2018-10-24 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob33c906afb094c06506b0c744caa3a273d2b3f67f
1 2018-10-24  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/87665
4         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
5         to reflect reality.
7 2018-10-12  Jeff Law  <law@redhat.com>
9         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
10         for H8/S.
12 2018-10-23  Richard Biener  <rguenther@suse.de>
14         * tree-vrp.c (add_assert_info): Guard dump_printf with
15         dump_enabled_p.
16         * gimple-ssa-evrp-analyze.c
17         (evrp_range_analyzer::record_ranges_from_incoming_edge):
18         Use value_range::ignore_equivs_equal_p.
20 2018-10-23  Richard Biener  <rguenther@suse.de>
22         PR tree-optimization/87105
23         PR tree-optimization/87608
24         * passes.def (pass_all_early_optimizations): Add early phi-opt
25         after dce.
26         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
27         addition to debug stmts.
28         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
29         and abs replacement early.
30         * tree-cfg.c (gimple_empty_block_p): Likewise.
32 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
34         PR target/86383
35         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
36         specified to configure.
37         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
39 2018-10-23  Richard Biener  <rguenther@suse.de>
41         PR tree-optimization/87700
42         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
44 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
46         PR target/87674
47         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
48         second argument from __mmask16 to __mmask8.
49         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
50         _mm_mask_packs_epi32): Likewise.
51         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
52         Likewise.
53         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
55 2018-10-23  Richard Biener  <rguenther@suse.de>
57         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
59 2018-10-23  Richard Biener  <rguenther@suse.de>
61         PR tree-optimization/86144
62         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
63         over simd attribute.
65 2018-10-23  Richard Biener  <rguenther@suse.de>
67         PR tree-optimization/87693
68         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
69         the case we do not find the taken edge.
71 2018-10-22  Paul Koning  <ni1d@arrl.net>
73         * symtab.c (symtab_node::increase_alignment): Correct max
74         alignment check.
76 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
78         PR tree-optimization/87633
79         * match.pd: Do not generate unordered integer comparisons.
81 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
83         PR rtl-optimization/87600
84         * combine.c: Add include of expr.h.
85         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
86         register to a pseudo.
87         (make_more_copies): New function, add a copy to a new pseudo after
88         the moves from hard registers into pseudos.
89         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
90         later.  Call make_more_copies.
92 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
94         * lra-constraints.c (process_alt_operands): New local array,
95         matching_early_clobber.  Check matching_early_clobber before
96         decrementing reject, and set matching_early_clobber after.
98 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
100         PR target/87598
101         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
102         call output_addr_const and hope for the best.
104 2018-10-22  Richard Biener  <rguenther@suse.de>
106         * gimple-ssa-evrp-analyze.c
107         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
108         smarter about what ranges to use.
109         * tree-vrp.c (add_assert_info): Dump here.
110         (register_edge_assert_for_2): Instead of here at multiple but
111         not all places.
113         * gcc.dg/tree-ssa/evrp12.c: New testcase.
114         * gcc.dg/predict-6.c: Adjust.
115         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
116         * gcc.dg/tree-ssa/vrp02.c: Likewise.
117         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
119 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
120         Richard Biener  <rguenther@suse.de>
122         * bitmap.h: Update data structure documentation, including a
123         description of bitmap views as either linked-lists or splay trees.
124         (struct bitmap_element_def): Update comments for splay tree bitmaps.
125         (struct bitmap_head_def): Likewise.
126         (bitmap_list_view, bitmap_tree_view): New prototypes.
127         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
128         tree_form fields.
129         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
130         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
131         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
132         released bitmap element here.
133         (bitmap_element_free): Remove.
134         (bitmap_elt_clear_from): Work on splay tree bitmaps.
135         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
136         this function similar ones such that linked-list bitmap implementation
137         functions are grouped.
138         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
139         and moved for grouping.
140         (bitmap_list_insert_element_after): Renamed from
141         bitmap_elt_insert_after, and moved for grouping.
142         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
143         (bitmap_tree_link_left, bitmap_tree_link_right,
144         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
145         bitmap_tree_link_element, bitmap_tree_unlink_element,
146         bitmap_tree_find_element): New functions for splay-tree bitmap
147         implementation.
148         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
149         Renamed and moved, see above entries.
150         (bitmap_tree_listify_from): New function to convert part of a splay
151         tree bitmap to a linked-list bitmap.
152         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
153         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
154         (bitmap_find_bit): Remove.
155         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
156         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
157         Handle splay tree bitmaps.
158         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
159         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
160         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
161         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
162         bitmap_intersect_compl_p, bitmap_ior_and_compl,
163         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
164         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
165         corresponding changes to use linked-list specific bitmap_element
166         manipulation functions as applicable for efficiency.
167         (bitmap_tree_to_vec): New function.
168         (debug_bitmap_elt_file): New function split out from ...
169         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
170         (bitmap_print): Likewise.
172         PR tree-optimization/63155
173         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
174         SSA edge worklists.
175         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
176         in tree-view.
178 2018-10-22  Martin Liska  <mliska@suse.cz>
180         PR tree-optimization/87686
181         Revert
182         2018-08-29  Martin Liska  <mliska@suse.cz>
184         * tree-switch-conversion.c (switch_conversion::expand):
185         Strenghten assumption about gswitch statements.
187 2018-10-22  Martin Liska  <mliska@suse.cz>
189         * ipa-icf.c (sem_item::compare_attributes): Remove.
190         (sem_item::compare_referenced_symbol_properties): Use
191         attribute_list_equal instead.
192         (sem_function::equals_wpa): Likewise.
193         * ipa-icf.h: Remove compare_attributes.
195 2018-10-22  Richard Biener  <rguenther@suse.de>
197         PR middle-end/87682
198         * mem-stats.h (mem_usage::operator==): Fix pasto.
200 2018-10-22  Richard Biener  <rguenther@suse.de>
202         PR tree-optimization/87640
203         * tree-vrp.c (set_value_range_with_overflow): Decompose
204         incomplete result.
205         (extract_range_from_binary_expr_1): Adjust.
207 2018-10-22  Martin Jambor  <mjambor@suse.cz>
209         * tree-eh.h (stmt_could_throw_p): Add function parameter.
210         (stmt_can_throw_external): Likewise.
211         (stmt_can_throw_internal): Likewise.
212         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
213         (lower_eh_constructs_2): Likewise.
214         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
215         (stmt_can_throw_external): Likewise.
216         (stmt_can_throw_internal): Likewise.
217         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
218         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
219         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
220         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
221         (pass_lower_eh_dispatch::execute): Pass cfun to
222         stmt_can_throw_external.
223         (cleanup_empty_eh): Likewise.
224         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
225         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
226         stmt_can_throw_external instead of pushing it to cfun.
227         (symbol_table::create_edge): Likewise.
228         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
229         stmt_can_throw_internal.
230         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
231         to stmt_could_throw_p.
232         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
233         stmt_can_throw_internal.
234         (pass_store_merging::execute): Likewise.
235         * gimple-ssa-strength-reduction.c
236         (find_candidates_dom_walker::before_dom_children): Pass cfun to
237         stmt_could_throw_p.
238         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
239         stmt_can_throw_internal.
240         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
241         to stmt_can_throw_external.
242         (check_stmt): Pass cfun to stmt_could_throw_p.
243         (check_stmt): Pass cfun to stmt_can_throw_external.
244         (pass_nothrow::execute): Likewise.
245         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
246         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
247         stmt_can_throw_internal.
248         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
249         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
250         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
251         * tree-complex.c (expand_complex_libcall): Pass cfun to
252         stmt_could_throw_p and to stmt_can_throw_internal.
253         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
254         * tree-inline.c (copy_edges_for_bb): Likewise.
255         (maybe_move_debug_stmts_to_successors): Likewise.
256         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
257         stmt_could_throw_p.
258         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
259         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
260         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
261         stmt_can_throw_internal.
262         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
263         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
264         stmt_could_throw_p.
265         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
266         stmt_can_throw_internal.
267         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
268         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
269         stmt_could_throw_p.
270         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
271         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
272         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
273         (convert_mult_to_fma_1): Likewise.
274         (convert_to_divmod): Likewise.
275         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
276         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
277         * tree-ssa-propagate.c
278         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
279         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
280         (maybe_optimize_range_tests): Likewise.
281         (linearize_expr_tree): Likewise.
282         (reassociate_bb): Likewise.
283         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
284         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
285         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
286         (handle_char_store): Likewise.
287         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
288         stmt_can_throw_internal.
289         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
290         stmt_could_throw_p.
291         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
292         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
293         (vectorizable_simd_clone_call): Likewise.
294         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
295         (gimple_stringop_fixed_value): Likewise.
297 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
299         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
300         literal pool references.
301         (s390_check_qrst_address): Adapt to the new behavior of
302         s390_loadrelative_operand_p ().
304 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
306         PR target/72782
307         * config/i386/sse.md (*andnot<mode>3_bcst): New.
309 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
311         PR target/72782
312         * config/i386/sse.md (*<code><mode>3_bcst): New.
314 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
316         PR target/72782
317         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
318         V4DI, V16SI and V8DI.
319         (*sub<mode>3<mask_name>_bcst): New.
320         (*add<mode>3<mask_name>_bcst): Likewise.
322 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
324         PR target/72782
325         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
326         __builtin_ia32_vfnmsubpd512_mask.
327         (_mm512_mask_fnmsub_round_pd): Likewise.
328         (_mm512_fnmsub_pd): Likewise.
329         (_mm512_mask_fnmsub_pd): Likewise.
330         (_mm512_maskz_fnmsub_round_pd): Use
331         __builtin_ia32_vfnmsubpd512_maskz.
332         (_mm512_maskz_fnmsub_pd): Likewise.
333         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
334         (_mm512_mask_fnmsub_round_ps): Likewise.
335         (_mm512_fnmsub_ps): Likewise.
336         (_mm512_mask_fnmsub_ps): Likewise.
337         (_mm512_maskz_fnmsub_round_ps): Use
338         __builtin_ia32_vfnmsubps512_maskz.
339         (_mm512_maskz_fnmsub_ps): Likewise.
340         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
341         __builtin_ia32_vfnmsubpd256_mask.
342         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
343         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
344         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
345         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
346         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
347         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
348         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
349         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
350         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
351         __builtin_ia32_vfnmsubpd.
352         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
353         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
354         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
355         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
356         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
357         * config/i386/i386-builtin.def: Add
358         __builtin_ia32_vfnmsubpd256_mask,
359         __builtin_ia32_vfnmsubpd256_maskz,
360         __builtin_ia32_vfnmsubpd128_mask,
361         __builtin_ia32_vfnmsubpd128_maskz,
362         __builtin_ia32_vfnmsubps256_mask,
363         __builtin_ia32_vfnmsubps256_maskz,
364         __builtin_ia32_vfnmsubps128_mask,
365         __builtin_ia32_vfnmsubps128_maskz,
366         __builtin_ia32_vfnmsubpd512_mask,
367         __builtin_ia32_vfnmsubpd512_maskz,
368         __builtin_ia32_vfnmsubps512_mask,
369         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
370         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
371         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
372         __builtin_ia32_vfnmsubpd256.
373         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
374         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
375         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
376         Likewise.
377         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
378         Likewise.
379         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
380         Likewise.
381         (fmai_vmfnmsub_<mode><round_name>): Likewise.
383 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
385         PR target/72782
386         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
387         __builtin_ia32_vfnmaddpd512_mask.
388         (_mm512_mask_fnmadd_round_pd): Likewise.
389         (_mm512_fnmadd_pd): Likewise.
390         (_mm512_mask_fnmadd_pd): Likewise.
391         (_mm512_maskz_fnmadd_round_pd): Use
392         __builtin_ia32_vfnmaddpd512_maskz.
393         (_mm512_maskz_fnmadd_pd): Likewise.
394         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
395         (_mm512_mask_fnmadd_round_ps): Likewise.
396         (_mm512_fnmadd_ps): Likewise.
397         (_mm512_mask_fnmadd_ps): Likewise.
398         (_mm512_maskz_fnmadd_round_ps): Use
399         __builtin_ia32_vfnmaddps512_maskz.
400         (_mm512_maskz_fnmadd_ps): Likewise.
401         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
402         __builtin_ia32_vfnmaddpd256_mask.
403         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
404         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
405         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
406         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
407         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
408         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
409         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
410         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
411         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
412         __builtin_ia32_vfnmaddpd.
413         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
414         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
415         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
416         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
417         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
418         * config/i386/i386-builtin.def: Add
419         __builtin_ia32_vfnmaddpd256_mask,
420         __builtin_ia32_vfnmaddpd256_maskz,
421         __builtin_ia32_vfnmaddpd128_mask,
422         __builtin_ia32_vfnmaddpd128_maskz,
423         __builtin_ia32_vfnmaddps256_mask,
424         __builtin_ia32_vfnmaddps256_maskz,
425         __builtin_ia32_vfnmaddps128_mask,
426         __builtin_ia32_vfnmaddps128_maskz,
427         __builtin_ia32_vfnmaddpd512_mask,
428         __builtin_ia32_vfnmaddpd512_maskz,
429         __builtin_ia32_vfnmaddps512_mask,
430         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
431         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
432         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
433         __builtin_ia32_vfnmaddpd256.
434         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
435         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
436         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
437         Likewise.
438         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
439         Likewise.
440         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
441         Likewise.
442         (fmai_vmfnmadd_<mode><round_name>): Likewise.
444 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
446         PR target/72782
447         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
448         __builtin_ia32_vfmsubpd512_mask.
449         (_mm512_mask_fmsub_round_pd): Likewise.
450         (_mm512_fmsub_pd): Likewise.
451         (_mm512_mask_fmsub_pd): Likewise.
452         (_mm512_maskz_fmsub_round_pd): Use
453         __builtin_ia32_vfmsubpd512_maskz.
454         (_mm512_maskz_fmsub_pd): Likewise.
455         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
456         (_mm512_mask_fmsub_round_ps): Likewise.
457         (_mm512_fmsub_ps): Likewise.
458         (_mm512_mask_fmsub_ps): Likewise.
459         (_mm512_maskz_fmsub_round_ps): Use
460         __builtin_ia32_vfmsubps512_maskz.
461         (_mm512_maskz_fmsub_ps): Likewise.
462         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
463         __builtin_ia32_vfmsubpd256_mask.
464         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
465         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
466         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
467         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
468         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
469         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
470         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
471         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
472         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
473         __builtin_ia32_vfmsubpd.
474         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
475         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
476         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
477         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
478         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
479         * config/i386/i386-builtin.def: Add
480         __builtin_ia32_vfmsubpd256_mask,
481         __builtin_ia32_vfmsubpd256_maskz,
482         __builtin_ia32_vfmsubpd128_mask,
483         __builtin_ia32_vfmsubpd128_maskz,
484         __builtin_ia32_vfmsubps256_mask,
485         __builtin_ia32_vfmsubps256_maskz,
486         __builtin_ia32_vfmsubps128_mask,
487         __builtin_ia32_vfmsubps128_maskz,
488         __builtin_ia32_vfmsubpd512_mask,
489         __builtin_ia32_vfmsubpd512_maskz,
490         __builtin_ia32_vfmsubps512_mask,
491         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
492         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
493         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
494         __builtin_ia32_vfmsubpd256.
495         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
496         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
497         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
498         Likewise.
499         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
500         Likewise.
501         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
502         Likewise.
503         (fmai_vmfmsub_<mode><round_name>): Likewise.
505 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
507         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
508         Remove plus.  Renamed to ...
509         (*sub<mode>3<mask_name>_bcst): This.
510         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
511         (*add<mode>3<mask_name>_bcst): This.
513 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
515         PR target/72782
516         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
518 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
520         PR target/87662
521         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
522         (_mm_or_epi32): Likewise.
523         (_mm256_xor_epi32): Likewise.
524         (_mm_xor_epi32): Likewise.
525         (_mm256_or_epi64): Likewise.
526         (_mm_or_epi64): Likewise.
527         (_mm256_xor_epi64): Likewise.
528         (_mm_xor_epi64): Likewise.
530 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
532         PR target/72782
533         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
535 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
537         PR middle-end/87647
538         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
540 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
542         * doc/ux.texi: Move @section directly after @node.
544 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
546         PR middle-end/85488
547         PR middle-end/87649
548         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
549         depend closely nested inside of loop with ordered clause with
550         a parameter.
552 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
554         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
555         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
556         * doc/ux.texi: New file.
558 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
560         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
561         be the first CR field allocated.
563 2018-10-19  Richard Biener  <rguenther@suse.de>
565         PR target/87657
566         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
567         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
569 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
571         PR target/72782
572         * config/i386/sse.md
573         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
574         (*add<mode>3<mask_name>_bcst_2): Likewise.
576 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
578         * config/i386/sse.md
579         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
580         Replace nonimmediate_operand with register_operand.
581         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
582         Likewise.
583         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
584         Likewise.
586 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
588         PR rtl-optimization/87596
589         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
590         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
591         for instructions in FROM..TO range.
593 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
595         * cfgexpand.c (expand_one_var): Use specific wording in error message
596         for non-local frame variables.
597         * stor-layout.c (layout_decl): Do not issue a warning for them.
599 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
601         * haifa-sched.c (priority): Add force_recompute parameter.
602         (apply_replacement): Call priority () with force_recompute = true.
603         (restore_pattern): Likewise.
605 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
607         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
608         HOST_BITS_PER_WIDE_INT.
609         (test_vector_ops_duplicate): Likewise.
611 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
613         PR target/72782
614         * config/i386/sse.md (VF_AVX512): New.
615         (avx512bcst): Likewise.
616         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
617         Likewise.
618         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
619         Likewise.
620         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
621         Likewise.
623 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
625         * doc/invoke.texi (-dumpversion): Improve grammar.
626         (-dumpfullversion): Make more consistent with -dumpversion.
628 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
630         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
631         Set JUMP_LABEL to the jump insn.
632         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
633         Predict emitted jump and add label to jump insn.
635 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
637         PR tree-optimization/87562
638         * input.c (get_substring_ranges_for_loc): Use
639         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
640         getting the linemap for the endpoint.  Verify that it's either
641         in the same linemap as the start point's spelling location, or
642         at least in the same file.
644 2018-10-18  Richard Biener  <rguenther@suse.de>
646         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
647         feed width-specific load/store costs through ix86_vec_cost.
648         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
649         (k8_cost): Likewise.
650         (bdver_cost): Likewise.
651         (znver1_cost): Likewise.
652         (btver1_cost): Likewise.
653         (btver2_cost): Likewise.
655 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
657         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
658         to simplify subreg of vec_merge.
660 2018-10-18  Richard Biener  <rguenther@suse.de>
662         * config/i386/i386.c: Fix costing of vector FMA.
664 2018-10-18  Richard Biener  <rguenther@suse.de>
666         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
667         and argument.
668         (ix86_builtin_vectorization_cost): For vec_construct properly
669         cost insertion into SSE regs.
670         (...): Adjust calls to ix86_vec_cost.
672 2018-10-18  Richard Biener  <rguenther@suse.de>
674         PR middle-end/87087
675         Revert
676         2018-02-07  Richard Biener  <rguenther@suse.de>
678         PR tree-optimization/84204
679         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
680         this place.
682 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
684         PR target/87537
685         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
686         of vec_duplicate.
687         (test_vector_ops_duplicate): Add test for a scalar subreg of a
688         VEC_MERGE of a VEC_DUPLICATE.
690 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
692         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
693         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
694         * doc/standards.texi (C Language): Document C2X.
695         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
696         (rl78_option_override): Handle "GNU C2X" language name.
698 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
700         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
701         Document C17 as published in 2018.
703 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
705         PR middle-end/87623
706         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
707         bail out if both sides do not have the same storage order.
709 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
711         * bitmap.c (bitmap_head::dump): New.
712         * bitmap.h (bitmap_head): Add dump().
713         * gimple-ssa-evrp-analyze.c
714         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
715         (evrp_range_analyzer::set_ssa_range_info): Same.
716         (evrp_range_analyzer::record_ranges_from_phis): Same.
717         (evrp_range_analyzer::record_ranges_from_stmt): Same.
718         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
719         * gimple-ssa-sprintf.c (get_int_range): Same.
720         (format_integer): Same.
721         (sprintf_dom_walker::handle_gimple_call): Same.
722         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
723         (ipcp_vr_lattice::top_p): Same.
724         (ipcp_vr_lattice::bottom_p): Same.
725         (ipcp_vr_lattice::set_to_bottom): Same.
726         (ipa_vr_operation_and_type_effects): Same.
727         (propagate_vr_across_jump_function): Same.
728         (ipcp_store_vr_results): Same.
729         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
730         (ipa_print_node_jump_functions_for_edge): Same.
731         (ipa_get_value_range): Same.
732         (ipa_compute_jump_functions_for_edge): Same.
733         (ipa_write_jump_function): Same.
734         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
735         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
736         Same.
737         * vr-values.c (set_value_range_to_nonnegative): Same.
738         (set_value_range_to_truthvalue): Same.
739         (vr_values::get_value_range): Same.
740         (vr_values::set_defs_to_varying): Same.
741         (vr_values::update_value_range): Same.
742         (symbolic_range_based_on_p): Same.
743         (vr_values::op_with_boolean_value_range_p): Same.
744         (vr_values::extract_range_for_var_from_comparison_expr): Same.
745         (vr_values::extract_range_from_ssa_name): Same.
746         (vr_values::extract_range_from_binary_expr): Same.
747         (vr_values::extract_range_from_unary_expr): Same.
748         (vr_values::extract_range_from_cond_expr): Same.
749         (vr_values::extract_range_from_comparison): Same.
750         (vr_values::check_for_binary_op_overflow): Same.
751         (vr_values::extract_range_basic): Same.
752         (vr_values::extract_range_from_assignment): Same.
753         (compare_ranges): Same.
754         (compare_range_with_value): Same.
755         (vr_values::adjust_range_with_scev): Same.
756         (vrp_valueize): Same.
757         (vrp_valueize_1): Same.
758         (vr_values::get_vr_for_comparison): Same.
759         (vr_values::compare_name_with_value): Same.
760         (vr_values::compare_names): Same.
761         (vr_values::vrp_evaluate_conditional): Same.
762         (find_case_label_ranges): Same.
763         (vr_values::vrp_visit_switch_stmt): Same.
764         (vr_values::extract_range_from_phi_node): Same.
765         (vr_values::simplify_div_or_mod_using_ranges): Same.
766         (vr_values::simplify_bit_ops_using_ranges): Same.
767         (test_for_singularity): Same.
768         (range_fits_type_p): Same.
769         (vr_values::simplify_cond_using_ranges_1): Same.
770         (vr_values::simplify_switch_using_ranges): Same.
771         (vr_values::simplify_float_conversion_using_ranges): Same.
772         (vr_values::two_valued_val_range_p): Same.
773         (vr_values::add_equivalence): Move to value_range::equiv_add.
774         * vr-values.h (vr_values::add_equivalence): Remove.
775         (VR_INITIALIZER): Remove.
776         * tree-vrp.c (value_range::set): New.
777         (value_range::equiv_add): New.
778         (value_range::value_range): New.
779         (value_range::deep_copy): New.
780         (value_range::check): New.
781         (value_range::equal_p): New.
782         (value_range::ignore_equivs_equal_p): New.
783         (value_range::operator==): New.
784         (value_range::operator!=): New.
785         (value_range::symbolic_p): New.
786         (value_range::numeric_p): New.
787         (value_range::set_undefined): New.
788         (value_range::set_varying): New.
789         (value_range::may_contain_p): New.
790         (value_range::equiv_clear): New.
791         (value_range::singleton_p): New.
792         (value_range::intersect): New.
793         (value_range::dump): New.
794         (value_range::set_and_canonicalize): New.
795         (set_value_range): Adjust for value_range API.
796         (set_value_range_to_undefined): Same.
797         (set_value_range_to_varying): Same.
798         (set_and_canonicalize_value_range): Same.
799         (set_value_range_to_nonnull): Same.
800         (set_value_range_to_null): Same.
801         (range_is_null): Same.
