmts.c: Likewise.
28951         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
28952         Delete.
28953         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
28954         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
28955         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
28956         (aarch64_vectorize_vec_perm_const): ...this new function.
28957         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28958         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28959         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
28960         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
28961         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28962         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28963         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
28964         into...
28965         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
28966         check for NEON modes.
28967         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
28968         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
28969         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
28970         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
28971         into...
28972         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
28973         the old VEC_PERM_CONST conditions.
28974         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
28975         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
28976         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
28977         (ia64_vectorize_vec_perm_const_ok): Merge into...
28978         (ia64_vectorize_vec_perm_const): ...this new function.
28979         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
28980         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
28981         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
28982         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
28983         * config/mips/mips.c (mips_expand_vec_perm_const)
28984         (mips_vectorize_vec_perm_const_ok): Merge into...
28985         (mips_vectorize_vec_perm_const): ...this new function.
28986         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
28987         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
28988         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
28989         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
28990         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
28991         (rs6000_expand_vec_perm_const): Delete.
28992         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
28993         Delete.
28994         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28995         (altivec_expand_vec_perm_const_le): Take each operand individually.
28996         Operate on constant selectors rather than rtxes.
28997         (altivec_expand_vec_perm_const): Likewise.  Update call to
28998         altivec_expand_vec_perm_const_le.
28999         (rs6000_expand_vec_perm_const): Delete.
29000         (rs6000_vectorize_vec_perm_const_ok): Delete.
29001         (rs6000_vectorize_vec_perm_const): New function.
29002         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29003         an element count and rtx array.
29004         (rs6000_expand_extract_even): Update call accordingly.
29005         (rs6000_expand_interleave): Likewise.
29006         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
29007         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
29008         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
29009         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
29010         (rs6000_expand_vec_perm_const): Delete.
29011         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29012         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29013         (altivec_expand_vec_perm_const_le): Take each operand individually.
29014         Operate on constant selectors rather than rtxes.
29015         (altivec_expand_vec_perm_const): Likewise.  Update call to
29016         altivec_expand_vec_perm_const_le.
29017         (rs6000_expand_vec_perm_const): Delete.
29018         (rs6000_vectorize_vec_perm_const_ok): Delete.
29019         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
29020         reference to the SPE evmerge intructions.
29021         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29022         an element count and rtx array.
29023         (rs6000_expand_extract_even): Update call accordingly.
29024         (rs6000_expand_interleave): Likewise.
29025         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
29026         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
29027         new function.
29028         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29030 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29032         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
29033         vector mode and that that mode matches the mode of the data
29034         being permuted.
29035         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
29036         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
29037         directly using expand_vec_perm_1 when forcing selectors into
29038         registers.
29039         (expand_vec_perm_var): New function, split out from expand_vec_perm.
29041 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29043         * optabs-query.h (can_vec_perm_p): Delete.
29044         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
29045         * optabs-query.c (can_vec_perm_p): Split into...
29046         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
29047         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
29048         particular selector is valid.
29049         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29050         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29051         (vect_grouped_load_supported): Likewise.
29052         (vect_shift_permute_load_chain): Likewise.
29053         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29054         (vect_transform_slp_perm_load): Likewise.
29055         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29056         (vectorizable_bswap): Likewise.
29057         (vect_gen_perm_mask_checked): Likewise.
29058         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
29059         implementations of variable permutation vectors into account
29060         when deciding which selector to use.
29061         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
29062         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
29063         with a false third argument.
29064         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
29065         to test whether the constant selector is valid and can_vec_perm_var_p
29066         to test whether a variable selector is valid.
29068 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29070         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
29071         * optabs-query.c (can_vec_perm_p): Likewise.
29072         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
29073         instead of vec_perm_indices.
29074         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
29075         (vect_gen_perm_mask_checked): Likewise,
29076         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
29077         (vect_gen_perm_mask_checked): Likewise,
29079 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29081         * optabs-query.h (qimode_for_vec_perm): Declare.
29082         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
29083         (qimode_for_vec_perm): ...this new function.
29084         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
29086 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29088         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
29089         does not have a conditional at the top.
29091 2018-01-02  Richard Biener  <rguenther@suse.de>
29093         * ipa-inline.c (big_speedup_p): Fix expression.
29095 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29097         PR target/81616
29098         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
29099         for generic 4->6.
29101 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29103         PR target/81616
29104         Generic tuning.
29105         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
29106         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
29107         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
29108         cond_taken_branch_cost 3->4.
29110 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
29112         PR tree-optimization/83581
29113         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
29114         TODO_cleanup_cfg if any changes have been made.
29116         PR middle-end/83608
29117         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
29118         convert_modes if target mode has the right side, but different mode
29119         class.
29121         PR middle-end/83609
29122         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
29123         last argument when extracting from CONCAT.  If either from_real or
29124         from_imag is NULL, use expansion through memory.  If result is not
29125         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
29126         the parts directly to inner mode, if even that fails, use expansion
29127         through memory.
29129         PR middle-end/83623
29130         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
29131         check for bswap in mode rather than HImode and use that in expand_unop
29132         too.
29134 Copyright (C) 2018 Free Software Foundation, Inc.
29136 Copying and distribution of this file, with or without modification,
29137 are permitted in any medium without royalty provided the copyright
29138 notice and this notice are preserved.