802         (range_is_nonnull): Same.
803         (range_int_cst_p): Same.
804         (range_int_cst_singleton_p): Same.
805         (symbolic_range_p): Same.
806         (range_includes_zero_p): Same.
807         (value_range_constant_singleton): Same.
808         (vrp_set_zero_nonzero_bits): Same.
809         (ranges_from_anti_range): Same.
810         (extract_range_into_wide_ints): Same.
811         (extract_range_from_multiplicative_op): Same.
812         (set_value_range_with_overflow): Same.
813         (extract_range_from_binary_expr_1): Same.
814         (extract_range_from_unary_expr): Same.
815         (dump_value_range): Same.
816         (debug_value_range): Same.
817         (vrp_prop::check_array_ref): Same.
818         (vrp_prop::check_mem_ref): Same.
819         (vrp_prop::vrp_initialize): Same.
820         (vrp_prop::visit_stmt): Same.
821         (intersect_ranges): Same.
822         (vrp_prop::visit_phi): Same.
823         (vrp_prop::vrp_finalize): Same.
824         (determine_value_range_1): Same.
825         (determine_value_range): Same.
826         (vrp_intersect_ranges_1): Rename to...
827         (vrp_intersect_1): this.
828         (vrp_intersect_ranges): Rename to...
829         (value_range::intersect_helper): ...this.
830         (vrp_meet_1): Rename to...
831         (value_range::union_helper): ...this.
832         (vrp_meet): Rename to...
833         (value_range::union_): ...this.
834         (copy_value_range): Remove.
835         * tree-vrp.h (struct value_range): Rewrite into a proper class.
836         (value_range::vrtype): New.
837         (value_range::type): New.
838         (value_range::equiv): New.
839         (value_range::min): New.
840         (value_range::max): New.
841         (value_range::varying_p): New.
842         (value_range::undefined_p): New.
843         (value_range::null_p): New.
844         (value_range::equiv_add): New.
845         (copy_value_range): Remove.
847 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
849         * Makefile.in (SELFTEST_TARGETS): New.
850         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
851         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
852         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
853         c/Make-lang.in.
854         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
855         (selftest-c++-gdb, selftest-c++-valgrind): Move to
856         cp/Make-lang.in.
857         * configure: Regenerate.
858         * configure.ac (selftest_languages): New.
860 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
862         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
863         overflow wraps argument.
864         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
865         wide_int_range_multiplicative_op.
866         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
867         overflow wraps argument.
868         (wide_int_range_multiplicative_op): Same.
869         (wide_int_range_lshift): Same.
870         (wide_int_range_div): Same.
871         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
872         (wide_int_range_lshift): Same.
873         (wide_int_range_div): Same.
875 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
877         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
878         use sign as argument.
879         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
880         wide_int_range_shift_undefined_p.
882 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
884         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
885         Rename to...
886         (@despeculate_copy<ALLI_TI:mode>): ... This.
887         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
888         switch statement.
890 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
892         * config.gcc: Obsolete *-*-solaris2.10*.
893         * doc/install.texi (Specific, *-*-solaris2*): Document it.
895 2018-10-12  Jeff Law  <law@redhat.com>
897         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
898         reg + sym +- const_int addressing modes.
900 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
902         * common.opt (fdiagnostics-minimum-margin-width=): New option.
903         * diagnostic-show-locus.c (layout::layout): Apply the minimum
904         margin width.
905         (layout::start_annotation_line): Only print up to 3 of the
906         margin character, to avoid touching the left-hand side.
907         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
908         minimum margin width, as set by test_diagnostic_context's ctor.
909         (selftest::test_fixit_insert_containing_newline): Likewise.
910         (selftest::test_fixit_insert_containing_newline_2): Likewise.
911         (selftest::test_line_numbers_multiline_range): Clear
912         dc.min_margin_width.
913         * diagnostic.c (diagnostic_initialize): Initialize
914         min_margin_width.
915         * diagnostic.h (struct diagnostic_context): Add field
916         "min_margin_width".
917         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
918         * opts.c (common_handle_option): Handle
919         OPT_fdiagnostics_minimum_margin_width_.
920         * selftest-diagnostic.c
921         (selftest::test_diagnostic_context::test_diagnostic_context):
922         Initialize min_margin_width to 6.
923         * toplev.c (general_init): Initialize global_dc->min_margin_width.
925 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
927         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
928         Fix usage of "error_at_rich_loc" in the comment.
930 2018-10-15  Renlin Li  <renlin.li@arm.com>
932         PR target/87563
933         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
934         if-conversioned loop when it contains ifn with types not
935         supported by backend.
936         * internal-fn.c (expand_direct_optab_fn): Add an assert.
937         (direct_internal_fn_supported_p): New helper function.
938         * internal-fn.h (direct_internal_fn_supported_p): Declare.
940 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
942         PR target/87572
943         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
944         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
945         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
947 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
949         PR tree-optimization/87022
950         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
951         bits in dist vector rather than the first one.
953 2018-10-15  Richard Biener  <rguenther@suse.de>
955         PR middle-end/87610
956         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
957         (visit_loadstore): When a used restrict tag escaped verify that
958         the points-to solution of "other" pointers do not include
959         escaped.
960         (compute_dependence_clique): If a used restrict tag escaped
961         communicated that down to visit_loadstore.
963 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
965         * config/s390/s390.c (s390_expand_vec_init): Force vector element
966         into reg if it isn't a general operand.
968 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
970         PR target/87599
971         * config/i386/sse.md (*vec_dupv2di): Add register source to
972         movddup.
974 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
976         PR target/87572
977         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
978         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
979         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
980         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
982 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
984         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
985         (notice_args_size): Set it in the current trace if no insn that can
986         throw internally has been seen yet.
987         (connect_traces): When connecting args_size between traces, allow the
988         incoming values not to match if there is an insn setting it before the
989         first insn that can throw internally; in that case, force the creation
990         of a CFI note on this latter insn.
992 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
994         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
995         the base class, because using the injected-class-name was not clearly
996         specified until DR 176.
998 2018-10-12  Paul Koning  <ni1d@arrl.net>
1000         * config/pdp11/pdp11.md (doloop_end): New expander.
1001         (doloop_end_insn): renamed from "doloop_end".
1002         (addqi3): New pattern.
1003         (subqi3): New pattern.
1004         * config/pdp11/predicates.md (incdec_operand): New predicate.
1006 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
1008         PR middle-end/81376
1009         * real.c (format_helper::can_represent_integral_type_p): New function
1010         * real.h (format_helper::can_represent_integral_type_p): Ditto.
1011         * match.pd: New pattern.
1013 2018-10-12  Alexandre Oliva <oliva@adacore.com>
1015         * configure.ac: Introduce --enable-large-address-aware
1016         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
1017         * doc/install.texi: Document it.
1018         * configure, config.in: Rebuilt.
1019         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
1020         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
1021         (LINK_SPEC): Insert it.
1022         * config/i386/mingw-264.h: Likewise.
1024         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
1026 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
1028         PR rtl-optimization/87600
1029         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
1031 2018-10-12  Paul Koning  <ni1d@arrl.net>
1033         * doc/md.texi (doloop_end): Document that the pattern code may
1034         need to check operand mode.
1036 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
1038         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
1039         to zero-extend between int and floating-point registers.
1040         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
1041         ldp into floating-point registers.  Add type and arch attributes.
1042         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
1043         Use f_loads for type attribute.
1045 2018-10-11  Martin Sebor  <msebor@redhat.com>
1047         * doc/extend.texi (attribute packed): Correct typos.
1049 2018-10-11  Martin Sebor  <msebor@redhat.com>
1051         * doc/extend.texi (attribute flatten): Mention interaction with
1052         noinline.
1054 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1056         PR target/87156
1057         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1058         Set new_decl virtual flag to zero.
1060 2018-10-11  Martin Sebor  <msebor@redhat.com>
1062         PR middle-end/87593
1063         * doc/extend.texi (attribute format_arg): Discuss using multiple
1064         attributes on a single function.
1066 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1068         PR tree-optimization/86829
1069         * match.pd (sin (atan (x))): New simplification rules.
1070         (cos (atan (x))): Likewise.
1071         * real.c (build_sinatan_real): New function.
1072         * real.h (build_sinatan_real): Prototype.
1074 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
1076         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
1077         function.
1078         (fold_mergeeo_helper): New helper function.
1079         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
1080         intrinsics.  Correct some whitespace indentation issues.
1082 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
1084         PR target/87511
1085         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
1086         Use HOST_WIDE_INT_1U for shift.
1088 2018-10-11  Doug Rupp  <rupp@adacore.com>
1089             Olivier Hainque  <hainque@adacore.com>
1091         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
1092         Pass --relax to the linker for RTPs.
1093         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
1095 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
1096             Jan Hubicka  <jh@suse.cz>
1097             Martin Jambor  <mjambor@suse.cz>
1099         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
1100         the same elements are repeated rather than printing all of them.
1101         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
1102         "repeated" elements.
1103         * read-rtl-function.c (test_loading_repeat): New function.
1104         (read_rtl_function_c_tests): Call test_loading_repeat.
1105         * rtl-tests.c (test_dumping_repeat): New function.
1106         (rtl_tests_c_tests): Call test_dumping_repeat.
1108 2018-10-11  Richard Biener  <rguenther@suse.de>
1110         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
1111         bdver?_cost): Unify to ...
1112         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
1113         * config/i386/i386.c (processor_cost_table): Adjust.
1115 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
1117         PR middle-end/87574
1118         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1119         the thunk when expanding to GIMPLE.
1121 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1123         * varasm.c (mergeable_string_section): Don't try to move zero-length
1124         strings to the merge section.
1126 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
1128         PR target/87573
1129         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1131 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
1133         PR target/87550
1134         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1135         to special_args set.
1137 2018-10-10  Richard Biener  <rguenther@suse.de>
1139         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1140         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1141         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1142         and recursing and pattern terminating the recursion on SSE
1143         vector width using ix86_expand_reduc.
1144         (reduc_sminmax_scal_<mode>): Split into part reducing to half
1145         width and recursing and SSE2 vector variant doing the final
1146         reduction with ix86_expand_reduc.
1147         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1148         with terminating the recursion at AVX level, splitting that
1149         to SSE there.
1151 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
1153         * genmatch.c (error_cb): Rename to...
1154         (diagnostic_cb): ...this, converting int params to enums.
1155         (fatal_at): Update for renaming.
1156         (warning_at): Likewise.
1157         (main): Likewise.
1158         * input.c (selftest::ebcdic_execution_charset::apply):
1159         Update for renaming of...
1160         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1161         to...
1162         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1163         converting level and reason to enums.
1164         (class selftest::lexer_error_sink): Rename to...
1165         (class selftest::lexer_test_options): ...this, renaming field
1166         "m_errors" to "m_diagnostics".
1167         (selftest::lexer_test_options::apply): Update for renaming of...
1168         (selftest::lexer_test_options::on_error): ...this, renaming to...
1169         (selftest::lexer_test_options::on_diagnostic): ...this
1170         converting level and reason to enums.
1171         (selftest::test_lexer_string_locations_raw_string_unterminated):
1172         Update for renamings.
1173         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
1174         "reason".
1176 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
1178         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1179         * config/rs6000/pmmintrin.h: New file.
1181 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
1183         PR tree-optimization/86659
1184         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1186 2018-10-09  Richard Biener  <rguenther@suse.de>
1188         PR tree-optimization/63155
1189         * tree-ssa-structalias.c: Include tree-ssa.h.
1190         (get_constraint_for_ssa_var): For undefs return nothing_id.
1191         (find_func_aliases): Cleanup PHI handling.
1193 2018-10-09  Richard Biener  <rguenther@suse.de>
1195         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1196         replacements.
1198 2018-10-09  Martin Liska  <mliska@suse.cz>
1200         * asan.c (asan_emit_stack_protection): If a stack variable
1201         is located in a same file as current function, then emit
1202         line info into variable definition string.
1204 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1206         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1207         information.
1209 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1211         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1212         on the thunk.
1214 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1216         PR tree-optimization/86659
1217         * gimple-match.h (struct gimple_match_op): Add reverse field.
1218         (gimple_match_op::set_op): New overloaded method.
1219         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1220         the REF_REVERSE_STORAGE_ORDER flag on the value.
1221         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
1222         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
1223         
1224 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
1226         PR middle-end/63155
1227         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
1228         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
1230 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
1232         PR target/87517
1233         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
1234         Defined with __builtin_ia32_vfmaddsubpd512_mask.
1236 2018-10-08  Richard Biener  <rguenther@suse.de>
1238         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
1239         cost the same as AVX128 ones.
1241 2018-10-08  Paul Koning  <ni1d@arrl.net>
1243         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
1244         (expand_block_move): New function.
1245         * config/pdp11/pdp11.c (output_block_move): Remove.
1246         (expand_block_move): New function.
1247         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
1248         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
1249         (*movmemhi1): Remove.
1251 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
1253         * config/s390/2827.md: Increase latencies for some FP instructions.
1255 2018-10-08  Richard Biener  <rguenther@suse.de>
1257         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1258         Open a dump scope.
1259         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
1260         * tree-vectorizer.h (dump_stmt_cost): Adjust.
1261         (add_stmt_cost): Dump return value of the hook.
1263 2018-10-08  Richard Biener  <rguenther@suse.de>
1265         PR tree-optimization/63155
1266         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
1267         (ssa_propagation_engine::ssa_propagate): Remove redundant
1268         bitmap bit clearing.
1270 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
1272         PR rtl-optimization/86939
1273         PR rtl-optimization/87479
1274         * ira.h (non_conflicting_reg_copy_p): New prototype.
1275         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
1276         (make_hard_regno_dead): Don't add conflicts for register
1277         ignore_reg_for_conflicts.
1278         (make_object_dead): Likewise.
1279         (non_conflicting_reg_copy_p): New function.
1280         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
1281         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
1282         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
1283         (make_hard_regno_dead): Don't add conflicts for register
1284         ignore_reg_for_conflicts.  Remove special conflict handling of
1285         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
1286         check_pic_pseudo_p and update callers.
1287         (mark_pseudo_dead): Don't add conflicts for register
1288         ignore_reg_for_conflicts.
1289         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
1291 2018-10-05  Andrew Waterman  <andrew@sifive.com>
1292             Jim Wilson  <jimw@sifive.com>
1294         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
1295         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
1296         new pattern using HONOR_SNANS that emits one extra instruction.
1298 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1300         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
1301         patterns): Merge SI and DI patterns to a GPR pattern.
1302         (unnamed define_insn and define_split for record form of that): Merge
1303         to a single define_insn_and_split pattern.
1305 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
1307         PR c++/56856
1308         * input.c (expand_location_to_spelling_point): Add param "aspect"
1309         and use rather than hardcoding LOCATION_ASPECT_CARET.
1310         (get_substring_ranges_for_loc): Handle the case of a single token
1311         within a macro expansion.
1312         * input.h (expand_location_to_spelling_point): Add "aspect" param,
1313         defaulting to LOCATION_ASPECT_CARET.
1315 2018-10-05  Paul Koning  <ni1d@arrl.net>
1317         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
1318         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
1319         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1320         (pdp11_guard_type): New function.
1322 2018-10-05  Paul Koning  <ni1d@arrl.net>
1324         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
1325         * config/pdp11/pdp11.opt (mfloat32): Remove.
1326         (mfloat64): Remove.
1327         * doc/invoke.texi (pdp11 -mfloat32): Remove:
1328         (pdp11 -mfloat64): Remove.
1330 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1332         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
1333         (*cmp<mode>_cc_i387): Ditto.
1334         (*cmpu<mode>_cc_i387): Ditto.
1335         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1336         * config/i386/i386.c (ix86_expand_fp_compare): Remove
1337         "scratch" argument.
1338         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
1339         Emit x86_sahf_1 pattern.
1340         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
1341         (ix86_expand_carry_flag_compare): Ditto.
1343 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1345         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
1346         to reg_or_0_operand.  Add "C" constraint.
1347         (*cmpxf_cc_i387): Ditto.
1348         (*cmp<mode>_i387): Change operand 2 predicate
1349         to nonimm_or_0_operand.  Add "C" constraint.
1350         (*cmp<mode>_cc_i387): Ditto.
1351         (*cmp<mode>_0_i387): Remove insn pattern.
1352         (*cmp<mode>_0_cc_i387): Ditto.
1354 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1356         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
1357         * config/i386/predicates.md (nonimm_or_0_operand): Rename
1358         from vector_move_operand.  Update all uses.
1360 2018-10-05  Martin Sebor  <msebor@redhat.com>
1362         PR tree-optimization/87490
1363         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
1364         consistently.
1366 2018-10-05  Richard Biener  <rguenther@suse.de>
1368         PR tree-optimization/63155
1369         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
1370         vertical space in dumpfiles.
1371         * tree-ssa-propagate.h
1372         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
1373         * tree-ssa-propagate.c (cfg_blocks_back): New global.
1374         (ssa_edge_worklist_back): Likewise.
1375         (curr_order): Likewise.
1376         (cfg_blocks_get): Remove abstraction.
1377         (cfg_blocks_add): Likewise.
1378         (cfg_blocks_empty_p): Likewise.
1379         (add_ssa_edge): Add to current or next worklist based on
1380         RPO index.
1381         (add_control_edge): Likewise.
1382         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
1383         into ...
1384         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
1385         iteration from CFG and SSA edge worklist so we process
1386         everything in RPO order, prioritizing forward progress
1387         over iteration.
1388         (ssa_prop_init): Allocate new worklists, do not dump
1389         immediate uses.
1390         (ssa_prop_fini): Free new worklists.
1392 2018-10-05  Richard Biener  <rguenther@suse.de>
1394         * tree-core.h (tree_block::abstract_flag): Remove.
1395         (tree_block::block_num): Make full 32bits.
1396         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
1397         * tree.h (BLOCK_ABSTRACT): Remove.
1398         * dwarf2out.c (gen_lexical_block_die): Remove dead code
1399         resulting from BLOCK_ABSTRACT being always false.
1400         (gen_inlined_subroutine_die): Likewise.
1401         (gen_block_die): Likewise.
1402         * tree.c (block_ultimate_origin): Likewise.
1403         * tree-pretty-print.c (dump_block_node): Remove code dealing
1404         with BLOCK_ABSTRACT.
1405         * tree-ssa-live.c (dump_scope_block): Likewise.
1406         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
1407         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
1409 2018-10-05   Richard Biener  <rguenther@suse.de>
1411         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
1412         is asked for initialize mode to the component mode of the
1413         vector type.
1415 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
1417         PR target/87522
1418         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
1419         assembler for -mavx.
1420         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
1422 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1424         PR target/87509
1425         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
1426         RS6000_BTM_DFP.
1427         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
1428         to be DImode.  When using mffscrn, force the operand to a register.
1430 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
1432         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
1433         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
1434         X87MODEF mode iterator.
1435         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
1436         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
1437         X87MODEF mode iterator.
1439 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
1441         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
1442         -Wno-prio-ctor-dtor.
1444 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1446         * Makefile.in (OBJS): Add opt-problem.o.
1447         * dump-context.h: Include "selftest.h.
1448         (selftest::temp_dump_context): New forward decl.
1449         (class dump_context): Make friend of class
1450         selftest::temp_dump_context.
1451         (dump_context::dump_loc_immediate): New decl.
1452         (class dump_pretty_printer): Move here from dumpfile.c.
1453         (class temp_dump_context): Move to namespace selftest.
1454         (temp_dump_context::temp_dump_context): Add param
1455         "forcibly_enable_dumping".
1456         (selftest::verify_dumped_text):
1457         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
1458         (selftest::verify_item):
1459         (ASSERT_IS_TEXT): Move here from dumpfile.c.
1460         (ASSERT_IS_TREE): Likewise.
1461         (ASSERT_IS_GIMPLE): Likewise.
1462         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
1463         to...
1464         (dump_context::dump_loc_immediate): ...this new function.
1465         (class dump_pretty_printer): Move to dump-context.h.
1466         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
1467         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
1468         (temp_dump_context::temp_dump_context): Move to "selftest"
1469         namespace.  Add param "forcibly_enable_dumping", and use it to
1470         conditionalize the use of m_pp;
1471         (selftest::verify_dumped_text): Make non-static.
1472         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
1473         (selftest::verify_item): Make non-static.
1474         (ASSERT_IS_TEXT): Move to dump-context.h.
1475         (ASSERT_IS_TREE): Likewise.
1476         (ASSERT_IS_GIMPLE): Likewise.
1477         (selftest::test_capture_of_dump_calls): Pass "true" for new
1478         param of temp_dump_context.
1479         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
1480         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
1481         TDF_COMPARE_DEBUG.
1482         * opt-problem.cc: New file.
1483         * opt-problem.h: New file.
1484         * optinfo-emit-json.cc
1485         (selftest::test_building_json_from_dump_calls): Pass "true" for
1486         new param of temp_dump_context.
1487         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
1488         (optinfo::emit_for_opt_problem): New function.
1489         (optinfo::emit): Clarity which emit_item is used.
1490         * optinfo.h (optinfo::get_dump_location): New accessor.
1491         (optinfo::emit_for_opt_problem): New decl.
1492         (optinfo::emit): Make const.
1493         * selftest-run-tests.c (selftest::run_tests): Call
1494         selftest::opt_problem_cc_tests.
1495         * selftest.h (selftest::opt_problem_cc_tests): New decl.
1496         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
1497         bool to opt_result, converting fprintf messages to
1498         opt_result::failure_at calls.  Add "stmt" param for use by the
1499         failure_at calls.
1500         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
1501         (runtime_alias_check_p): Convert return type from bool to
1502         opt_result, converting dump_printf calls to
1503         opt_result::failure_at, using the statement DDR_A for their
1504         location.
1505         (find_data_references_in_stmt): Convert return type from bool to
1506         opt_result, converting "return false" to opt_result::failure_at
1507         with a new message.
1508         * tree-data-ref.h: Include "opt-problem.h".
1509         (dr_analyze_innermost): Convert return type from bool to opt_result,
1510         and add a const gimple * param.
1511         (find_data_references_in_stmt): Convert return type from bool to
1512         opt_result.
1513         (runtime_alias_check_p): Likewise.
1514         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
1515         dr_analyze_innermost.
1516         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
1517         Convert return type from bool to opt_result, adding a message for
1518         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
1519         (vect_analyze_data_ref_dependence): Convert return type from bool
1520         to opt_result.  Change sense of return type from "false"
1521         effectively meaning "no problems" to "false" meaning a problem,
1522         so that "return false" becomes "return opt_result::success".
1523         Convert "return true" calls to opt_result::failure_at, using
1524         the location of statement A rather than vect_location.
1525         (vect_analyze_data_ref_dependences): Convert return type from bool
1526         to opt_result.
1527         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
1528         calls to opt_result::failure_at, using the stmt location rather
1529         than vect_location.
1530         (vect_verify_datarefs_alignment): Convert return type from bool
1531         to opt_result.
1532         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
1533         into multiple more-tightly-scoped copies.
1534         (vect_analyze_data_refs_alignment): Convert return type from bool
1535         to opt_result.
1536         (vect_analyze_data_ref_accesses): Likewise, converting a
1537         "return false" to a "return opt_result::failure_at", adding a
1538         new message.
1539         (vect_prune_runtime_alias_test_list): Convert return type from
1540         bool to opt_result, converting dump_printf_loc to
1541         opt_result::failure_at.  Add a %G to show the pertinent statement,
1542         and use the stmt's location rather than vect_location.
1543         (vect_find_stmt_data_reference): Convert return type from
1544         bool to opt_result, converting dump_printf_loc to
1545         opt_result::failure_at, using stmt's location.
1546         (vect_analyze_data_refs):  Convert return type from bool to
1547         opt_result.  Convert "return false" to "return
1548         opt_result::failure_at", adding messages as needed.
1549         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
1550         type from bool to opt_result.
1551         (vect_determine_vf_for_stmt): Likewise.
1552         (vect_determine_vectorization_factor): Likewise, converting
1553         dump_printf_loc to opt_result::failure_at, using location of phi
1554         rather than vect_location.
1555         (vect_analyze_loop_form_1): Convert return type from bool to
1556         opt_result, converting dump_printf_loc calls, retaining the use of
1557         vect_location.
1558         (vect_analyze_loop_form): Convert return type from loop_vec_info
1559         to opt_loop_vec_info.
1560         (vect_analyze_loop_operations): Convert return type from bool to
1561         opt_result, converting dump_printf_loc calls, using the location
1562         of phi/stmt rather than vect_location where available.  Convert
1563         various "return false" to "return opt_result::failure_at" with
1564         "unsupported phi" messages.
1565         (vect_get_datarefs_in_loop): Convert return type from bool to
1566         opt_result.  Add a message for the
1567         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
1568         (vect_analyze_loop_2): Convert return type from bool to
1569         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
1570         each "goto again;", adding new messages where needed.
1571         Add "unsupported grouped {store|load}" messages.
1572         (vect_analyze_loop): Convert return type from loop_vec_info to
1573         opt_loop_vec_info.
1574         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
1575         bool to opt_result.
1576         * tree-vect-stmts.c (process_use): Likewise, converting
1577         dump_printf_loc call and using stmt location, rather than
1578         vect_location.
1579         (vect_mark_stmts_to_be_vectorized): Likeise.
1580         (vect_analyze_stmt): Likewise, adding a %G.
1581         (vect_get_vector_types_for_stmt): Convert return type from bool to
1582         opt_result, converting dump_printf_loc calls and using stmt
1583         location, rather than vect_location.
1584         (vect_get_mask_type_for_stmt): Convert return type from tree to
1585         opt_tree, converting dump_printf_loc calls and using stmt location.
1586         * tree-vectorizer.c: Include "opt-problem.h.
1587         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
1588         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
1589         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
1590         enabled, use it to report at the top level "couldn't vectorize
1591         loop" followed by the problem.
1592         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
1593         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
1594         to opt_result.
1595         (vect_analyze_stmt): Likewise.
1596         (vect_get_vector_types_for_stmt): Likewise.
1597         (tree vect_get_mask_type_for_stmt): Likewise.
1598         (vect_analyze_data_ref_dependences): Likewise.
1599         (vect_enhance_data_refs_alignment): Likewise.
1600         (vect_analyze_data_refs_alignment): Likewise.
1601         (vect_verify_datarefs_alignment): Likewise.
1602         (vect_analyze_data_ref_accesses): Likewise.
1603         (vect_prune_runtime_alias_test_list): Likewise.
1604         (vect_find_stmt_data_reference): Likewise.
1605         (vect_analyze_data_refs): Likewise.
1606         (vect_analyze_loop): Convert return type from loop_vec_info to
1607         opt_loop_vec_info.
1608         (vect_analyze_loop_form): Likewise.
1609         (vect_analyze_slp): Convert return type from bool to opt_result.
1611 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1613         * doc/invoke.texi (-fopt-info): Document new "internals"
1614         sub-option.
1615         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
1616         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
1617         MSG_ALL_KINDS.
1618         (optinfo_verbosity_options): Add "internals".
1619         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
1620         (dump_context::apply_dump_filter_p): New member function.
1621         (dump_context::dump_loc): Use apply_dump_filter_p rather than
1622         explicitly masking the dump_kind.
1623         (dump_context::begin_scope): Increment the scope depth first.  Use
1624         apply_dump_filter_p rather than explicitly masking the dump_kind.
1625         (dump_context::emit_item): Use apply_dump_filter_p rather than
1626         explicitly masking the dump_kind.
1627         (dump_dec): Likewise.
1628         (dump_hex): Likewise.
1629         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
1630         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
1631         (opt_info_switch_p): Update handling of default
1632         MSG_OPTIMIZED_LOCATIONS to cope with default of
1633         MSG_PRIORITY_USER_FACING.
1634         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
1635         masking the dump_kind.
1636         (selftest::test_capture_of_dump_calls): Update test_dump_context
1637         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
1638         than MSG_ALL.  Generalize scope test to be run at all four
1639         combinations of with/without MSG_PRIORITY_USER_FACING and
1640         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
1641         for each of the two values.
1642         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
1643         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
1644         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
1645         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
1646         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
1647         with MSG_PRIORITY_*.
1648         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
1649         dump messages as MSG_PRIORITY_USER_FACING.
1650         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
1651         about the interaction with MSG_PRIORITY_*.
1653 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1655         * varasm.c (output_constant): Add new parameter merge_strings.
1656         Make strings properly zero terminated in merge string sections.
1657         (mergeable_string_section): Don't fail if the last char is non-zero.
1658         (assemble_variable_contents): Handle merge string sections.
1659         (assemble_variable): Likewise.
1660         (assemble_constant_contents): Likewise.
1661         (output_constant_def_contents): Likewise.
1662         (output_constructor_array_range,
1663         output_constructor_regular_field): Adjust call to output_constant.
1664         (output_object_block): Adjust call to assemble_constant_contents
1665         and assemble_variable_contents.
1667 2018-10-04  Martin Liska  <mliska@suse.cz>
1669         PR c/87483
1670         * cgraphunit.c (process_function_and_variable_attributes):
1671         Warn about a function with alias attribute and a body.
1673 2018-10-04  Martin Liska  <mliska@suse.cz>
1675         PR ipa/82625
1676         * multiple_target.c (redirect_to_specific_clone): New function.
1677         (ipa_target_clone): Use it.
1678         * tree-inline.c: Fix comment.
1680 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1682         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
1683         fields.
1684         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
1685         (gcc::dump_manager::register_pass): New member function, adapted
1686         from loop body in gcc::pass_manager::register_pass, adding a
1687         call to update_dfi_for_opt_info.
1688         (gcc::dump_manager::opt_info_enable_passes): Store the
1689         -fopt-info options into the new fields.  Move the loop
1690         bodies into...
1691         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
1692         function.
1693         * dumpfile.h (struct opt_pass): New forward decl.
1694         (gcc::dump_manager::register_pass): New decl.
1695         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
1696         (class gcc::dump_manager): Add fields "m_optgroup_flags",
1697         "m_optinfo_flags", and "m_optinfo_filename".
1698         * passes.c (gcc::pass_manager::register_pass): Move all of the
1699         dump-handling code to gcc::dump_manager::register_pass.
1701 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
1703         PR rtl-optimization/87466
1704         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
1705         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
1706         * doc/tm.texi: Regenerate.
1707         * ira-lives.c (process_bb_node_lives): Use the new target hook.
1708         * lra-lives.c (process_bb_lives): Likewise.
1709         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
1710         Define.
1712 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
1714         * params.c (add_params): Fix initialization.
1716 2018-10-04  Martin Liska  <mliska@suse.cz>
1718         PR gcov-profile/84107
1719         * tree-profile.c (init_ic_make_global_vars):
1720         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
1721         Come up with new ic_tuple* variables.  Emit
1722         __gcov_indirect_call{,_topn} variables.
1723         (gimple_gen_ic_profiler): Access the variable
1724         and emit gimple.
1725         (gimple_gen_ic_func_profiler): Access
1726         __gcov_indirect_call.callee field.
1727         (gimple_init_gcov_profiler): Use ptr_type_node.
1728         * value-prof.c (gimple_ic): Use ptr_type_node.
1730 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1732         PR tree-optimization/85787
1733         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
1734         into this function and add support for detecting multiple phis.
1735         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
1737 2018-10-04  Martin Liska  <mliska@suse.cz>
1739         PR ipa/87491
1740         * ipa-inline.c (inline_to_all_callers_1):
1741         Call ultimate_alias_target for node being inlined.
1743 2018-10-03  Jeff Law  <law@redhat.com>
1745         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
1746         target's wchar_t.
1747         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
1748         * tree.h (get_typenode_from_name): Prototype.
1750 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
1752         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
1753         Change operand 2 predicate to nonimmediate_operand.
1754         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1756 2018-10-03  Martin Sebor  <msebor@redhat.com>
1757             Jeff Law  <law@redhat.com>
1759         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
1760         initialize it.
1761         (get_string_length): Detect unterminated arrays.
1762         (format_string): Same.
1763         (format_directive): Warn about unterminated arrays.
1764         (handle_gimple_call): Mark statements with no_warning as needed.
1766 2018-10-03  Jim Wilson  <jimw@sifive.com>
1768         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
1769         also define __riscv_abi_rve.  Delete trailing white space.
1771 2018-10-03  Paul Koning  <ni1d@arrl.net>
1773         Enable LRA register allocator for PDP11.
1774         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
1775         (R): Likewise.
1776         (D): Likewise.
1777         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
1778         * config/pdp11/pdp11.opt (-mlra): New option.
1779         * doc/invoke.texi (PDP-11 Options): Document -mlra.
1781 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
1783         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
1784         (*<absneg:code>extend<mode>xf2): Ditto.
1786 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
1788         PR tree-optimization/87415
1789         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
1790         precision fields.
1792 2018-10-02  Jeff Law  <law@redhat.com>
1794         * gimple-fold.c (get_range_strlen): Only set *nonstr when
1795         an unterminated string is discovered.  Bubble up range
1796         even for unterminated strings.
1797         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
1798         indicates the string was not terminated via NONSTR.
1800 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
1802         * tree-vrp.c (extract_range_from_unary_expr): Special case all
1803         pointer conversions.
1804         Do not do anything special for anti-ranges.
1806 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
1808         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
1809         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
1811 2018-10-03  Martin Liska  <mliska@suse.cz>
1813         PR gcov-profile/86109
1814         * coverage.c (coverage_begin_function): Do not
1815         mark lambdas as artificial.
1816         * tree-core.h (struct GTY): Remove tm_clone_flag
1817         and introduce new lambda_function.
1818         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
1820 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
1822         PR target/87474
1823         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
1824         P8_VECTOR and VSX are enabled.
1826 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
1828         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
1829         0x3907 as CPU model number.
1831 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
1833         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
1834         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
1835         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
1836         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
1837         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
1838         * config/s390/s390.md: Likewise. Rename also the cpu attribute
1839         value from arch12 to z14.
1841 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
1843         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
1844         (isinfxf2): Ditto.
1845         (isinf<mode>2): Ditto.
1847 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
1849         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
1850         before emitting fxam.  Perform calculations in XFmode.
1852 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
1854         * match.pd (((X /[ex] A) +- B) * A): New transformation.
1856 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
1858         PR middle-end/87319
1859         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
1860         * tree.c (signed_or_unsigned_type_for): Handle complex.
1862 2018-10-02  Jeff Law  <law@redhat.com>
1864         * gimple-fold.c (get_range_strlen): Remove dead code.
1866 2018-10-02  Martin Sebor  <msebor@redhat.com>
1867             Jeff Law  <law@redhat.com>
1869         * builtins.c (unterminated_array): Add new arguments.
1870         If argument is not terminated, bubble up size and exact
1871         state to callers.
1872         (expand_builtin_strnlen): Detect, avoid expanding
1873         and diagnose unterminated arrays.
1874         (c_strlen): Fill in offset of start of unterminated strings.
1875         * builtins.h (unterminated_array): Update prototype.
1877 2018-10-02  Richard Biener  <rguenther@suse.de>
1879         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
1880         of haddv4df, first reduce to SSE width and exploit the fact
1881         that we only need element zero with the reduction result.
1882         (reduc_plus_scal_v2df): Likewise.
1884 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
1886         * dojump.h (do_jump): Delete.
1887         (do_jump_1): Likewise.
1888         (split_comparison): Move around.
1889         * dojump.c (do_jump): Make static.
1890         (do_jump_1): Likewise.
1891         (jumpifnot): Move around.
1892         (jumpifnot_1): Likewise.
1893         (jumpif): Likewise.
1894         (jumpif_1): Likewise.
1895         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
1897 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
1899         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
1900         insns in the delay slot and add_insn_after for the jump insn.
1902 2018-10-02  Richard Biener  <rguenther@suse.de>
1904         * tree-inline.c (expand_call_inline): Use the location of
1905         the callee declaration for the inline-entry marker.
1906         * final.c (notice_source_line): Remove special-casing of
1907         NOTE_INSN_INLINE_ENTRY.
1909 2018-10-01  Carl Love  <cel@us.ibm.com>
1911         PR 69431
1912         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
1913         (__builtin_mtfsb0): New.
1914         (__builtin_mtfsb1): New.
1915         ( __builtin_set_fpscr_rn): New.
1916         (__builtin_set_fpscr_drn): New.
1917         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
1918         (rs6000_expand_set_fpscr_rn_builtin): Add.
1919         (rs6000_expand_set_fpscr_drn_builtin): Add.
1920         (rs6000_expand_builtin): Add case statement entries for
1921         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
1922         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
1923         RS6000_BUILTIN_MFFSL.
1924         (rs6000_init_builtins): Add ftype initialization and def_builtin
1925         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1926         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1927         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1928         rs6000_mffscdrn): Add define_insn.
1929         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1930         * doc/extend.texi: Add documentation for the builtins.
1932 2018-10-01  Richard Biener  <rguenther@suse.de>
1934         PR tree-optimization/87465
1935         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1936         causing branch miscounts.
1938 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1940         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1941         aarch64_option_default_param):  New.
1942         (params.h): Include.
1943         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1944         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1945         stack-clash protection validation code.
1947 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1949         * params.c (validate_param): New.
1950         (add_params): Use it.
1951         (set_param_value): Refactor param validation into validate_param.
1952         (diagnostic.h): Include.
1953         * diagnostic.h (diagnostic_ready_p): New.
1955 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1957         * params.c (set_param_value):
1958         Add index of parameter being validated.
1959         * common/common-target.def (option_validate_param): New.
1960         * common/common-targhooks.h (default_option_validate_param): New.
1961         * common/common-targhooks.c (default_option_validate_param): New.
1962         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1963         * doc/tm.texi: Regenerate.
1965 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1967         PR target/86486
1968         * config/aarch64/aarch64.c (aarch64_override_options_internal):
1969         Add validation for stack-clash parameters and set defaults.
1971 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1973         PR target/86486
1974         * configure.ac: Add stack-clash-protection-guard-size.
1975         * doc/install.texi: Document it.
1976         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1977         * params.def: Update comment for guard-size.
1978         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1979         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1980         * configure: Regenerate.
1982 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1984         PR target/86486
1985         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1986         STACK_DYNAMIC_OFFSET): New.
1987         * config/aarch64/aarch64.c (aarch64_layout_frame):
1988         Update outgoing args size.
1989         (aarch64_stack_clash_protection_alloca_probe_range,
1990         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1992 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1994         PR target/86486
1995         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1996         probe ranges.
1997         * target.def (stack_clash_protection_alloca_probe_range): New.
1998         (stack_clash_protection_final_dynamic_probe): Remove.
1999         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
2000         (default_stack_clash_protection_final_dynamic_probe): Remove.
2001         * targhooks.c: Likewise.
2002         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2003         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
2004         * doc/tm.texi: Regenerate.
2006 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2008         PR target/86486
2009         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
2010         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
2011         aarch64_clamp_to_uimm12_shift): New.
2012         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
2013         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
2015 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2017         PR target/86486
2018         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
2020 2018-10-01  Jeff Law  <law@redhat.com>
2021             Richard Sandiford <richard.sandiford@linaro.org>
2022             Tamar Christina  <tamar.christina@arm.com>
2024         PR target/86486
2025         * config/aarch64/aarch64.md
2026         (probe_stack_range): Add k (SP) constraint.
2027         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
2028         STACK_CLASH_MAX_UNROLL_PAGES): New.
2029         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
2030         stack probes for stack clash.
2031         (aarch64_allocate_and_probe_stack_space): New.
2032         (aarch64_expand_prologue): Use it.
2033         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
2034         (aarch64_sub_sp): Add emit_move_imm optional param.
2036 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
2038         PR tree-optimization/87261
2039         * match.pd: Remove trailing whitespace.
2040         Add (x & y) | ~(x | y) -> ~(x ^ y),
2041         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
2043 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2045         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
2046         constraints.
2047         (ashlsi3_insn): Update instruction constraints.
2048         (ashrsi3_insn): Likewise.
2049         (rotrsi3): Likewise.
2050         (add_shift): Likewise.
2051         * config/arc/constraints.md (Csz): New 32 bit constraint. It
2052         avoids placing in the limm field small constants which, otherwise,
2053         could end into a small instruction.
2055 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2057         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
2058         destination register is not odd-even.
2059         (umaddsidi4_split): Likewise.
2061 2018-10-01  Richard Biener  <rguenther@suse.de>
2063         * tree-inline.c (expand_call_inline): Store origin of fn
2064         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
2065         * tree.c (block_ultimate_origin): Simplify and do some
2066         checking.
2068 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
2070         * config/i386/mmx.md (EMMS): New int iterator.
2071         (emms): New int attribute.
2072         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
2073         EMMS int iterator.  Explicitly declare clobbers.
2074         (mmx_emms): Remove expander.
2075         (mmx_femms): Ditto.
2076         * config/i386/predicates.md (emms_operation): Remove predicate.
2077         (vzeroall_pattern): New predicate.
2078         (vzeroupper_pattern): Rename from vzeroupper_operation.
2079         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
2080         vzeroupper_pattern and vzeroall_pattern predicates.
2082 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
2084         PR rtl-optimization/86939
2085         * ira-lives.c (make_hard_regno_born): Rename from this...
2086         (make_hard_regno_live): ... to this.  Remove update to conflict
2087         information.  Update function comment.
2088         (make_hard_regno_dead): Add conflict information update.  Update
2089         function comment.
2090         (make_object_born): Rename from this...
2091         (make_object_live): ... to this.  Remove update to conflict information.
2092         Update function comment.
2093         (make_object_dead):  Add conflict information update.  Update function
2094         comment.
2095         (mark_pseudo_regno_live): Call make_object_live.
2096         (mark_pseudo_regno_subword_live): Likewise.
2097         (mark_hard_reg_dead): Update function comment.
2098         (mark_hard_reg_live): Call make_hard_regno_live.
2099         (process_bb_node_lives): Likewise.
2100         * lra-lives.c (make_hard_regno_born): Rename from this...
2101         (make_hard_regno_live): ... to this.  Remove update to conflict
2102         information.  Remove now uneeded check_pic_pseudo_p argument.
2103         Update function comment.
2104         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
2105         to conflict information.  Update function comment.
2106         (mark_pseudo_live): Remove update to conflict information.  Update
2107         function comment.
2108         (mark_pseudo_dead): Add conflict information update.
2109         (mark_regno_live): Call make_hard_regno_live.
2110         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
2111         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
2113 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
2115         PR target/87370
2116         * config/i386/i386.c (construct_container): Use TImode for
2117         BLKmode values in 2 integer registers.
2119 2018-09-29  Jeff Law  <law@redhat.com>
2121         * builtins.c (unterminated_array): Pass in c_strlen_data * to
2122         c_strlen rather than just a tree *.
2123         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2124         Update recursive calls appropriately.  If caller did not provide a
2125         suitable data pointer, create a local one.  When a non-terminated
2126         string is discovered, bubble up information about the string via the
2127         c_strlen_data object.
2128         * builtins.h (c_strlen): Update prototype.
2129         (c_strlen_data): New structure.
2130         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2131         For a type 2 call, if c_strlen indicates a non-terminated string
2132         use the length of the non-terminated string.
2133         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2135 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
2137         PR target/87467
2138         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2139         __m512d type for __A argument rather than __m512.
2141 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
2143         * match.pd (simple_comparison): Don't optimize if either operand is
2144         a function pointer when target needs function pointer canonicalization.
2146 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2148         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2149         power5 .. power9 to remove indirection.
2150         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2151         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2152         ASM_CPU_476_SPEC): Delete.
2153         (ASM_CPU_SPEC): Adjust.
2154         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2155         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2157 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2159         * config.in: Delete HAVE_AS_DCI.
2160         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2161         * config/rs6000/rs6000.h: Ditto.
2162         * configure.ac: Delete HAVE_AS_DCI.
2163         * configure: Regenerate.
2165 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2167         * config.in (HAVE_AS_LWSYNC): Delete.
2168         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2169         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
2170         do it as a .long .
2171         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2172         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
2173         as a .long .
2174         * configure.ac: Delete HAVE_AS_LWSYNC.
2175         * configure: Regenerate.
2177 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2178             Pierre-Marie de Rodat  <derodat@adacore.com>
2180         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2181         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2182         (cgraph_node::create_thunk): Add indirect_offset parameter.
2183         (thunk_adjust): Likewise.
2184         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2185         and initialize the corresponding field with it.
2186         (cgraph_node::dump): Dump indirect_offset field.
2187         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2188         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2189         (thunk_adjust): Add indirect_offset parameter and deal with it.
2190         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2191         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
2192         or if the thunk is external or local.  Fix formatting.  Do not chain
2193         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
2194         if any, in the GIMPLE representation.
2195         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2196         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2197         (input_node): Read indirect_offset field.
2198         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2199         call to thunk_adjust.
2200         * tree-nested.c (struct nesting_info): Add thunk_p field.
2201         (create_nesting_tree): Set it.
2202         (convert_all_function_calls): Copy static chain from targets to thunks.
2203         (finalize_nesting_tree_1): Return early for thunks.
2204         (unnest_nesting_tree_1): Do not finalize thunks.
2205         (gimplify_all_functions): Do not gimplify thunks.
2207 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2209         * opt-suggestions.c (option_proposer::build_option_suggestions):
2210         Release "option_values".
2212 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2214         * coverage.c (get_coverage_counts): Convert problem-reporting dump
2215         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2216         * dumpfile.c (kind_as_string): New function.
2217         (dump_loc): Rather than a hardcoded prefix of "note: ", use
2218         kind_as_string to vary the prefix based on dump_kind.
2219         (selftest::test_capture_of_dump_calls): Update for above.
2221 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2223         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2224         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
2226 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2228         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
2229         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
2230         INVALID_REGNUM instead of FPSR_REG.
2231         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
2232         * config/i386/i386.md: Update comment of FP compares.
2233         (fldenv): Do not clobber FPSR_REG.
2235 2018-09-28  Richard Biener  <rguenther@suse.de>
2237         * tree.h (BLOCK_ORIGIN): New.
2238         * omp-expand.c (grid_expand_target_grid_body): Assign
2239         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
2240         * tree-inline.c (remap_block): Likewise.
2241         * auto-profile.c (get_function_decl_from_block): Simplify
2242         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
2243         * langhooks.c (lhd_print_error_function): Likewise.
2244         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
2245         Likewise.
2246         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2247         * tree.c (block_nonartificial_location): Likewise.
2248         (block_ultimate_origin): Likewise.
2249         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
2250         no longer needed LTO case.
2252 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
2253             Jan Hubicka  <jh@suse.cz>
2254             Martin Jambor  <mjambor@suse.cz>
2256         * simplify-rtx.c (simplify_merge_mask): New function.
2257         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
2258         same masks are used in op1 or op2.
2259         (test_vec_merge): New function.
2260         (test_vector_ops): Call test_vec_merge.
2262 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2264         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
2265         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
2266         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
2267         (cypress_costs): Set it.
2268         (supersparc_costs): Likewise.
2269         (hypersparc_costs): Likewise.
2270         (leon_cost): Likewise.
2271         (leon3_costs): Likewise.
2272         (sparclet_costs): Likewise.
2273         (ultrasparc_costs): Likewise.
2274         (ultrasparc_costs): Likewise.
2275         (niagara_costs): Likewise.
2276         (niagara2_costs): Likewise.
2277         (niagara3_costs): Likewise.
2278         (niagara4_costs): Likewise.
2279         (niagara7_costs): Likewise.
2280         (m8_costs): Likewise.
2281         (TARGET_CAN_FOLLOW_JUMP): Define.
2282         (pass_work_around_errata::gate): Minor tweak.
2283         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
2284         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
2285         Automaitcally clear MASK_FSMULD mask for V7 processors.
2286         (sparc_can_follow_jump): New static function.
2287         (output_ubranch): Deal with CROSSING_JUMP_P.
2288         (sparc_use_sched_lookahead): Rewrite using switch statement.
2289         (sparc_issue_rate): Reorder.
2290         (sparc_branch_cost): New function.
2292 2018-09-27  Martin Sebor  <msebor@redhat.com>
2294         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
2295         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
2296         (int_fits_type_p): Same.
2298 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
2300         * config/i386/i386.md (FPCR_REG): Remove.
2301         (UNSPEC_FLDCW): Remove.
2302         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
2303         (x86_fldcw_1): Remove insn pattern.
2304         (fnstenv): Do not clobber FPCR_REG.
2305         (fldenv): Ditto.
2306         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
2307         (CALL_USED_REGISTERS): Ditto.
2308         (REG_ALLOC_ORDER): Ditto.
2309         (REG_CLASS_CONTENTS): Ditto.
2310         (HI_REGISTER_NAMES): Ditto.
2311         (ADDITIONAL_REGISTER_NAMES): Use defines instead
2312         of numerical constants.
2313         * config/i386/i386.c (regclass_map): Remove fpsr register.
2314         (dbx_register_map): Ditto.
2315         (dbx64_register_map): Ditto.
2316         (svr4_dbx_register_map): Ditto.
2317         (print_reg): Do not handle FPCR_REG.
2319 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
2321         PR target/87149
2322         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
2323         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
2324         Delete, always treat as true.
2325         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2326         Ditto.  Simplify remaining code.
2327         * config/powerpcspe/powerpcspe.h: Ditto.
2328         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
2329         Simplify remaining code.
2330         (rs6000_expand_builtin): Ditto.
2331         * config/rs6000/rs6000.h: Ditto.
2332         * configure.ac: Ditto.
2333         * configure: Regenerate.
2335 2018-09-27  Martin Liska  <mliska@suse.cz>
2337         * coverage.c (get_coverage_counts): Revert the formatting
2338         of missing profile opt info.
2340 2018-09-27  Richard Biener  <rguenther@suse.de>
2342         PR debug/37801
2343         PR debug/87440
2344         * dwarf2out.c (set_block_origin_self): Do not mark outermost
2345         block as we do not output that.
2346         (gen_inlined_subroutine_die): Elide the originally outermost
2347         block, matching what we do for concrete instances.
2348         (decls_for_scope): Add parameter specifying whether to recurse
2349         to subblocks.
2351 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
2352             Tom de Vries  <tom@codesourcery.com>
2354         PR 82089
2356         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
2357         STORE_FLAG_VALUE == 1.
2359 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
2361         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
2362         constant definitions.
2363         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
2364         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
2365         ("speculation_barrier"): New expander definition.
2367 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
2369         PR gcov-profile/86957
2370         * common.opt: New warning option -Wmissing-profile.
2371         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
2372         * doc/invoke.texi: Document -Wmissing-profile.
2374 2018-09-26  Jim Wilson  <jimw@sifive.com>
2376         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
2377         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
2379 2018-09-26  Martin Sebor  <msebor@redhat.com>
2381         * tree.c (zerop): Change return type to bool.
2382         (integer_zerop, integer_onep, integer_each_onep): Same.
2383         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2384         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2385         (real_onep, real_minus_onep, chain_index): Same.
2386         (print_type_hash_statistics, type_list_equal): Same.
2387         * tree.h (zerop): Same.
2388         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
2389         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2390         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2391         (real_onep, real_minus_onep, chain_index): Same.
2392         (print_type_hash_statistics, type_list_equal): Same.
2394 2018-09-26  Jim Wilson  <jimw@sifive.com>
2396         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
2398 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
2400         PR target/87414
2401         * config/i386/i386.c: Include debug.h and dwarf2out.h.
2402         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2403         call.
2405 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
2407         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
2409 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2411         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
2412         and FP_SECOND_SSE_REGS.
2413         (REG_CLASS_NAMES): Ditto.
2414         (REG_CLASS_CONTENTS): Ditto.
2415         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
2416         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
2417         (ix86_preferred_output_reload_class): Ditto.
2418         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
2419         clobber constraint to "=&f".
2420         (fix_truncdi_i387): Ditto.
2421         (lrintxfdi2): Ditto.
2422         (fistdi2_<rounding>): Ditto.
2423         (fpremxf4_i387): Change "=u" constraint to "=f".
2424         (fprem1xf4_i387): Ditto.
2425         (sincosxf3): Ditto.
2426         (fptanxf4_i387): Ditto.
2427         (fxtractxf3_i387): Ditto.
2428         (fscalexf4_i387): Ditto.
2429         (atan2xf3): Change "u" constraint to "f".
2430         (fyl2xxf3_i387): Ditto.
2431         (fyl2xp1xf3_i387): Ditto.
2433 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2435         PR target/87439
2436         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
2437         for removed I387_MASK_PM entity.
2440 2018-09-26  Jeff Law  <law@redhat.com>
2441         Revert
2442         2018-09-26  Alexey Neyman  <stilor@att.net>
2444         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2445         headers are no longer pulled in by <isl/val.h>.
2447 2018-09-26  Richard Biener  <rguenther@suse.de>
2449         PR debug/87443
2450         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
2451         or concrete instance DIE to the tree.  Create abstract origin
2452         attributes also for concrete instances.
2454 2018-09-26  Alexey Neyman  <stilor@att.net>
2456         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2457         headers are no longer pulled in by <isl/val.h>.
2459 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
2461         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
2462         Use new helper functions.
2463         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
2464         Use new helper functions.
2465         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
2466         aarch_mm_needs_release): New declarations.
2467         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
2468         aarch_mm_needs_release): New.
2470 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
2472         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
2473         (arm32_output_mi_thunk): Deal with long calls.
2475 2018-09-26  Richard Biener  <rguenther@suse.de>
2477         PR debug/87428
2478         PR debug/87362
2479         * tree-inline.c (expand_call_inline): When the location
2480         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2481         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2482         the inserted BLOCK to make inlined_function_outer_scope_p
2483         recognize it.
2484         * dwarf2out.c (add_call_src_coords_attributes): Do not add
2485         coords for reserved locations.
2487 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2489         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
2490         (*call_indirect_nonlocal_sysv<mode>): Ditto.
2491         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
2492         (*sibcall_nonlocal_sysv<mode>): Ditto.
2493         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
2494         (<bd>_<mode>): Ditto.
2495         (<bd>tf_<mode>): Ditto.
2497 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2499         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
2500         control string as a list of templates instead of as C code.
2501         (*altivec_movti): Ditto.
2502         * config/rs6000/darwin.md (movdf_low_di): Ditto.
2504 2018-09-25  Jim Wilson  <jimw@sifive.com>
2506         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
2507         when target symbol is weak.
2509 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2511         PR c/87387
2512         * builtins.c (unterminated_array): Simplify.
2513         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
2514         where pointer arithmetic is safe.
2516 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2518         PR target/86987
2519         * config/rs6000/altivec.md (altivec_vspltb): Use
2520         const_0_to_15_operand instead of u5bit_cint_operand.
2521         (*altivec_vspltb_internal): Ditto.
2522         (altivec_vspltb_direct): Ditto.
2523         (altivec_vsplth): Use const_0_to_7_operand instead of
2524         u5bit_cint_operand.
2525         (*altivec_vsplth_internal): Ditto.
2526         (altivec_vsplth_direct): Ditto.
2527         (altivec_vspltw): Use const_0_to_3_operand instead of
2528         u5bit_cint_operand.
2529         (*altivec_vspltw_internal): Ditto.
2530         (altivec_vspltw_direct): Ditto.
2531         (altivec_vspltsf): Ditto.
2532         (*altivec_vspltsf_internal): Ditto.
2533         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
2534         various splats with the proper size immediate.  Reorder the various
2535         cases by ascending size of immediate, and put all such together.
2537 2018-09-25  Richard Biener  <rguenther@suse.de>
2539         PR debug/83941
2540         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
2541         GC-ification.
2542         (maybe_create_die_with_external_ref): Do not create
2543         DW_TAG_imported_unit here.
2544         (add_abstract_origin_attribute): Handle external BLOCK refs.
2545         (dwarf2out_abstract_function): Simplify LTO case.
2546         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
2547         rather than using maybe_create_die_with_external_ref.
2549 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
2551         PR target/71278
2552         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
2553         (frndintxf2_mask_pm_i387): Ditto.
2554         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
2555         Enable for !flag_trapping_math.
2556         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
2557         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
2558         Change operand 1 predicate to nonimmediate_operand.
2559         (attr "i387_cw"): Remove mask_pm.
2560         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
2561         (enum ix86_entity): Remove I387_MASK_PM.
2562         * config/i386/i386.c (ix86_i387_mode_needed): Do not
2563         handle I387_MASK_PM.
2564         (ix86_mode_needed): Ditto.
2565         (ix86_mode_after): Ditto.
2566         (ix86_mode_entry): Ditto.
2567         (ix86_mode_exit): Ditto.
2568         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
2570 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
2572         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
2573         to_update_switch_stmts to vNULL instead of calling create on them
2574         immediately.
2576 2018-09-25  Richard Biener  <rguenther@suse.de>
2578         PR tree-optimization/87402
2579         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
2580         (visit_phi): Re-instantiate handling of supposed to be VARYING
2581         but non-VARYING backedge value.
2583 2018-09-25  Richard Biener  <rguenther@suse.de>
2585         PR debug/83941
2586         * dwarf2out.c (struct sym_off_pair): New.
2587         (external_die_map): New global.
2588         (lookup_decl_die): When in LTO create DIEs lazily from the
2589         external_die_map.
2590         (lookup_block_die): New function, create DIEs lazily in LTO.
2591         (equate_block_to_die): New function.
2592         (dwarf2out_die_ref_for_decl): During WPA get the association
2593         from the external DIE map.
2594         (dwarf2out_register_external_die): Record mapping into the
2595         external DIE map.
2596         (maybe_create_die_with_external_ref): New function split out from
2597         DIE generation part of old dwarf2out_register_external_die.
2598         (add_abstract_origin_attribute): Do not return the DIE.  When
2599         in LTO reference externals directly.
2600         (dwarf2out_abstract_function): When in LTO ignore calls for
2601         decls with external DIEs (already present abstract instances).
2602         (gen_call_site_die): Adjust.
2603         (add_high_low_attributes): Likewise.
2604         (gen_lexical_block_die): Likewise.
2605         (gen_inlined_subroutine_die): Likewie.
2606         (gen_block_die): Likewise.
2607         (dwarf2out_inline_entry): Likewise.
2608         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
2609         DIEs.
2611 2018-09-25  Martin Liska  <mliska@suse.cz>
2613         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
2614         integers and not by a float value.
2616 2018-09-25  Martin Liska  <mliska@suse.cz>
2618         PR fortran/87394
2619         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
2620         instead of NULL.
2621         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
2622         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
2623         * godump.c (go_format_type): Remove extra parenthesis.
2625 2018-09-25  Martin Liska  <mliska@suse.cz>
2627         * alias.c (set_dest_equal_p): Remove unused function.
2628         * config/i386/i386.c (def_builtin_pure2): Likewise.
2629         * diagnostic-show-locus.c (class layout): Remove
2630         unused field.
2631         (layout::layout): Likewise here.
2632         * dump-context.h (class temp_dump_context): Likewise.
2633         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
2634         (add_AT_loclistsptr): Likewise.
2635         (add_AT_offset): Likewise.
2636         (get_AT_hi_pc): Likewise.
2637         (is_comdat_die): Likewise.
2638         (type_is_enum): Likewise.
2639         (ceiling): Likewise.
2640         (add_AT_vms_delta): Likewise.
2641         (is_class_die): Likewise.
2642         * edit-context.c (class line_event): Remove unused field.
2643         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
2644         unused function.
2645         * ipa-cp.c (ipa_get_vr_lat): Likewise.
2646         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
2647         (ok_for_base_p_nonstrict): Likewise.
2648         * tree-chrec.c (is_not_constant_evolution): Likewise.
2649         (chrec_fold_poly_cst): Likewise.
2650         * tree-if-conv.c (has_pred_critical_p): Likewise.
2651         * tree-ssa-coalesce.c (print_exprs): Likewise.
2652         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
2653         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
2654         * tree-vrp.c (value_ranges_intersect_p): Likewise.
2655         (value_range_nonnegative_p): Likewise.
2657 2018-09-25  Martin Liska  <mliska@suse.cz>
2659         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2660         Do not handle "GNU Pascal".
2661         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2662         Likewise.
2663         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
2664         from documentation. Likewise.
2665         * dbxout.c (dbxout_range_type): Likewise.
2666         * doc/cpp.texi: Likewise.
2667         * doc/extend.texi: Likewise.
2668         * doc/frontends.texi: Likewise.
2669         * doc/invoke.texi: Remove Pascal entry.
2670         * tree.def (CLEANUP_POINT_EXPR): Likewise.
2671         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
2673 2018-09-25  Martin Liska  <mliska@suse.cz>
2675         PR middle-end/86078
2676         * doc/invoke.texi: Document all parameters and remove default
2677         of the parameters.
2679 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2681         PR bootstrap/87417
2682         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
2683         contains HOST_WIDE_INTs when computing its size.
2685 2018-09-24  Jim Wilson  <jimw@sifive.com>
2687         PR target/87391
2688         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
2689         not TARGET_RVE.
2690         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
2692 2018-09-24  Andrew Pinski  <apinski@marvell.com>
2694         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2695         access prev before checking it for NULLness in the
2696         AARCH64_FUSE_CMP_BRANCH case.
2698 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
2700         PR target/82699
2701         * config/i386/i386.c (rest_of_insert_endbranch): Set
2702         endbr_queued_at_entrance to true and don't insert ENDBR if
2703         x86_function_profiler will be called.
2704         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2705         is true.
2706         * config/i386/i386.h (machine_function): Add
2707         endbr_queued_at_entrance.
2709 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2711         * genattrtab.c (mk_attr_alt): Use alternative_mask.
2712         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
2713         types.
2714         (check_attr_test): Use alternative_mask.
2715         (get_attr_value): Likewise.
2716         (compute_alternative_mask): Use alternative_mask and XWINT.
2717         (make_alternative_compare): Use alternative_mask.
2718         (attr_alt_subset_p): Use XWINT.
2719         (attr_alt_subset_of_compl_p): Likewise.
2720         (attr_alt_intersection): Use alternative_mask and XWINT.
2721         (attr_alt_union): Likewise.
2722         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
2723         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
2724         (simplify_test_exp): Use alternative_mask and XWINT.
2725         (write_test_expr): Use alternative_mask and XWINT, adjust bit
2726         number calculation to support 64 bits.  Generate code that
2727         checks 64-bit masks.
2728         (main): Use alternative_mask.
2729         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
2731 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2733         PR target/80080
2734         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
2735         RETURN+USE when returning via %r14.
2737 2018-09-24  Martin Liska  <mliska@suse.cz>
2739         * gcov.c (output_lines): Print colorization legend
2740         for both flag_use_colors and flag_use_hotness_colors.
2741         Reword the help.
2743 2018-09-24  Martin Liska  <mliska@suse.cz>
2745         * coverage.c (get_coverage_counts): Use warning_at
2746         with current_function_decl location. Use %qD in warning
2747         message.
2749 2018-09-24  Martin Liska  <mliska@suse.cz>
2751         * memory-block.h (memory_block_pool::release): Annotate with
2752         valgrind that the memory is not accessible.
2754 2018-09-24  Martin Liska  <mliska@suse.cz>
2756         PR sanitizer/85774
2757         * asan.c: Make asan_handled_variables extern.
2758         * asan.h: Likewise.
2759         * cfgexpand.c (expand_stack_vars): Make sure
2760         a representative is unpoison if another
2761         variable in the partition is handled by
2762         use-after-scope sanitization.
2764 2018-09-24  Richard Biener  <rguenther@suse.de>
2766         PR tree-optimization/63155
2767         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2768         the worklist when the edge of the respective argument isn't
2769         executable.
2771 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2773         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
2774         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
2775         (MASK_CLASS_P): Update for rename.
2776         (MAYBE_MASK_CLASS_P): Ditto.
2777         (REG_CLASS_NAMES): Update.
2778         (REG_CLASS_CONTENT): Update.
2779         * config/i386/i386.c (regclass_map): Update for MASK_REG
2780         and ALL_MASK_REGS rename.
2781         * config/i386/constraints.md (Yk): Update for rename.
2782         (k): Ditto.
2784 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2786         * config/i386/i386.h (enum reg_class): Remove
2787         EVEX_SSE_REGS and MOD4_SSE_REGS.
2788         (REG_CLASS_NAMES): Update.
2789         (REG_CLASS_CONTENT): Update.
2790         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
2791         registers as ALL_SSE_REGS.
2792         (ix86_additional_allocno_class_p): Remove.
2793         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
2794         (ix86_register_priority): Lower priority of EVEX SSE registers.
2795         Use IN_RANGE macro where appropriate.
2796         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
2797         AVX-5124VNNIW checks.
2798         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
2799         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
2800         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
2801         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
2802         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
2803         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
2804         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
2805         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
2806         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
2807         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
2808         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
2809         * config/i386/constraints.md (Yh): Remove.
2811 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2813         * config/i386/i386.c (regclass_map): Declare integer REX registers
2814         as GENERAL_REGS.
2816 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
2818         * doc/service.texi (Service): Switch the fsf.org link to https.
2820 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2822         PR target/86798
2823         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2824         Define to speculation_safe_value_not_needed.
2826 2018-09-21  Florian Weimer  <fweimer@redhat.com>
2828         PR middle-end/81035
2829         * doc/extend.texi (Common Function Attributes): Mention that
2830         noreturn suppresses tail call optimization.
2832 2018-09-21  Jeff Law  <law@redhat.com>
2834         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
2835         vr_values::cleanup_edges_and_switches.
2836         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
2837         vr_values class.
2838         (identify_jump_threads): Remove EDGE_IGNORE handling.
2839         (execute_vrp): Move handling of to_remove_edges and
2840         to_update_switch_stmts into vr_values class member functions.
2841         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
2842         (to_update_switch_stmts): Likewise.
2843         * vr-values.c: Include cfghooks.h.
2844         (vr_values::vr_values): Initialize to_remove_edges and
2845         to_update_switch_stmts.
2846         (vr_values::~vr_values): Verify to_remove_edges and
2847         to_update_switch_stmts are empty.
2848         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
2849         (vr_values::cleanup_edges_and_switches): New member function.
2850         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
2851         function.  Add new data members.
2853 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
2855         PR tree-optimization/87309
2856         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
2857         calls with pflags and alt_flags.
2858         (selftest::test_capture_of_dump_calls): Add test of interaction of
2859         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
2861 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2863         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
2865 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2867         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
2869 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2871         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
2872         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
2873         (VXWORKS_LIBS_RTP): Minor reordering.
2875 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2877         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
2878         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
2880 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2882         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
2883         (PTRDIFF_TYPE): Likewise.
2885 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2887         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
2888         triplet, similar to support for VxWorks7.
2889         * config/vxworks-dummy.h: Provide a default definition
2890         of TARGET_VXWORKS64 to 0.
2892 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2894         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
2895         * config/vxworks-dummy.h: here.
2897 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2899         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
2901 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
2902             Bo Zhou  <zbo.zhou@hisilicon.com>
2904         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
2905         * config/aarch64/aarch64-tune.md: Regenerated.
2906         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
2907         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
2908         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
2910 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
2911             Julian Brown  <julian@codesourcery.com>
2913         * builtins.c (get_builtin_sync_mem): Handle address spaces.
2915 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2917         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2918         if the call takes a static chain.
2920 2018-09-21  Martin Liska  <mliska@suse.cz>
2922         * auto-profile.c (autofdo_source_profile::read): Do not
2923         set sum_all.
2924         (read_profile): Do not add working sets.
2925         (read_autofdo_file): Remove sum_all.
2926         (afdo_callsite_hot_enough_for_early_inline): Remove const
2927         qualifier.
2928         * coverage.c (struct counts_entry): Remove gcov_summary.
2929         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2930         do not support GCOV_TAG_PROGRAM_SUMMARY.
2931         (get_coverage_counts): Remove summary and expected
2932         arguments.
2933         * coverage.h (get_coverage_counts): Likewise.
2934         * doc/gcov-dump.texi: Remove -w option.
2935         * gcov-dump.c (dump_working_sets): Remove.
2936         (main): Do not support '-w' option.
2937         (print_usage): Likewise.
2938         (tag_summary): Likewise.
2939         * gcov-io.c (gcov_write_summary): Do not dump
2940         histogram.
2941         (gcov_read_summary): Likewise.
2942         (gcov_histo_index): Remove.
2943         (gcov_histogram_merge): Likewise.
2944         (compute_working_sets): Likewise.
2945         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2946         it not obsolete.
2947         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2948         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2949         (GCOV_HISTOGRAM_SIZE): Remove.
2950         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2951         (struct gcov_summary): Simplify rapidly just
2952         to runs and sum_max fields.
2953         (gcov_histo_index): Remove.
2954         (NUM_GCOV_WORKING_SETS): Likewise.
2955         (compute_working_sets): Likewise.
2956         * gcov-tool.c (print_overlap_usage_message): Remove
2957         trailing empty line.
2958         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2959         (output_lines): Remove program related line.
2960         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2961         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2962         histogram.
2963         (input_profile_summary): Do not read it.
2964         (merge_profile_summaries): And do not merge it.
2965         (input_symtab): Do not call removed function.
2966         * modulo-sched.c (sms_schedule): Do not print sum_max.
2967         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2968         removed when histogram method was invented.
2969         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2970         mode.
2971         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2972         GCOV coding style.
2973         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2974         and dump selected value.
2975         * profile.c (add_working_set): Remove.
2976         (get_working_sets): Likewise.
2977         (find_working_set): Likewise.
2978         (get_exec_counts): Do not work with working sets.
2979         (read_profile_edge_counts): Do not inform as sum_max is removed.
2980         (compute_branch_probabilities): Likewise.
2981         (compute_value_histograms): Remove argument for call of
2982         get_coverage_counts.
2983         * profile.h: Do not make gcov_summary const.
2985 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
2987         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2989 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2991         PR tree-optimization/86990
2992         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2993         Check that the entire merged store group is made of constants only for
2994         overlapping stores.
2996 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2998         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2999         (VTABLE_VERIFICATION_SPEC): Likewise.
3000         (SANITIZER_EARLY_SPEC): Likewise.
3001         (SANITIZER_SPEC): Likewise.
3002         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
3003         * doc/invoke.texi (Link Options): Document -r.
3005 2018-09-20  Richard Biener <rguenther@suse.de>
3007         PR middle-end/87054
3008         * gimplify.c (gimplify_expr): Retain alignment of
3009         addressable lvalue in dereference.
3011 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
3013         PR bootstrap/87013
3014         * configure.ac: Check for .loc is_stmt support.
3015         * configure, config.in: Rebuilt.
3016         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
3017         if not supported.
3019 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3021         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
3022         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
3023         -misel=no.
3025 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3027         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3028         VECTOR_OTHER.
3029         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
3030         case VECTOR_OTHER.
3032 2018-09-20  Marek Polacek  <polacek@redhat.com>
3034         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
3036 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3038         PR tree-optimization/87288
3039         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
3040         into account when determining PEELING_FOR_NITERS.
3042 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3044         PR tree-optimization/86877
3045         * tree-vect-loop.c (vect_analyze_loop_2): Call
3046         vect_verify_datarefs_alignment.
3048 2018-09-19  Marek Polacek  <polacek@redhat.com>
3050         * doc/invoke.texi: Document -Wclass-conversion.
3052 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
3054         * config/pa/pa.c (pa_adjust_priority): Delete.
3055         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
3057         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3058         (atomic_storehi): Likewise.
3059         (atomic_storesi): Likewise.
3060         (atomic_loaddi): Restore compare and swap exchange loop code.
3062 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
3064         PR rtl-optimization/86902
3065         * combine.c (try_combine): When changing the CC mode used, don't change
3066         an unrelated mode in other_insn to that new CC mode.
3068 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
3070         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
3071         with %T in place of calls to dump_generic_expr.
3072         (prune_runtime_alias_test_list): Likewise.
3073         (create_runtime_alias_checks): Likewise.
3074         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
3075         (vect_analyze_data_ref_dependence): Likewise.
3076         (vect_slp_analyze_data_ref_dependence): Likewise.
3077         (vect_record_base_alignment): Likewise.  Use %G in place of call
3078         to dump_gimple_stmt.
3079         (vect_compute_data_ref_alignment): Likewise.
3080         (verify_data_ref_alignment): Likewise.
3081         (vect_find_same_alignment_drs): Likewise.
3082         (vect_analyze_group_access_1): Likewise.
3083         (vect_analyze_data_ref_accesses): Likewise.
3084         (dependence_distance_ge_vf): Likewise.
3085         (dump_lower_bound): Likewise.
3086         (vect_prune_runtime_alias_test_list): Likewise.
3087         (vect_find_stmt_data_reference): Likewise.
3088         (vect_analyze_data_refs): Likewise.
3089         (vect_create_addr_base_for_vector_ref): Likewise.
3090         (vect_create_data_ref_ptr): Likewise.
3091         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
3092         (vect_can_advance_ivs_p): Likewise.
3093         (vect_update_ivs_after_vectorizer): Likewise.
3094         (vect_gen_prolog_loop_niters): Likewise.
3095         (vect_prepare_for_masked_peels): Likewise.
3096         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3097         (vect_determine_vectorization_factor): Likewise.
3098         (vect_is_simple_iv_evolution): Likewise.
3099         (vect_analyze_scalar_cycles_1): Likewise.
3100         (vect_analyze_loop_operations): Likewise.
3101         (report_vect_op): Likewise.
3102         (vect_is_slp_reduction): Likewise.
3103         (check_reduction_path): Likewise.
3104         (vect_is_simple_reduction): Likewise.
3105         (vect_create_epilog_for_reduction): Likewise.
3106         (vect_finalize_reduction:): Likewise.
3107         (vectorizable_induction): Likewise.
3108         (vect_transform_loop_stmt): Likewise.
3109         (vect_transform_loop): Likewise.
3110         (optimize_mask_stores): Likewise.
3111         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
3112         (vect_split_statement): Likewise.
3113         (vect_recog_over_widening_pattern): Likewise.
3114         (vect_recog_average_pattern): Likewise.
3115         (vect_determine_min_output_precision_1): Likewise.
3116         (vect_determine_precisions_from_range): Likewise.
3117         (vect_determine_precisions_from_users): Likewise.
3118         (vect_mark_pattern_stmts): Likewise.
3119         (vect_pattern_recog_1): Likewise.
3120         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3121         (vect_record_max_nunits): Likewise.
3122         (vect_build_slp_tree_1): Likewise.
3123         (vect_build_slp_tree_2): Likewise.
3124         (vect_print_slp_tree): Likewise.
3125         (vect_analyze_slp_instance): Likewise.
3126         (vect_detect_hybrid_slp_stmts): Likewise.
3127         (vect_detect_hybrid_slp_1): Likewise.
3128         (vect_slp_analyze_operations): Likewise.
3129         (vect_slp_analyze_bb_1): Likewise.
3130         (vect_transform_slp_perm_load): Likewise.
3131         (vect_schedule_slp_instance): Likewise.
3132         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3133         (vect_mark_stmts_to_be_vectorized): Likewise.
3134         (vect_init_vector_1): Likewise.
3135         (vect_get_vec_def_for_operand): Likewise.
3136         (vect_finish_stmt_generation_1): Likewise.
3137         (vect_check_load_store_mask): Likewise.
3138         (vectorizable_call): Likewise.
3139         (vectorizable_conversion): Likewise.
3140         (vectorizable_operation): Likewise.
3141         (vectorizable_load): Likewise.
3142         (vect_analyze_stmt): Likewise.
3143         (vect_is_simple_use): Likewise.
3144         (vect_get_vector_types_for_stmt): Likewise.
3145         (vect_get_mask_type_for_stmt): Likewise.
3146         * tree-vectorizer.c (increase_alignment): Likewise.
3148 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
3150         * doc/rtl.texi: Adjust vec_select description.
3151         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3152         non-constant selectors.
3154 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
3156         * config/aarch64/aarch64-protos.h
3157         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3158         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3159         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3160         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3161         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3162         (AARCH64_FL_PROFILE): Move index so flags are ordered.
3163         (AARCH64_ISA_RCPC8_4): New flag.
3164         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
3165         to aarch64_offset_9bit_signed_unscaled_p.
3166         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
3167         and use stlur.
3168         * config/aarch64/constraints.md (Ust): New constraint.
3169         * config/aarch64/predicates.md.
3170         (aarch64_9bit_offset_memory_operand): New predicate.
3171         (aarch64_rcpc_memory_operand): New predicate.
3173 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
3175         PR rtl-optimization/87361
3176         * rtlanal.c (nonzero_bits1): Revert accidental change.
3178 2018-09-19  Richard Biener  <rguenther@suse.de>
3180         PR tree-optimization/87349
3181         PR tree-optimization/87342
3182         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3184 2018-09-18  Marek Polacek  <polacek@redhat.com>
3186         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3187         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3189 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3191         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3193 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3195         PR rtl-optimization/86882
3196         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3198 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
3200         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3201         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3203 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
3205         PR other/87353
3206         * doc/invoke.texi (Link Options): Fix formatting and grammar.
3208 2018-09-18  Richard Biener  <rguenther@suse.de>
3210         PR middle-end/63155
3211         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3212         (compute_samebase_partition_bases): Likewise.
3213         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3214         (gimple_can_coalesce_p): Simplify.
3216 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
3218         Handle a library implementation of ffs calling __builtin_ffs.
3219         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3220         (mmix_init_libfuncs): New function: make __builtin_ffs expand
3221         to __ffsdi2.
3223 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3225         * diagnostic-show-locus.c (class layout_range): Add field
3226         "m_original_idx".
3227         (layout_range::layout_range): Add "original_idx" param and use it
3228         to initialize new field.
3229         (make_range): Use 0 for original_idx.
3230         (layout::layout): Pass in index to calls to
3231         maybe_add_location_range.
3232         (layout::maybe_add_location_range): Add param "original_idx" and
3233         pass it on to layout_range.
3234         (layout::print_any_labels): Pass on range->m_original_idx to
3235         get_text call.
3236         (gcc_rich_location::add_location_if_nearby): Use 0 for
3237         original_idx.
3238         * gcc-rich-location.h (text_range_label::get_text): Update for new
3239         param.
3240         (range_label_for_type_mismatch::get_text): Likewise.
3242 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3244         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
3246 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3248         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
3249         format_string_diagnostic_t.
3250         (fmtwarn_n): Likewise.
3251         * substring-locations.c
3252         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
3253         (format_warning_n_va): Convert to...
3254         (format_string_diagnostic_t::emit_warning_n_va): ...this.
3255         (format_warning_va): Convert to...
3256         (format_string_diagnostic_t::emit_warning_va): ...this.
3257         (format_warning_at_substring): Convert to...
3258         (format_string_diagnostic_t::emit_warning): ...this.
3259         (format_warning_at_substring_n): Convert to...
3260         (format_string_diagnostic_t::emit_warning_n): ...this.
3261         * substring-locations.h (class format_string_diagnostic_t): New
3262         class.
3263         (format_warning_va): Convert to
3264         format_string_diagnostic_t::emit_warning_va.
3265         (format_warning_n_va): Convert to
3266         format_string_diagnostic_t::emit_warning_n_va.
3267         (format_warning_at_substring): Convert to
3268         format_string_diagnostic_t::emit_warning.
3269         (format_warning_at_substring_n): Convert to
3270         format_string_diagnostic_t::emit_warning_n.
3272 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
3273             Bernd Schmidt <bernds_cb1@t-online.de>
3275         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
3276         SImode args.
3278 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3280         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
3281         operand 0 predicate to nonimmediate operand.
3282         (rint<mode>2_frndint): Remove insn pattern.
3283         (rint<mode>2): Change operand 1 predicate to general_operand.
3284         Extend operand 1 to XFmode and generate rintxf2 insn.
3285         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
3286         Do not use X87MODEF mode macro.
3287         (frndintxf2_<rounding>_i387): Rename from
3288         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
3289         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
3290         to XFmode and generate significandxf3 insn.
3292 2018-09-17  Richard Biener  <rguenther@suse.de>
3294         PR tree-optimization/87328
3295         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
3296         visiting unexecutable backedges when not iterating.
3297         (do_rpo_vn): Mark all edges not executable even when not
3298         iterating.
3300 2018-09-17  Martin Jambor  <mjambor@suse.cz>
3302         PR c/63886
3303         * doc/invoke.texi (Warning Options): Likewise.
3305 2018-09-17  Richard Biener  <rguenther@suse.de>
3307         PR tree-optimization/87301
3308         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
3309         clean EH info from leftover copy assignments.
3311 2018-09-17  Martin Liska  <mliska@suse.cz>
3313         PR gcov-profile/85871
3314         * gcov.c (output_intermediate_file): Fix out of bounds
3315         access.
3317 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
3319         * config/arc/arc.c: Object attributes for core4 not reflected
3320         correctly.
3321         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
3322         core3).
3324 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
3326         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
3328 2018-09-17  Martin Liska  <mliska@suse.cz>
3330         * doc/gcov.texi: Document new option --use-hotness-colors.
3331         * gcov.c (struct source_info): Declare new field.
3332         (source_info::source_info): Set default for maximum_count.
3333         (print_usage): Add new -q option.
3334         (process_args): Process it.
3335         (accumulate_line_info): Save src->maximum_count.
3336         (output_line_beginning): Make color line number if
3337         flag_use_hotness_colors is set.
3338         (output_line_details): Pass default argument value.
3339         (output_lines): Pass src->maximum_count.
3341 2018-09-17  Martin Liska  <mliska@suse.cz>
3343         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
3344         Use processor_names table.
3345         * config/i386/i386.c (ix86_default_align): Use
3346         processor_cost_table for alignment values.
3347         (ix86_option_override_internal): Use processor_names.
3348         (ix86_function_specific_print): Likewise.
3349         * config/i386/i386.h (struct processor_costs):
3350         Add alignment values.
3351         (struct ptt): Remove and replace with const char *.
3352         * config/i386/x86-tune-costs.h (struct processor_costs):
3353         Declare default alignments for all costs.
3355 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
3357         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
3358         symbolics or VR_VARYING ranges for ABS_EXPR.
3359         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
3360         when range will wrap.
3362 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
3364         PR middle-end/86864
3365         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
3366         before and after a JUMP_TABLE_DATA.
3368 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
3370         PR middle-end/87188
3371         * dojump.c (do_compare_and_jump): Canonicalize function pointers
3372         when one operand is a function pointer.  Use POINTER_TYPE_P and
3373         FUNC_OR_METHOD_TYPE_P.
3374         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3375         * fold-const.c (build_range_check): Likewise.
3376         * match.pd (simple_comparison): Likewise.
3378 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
3380         PR c/82967
3381         * spellcheck.c (get_edit_distance_cutoff): New function.
3382         (selftest::test_edit_distance_unit_test_oneway): Rename to...
3383         (selftest::test_get_edit_distance_one_way): ...this.
3384         (selftest::test_get_edit_distance_unit): Rename to...
3385         (selftest::test_get_edit_distance_both_ways): ...this.
3386         (selftest::test_edit_distances): Move tests to this new function,
3387         and test some more pairs of strings.  Update for above renaming.
3388         (selftest::get_old_cutoff): New function.
3389         (selftest::test_get_edit_distance_cutoff): New function.
3390         (selftest::assert_suggested_for): New function.
3391         (ASSERT_SUGGESTED_FOR): New macro.
3392         (selftest::assert_not_suggested_for): New function.
3393         (ASSERT_NOT_SUGGESTED_FOR): New macro.
3394         (selftest::test_suggestions): New function.
3395         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
3396         tests to selftest::test_edit_distances and call it.  Add calls to
3397         selftest::test_get_edit_distance_cutoff and
3398         selftest::test_suggestions.
3399         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
3400         (best_match::consider): Replace hard-coded cutoff calculation with
3401         a call to...
3402         (best_match::get_cutoff): New declaration.
3403         (best_match::get_best_meaningful_candidate): Likewise.
3405 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3407         * builtins.c (fold_builtin_strlen): Remove TODO comment.
3409 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3411         revert:
3412         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3414         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3415         terminated string literal.
3417 2018-09-14  Martin Sebor  <msebor@redhat.com>
3419         * builtins.c (unterminated_array): Handle ARRAY_REF.
3420         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
3421         * builtins.h (unterminated_array): Declare extern.
3422         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
3423         arrays.
3424         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
3425         calls.
3427 2018-09-14  Martin Sebor  <msebor@redhat.com>
3428             Jeff Law  <law@redhat.com>
3430         * builtins.c (unterminated_array): New.
3431         (expand_builtin_strcpy): Adjust.
3432         (expand_builtin_strcpy_args): Detect unterminated arrays.
3433         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
3434         unterminated arrays.
3435         * gimple-fold.h (get_maxval_strlen): Add argument.
3436         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
3438         * gimple-fold.c (get_range_strlen): Add argument.
3439         (get_maxval_strlen): Adjust.
3440         * gimple-fold.h (get_range_strlen): Add argument.
3442 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
3444         * config/i386/movdirintrin.h: Fix copyright year.
3446 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
3448         * reg-stack.c: Include regs.h.
3449         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
3450         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
3451         FIRST_STACK_REG, not DFmode.
3452         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
3453         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
3454         (change stack): Default register mode to the reg_raw_mode of
3455         FIRST_STACK_REG, not DFmode.
3456         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
3457         (*swapxf): Rename from swapxf.
3459 2018-09-14  Carl Love  <cel@us.ibm.com>
3461         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
3462         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
3464 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
3466         PR target/87224
3467         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
3468         alternatives.
3470 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
3472         PR target/85628
3473         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
3475 2018-09-14  Jason Merrill  <jason@redhat.com>
3477         Fix --enable-gather-detailed-mem-stats.
3478         * hash-table.c (hash_table_usage): Change from variable to function.
3479         * hash-table.h: Adjust.
3480         * Makefile.in: Add missing dependencies on hash-table.h.
3482 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3484         PR tree-optimization/87259
3485         PR lto/87283
3486         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
3487         execute_cse_reciprocals_1 has tried transforming.
3489 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
3491         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
3492         VR_VARYING for PLUS/MINUS_EXPR.
3494 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
3496         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
3497         formatting.
3499 2018-09-14  Richard Biener  <rguenther@suse.de>
3501         PR middle-end/63155
3502         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
3503         bits for the merged partition.
3505 2018-09-13  Martin Sebor  <msebor@redhat.com>
3506             Bernd Edlinger  <bernd.edlinger@hotmail.de>
3508         * builtins.h (c_srlen): Add argument.
3509         * builtins.c (warn_string_no_nul): New function.
3510         (c_strlen): Add argument and use it.  Update recursive calls.
3511         Pass DECL argument to string_constant to get info on non
3512         terminated strings.  Update *NONSTR as needed.
3513         (fold_builtin_strlen): Add argument to calls to c_strlen.
3514         Warn for unterminated arrays.
3515         (warn_string_no_null): Add prototype.
3516         * expr.c (string_constant): Update arguments.  Update recursive
3517         calls appropriately.  Detect missing NUL terminator and outermost
3518         declaration its missing in.
3519         Improve checks for arrays with nonzero lower bound or elements
3520         that are not a single byte.  Simplify offset computation.
3521         Simplify checks for non-NUL terminated strings.
3522         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
3523         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
3525 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3527         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
3528         correctly.
3529         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
3530         argument.  Fix range checks.
3531         * fold-const.h (c_getstr): Adjust protoype.
3532         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
3533         string is constant but contains no NUL byte.
3535         * expr.c (string_constant): Adjust function comment.
3536         Remove bogus check for zero termination.
3538         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
3540         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
3541         (get_constant_size): Don't make STRING_CSTs larger than they are.
3542         (check_string_literal): New check function for STRING_CSTs.
3543         (output_constant): Use it.
3545 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
3547         PR target/86812
3548         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
3550 2018-09-13  Richard Biener  <rguenther@suse.de>
3552         PR tree-optimization/87263
3553         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
3554         (struct unwind_state): Add max_rpo field.
3555         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
3556         Compute max_rpo, the max RPO number a block can be backwards reached
3557         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
3558         separating it from the iterating mode.
3560 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
3562         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
3563         (rfs_decision): New scheduling decision.
3565 2018-09-13  Richard Biener  <rguenther@suse.de>
3567         PR bootstrap/87134
3568         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
3569         (vn_nary_op_insert_pieces_predicated): Do not write useless
3570         valid_dominated_by_p entry outside of the allocated storage.
3572 2018-09-13  Omar Sandoval  <osandov@osandov.com>
3573             Tom de Vries  <tdevries@suse.de>
3575         PR debug/86985
3576         * dwarf2out.c (is_c): New function.
3577         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
3579 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
3581         PR target/85628
3582         * config/aarch64/aarch64.md (*aarch64_bfxil):
3583         Define.
3584         * config/aarch64/constraints.md (Ulc): Define.
3585         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
3586         Define.
3587         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
3588         New function.
3590 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
3592         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
3593         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
3594         aarch64_layout_frame call.
3595         (aarch64_expand_epilogue): Likewise.
3596         (aarch64_initial_elimination_offset): Likewise.
3597         (aarch64_get_separate_components): Likewise.
3598         (aarch64_use_return_insn_p): Likewise.
3599         (aarch64_layout_frame): Remove unneeded check.
3601 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3603         * configure.ac: Only append
3604         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
3605         gcc_config_arguments if it was never reconfigured or last reconfigure
3606         was with different arguments.
3607         * configure: Regenerated.
3609 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3610             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3612         PR middle-end/87290
3613         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
3614         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
3616 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3618         PR tree-optimization/87287
3619         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
3620         X % C == 0 to X % (unsigned) C == 0 optimization to ...
3621         * match.pd (X % C == 0): ... here.  New optimization.
3623 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3625         PR middle-end/82853
3626         * expr.h (maybe_optimize_mod_cmp): Declare.
3627         * expr.c (mod_inv): New function.
3628         (maybe_optimize_mod_cmp): New function.
3629         (do_store_flag): Use it.
3630         * cfgexpand.c (expand_gimple_cond): Likewise.
3632 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
3633             Julian Brown  <julian@codesourcery.com>
3635         PR middle-end/86336
3636         * gimplify.c (gimplify_scan_omp_clauses): Set
3637         target_firstprivatize_array_bases in OpenACC parallel and kernels
3638         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
3639         OpenACC data regions.
3641 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
3643         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
3644         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
3646 2018-09-12  Richard Biener  <rguenther@suse.de>
3648         PR tree-optimization/87280
3649         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
3650         edge but unreachable target.
3651         (do_rpo_vn): For conservatively handling a PHI only mark
3652         the backedge executable but not the block reachable.
3654 2018-09-12  Richard Biener  <rguenther@suse.de>
3656         PR tree-optimization/87266
3657         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
3658         visited blocks.
3660 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
3662         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
3663         constants.
3664         ("trunc<BFP:mode><DFP_ALL:mode>2")
3665         ("trunc<DFP_ALL:mode><BFP:mode>2")
3666         ("extend<BFP:mode><DFP_ALL:mode>2")
3667         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
3668         according to the target operand type.
3670 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3671             Andreas Krebbel  <krebbel@linux.ibm.com>
3673         PR tree-optimization/86844
3674         * gimple-ssa-store-merging.c
3675         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
3676         there are any overlapping stores in between them, make sure they are
3677         also coalesced or we give up completely.
3679 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3681         PR middle-end/87248
3682         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
3683         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
3685 2018-09-12  Tom de Vries  <tdevries@suse.de>
3687         * common.opt (gdescribe-dies): Add option.
3688         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
3689         attribute for artifical and nameless decls.
3690         (dwarf2out_register_external_die): Add description attribute to
3691         external reference die.
3692         (add_desc_attribute): New functions.
3693         (gen_subprogram_die): Add description attribute to
3694         DW_TAG_call_site_parameter.
3695         * tree-pretty-print.c (print_generic_expr_to_str): New function.
3696         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
3697         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
3698         -gno-describe-dies.
3699         (@item -gdescribe-dies): Add.
3701 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
3703         * tree-vrp.c (vrp_shift_undefined_p): Remove.
3704         (extract_range_from_binary_expr_1: Call
3705         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
3706         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
3707         depend on sign.
3709 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
3711         * gimple-ssa-warn-alloca.c
3712         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
3713         field for ALLOCA_BOUND_*_LARGE.
3715 2018-09-11  Nathan Sidwell  <nathan@acm.org>
3717         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
3719 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3721         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
3722         for clobbers.  Remove obsolete comment.
3724 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3726         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
3727         mpxchk, mpxld and mpxst types.
3728         (define_attr length_immediate): Remove all processing of mpx types.
3729         (define_attr prefix_0f): Ditto.
3730         (define_attr memory): Ditto.
3732 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3734         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
3735         (log<mode>2): Change operand 1 predicate to general_operand.
3736         Extend operand 1 to XFmode and generate logxf3 insn.
3737         (log10<mode>2): Change operand 1 predicate to general_operand.
3738         Extend operand 1 to XFmode and generate log10xf3 insn.
3739         (log2<mode>2): Change operand 1 predicate to general_operand.
3740         Extend operand 1 to XFmode and generate log2xf3 insn.
3741         (fyl2xp1_extend<mode>xf3_i387): Remove.
3742         (log1p<mode>2): Change operand 1 predicate to general_operand.
3743         Extend operand 1 to XFmode and generate log1pxf3 insn.
3744         (fxtract_extend<mode>xf3_i387): Remove.
3745         (logb<mode>2): Change operand 1 predicate to general_operand.
3746         Extend operand 1 to XFmode and generate logbxf3 insn.
3747         (ilogb<mode>2): Change operand 1 predicate to general_operand.
3748         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
3749         (significand<mode>2): Change operand 1 predicate to general_operand.
3750         Extend operand 1 to XFmode and generate significandxf3 insn.
3752 2018-09-11  Nathan Sidwell  <nathan@acm.org>
3754         * gcc.c (perror_with_name, pfatal_with_name): Delete.
3755         (load_specs): Use fatal_error.
3756         (DELETE_IF_ORDINARY, process_command): Use error.
3757         (execute, run_attempt): Use fatal_error.
3759 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
3761         * diagnostic-core.h (sorry_at): New prototype.
3762         * diagnostic.c (sorry_at): New function.
3764 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
3766         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
3767         by zero as VR_UNDEFINED.
3769 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
3771         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
3772         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
3773         (<sincos>mode2): New expander.
3774         (sincos_extend<mode>xf3_i387): Remove insn pattern.
3775         (sincos -> sin, cos splitters): Remove splitter patterns.
3776         (sincos<mode>3): Change operand 2 predicate to general_operand.
3777         Extend operand 2 to XFmode and generate sincosxf3 insn.
3778         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
3779         Change operand 3 predicate to const1_operand.
3780         (fptan_extend<mode>xf4_i387): Remove insn pattern.
3781         (tanxf2): Update operands in the call to fptanxf4_i387.
3782         (tan<mode>2): Change operand 1 predicate to general_operand.
3783         Extend operand 1 to XFmode and generate tanxf3 insn.
3784         (atan2xf3): Rename from *fpatanxf3_i387.
3785         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
3786         (atan2xf3): Remove expander.
3787         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
3788         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
3789         (atan<mode>2): Change operand 1 predicate to general_operand.
3790         Extend operand 1 to XFmode and generate atanxf3 insn.
3792 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
3794         * config/i386/i386.md (x87/SSE constant load splitter): Use
3795         memory_operand instead of nonimmediate_operand for input operand
3796         predicate.
3798 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
3800         * config/i386/i386.md (float partial SSE register stall splitter): Move
3801         splitter near its instruction pattern.
3802         (float_extend partial SSE register stall splitter): Ditto.
3803         (float_truncate partial SSE register stall splitter): Ditto.
3805 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
3807         PR target/86794
3808         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3809         to speculation_safe_value_not_needed.
3811         PR target/85666
3812         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
3813         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
3814         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
3815         leaf_function_p, instead use has_hard_reg_initial_val.
3817 2018-09-09  Nathan Sidwell  <nathan@acm.org>
3819         * gcc.h (pfatal_with_name): Don't declare here.
3820         * gcc.c (pfatal_with_name): Make static.
3822 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3824         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
3825         earlyclobber.
3827 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
3829         PR rtl-optimization/85458
3830         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
3831         priority hook to reduce the priority of EXPR.
3833 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
3835         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
3836         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
3837         with X87_ENABLE_FLOAT.  Remove preparation code.
3838         (*float<SWI48:mode><MODEF:mode>2): Rename from
3839         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
3840         math using "enabled" attribute.
3841         (*floatdi<MODEF:mode>2_i387): Rename from
3842         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
3843         enable for 32bit targets only.
3844         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
3845         splitter.
3846         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
3847         as operand 1 predicate.  Rewrite as define_insn_and_split.
3848         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
3850 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
3852         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
3853         to fallthru to FLOAT case.
3855 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3857         PR target/86731
3858         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
3859         around folding of vec_sl to handle out of range shift values.
3861 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
3863         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
3864         Update callers to gen_fix_trunc<mode>_i387_fisttp
3865         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
3866         nonimmediate_operand.
3867         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
3868         and corresponding splitters.
3869         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
3870         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
3871         (fix_truncdi_i387_with_temp): Remove insn pattern
3872         and corresponding splitters.
3873         (fix_trunc<mode>_i387): Change operand 0 predicate to
3874         nonimmediate_operand.
3875         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
3876         and corresponding splitters.
3877         (*fistdi2_1): Remove.
3878         (fistdi2): Ditto.
3879         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
3880         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
3881         (*fist<mode>2_1): Remove.
3882         (fist<mode>2): Ditto.
3883         (fist<mode>2_with_temp): Remove insn pattern and corresponding
3884         splitters.
3885         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
3886         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
3887         (fistdi2_<rounding>): Change operand 0 predicate to
3888         nonimmediate_operand.
3889         (fistdi2_<rounding>_with_temp): Remove insn pattern
3890         and corresponding splitters.
3891         (fist<mode>2_<rounding>): Change operand 0 predicate to
3892         nonimmediate_operand.
3893         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
3894         and corresponding splitters.
3896         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
3898 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3900         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
3901         the init value.
3903 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3905         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
3906         early gimple folding of vec_splat().
3907         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
3908         * gimple-fold.h: Add an extern define for tree_vec_extract().
3910 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3912         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
3913         wrappers around TREE_TYPE comparisons.
3915 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3917         PR target/80080
3918         * config/s390/predicates.md: Add nonsym_memory_operand.
3919         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
3920         contains a SYMBOL_REF, load it into an intermediate pseudo.
3921         (s390_emit_compare_and_swap): Legitimize operand.
3922         * config/s390/s390.md: Use the new nonsym_memory_operand
3923         with UNSPECV_CAS patterns.
3925 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3927         PR target/80080
3928         * config/s390/s390-passes.def: New file.
3929         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3930         declaration.
3931         (make_pass_s390_early_mach): Add declaration.
3932         * config/s390/s390.c (make_pass_s390_early_mach):
3933         (s390_option_override): Remove dynamic registration.
3934         * config/s390/t-s390: Add s390-passes.def.
3936 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3938         * config/s390/s390.c (s390_decompose_constant_pool_ref):
3939         Remove UNSPEC_LTREL_BASE check.
3940         (annotate_constant_pool_refs): Likewise.
3941         (find_constant_pool_ref): Likewise.
3942         (find_ltrel_base): Removed.
3943         (replace_ltrel_base): Removed.
3944         (s390_mainpool_finish): Remove replace_ltrel_base call.
3945         (s390_chunkify_start): Remove pending LTREL_BASE logic.
3946         (s390_chunkify_finish): Remove replace_ltrel_base call.
3947         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3949 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
3951         PR target/86779
3952         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3953         to speculation_safe_value_not_needed.
3955 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
3956             Bernd Schmidt  <bernds_cb1@t-online.de>
3958         * config/nvptx/nvptx-opts.h: New file.
3959         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3960         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3961         (ASM_SPEC): Define.
3962         (TARGET_SM35): New macro.
3963         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3964         correct predicate.
3965         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3966         values.
3967         (misa=): New option.
3968         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3970 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
3972         * config/i386/i386.md (truncdfsf2): Remove expander.
3973         (truncdfsf2_with_temp): Ditto.
3974         (truncxf<mode>2): Ditto.
3975         (*truncdfsf_fast_mixed): Remove insn pattern.
3976         (*truncdfsf_fast_i387): Ditto.
3977         (*truncdfsf_mixed): Ditto.
3978         (*truncdfsf_i387): Ditto.
3979         (*truncdfsf2_i387_1): Ditto.
3980         (*truncxfsf2_mixed): Ditto.
3981         (*truncxfdf2_mixed): Ditto.
3982         (*truncxf<mode>2_i387_noop): Ditto. Update callers
3983         to call gen_truncxf<mode>2 instead.
3984         (*truncxf<mode>2_i387): Remove.
3985         (reg->reg splitters): Remove splitter pattern.
3986         (reg->mem splitters): Ditto.
3988         (truncdfsf2): New insn pattern.
3989         (truncxf<mode>2): Ditto.
3991 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3993         * tree-ssa-math-opts.c (is_mult_by): New function.
3994         (is_square_of): Use the above.
3995         (optimize_recip_sqrt): New function.
3996         (pass_cse_reciprocals::execute): Use the above.
3998 2018-09-05  Richard Biener  <rguenther@suse.de>
4000         PR bootstrap/87134
4001         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
4002         to zero-init the emplaced vec.
4004 2018-09-05  Martin Liska  <mliska@suse.cz>
4006         PR tree-optimization/87205
4007         * tree-switch-conversion.c (pass_lower_switch::execute):
4008         Group cases for switch statements.
4010 2018-09-05  Richard Biener  <rguenther@suse.de>
4012         PR tree-optimization/87217
4013         * tree-ssa-sccvn.c (vuse_valueize): New.
4014         (vn_reference_lookup_pieces): Use it.
4015         (vn_reference_lookup): Likewise.
4017 2018-09-05  Nathan Sidwell  <nathan@acm.org>
4019         PR c++/87137
4020         * stor-layout.c (place_field): Scan forwards to check last
4021         bitfield when ms_bitfield_placement is in effect.
4023 2018-09-05  Richard Biener  <rguenther@suse.de>
4025         PR bootstrap/87225
4026         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
4027         return.
4029 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
4030             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
4032         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
4033         * config.gcc (extra_objs): Build it.
4034         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
4035         Likewise.
4036         * config/aarch64/aarch64-passes.def
4037         (pass_tag_collision_avoidance): New pass.
4038         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
4039         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
4040         (aarch64_classify_address): Remove static qualifier.
4041         (aarch64_address_info, aarch64_address_type): Move to...
4042         * config/aarch64/aarch64-protos.h: ... here.
4043         (make_pass_tag_collision_avoidance): New function.
4044         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
4045         New tuning flag.
4047 2018-09-05  Martin Liska  <mliska@suse.cz>
4049         * doc/gcov.texi: Update documentation of humar
4050         readable mode.
4051         * gcov.c (format_count): Print one decimal place, it provides
4052         more fine number of situations like '1G' vs. '1.4G'.
4054 2018-09-05  Martin Liska  <mliska@suse.cz>
4056         PR target/87164
4057         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
4058         * optc-gen.awk: Allow 'Var' for Deprecated options in order
4059         to generate a MASK value.
4061 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
4063         PR debug/86593
4064         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
4065         if frame pointer isn't used.
4066         (compute_frame_pointer_to_fb_displacement): Likewise.
4068 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
4070         PR target/87198
4071         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
4072         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
4073         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
4074         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
4075         and OPTION_MASK_ISA_XSAVEC_UNSET.
4077 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
4079         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
4080         XOR operations in NAND case.
4082 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4084         * wide-int-range.cc (wide_int_range_convert): New.
4085         * wide-int-range.h (wide_int_range_convert): New.
4086         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
4087         code into wide_int_range_convert.
4088         (extract_range_into_wide_ints): Do not munge anti range constants
4089         into the entire domain.  Just return the range back.
4091 2018-09-04  Martin Liska  <mliska@suse.cz>
4093         * genmatch.c (output_line_directive): Add new argument
4094         fnargs.
4095         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
4097 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4099         * doc/invoke.texi (Option Summary): Add whitespace.
4101         * doc/invoke.texi (Option Summary): Add -Waligned-new.
4103 2018-09-04  Richard Biener  <rguenther@suse.de>
4105         PR tree-optimization/87211
4106         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
4107         backedge value we're supposed to treat as VARYING also number
4108         the PHI to VARYING in case it got a different value-number already.
4110 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4112         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
4113         (extract_range_from_binary_expr_1): Do not call
4114         vrp_can_optimize_bit_op.
4115         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
4116         static.
4117         (wide_int_range_get_mask_and_bounds): New.
4118         (wide_int_range_optimize_bit_op): New.
4119         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4120         (wide_int_range_bit_and): Same.
4121         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4122         (wide_int_range_optimize_bit_op): New.
4123         (wide_int_range_get_mask_and_bounds): New.
4125 2018-09-04  Richard Biener  <rguenther@suse.de>
4127         PR tree-optimization/87176
4128         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4129         variable.  When value-numbering a virtual PHI node make sure
4130         to not value-number to the backedge value.
4132 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4134         * doc/extend.texi (Long Long, Hex Floats): Document support for
4135         long long and hex floats in more recent versions of ISO C++.
4137 2018-09-03  Richard Biener  <rguenther@suse.de>
4139         PR tree-optimization/87177
4140         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4141         cleanup.
4143 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4145         * bb-reorder.c (edge_order): Convert to C-qsort-style
4146         tri-state comparator.
4147         (reorder_basic_blocks_simple): Change std::stable_sort to
4148         gcc_stablesort.
4150 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4152         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4153         tri-state comparator.
4154         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4156 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4158         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
4159         (mergesort): ... here as maximum count for using netsort.
4160         (gcc_qsort): Set nlim to 3 if stable sort is requested.
4161         (gcc_stablesort): New.
4162         * system.h (gcc_stablesort): Declare.
4164 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4166         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
4167         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
4168         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
4170 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
4172         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
4173         lxsdx and stxsdx alternatives.
4174         (*mov<mode>_hardfloat64): Ditto.
4175         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4177 2018-09-03  Richard Biener  <rguenther@suse.de>
4179         PR tree-optimization/87200
4180         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4181         simplify result.
4183 2018-09-03  Martin Liska  <mliska@suse.cz>
4185         PR tree-optimization/87201
4186         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4187         Fix parenthesis in an expression.
4189 2018-09-03  Richard Biener  <rguenther@suse.de>
4191         PR tree-optimization/87197
4192         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4193         visited.  CSE the VN_INFO hashtable lookup.
4195         PR tree-optimization/87169
4196         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4197         iterating make sure there's no extra backedges from irreducible
4198         regions feeding the header.  Mark the destination block
4199         executable.
4201 2018-09-03  Martin Liska  <mliska@suse.cz>
4203         PR driver/83193
4204         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4205         * common/common-targhooks.c (default_get_valid_option_values):
4206         New function.
4207         * common/common-targhooks.h (default_get_valid_option_values):
4208         Likewise.
4209         * common/config/i386/i386-common.c: Move processor_target_table
4210         from i386.c.
4211         (ix86_get_valid_option_values): New function.
4212         (TARGET_GET_VALID_OPTION_VALUES): New macro.
4213         * config/i386/i386.c (struct ptt): Move to i386-common.c.
4214         (PTA_*): Move all defined masks into i386-common.c.
4215         (ix86_function_specific_restore): Use new processor_cost_table.
4216         * config/i386/i386.h (struct ptt): Moved from i386.c.
4217         (struct pta): Likewise.
4218         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4219         * doc/tm.texi.in: Likewise.
4220         * opt-suggestions.c (option_proposer::suggest_option):
4221         Pass prefix to build_option_suggestions.
4222         (option_proposer::get_completions): Likewise.
4223         (option_proposer::build_option_suggestions): Use the new target
4224         hook.
4225         * opts.c (struct option_help_tuple): New struct.
4226         (print_filtered_help): Use the new target hook.
4228 2018-09-03  Martin Liska  <mliska@suse.cz>
4230         PR middle-end/59521
4231         * predict.c (set_even_probabilities): Add likely_edges
4232         argument and handle cases where we have precisely one
4233         likely edge.
4234         (combine_predictions_for_bb): Catch also likely_edges.
4235         (tree_predict_by_opcode): Handle gswitch statements.
4236         * tree-cfg.h (find_case_label_for_value): New declaration.
4237         (find_taken_edge_switch_expr): Likewise.
4238         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4239         Find pivot in decision tree based on probabily, not by number of
4240         nodes.
4242 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
4244         * doc/standards.texi (Standards): Update Objective-C reference.
4246 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4248         * doc/install.texi (Prerequisites): Update link for MPC.
4250 2018-09-01  Michael Matz  <matz@suse.de>
4252         PR tree-optimization/87074
4253         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
4254         PHIs for outer-loop uses.
4256 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4258         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
4259         * doc/invoke.texi (C Dialect Options): Ditto.
4261 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4263         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
4265 2018-08-31  Richard Biener  <rguenther@suse.de>
4267         PR tree-optimization/87168
4268         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
4269         (rpo_elim::eliminate_avail): When OP was not visited it must
4270         be available.
4272 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
4274         * tree-vrp.c (copy_value_range): Convert param "from" from
4275         "value_range *" to "const value_range *".
4276         (range_is_null): Likewise for param "vr".
4277         (range_int_cst_p): Likewise.
4278         (range_int_cst_singleton_p): Likewise.
4279         (symbolic_range_p): Likewise.
4280         (value_ranges_intersect_p): Likewise for both params.
4281         (value_range_nonnegative_p): Likewise for param "vr".
4282         (value_range_constant_singleton): Likewise.
4283         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
4284         (extract_range_into_wide_ints): Likewise for param "vr".
4285         (extract_range_from_multiplicative_op): Likewise for params "vr0"
4286         and "vr1".
4287         (vrp_can_optimize_bit_op): Likewise.
4288         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
4289         "vr1_".
4290         (extract_range_from_unary_expr): Likewise.
4291         (debug_value_range): Likewise for param "vr".
4292         (value_range::dump): Add "const" qualifier.
4293         (vrp_prop::check_array_ref): Convert local "vr" from
4294         "value_range *" to "const value_range *".
4295         (vrp_prop::check_mem_ref): Likewise.
4296         (vrp_prop::visit_stmt): Likewise for local "old_vr".
4297         (vrp_intersect_ranges_1): Likewise for param "vr_1".
4298         (vrp_intersect_ranges): Likewise.
4299         (simplify_stmt_for_jump_threading): Likewise for local "vr".
4300         (vrp_prop::vrp_finalize): Likewise.
4301         * tree-vrp.h (value_range::dump): Add "const" qualifier.
4302         (vrp_intersect_ranges): Add "const" qualifier to params as above.
4303         (extract_range_from_unary_expr): Likewise.
4304         (value_range_constant_singleton): Likewise.
4305         (symbolic_range_p): Likewise.
4306         (copy_value_range): Likewise.
4307         (extract_range_from_binary_expr_1): Likewise.
4308         (range_int_cst_p): Likewise.
4309         (vrp_set_zero_nonzero_bits): Likewise.
4310         (range_int_cst_singleton_p): Likewise.
4312 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
4314         * config/aarch64/arm_neon.h (vabsd_s64): New.
4315         (vnegd_s64): Likewise.
4317 2018-08-31  Martin Jambor  <mjambor@suse.cz>
4319         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
4321 2018-08-31  Martin Liska  <mliska@suse.cz>
4323         * ipa-icf.c (sem_item::add_type): Use
4324         sem_item::m_type_hash_cache.
4325         * ipa-icf.h: Move the cache from sem_item_optimizer
4326         to sem_item.
4328 2018-08-31  Nathan Sidwell  <nathan@acm.org>
4330         * doc/extend.texi (Backwards Compatibility): Remove implicit
4331         extern C leeway of () being (...).
4333 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4335         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
4337 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
4339         PR target/86684
4340         PR target/87149
4341         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
4343 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
4345         PR middle-end/87138
4346         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
4347         gen_int_mode.  Formatting fixes.
4349 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
4351         * target.def (custom_function_descriptors): Improve documentation.
4352         * doc/tm.texi.in (Trampolines): Expand discussion of function
4353         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
4354         beginning of the section.
4355         * doc/tm.texi: Regenerated.
4357 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
4359         * cfg.h (class auto_edge_flag): Spell out the template-id of the
4360         base class in the initializer list.  This is a workaround for
4361         building with older GCC.
4362         (class auto_bb_flag): Likewise.
4364 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
4366         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
4367         (altivec_vcmpequ<VI_char>_p): Remove star.
4368         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
4369         vector load modes.
4370         (expand_strncmp_vec_sequence): New function.
4371         (emit_final_str_compare_vec): New function.
4372         (expand_strn_compare): Add support for vector strncmp.
4373         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
4374         length specification to bytes.
4375         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
4376         (vcmpnezb_p): New pattern.
4377         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
4378         for option -mstring-compare-inline-limit.
4380 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
4382         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
4383         (PTA_SKYLAKE): Add PTA_AES.
4384         (PTA_GOLDMONT): Likewise.
4386 2018-08-29  Jan Hubicka  <jh@suse.cz>
4388         PR lto/86517
4389         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
4390         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
4392 2018-08-29  Jan Hubicka  <jh@suse.cz>
4394         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
4395         TYPE_STUB_DECL.
4396         (hash_tree): Do not visit TYPE_STUB_DECL.
4397         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
4398         stream TYPE_STUB_DECL.
4399         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
4400         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
4401         after free_lang_data.
4402         (type_in_anonymous_namespace_p): Likewise.
4404 2018-08-29  Jan Hubicka  <jh@suse.cz>
4406         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4407         comment that it has to be even number.
4408         (class sreal): Change m_sig type to int32_t.
4409         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4410         int64_t for temporary calculations.
4411         (sreal_verify_basics): Drop one bit from minimum and maximum.
4413 2018-08-30  Richard Biener  <rguenther@suse.de>
4415         PR tree-optimization/87147
4416         * tree-ssa-sccvn.c (SSA_VISITED): New function.
4417         (visit_phi): When the degenerate result is from the backedge and
4418         we didn't visit its definition yet drop to VARYING.
4419         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
4421 2018-08-29  Jan Hubicka  <jh@suse.cz>
4423         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
4424         DECL_VINDEX.
4425         (hash_tree): Likewise.
4427 2018-08-29  Jan Hubicka  <jh@suse.cz>
4429         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
4430         and TYPE_NEXT_REF_TO.
4432 2018-08-29  Jan Hubicka  <jh@suse.cz>
4434         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4435         comment that it has to be even number.
4436         (class sreal): Change m_sig type to int32_t.
4437         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4438         int64_t for temporary calculations.
4439         (sreal_verify_basics): Drop one bit from minimum and maximum.
4441 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
4443         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
4445 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
4447         PR middle-end/86995
4448         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
4449         if to_add is negative.
4451 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4453         PR middle-end/87053
4454         * builtins.c (c_strlen): Improve range checks.
4456 2018-08-29  Martin Sebor  <msebor@redhat.com>
4457             Jeff Law  <law@redhat.com>
4459         PR tree-optimization/86714
4460         PR tree-optimization/86711
4461         * builtins.c (c_strlen): Add arguments to call to string_constant.
4462         * expr.c (string_constant): Add argument.  Detect missing nul
4463         terminator and outermost declaration it's missing in.
4464         * expr.h (string_constant): Add argument.
4465         * fold-const.c (read_from_constant_string): Add arguments to call to
4466         string_constant.
4467         (c_getstr): Likewise.
4468         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4469         to string_constant.
4470         * tree-ssa-strlen.c (get_stridx): Likewise.
4472 2018-08-29  Jan Hubicka  <jh@suse.cz>
4474         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
4475         Do not stream DECL_VINDEX.
4476         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
4477         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
4478         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
4480 2018-08-29  Richard Biener  <rguenther@suse.de>
4482         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
4483         virtual operands that are not default defs to honor region
4484         boundaries.
4485         (rpo_vn_valueize): Remove ineffective code here.
4487 2018-08-29  Richard Biener  <rguenther@suse.de>
4489         PR tree-optimization/87132
4490         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
4491         when skipping defs reachable over backedges.
4493 2018-08-29  Richard Biener  <rguenther@suse.de>
4495         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
4496         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
4497         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
4498         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
4499         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
4500         (vn_reference_lookup_3): Remove use of const_parms.
4501         (free_rpo_vn): Do not free const_parms.
4502         (do_rpo_vn): Do not call init_const_parms.
4503         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
4504         SSA_NAME_POINTS_TO_READONLY_MEMORY.
4505         (call_may_clobber_ref_p_1): Likewise.
4507 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
4509         PR other/86726
4510         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
4511         (-O): Ditto.
4512         (-ftree-scev-cprop): Document.
4514 2018-08-29  Jan Hubicka  <jh@suse.cz>
4516         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
4517         parameters.
4518         (sreal constructor): Update.
4519         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
4520         sreal:operator/): Update.
4522 2018-08-29  Martin Liska  <mliska@suse.cz>
4524         * tree-switch-conversion.c (switch_conversion::expand):
4525         Strenghten assumption about gswitch statements.
4527 2018-08-29  Richard Biener  <rguenther@suse.de>
4529         PR tree-optimization/87117
4530         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
4531         re-value-number released SSA VDEFs.
4533 2018-08-29  Richard Biener  <rguenther@suse.de>
4535         PR tree-optimization/87126
4536         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
4538 2018-08-28  Jim Wilson  <jimw@sifive.com>
4540         * config/riscv/pic.md: Rewrite.
4541         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
4542         invalid address.
4543         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
4544         (SOFTF, default_load, softload, softstore): New.
4546 2018-08-28  Jeff Law  <law@redhat.com>
4548         * fold-const.c (fold_binary_loc): Remove recently added assert.
4550 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
4552         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
4553         to OP parmeter of generated function.
4555 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
4557         PR tree-optimization/87009
4558         * match.pd: Add boolean optimizations.
4560 2018-08-28  Martin Sebor  <msebor@redhat.com>
4562         PR middle-end/86631
4563         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
4564         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
4565         (pass_walloca::gate): Use it.
4566         (alloca_call_type): Same.
4567         (pass_walloca::execute): Same.
4568         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
4570 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
4572         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
4573         GCC_VERSION for usage of "__gcc_dump_printf__" format from
4574         >= 3005 to >= 9000.
4576 2018-08-28  Richard Biener  <rguenther@suse.de>
4578         PR tree-optimization/87124
4579         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
4580         constants before looking up avail.
4582 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
4584         PR middle-end/87099
4585         * calls.c (maybe_warn_nonstring_arg): Punt early if
4586         warn_stringop_overflow is zero.  Don't call get_range_strlen
4587         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
4588         Swap comparison operands to have constants on rhs.  Only use
4589         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
4590         increment lenrng[0].
4592 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
4594         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
4595         use of tree_to_shwi.  Remove duplicated test for the size being
4596         a whole number of bytes.
4598 2018-08-28  Richard Biener  <rguenther@suse.de>
4600         PR tree-optimization/87117
4601         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
4602         Handle removed stmt without LHS (GIMPLE_NOP).
4604 2018-08-28  Richard Biener  <rguenther@suse.de>
4606         PR tree-optimization/87117
4607         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
4608         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
4610 2018-08-28  Richard Biener  <rguenther@suse.de>
4612         PR tree-optimization/87117
4613         * tree-ssa-pre.c (compute_avail): Do not make expressions
4614         with predicated values available.
4615         (get_expr_value_id): Assert we do not run into predicated value
4616         expressions.
4618 2018-08-28  Richard Biener  <rguenther@suse.de>
4620         PR tree-optimization/87117
4621         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
4622         get virtual operands.
4623         (get_expr_operands): Handle STRING_CST like other decls.
4625 2018-08-28  Martin Liska  <mliska@suse.cz>
4627         * tree.h: Update documentation of fndecl_built_in_p
4628         functions.
4631 2018-08-27  Jeff Law  <law@redhat.com>
4632         PR tree-optimization/87110
4633         * tree-ssa-dse.c (compute_trims): Handle non-constant
4634         TYPE_SIZE_UNIT.
4636 2018-08-27  Martin Sebor  <msebor@redhat.com>
4638         PR tree-optimization/86914
4639         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
4641 2018-08-27  Martin Sebor  <msebor@redhat.com>
4643         PR tree-optimization/87112
4644         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
4645         the type of the bound argument.
4647 2018-08-27  Jeff Law  <law@redhat.com>
4649         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
4650         type does not have a TYPE_SIZE_UNIT.
4652 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
4654         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
4655         with include of backend.h.
4657 2018-08-27  Richard Biener  <rguenther@suse.de>
4659         PR tree-optimization/86927
4660         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
4661         use const cond reduction code.
4663 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
4665         PR tree-optimization/85758
4666         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
4668 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
4670         PR c++/87091
4671         * diagnostic-show-locus.c (class layout_range): Update for
4672         conversion of show_caret_p to a tri-state.
4673         (layout_range::layout_range): Likewise.
4674         (make_range): Likewise.
4675         (layout::maybe_add_location_range): Likewise.
4676         (layout::should_print_annotation_line_p): Don't show annotation
4677         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
4678         (layout::get_state_at_point): Update for conversion of
4679         show_caret_p to a tri-state.  Bail out early for
4680         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
4681         underlining or source colorization.
4682         (gcc_rich_location::add_location_if_nearby): Update for conversion
4683         of show_caret_p to a tri-state.
4684         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
4685         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
4686         Likewise.
4687         (selftest::test_one_liner_labels): Likewise.
4688         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
4689         conversion of show_caret_p to a tri-state.
4690         * pretty-print.c (text_info::set_location): Likewise.
4691         * pretty-print.h (text_info::set_location): Likewise.
4692         * substring-locations.c (format_warning_n_va): Likewise.
4693         * tree-diagnostic.c (default_tree_printer): Likewise.
4694         * tree-pretty-print.c (newline_and_indent): Likewise.
4696 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
4698         PR c++/87091
4699         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
4700         line above for line-insertion fix-it hints.
4701         (selftest::test_fixit_insert_containing_newline): Update the
4702         expected results, and add a test with line-numbering enabled.
4704 2018-08-27  Martin Liska  <mliska@suse.cz>
4706         PR sanitizer/86962
4707         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
4708         params with DECL_HAS_VALUE_EXPR_P.
4710 2018-08-27  Martin Liska  <mliska@suse.cz>
4712         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
4713         selected expansion strategy.
4715 2018-08-27  Martin Liska  <mliska@suse.cz>
4717         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
4718         * builtins.c (is_builtin_fn): Likewise.
4719         * attribs.c (diag_attr_exclusions): Use new function
4720         fndecl_built_in_p and remove check for FUNCTION_DECL if
4721         possible.
4722         (builtin_mathfn_code): Likewise.
4723         (fold_builtin_expect): Likewise.
4724         (fold_call_expr): Likewise.
4725         (fold_builtin_call_array): Likewise.
4726         (fold_call_stmt): Likewise.
4727         (set_builtin_user_assembler_name): Likewise.
4728         (is_simple_builtin): Likewise.
4729         * calls.c (gimple_alloca_call_p): Likewise.
4730         (maybe_warn_nonstring_arg): Likewise.
4731         * cfgexpand.c (expand_call_stmt): Likewise.
4732         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
4733         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
4734         (cgraph_node::verify_node): Likewise.
4735         * cgraphclones.c (build_function_decl_skip_args): Likewise.
4736         (cgraph_node::create_clone): Likewise.
4737         * config/arm/arm.c (arm_insert_attributes): Likewise.
4738         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
4739         * dse.c (scan_insn): Likewise.
4740         * expr.c (expand_expr_real_1): Likewise.
4741         * fold-const.c (operand_equal_p): Likewise.
4742         (fold_binary_loc): Likewise.
4743         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
4744         * gimple-low.c (lower_stmt): Likewise.
4745         * gimple-pretty-print.c (dump_gimple_call): Likewise.
4746         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
4747         Likewise.
4748         * gimple.c (gimple_build_call_from_tree): Likewise.
4749         (gimple_call_builtin_p): Likewise.
4750         (gimple_call_combined_fn): Likewise.
4751         * gimplify.c (gimplify_call_expr): Likewise.
4752         (gimple_boolify): Likewise.
4753         (gimplify_modify_expr): Likewise.
4754         (gimplify_addr_expr): Likewise.
4755         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
4756         * ipa-cp.c (determine_versionability): Likewise.
4757         * ipa-fnsummary.c (compute_fn_summary): Likewise.
4758         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
4759         * ipa-split.c (visit_bb): Likewise.
4760         (split_function): Likewise.
4761         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4762         * lto-cgraph.c (input_node): Likewise.
4763         * lto-streamer-out.c (write_symbol): Likewise.
4764         * omp-low.c (setjmp_or_longjmp_p): Likewise.
4765         (lower_omp_1): Likewise.
4766         * predict.c (strip_predict_hints): Likewise.
4767         * print-tree.c (print_node): Likewise.
4768         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
4769         * trans-mem.c (is_tm_irrevocable): Likewise.
4770         (is_tm_load): Likewise.
4771         (is_tm_simple_load): Likewise.
4772         (is_tm_store): Likewise.
4773         (is_tm_simple_store): Likewise.
4774         (is_tm_abort): Likewise.
4775         (tm_region_init_1): Likewise.
4776         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
4777         * tree-cfg.c (verify_gimple_call): Likewise.
4778         (move_stmt_r): Likewise.
4779         (stmt_can_terminate_bb_p): Likewise.
4780         * tree-eh.c (lower_eh_constructs_2): Likewise.
4781         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
4782         * tree-inline.c (remap_gimple_stmt): Likewise.
4783         (copy_bb): Likewise.
4784         (estimate_num_insns): Likewise.
4785         (fold_marked_statements): Likewise.
4786         * tree-sra.c (scan_function): Likewise.
4787         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
4788         (optimize_stack_restore): Likewise.
4789         (pass_fold_builtins::execute): Likewise.
4790         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
4791         (mark_all_reaching_defs_necessary_1): Likewise.
4792         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
4793         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4794         (pass_forwprop::execute): Likewise.
4795         * tree-ssa-loop-im.c (stmt_cost): Likewise.
4796         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4797         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
4798         * tree-ssa-strlen.c (get_string_length): Likewise.
4799         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
4800         (find_func_aliases_for_call): Likewise.
4801         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
4802         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
4803         * tree-tailcall.c (find_tail_calls): Likewise.
4804         * tree.c (need_assembler_name_p): Likewise.
4805         (free_lang_data_in_decl): Likewise.
4806         (get_call_combined_fn): Likewise.
4807         * ubsan.c (is_ubsan_builtin_p): Likewise.
4808         * varasm.c (incorporeal_function_p): Likewise.
4809         * tree.h (DECL_BUILT_IN): Remove and replace with
4810         fndecl_built_in_p.
4811         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
4812         (fndecl_built_in_p): New.
4814 2018-08-27  Martin Liska  <mliska@suse.cz>
4816         PR tree-optimization/86847
4817         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
4818         Dump also subtree probability.
4819         (switch_decision_tree::do_jump_if_equal): New function.
4820         (switch_decision_tree::emit_case_nodes): Handle special
4821         situations in balanced tree that can be emitted much simpler.
4822         Fix calculation of probabilities that happen in tree expansion.
4823         * tree-switch-conversion.h (struct cluster): Add
4824         is_single_value_p.
4825         (struct simple_cluster): Likewise.
4826         (struct case_tree_node): Add new function has_child.
4827         (do_jump_if_equal): New.
4829 2018-08-27  Martin Liska  <mliska@suse.cz>
4831         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
4832         Add new argument to bit_test_cluster constructor.
4833         (bit_test_cluster::emit): Set bits really number of values
4834         handlel by a test.
4835         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
4836         probability argument.
4837         * tree-switch-conversion.h (struct bit_test_cluster):
4838         Add m_handles_entire_switch.
4840 2018-08-27  Martin Liska  <mliska@suse.cz>
4842         PR tree-optimization/86702
4843         * tree-switch-conversion.c (jump_table_cluster::emit):
4844         Make probabilities even for values in jump table
4845         according to number of cases handled.
4846         (switch_decision_tree::compute_cases_per_edge): Pass
4847         argument to reset_out_edges_aux function.
4848         (switch_decision_tree::analyze_switch_statement): Likewise.
4849         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
4850         Make it static.
4852 2018-08-27  Martin Liska  <mliska@suse.cz>
4854         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
4855         cfun argument explicitly.
4856         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
4857         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
4858         function gimple_switch_default_bb.
4859         (convert_switch_statements):
4860         (expand_builtins):
4861         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
4862         * stmt.c (label_to_block_fn): Use label_to_block and pass
4863         cfun argument explicitly and use gimple_switch_label_bb.
4864         (expand_case): Likewise.
4865         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
4866         cfun argument explicitly. Likewise.
4867         (make_edges_bb): Likewise.
4868         (make_cond_expr_edges): Likewise.
4869         (get_cases_for_edge): Likewise.
4870         (make_gimple_switch_edges): Likewise.
4871         (label_to_block_fn): Likewise.
4872         (label_to_block): Likewise.
4873         (make_goto_expr_edges): Likewise.
4874         (make_gimple_asm_edges): Likewise.
4875         (main_block_label): Likewise.
4876         (group_case_labels_stmt): Likewise.
4877         (find_taken_edge_computed_goto): Likewise.
4878         (find_taken_edge_switch_expr): Likewise.
4879         (gimple_verify_flow_info): Likewise.
4880         (gimple_redirect_edge_and_branch): Likewise.
4881         (gimple_switch_label_bb): New function.
4882         (gimple_switch_default_bb): Likewise.
4883         (gimple_switch_edge): Likewise.
4884         (gimple_switch_default_edge): Likewise.
4885         * tree-cfg.h (label_to_block_fn): Remove and replace ...
4886         (label_to_block): ... with this.
4887         (gimple_switch_label_bb): New.
4888         (gimple_switch_default_bb): Likewise.
4889         (gimple_switch_edge): Likewise.
4890         (gimple_switch_default_edge): Likewise.
4891         * tree-cfgcleanup.c (convert_single_case_switch): Use
4892         new gimple functions and pass new argument to label_to_block.
4893         (cleanup_control_flow_bb):
4894         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
4895         cfun argument explicitly.
4896         (make_eh_edges): Likewise.
4897         (redirect_eh_dispatch_edge): Likewise.
4898         (lower_resx): Likewise.
4899         (lower_eh_dispatch): Likewise.
4900         (maybe_remove_unreachable_handlers): Likewise.
4901         (unsplit_eh): Likewise.
4902         (cleanup_empty_eh): Likewise.
4903         (verify_eh_edges): Likewise.
4904         (verify_eh_dispatch_edge): Likewise.
4905         * tree-ssa-dom.c (record_edge_info): Likewise.
4906         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
4907         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
4908         (thread_through_normal_block): Likewise.
4909         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
4910         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
4911         * tree-switch-conversion.c (switch_conversion::collect): Use new
4912         gimple functions.
4913         (switch_conversion::check_final_bb): Likewise.
4914         (switch_conversion::gather_default_values): Pass new argument
4915         to label_to_block.
4916         (switch_conversion::build_constructors): Likewise.
4917         (switch_decision_tree::compute_cases_per_edge): Use new
4918         gimple_switch_edge function.
4919         (switch_decision_tree::analyze_switch_statement): Pass new argument
4920         to label_to_block.
4921         (switch_decision_tree::try_switch_expansion): Use
4922         gimple_switch_default_edge.
4923         * tree-vrp.c (find_switch_asserts): Pass new argument
4924         to label_to_block.
4925         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4926         (vr_values::simplify_switch_using_ranges): Likewise.
4928 2018-08-27  Richard Biener  <rguenther@suse.de>
4930         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4931         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4933         * tree-ssa-sccvn.h (struct vn_pval): New structure.
4934         (struct vn_nary_op_s): Add unwind_to member.  Add
4935         predicated_values flag and put result into a union together
4936         with a linked list of vn_pval.
4937         (struct vn_ssa_aux): Add name member to make maintaining
4938         a map of SSA name to vn_ssa_aux possible.  Remove no longer
4939         needed info, dfsnum, low, visited, on_sccstack, use_processed
4940         and range_info_anti_range_p members.
4941         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4942         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4943         New functions.
4944         (vn_valueize): New global.
4945         (vn_context_bb): Likewise.
4946         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4947         VN_INFO_PTR_INFO): Remove.
4948         * tree-ssa-sccvn.c: ... (rewrite)
4949         (pass_fre::execute): For -O2+ initialize loops and run
4950         RPO VN in optimistic mode (iterating).  For -O1 and -Og
4951         run RPO VN in non-optimistic mode.
4952         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4953         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4954         * doc/invoke.texi (sccvn-max-scc-size): Remove.
4955         (rpo-vn-max-loop-depth): Document.
4956         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4957         when valuezing the VUSE signals we walked out of the region.
4958         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4959         (phi_translate): Set VN context block to use for availability
4960         lookup.
4961         (compute_avail): Likewise.
4962         (pre_valueize): New function.
4963         (pass_pre::execute): Adjust to the RPO VN API.
4965         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4966         (propagate_constants_for_unrolling): Remove.
4967         (tree_unroll_loops_completely): Perform value-numbering
4968         on the unrolled bodies loop parent.
4970 2018-08-27  Richard Biener  <rguenther@suse.de>
4972         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4973         for partial antic compute.
4975 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
4977         PR rtl-optimization/87065
4978         * combine.c (simplify_if_then_else): Formatting fix.
4979         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4980         check.
4981         (known_cond): Don't return const_true_rtx for vector modes.  Use
4982         CONST0_RTX instead of const0_rtx.  Formatting fixes.
4984 2018-08-27  Martin Liska  <mliska@suse.cz>
4986         PR gcov-profile/87069
4987         * gcov.c (process_file): Record files already processed
4988         and warn about a file being processed multiple times.
4990 2018-08-27  Martin Liska  <mliska@suse.cz>
4992         PR driver/83193
4993         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4994         Set default values for x_aarch64_*_string strings.
4995         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4996         prefix.  For -mabi do not print '=ABI' in help and use
4997         <option_value> format for -msve-vector-bits and -moverride
4998         options.
5000 2018-08-26  Jeff Law  <law@redhat.com>
5002         * config/mips/frame-header-opt.c: Include "backend.h" rather than
5003         "cfg.h"
5005 2018-08-26  Marek Polacek  <polacek@redhat.com>
5007         PR c++/87029, Implement -Wredundant-move.
5008         * doc/invoke.texi: Document -Wredundant-move.
5010 2018-08-25  Martin Sebor  <msebor@redhat.com>
5012         PR tree-optimization/87059
5013         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
5014         to the same type as the other.
5015         * fold-const.c (fold_binary_loc): Assert expectation.
5017 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
5019         * config/darwin.c (machopic_legitimize_pic_address): Clean up
5020         extraneous parentheses, dead code section and formatting.
5022 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
5024         PR c++/87091
5025         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
5026         wide enough for jumps in the line-numbering to be visible.
5027         (layout::print_gap_in_line_numbering): New member function.
5028         (layout::calculate_line_spans): When using line numbering, merge
5029         line spans that are only 1 line apart.
5030         (diagnostic_show_locus): When printing line numbers, show gaps in
5031         line numbering directly, rather than printing headers.
5032         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
5033         line-numbering with multiple line spans.
5034         (selftest::test_fixit_insert_containing_newline_2): Add test of
5035         line-numbering, in which the spans are close enough to be merged.
5037 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5039         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
5040         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
5041         * tree-vrp.c (range_is_nonnull): Remove.
5042         (range_includes_zero_p): Accept value_range instead of min/max.
5043         (extract_range_from_binary_expr_1): Do not early bail on
5044         POINTER_PLUS_EXPR.
5045         Use range_includes_zero_p instead of range_is_nonnull.
5046         (extract_range_from_unary_expr): Use range_includes_zero_p instead
5047         of range_is_nonnull.
5048         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
5049         special case VR_ANTI_RANGE.
5050         (vrp_finalize): Same.
5051         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
5052         instead of min/max.
5053         (range_is_nonnull): Remove.
5054         * vr-values.c (vrp_stmt_computes_nonzero): Use
5055         range_includes_zero_p instead of range_is_nonnull.
5056         (extract_range_basic): Pass value_range to range_includes_zero_p
5057         instead of range_is_nonnull.
5059 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
5061         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
5062         * emit-rtl.h (rtl_data): Remove return_bnd.
5063         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
5064         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
5065         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
5066         (POINTER_BOUNDS_MODE): Remove definition.
5067         (make_pointer_bounds_mode): Remove.
5068         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
5069         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
5070         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
5071         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
5072         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
5073         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
5074         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
5076         * config/i386/i386-modes.def (BND32, BND64): Remove.
5077         * config/i386/i386.c (dbx_register_map): Remove bound registers.
5078         (dbx64_register_map): Ditto.
5079         (svr4_dbx_register_map): Ditto.
5080         (indirect_thunk_bnd_needed): Remove.
5081         (indirect_thunks_bnd_used): Ditto.
5082         (indirect_return_bnd_needed): Ditto.
5083         (indirect_return_via_cx_bnd): Ditto.
5084         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
5085         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
5086         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
5087         (output_indirect_thunk_function): Remove handling of
5088         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
5089         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
5090         (ix86_save_reg): Remove handling of crtl->return_bnd.
5091         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
5092         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
5093         and UNSPEC_BNDLX_ADDR.
5094         (ix86_output_indirect_branch_via_reg): Remove handling of
5095         indirect_thunk_prefix_bnd.
5096         (ix86_output_indirect_branch_via_push): Ditto.
5097         (ix86_output_function_return): Ditto.
5098         (ix86_output_indirect_function_return): Ditto.
5099         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
5100         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
5101         (CALL_USED_REGISTERS): Ditto.
5102         (REG_ALLOC_ORDER): Update for removal of bound registers.
5103         (HI_REGISTER_NAMES): Ditto.
5104         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
5105         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
5106         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
5107         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
5108         (FIRST_PSEUDO_REG): Update.
5109         (BND): Remove mode iterator.
5110         * config/i386/predicates.md (bnd_mem_operator): Remove.
5112 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5114         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
5115         vectors.
5117 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5119         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5120         the case in which the permute needs only a single element and
5121         repeats for every vector of the result.  Extend that case to
5122         handle variable-length vectors.
5123         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5125 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
5127         PR debug/79342
5128         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5129         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5131 2018-08-24  Richard Biener  <rguenther@suse.de>
5133         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5134         bb_flags_allocated members.
5135         (auto_flag): New RAII class for allocating flags.
5136         (auto_edge_flag): New RAII class for allocating edge flags.
5137         (auto_bb_flag): New RAII class for allocating bb flags.
5138         * cfgloop.c (verify_loop_structure): Allocate temporary edge
5139         flag dynamically.
5140         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5141         in favor of temporarily allocated BB flag.
5142         * hsa-brig.c: Re-order includes.
5143         * hsa-dump.c: Likewise.
5144         * hsa-regalloc.c: Likewise.
5145         * print-rtl.c: Likewise.
5146         * profile-count.c: Likewise.
5148 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
5150         PR target/86989
5151         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5152         the TOC register.
5154 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5156         PR 87073/bootstrap
5157         * wide-int-range.cc (wide_int_range_div): Do not ignore result
5158         from wide_int_range_multiplicative_op.
5160 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5162         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5163         "permutaion".
5165 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5167         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
5168         to 'expanded'.
5170 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
5172         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
5173         full GENERIC expression used for replacement.
5175 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
5177         * tree-vrp.c (abs_extent_range): Remove.
5178         (extract_range_into_wide_ints): Pass wide ints by reference.
5179         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5180         Pass wide ints by reference in all calls to
5181         extract_range_into_wide_ints.
5182         * wide-int-range.cc (wide_int_range_div): New.
5183         * wide-int-range.h (wide_int_range_div): New.
5184         (wide_int_range_includes_zero_p): New.
5185         (wide_int_range_zero_p): New.
5187 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
5189         * config/aarch64/aarch64.md (arches): New enum.
5190         (arch): New enum attr.
5191         (arch_enabled): New attr.
5192         (enabled): Now uses arch_enabled only.
5193         (simd, sve, fp16): Removed attribute.
5194         (fp): Attr now defined in terms of 'arch'.
5195         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5196         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5197         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5198         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5199         attributes into 'arch'.
5200         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5201         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5202         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5203         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5204         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5205         'simd' attribute into 'arch'.
5206         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5207         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5208         Convert use of 'fp' attribute to 'arch'.
5209         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5210         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5211         into 'arch'.
5212         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5213         (different modes) Merge 'fp' and 'simd' into 'arch'.
5214         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5215         'simd' into 'arch'.
5217 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
5219         PR rtl-optimization/87026
5220         * expmed.c (canonicalize_comparison): If we can no longer create
5221         pseudoregisters, don't.
5223 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
5225         PR target/86951
5226         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
5227         prototype.
5228         * config/arm/arm.c (speculation_barrier_libfunc): New static
5229         variable.
5230         (arm_init_libfuncs): Initialize it.
5231         (arm_emit_speculation_barrier): New function.
5232         * config/arm/arm.md (speculation_barrier): Call
5233         arm_emit_speculation_barrier for architectures that do not have
5234         DSB or ISB.
5235         (speculation_barrier_insn): Only match on Armv7 or later.
5237 2018-08-23  Richard Biener  <rguenther@suse.de>
5239         PR middle-end/87024
5240         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
5241         calls.
5243 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5245         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
5246         of single-vector TBLs.
5247         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
5248         one input is given.
5250 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5252         PR target/85910
5253         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
5254         aarch64_evpc_tbl guard.
5256 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5258         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
5259         behaviour.
5261 2018-08-22  Martin Sebor  <msebor@redhat.com>
5263         PR middle-end/87052
5264         * tree-pretty-print.c (pretty_print_string): Add argument.
5265         (dump_generic_node): Call to pretty_print_string with string size.
5267 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
5269         PR rtl-optimization/86771
5270         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
5271         of two SETs into those two SETs, one to be placed at i2, if that SETs
5272         destination is modified between i2 and i3.
5274 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5276         PR tree-optimization/86725
5277         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
5278         function.
5279         (vect_analyze_scalar_cycles_1): Check it.
5281 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5283         PR tree-optimization/86725
5284         * tree-vect-loop.c (vect_is_simple_reduction): When treating
5285         an outer loop phi as a double reduction, make sure that the
5286         single user of the phi result is an inner loop phi.
5288 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5290         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
5291         grouped stores with gaps to a strided group.
5293 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5295         * tree-vect-stmts.c (get_group_load_store_type)
5296         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
5297         first statement in a group.
5299 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5301         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
5302         the sequence used in gcc/gcc.c.
5304 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5306         PR other/704
5307         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
5308         building it.
5310 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5312         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
5313         Darwin10-specific unwinder-shim.
5314         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
5315         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
5316         New to cater for Darwin10 Rosetta.
5318 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5320         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
5321         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
5323 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5325         PR bootstrap/81033
5326         PR target/81733
5327         PR target/52795
5328         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
5329         (dwarf2out_switch_text_section): Generate a local label for the
5330         second function sub-section and apply it as the second FDE start
5331         label.
5332         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
5333         second sub-section start.
5335 2018-08-22  Richard Biener  <rguenther@suse.de>
5337         PR tree-optimization/86988
5338         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
5340 2018-08-22  Richard Biener  <rguenther@suse.de>
5342         PR tree-optimization/86945
5343         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
5345 2018-08-22  Alexandre Oliva <oliva@adacore.com>
5347         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
5348         a comment about how uses of r2 for .sdata2 come about.
5350 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
5352         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
5354 2018-08-21  Marek Polacek  <polacek@redhat.com>
5356         PR c++/86981, Implement -Wpessimizing-move.
5357         * doc/invoke.texi: Document -Wpessimizing-move.
5359 2018-08-21  Jan Hubicka  <jh@suse.cz>
5361         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
5362         * tree.h (is_redundant_typedef): Remove.
5363         * dwarf2out.c (is_redundant_typedef): Turn into static function.
5365 2018-08-21  Jan Hubicka  <jh@suse.cz>
5367         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
5368         when possible.
5370 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
5372         * expmed.c (extract_low_bits): Reject invalid subregs early.
5374 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5376         PR middle-end/86121
5377         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
5378         behaviour.
5380 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5382         * config/vxworks.h: Guard vxworks_asm_out_constructor and
5383         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
5384         * config/vxworks.c: Likewise.
5386 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5388         * config/vxworks.c: Set targetm.have_ctors_dtors
5389         if HAVE_INITFINI_ARRAY_SUPPORT.
5390         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
5391         if HAVE_INITFINI_ARRAY_SUPPORT.
5393 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5395         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
5396         default search path for VxWorks < 7.
5398 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5400         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
5401         (get_format_string): Refer to c_getstr.
5403 2018-08-21  Tom de Vries  <tdevries@suse.de>
5405         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5406         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5407         (debuginfo_early_stop): Declare.
5408         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5409         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5410         (debuginfo_early_stop): New function.
5411         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
5412         and debuginfo_early_stop.
5413         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
5414         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
5415         (general_init): Call debuginfo_early_init.
5416         (finalize): Call debuginfo_fini.
5417         (do_compile): Call debuginfo_init.
5418         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
5419         -fdump-early-debug.
5420         (@item -fdump-debug, @item -fdump-earlydebug): Add.
5422 2018-08-21  Tom de Vries  <tdevries@suse.de>
5424         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
5425         flag_dump_noaddr and flag_dump_unnumbered.
5427 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
5429         * wide-int-range.cc (wide_int_range_abs): New.
5430         (wide_int_range_order_set): Rename from wide_int_range_min_max.
5431         * wide-int-range.h (wide_int_range_abs): New.
5432         (wide_int_range_min_max): New.
5433         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
5434         case to call wide_int_range_abs.
5435         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
5436         (extract_range_from_abs_expr): Delete.
5438 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
5440         PR target/87033
5441         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
5442         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
5443         for indexed loads.
5445 2018-08-20  Nathan Sidwell  <nathan@acm.org>
5446             Jeff Law <law@redhat.com>
5448         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
5449         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
5451 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
5453         PR other/84889
5454         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
5455         (decl_attributes): Likewise.
5456         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
5457         instance.
5458         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5459         * diagnostic-core.h (class auto_diagnostic_group): New class.
5460         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
5461         (diagnostic_report_diagnostic): Handle the first diagnostics within
5462         a group.
5463         (emit_diagnostic): Add auto_diagnostic_group instance.
5464         (inform): Likewise.
5465         (inform_n): Likewise.
5466         (warning): Likewise.
5467         (warning_at): Likewise.
5468         (warning_n): Likewise.
5469         (pedwarn): Likewise.
5470         (permerror): Likewise.
5471         (error): Likewise.
5472         (error_n): Likewise.
5473         (error_at): Likewise.
5474         (sorry): Likewise.
5475         (fatal_error): Likewise.
5476         (internal_error): Likewise.
5477         (internal_error_no_backtrace): Likewise.
5478         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
5479         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
5480         * diagnostic.h (struct diagnostic_context): Add fields
5481         "diagnostic_group_nesting_depth",
5482         "diagnostic_group_emission_count", "begin_group_cb",
5483         "end_group_cb".
5484         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
5485         Add auto_diagnostic_group instance(s).
5486         (find_explicit_erroneous_behavior): Likewise.
5487         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
5488         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
5489         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
5490         (gimplify_va_arg_expr): Likewise.
5491         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
5492         (HSA_SORRY_AT): Likewise.
5493         * ipa-devirt.c (compare_virtual_tables): Likewise.
5494         (warn_odr): Likewise.
5495         * multiple_target.c (expand_target_clones): Likewise.
5496         * opts-common.c (cmdline_handle_error): Likewise.
5497         * reginfo.c (globalize_reg): Likewise.
5498         * substring-locations.c (format_warning_n_va): Likewise.
5499         * tree-inline.c (expand_call_inline): Likewise.
5500         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
5501         * tree-ssa-loop-niter.c
5502         (do_warn_aggressive_loop_optimizations): Likewise.
5503         * tree-ssa-uninit.c (warn_uninit): Likewise.
5504         * tree.c (warn_deprecated_use): Likewise.
5506 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
5508         PR target/87014
5509         * config/i386/i386.md (eh_return): Always update EH return
5510         address in word_mode.
5512 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
5514         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
5515         TARGET_SPLIT_COMPLEX_ARG is defined.
5517 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5519         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
5521 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5523         PR target/86984
5524         * expr.c (expand_assignment): Assert that bitpos is positive.
5525         (store_field): Likewise
5526         (expand_expr_real_1): Make sure that bitpos is positive.
5527         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
5528         integer overflow.
5530 2018-08-20  Nathan Sidwell  <nathan@acm.org>
5532         * Makefile.in (CPP_ID_DATA_H): Delete.
5533         (CPP_INTERNAL_H): Don't add it.
5534         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
5535         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
5537 2018-08-20  Richard Biener  <rguenther@suse.de>
5539         PR tree-optimization/78655
5540         * tree-vrp.c (extract_range_from_binary_expr_1): Make
5541         pointer + offset nonnull if either operand is nonnull work.
5543 2018-08-20  Tom de Vries  <tdevries@suse.de>
5545         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
5546         unless the referenced die describes the added property using
5547         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
5548         Otherwise, add a DW_AT_location to the referenced die.
5550 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
5552         PR target/86994
5553         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
5554         register_operand when calling ix86_set_reg_reg_cost.
5555         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
5556         Set *total to 0 for operands that satisfy x86_64_immediate_operand
5557         predicate and to 1 otherwise.
5559 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
5561         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
5562         emit a diagnostic that it is not supported and reset the option.
5563         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
5564         supported and consume the option.  (ASM_FINAL_SPEC): New.
5566 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
5568         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
5569         a sentence.
5571 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
5573         C-SKY port: Documentation
5575         * doc/extend.texi (C-SKY Function Attributes): New section.
5576         * doc/invoke.texi (Option Summary): Add C-SKY options.
5577         (C-SKY Options): New section.
5578         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
5580 2018-08-17  Jojo  <jijie_rong@c-sky.com>
5581             Huibin Wang  <huibin_wang@c-sky.com>
5582             Sandra Loosemore  <sandra@codesourcery.com>
5583             Chung-Lin Tang  <cltang@codesourcery.com>
5585         C-SKY port: Backend implementation
5587         * config/csky/*: New.
5588         * common/config/csky/*: New.
5590 2018-08-17  Jojo  <jijie_rong@c-sky.com>
5591             Huibin Wang  <huibin_wang@c-sky.com>
5592             Sandra Loosemore  <sandra@codesourcery.com>
5593             Chung-Lin Tang  <cltang@codesourcery.com>
5594             Andrew Jenner  <andrew@codesourcery.com>
5596         C-SKY port: Configury
5598         * config.gcc (csky-*-*): New.
5599         * configure.ac: Add csky to targets for dwarf2 debug_line support.
5600         * configure: Regenerated.
5602 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
5604         * dump-context.h: Include "dumpfile.h".
5605         (dump_context::dump_printf_va): Convert final param from va_list
5606         to va_list *.  Convert from ATTRIBUTE_PRINTF to
5607         ATTRIBUTE_GCC_DUMP_PRINTF.
5608         (dump_context::dump_printf_loc_va): Likewise.
5609         * dumpfile.c: Include "stringpool.h".
5610         (make_item_for_dump_printf_va): Delete.
5611         (make_item_for_dump_printf): Delete.
5612         (class dump_pretty_printer): New class.
5613         (dump_pretty_printer::dump_pretty_printer): New ctor.
5614         (dump_pretty_printer::emit_items): New member function.
5615         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
5616         function.
5617         (dump_pretty_printer::emit_item): New member function.
5618         (dump_pretty_printer::stash_item): New member function.
5619         (dump_pretty_printer::format_decoder_cb): New member function.
5620         (dump_pretty_printer::decode_format): New member function.
5621         (dump_context::dump_printf_va): Reimplement in terms of
5622         dump_pretty_printer.
5623         (dump_context::dump_printf_loc_va): Convert final param from va_list
5624         to va_list *.
5625         (dump_context::begin_scope): Reimplement call to
5626         make_item_for_dump_printf.
5627         (dump_printf): Update for change to dump_printf_va.
5628         (dump_printf_loc): Likewise.
5629         (selftest::test_capture_of_dump_calls): Convert "stmt" from
5630         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
5631         with %T, %E, and %G.
5632         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
5633         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
5634         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
5635         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
5636         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
5637         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
5638         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
5639         within a dump_printf_loc call to "%wu".
5640         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
5641         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
5642         missing space after "=".
5643         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
5644         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
5645         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
5646         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
5647         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
5648         duplicate "vectorized" from message.
5650 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
5652         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
5653         polyNxK_t element's TYPE_STRING_FLAG.
5655 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
5657         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
5658         (they were unnamed before).  Fix comments.
5660 2018-08-17  Nathan Sidwell  <nathan@acm.org>
5662         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
5664 2018-08-17  Richard Biener  <rguenther@suse.de>
5666         PR tree-optimization/86841
5667         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
5669 2018-08-17  Martin Liska  <mliska@suse.cz>
5671         * common.opt: Remove Warn, Init and Report for options with
5672         Ignore/Deprecated flag. Warning is done automatically for
5673         Deprecated flags.
5674         * config/i386/i386.opt: Likewise.
5675         * config/ia64/ia64.opt: Likewise.
5676         * config/rs6000/rs6000.opt: Likewise.
5677         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5678         Remove usage of flag_check_pointer_bounds.
5679         * lto-wrapper.c (merge_and_complain): Do not handle
5680         OPT_fcheck_pointer_bounds.
5681         (append_compiler_options): Likewise.
5682         * opt-functions.awk: Do not handle Deprecated.
5683         * optc-gen.awk: Check that Var, Report and Init are not
5684         used for an option with Ignore/Deprecated flag.
5685         * opts-common.c (decode_cmdline_option): Do not report
5686         CL_ERR_DEPRECATED.
5687         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
5688         options.
5689         * opts.h (struct cl_option): Remove cl_deprecated flag.
5690         (CL_ERR_DEPRECATED): Remove error enum value.
5692 2018-08-17  Richard Biener  <rguenther@suse.de>
5694         PR middle-end/86505
5695         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
5696         across a va-arg-pack using call adjust its return value accordingly.
5698 2018-08-16  Martin Sebor  <msebor@redhat.com>
5700         PR tree-optimization/86853
5701         * gimple-ssa-sprintf.c (struct format_result): Rename member.
5702         (struct fmtresult): Add member and initialize it in ctors.
5703         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
5704         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
5705         (format_directive): Set POSUNDER4K when MAYFAIL is set.
5706         (parse_directive): Handle %C same as %c.
5707         (sprintf_dom_walker::compute_format_length): Adjust.
5708         (is_call_safe): Adjust.
5710 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5712         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
5713         for determining how to count the elements.
5714         * builtins.h (c_strlen): Adjust prototype.
5715         * expr.c (string_constant): Add new parameter mem_size.
5716         Set *mem_size appropriately.
5717         * expr.h (string_constant): Adjust protoype.
5718         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
5719         * gimple-fold.h (get_range_strlen): Adjust prototype.
5720         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
5721         (format_string): Call get_string_length with eltsize.
5723 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
5725         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
5726         to emit the span, rather than setting it as the prefix.
5728 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
5730         * diagnostic-show-locus.c (layout::start_annotation_line): Add
5731         "margin_char" parameter, defaulting to space.  Use it in place
5732         of pp_space for the initial part of the margin.
5733         (layout::print_leading_fixits): Use '+' when filling the margin
5734         of line-insertion fix-it hints.
5736 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5738         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
5739         Delete.
5741 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5743         * config/rs6000/altivec.md: Don't set length attribute to the default
5744         value.
5745         * config/rs6000/darwin.md: Ditto.
5746         * config/rs6000/dfp.md: Ditto.
5747         * config/rs6000/htm.md: Ditto.
5748         * config/rs6000/rs6000.md: Ditto.
5749         * config/rs6000/sync.md: Ditto.
5750         * config/rs6000/vsx.md: Ditto.
5752 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5754         * config/rs6000/altivec.md: Don't set length attribute to the default
5755         value, for branch instructions.
5756         * config/rs6000/darwin.md: Ditto.
5757         * config/rs6000/rs6000.md: Ditto.
5759 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5761         * config/rs6000/rs6000.md (length): Always define as const_int 4.
5762         (unnamed conditional branch define_insn): Set length to 4 or 8
5763         depending on offset.
5764         (<bd>_<mode>): Similar, for alternative 0.
5765         (<bd>tf_<mode>): Ditto.
5767 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5769         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
5771 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
5773         * doc/rtl.texi: Replace old RTX class names with new names.
5776 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
5778         * expmed.h (canonicalize_comparison): New declaration.
5779         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
5780         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
5781         * optabs.c (prepare_cmp_insn): Likewise.
5782         * rtl.h (unsigned_condition_p): New function which checks if a
5783         comparison operator is unsigned.
5785 2018-08-16  Nathan Sidwell  <nathan@acm.org>
5787         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
5788         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
5790 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5792         PR target/84711
5793         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
5794         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
5795         (mov<mov>): ..this and enable unconditionally.
5797 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5799         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
5801 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
5803         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
5804         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
5805         "Common" with "Target".
5807 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
5809         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
5810         * doc/invoke.texi (mmitigate-rop): Remove.
5811         * config/i386/i386.c: Do not include "regrename.h".
5812         (ix86_rop_should_change_byte_p, reg_encoded_number)
5813         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
5814         Remove.
5815         (ix86_reorg): Remove call to ix86_mitigate_rop.
5816         * config/i386/i386.md (attr "modrm_class"): Remove.
5817         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
5818         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
5819         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
5821 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
5823         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
5824         allow folding of mergeh() and mergel() for the float and double types.
5825         (fold_mergehl_helper): Rework to handle building a permute tree
5826         for float vectors.
5828 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
5830         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
5831         for TARGET_SSE.
5833 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
5835         * common.opt (fdiagnostics-show-labels): New option.
5836         * diagnostic-show-locus.c (class layout_range): Add field
5837         "m_label".
5838         (class layout): Add field "m_show_labels_p".
5839         (layout_range::layout_range): Add param "label" and use it to
5840         initialize m_label.
5841         (make_range): Pass in NULL for new "label" param of layout_range's
5842         ctor.
5843         (layout::layout): Initialize m_show_labels_p.
5844         (layout::maybe_add_location_range): Pass in loc_range->m_label
5845         when constructing layout_range instances.
5846         (struct line_label): New struct.
5847         (layout::print_any_labels): New member function.
5848         (layout::print_line): Call it if label-printing is enabled.
5849         (selftest::test_one_liner_labels): New test.
5850         (selftest::test_diagnostic_show_locus_one_liner): Call it.
5851         * diagnostic.c (diagnostic_initialize): Initialize
5852         context->show_labels_p.
5853         * diagnostic.h (struct diagnostic_context): Add field
5854         "show_labels_p".
5855         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5856         -fno-diagnostics-show-labels.
5857         * dwarf2out.c (gen_producer_string): Add
5858         OPT_fdiagnostics_show_labels to the ignored options.
5859         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
5860         param.
5861         (gcc_rich_location::maybe_add_expr): Likewise.
5862         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
5863         label" param, defaulting to NULL.
5864         (gcc_rich_location::add_expr): Add "label" param.
5865         (gcc_rich_location::maybe_add_expr): Likewise.
5866         (class text_range_label): New class.
5867         (class range_label_for_type_mismatch): New class.
5868         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
5869         of format_warning_va.
5870         (fmtwarn_n): Likewise for new params of format_warning_n_va.
5871         * lto-wrapper.c (merge_and_complain): Add
5872         OPT_fdiagnostics_show_labels to the "pick one setting" options.
5873         (append_compiler_options): Likewise to the dropped options.
5874         (append_diag_options): Likewise to the passed-on options.
5875         * opts.c (common_handle_option): Handle the new option.
5876         * selftest-diagnostic.c
5877         (test_diagnostic_context::test_diagnostic_context): Enable
5878         show_labels_p.
5879         * substring-locations.c: Include "gcc-rich-location.h".
5880         (format_warning_n_va): Add "fmt_label" and "param_label" params
5881         and use them as appropriate.
5882         (format_warning_va): Add "fmt_label" and "param_label" params,
5883         passing them on to format_warning_n_va.
5884         (format_warning_at_substring): Likewise.
5885         (format_warning_at_substring_n): Likewise.
5886         * substring-locations.h (format_warning_va): Add "fmt_label" and
5887         "param_label" params.
5888         (format_warning_n_va): Likewise.
5889         (format_warning_at_substring): Likewise.
5890         (format_warning_at_substring_n): Likewise.
5891         * toplev.c (general_init): Initialize global_dc->show_labels_p.
5893 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
5895         PR testsuite/86519
5896         * builtins.c (expand_builtin_memcmp): Do not expand the call
5897         when overflow is detected.
5899 2018-08-15  Martin Sebor  <msebor@redhat.com>
5901         PR tree-optimization/71625
5902         * config/aarch64/aarch64-builtins.c
5903         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
5905 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
5907         * config/s390/s390.c (s390_reorg): Remove loop.
5909 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
5911         * config/darwin.c
5912          (darwin_function_switched_text_sections): Delete.
5913         * gcc/config/darwin.h
5914          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
5916 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
5918         PR target/81685
5919         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
5920         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
5921         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
5923 2018-08-15  Martin Liska  <mliska@suse.cz>
5925         PR tree-optimization/86925
5926         * predict.c (expr_expected_value_1): When taking
5927         later predictor, assign also probability.
5928         Use fold_build2_initializer_loc in order to fold
5929         the expression in -frounding-math.
5931 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
5933         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5934         patterns.
5935         (expand_vec_perm_1): Try the new method.
5937 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
5939         PR target/86547
5940         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5941         Check whether lra_live_max_point is 0 before dividing.
5943 2018-08-14  Martin Sebor  <msebor@redhat.com>
5945         PR tree-optimization/86650
5946         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5947         (vrp_prop::check_mem_ref): Same.
5949 2018-08-13  Liu Hao <lh_mouse@126.com>
5951         * pretty-print.c (eat_esc_sequence): Swap the foreground and
5952         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5953         and clear it thereafter, as it only works for DBCS.
5955 2018-08-13  Liu Hao <lh_mouse@126.com>
5957         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5958         handle returned by _get_osf_handle().
5960 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5962         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5963         for folding vec_perm.
5965 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5967         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5968         Add support for gimple-folding of vec_pack() and vec_unpack()
5969         intrinsics.
5971 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
5973         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5974         vec_xst variants to the list.
5975         (rs6000_gimple_fold_builtin): Add support for folding unaligned
5976         vector loads and stores.
5978 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
5980         * config.gcc (rs6000-ibm-aix4.x): Delete.
5981         (rs6000-ibm-aix5.1): Delete.
5982         (rs6000-ibm-aix5.2): Delete.
5983         (rs6000-ibm-aix5.3): Delete.
5984         * config/rs6000/aix43.h: Delete.
5985         * config/rs6000/aix51.h: Delete.
5986         * config/rs6000/aix52.h: Delete.
5987         * config/rs6000/t-aix43: Delete.
5989 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
5991         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5992         New function.
5993         (s390_decompose_address): Factor out constant pool ref
5994         decomposition.
5996 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5998         * config/nds32/nds32-predicates.c
5999         (nds32_can_use_bclr_p): Change return type as bool.
6000         (nds32_can_use_bset_p): Ditto.
6001         (nds32_can_use_btgl_p): Ditto.
6002         (nds32_can_use_bitci_p): Ditto.
6003         * config/nds32/nds32-protos.h
6004         (nds32_can_use_bclr_p): Change declaration.
6005         (nds32_can_use_bset_p): Ditto.
6006         (nds32_can_use_btgl_p): Ditto.
6007         (nds32_can_use_bitci_p): Ditto.
6009 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6011         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
6012         Support -msched-prolog-epilog option.
6013         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
6015 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6017         * common/config/nds32/nds32-common.c
6018         (nds32_option_optimization_table): Enalbe -malways-align.
6020 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6022         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
6023         extra_headers.
6024         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
6025         OPT_misr_secure_ case.
6026         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
6027         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
6028         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
6029         secure attribute.
6030         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
6031         (nds32_isr_info): New field security_level.
6032         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
6033         * config/nds32/nds32.md (return_internal): Consider critical attribute.
6034         * config/nds32/nds32.opt (misr-secure): New option.
6035         * config/nds32/nds32_init.inc: New file.
6036         * config/nds32/nds32_isr.h: New file.
6038 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
6040         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
6041         Update comment for atomic instructions.
6042         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
6043         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
6044         Remove.
6045         (atomic_loaddi): Revise fence expansion to only emit fence prior to
6046         load for __ATOMIC_SEQ_CST model.
6047         (atomic_loaddi_1): Remove float register target.
6048         (atomic_storedi): Handle CONST_INT values.
6049         (atomic_storedi_1): Remove float register source.  Add special case
6050         for zero value.
6051         (memory_barrier): New expander and insn.
6053 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
6055         PR tree-optimization/86835
6056         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
6057         new_stmt after def_gsi, make sure to insert new_square_stmt after
6058         that stmt, not 2 stmts before it.
6060 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
6062         PR target/82418
6063         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
6064         instead of SWI48.
6066 2018-08-10  Martin Liska  <mliska@suse.cz>
6068         PR target/83610
6069         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
6070         function type.
6071         * builtins.c (expand_builtin_expect_with_probability):
6072         New function.
6073         (expand_builtin_expect_with_probability): New function.
6074         (build_builtin_expect_predicate): Add new argumnet probability
6075         for BUILT_IN_EXPECT_WITH_PROBABILITY.
6076         (fold_builtin_expect):
6077         (fold_builtin_2):
6078         (fold_builtin_3):
6079         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
6080         * builtins.h (fold_builtin_expect): Set new argument.
6081         * doc/extend.texi: Document __builtin_expect_with_probability.
6082         * doc/invoke.texi: Likewise.
6083         * gimple-fold.c (gimple_fold_call): Pass new argument.
6084         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
6085         also BUILT_IN_EXPECT_WITH_PROBABILITY.
6086         * predict.c (get_predictor_value): New function.
6087         (expr_expected_value): Add new argument probability. Assume
6088         that predictor and probability are always non-null.
6089         (expr_expected_value_1): Likewise.  For __builtin_expect and
6090         __builtin_expect_with_probability set probability.  Handle
6091         combination in binary expressions.
6092         (tree_predict_by_opcode): Simplify code by simply calling
6093         get_predictor_value.
6094         (pass_strip_predict_hints::execute): Add handling of
6095         BUILT_IN_EXPECT_WITH_PROBABILITY.
6096         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
6097         new predictor.
6098         * tree.h (DECL_BUILT_IN_P): New function.
6100 2018-08-10  Martin Liska  <mliska@suse.cz>
6102         PR tree-optimization/85799
6103         * passes.def: Add argument for pass_strip_predict_hints.
6104         * predict.c (class pass_strip_predict_hints): Add new argument
6105         early_p.
6106         (strip_predictor_early): New function.
6107         (pass_strip_predict_hints::execute): Call the function to
6108         strip predictors.
6109         (strip_predict_hints): New function.
6110         * predict.def: Fix comment.
6112 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
6114         * Makefile.in: Clarify which tm.texi to copy over to assert the
6115         right to grant a GFDL license for all.
6117 2018-08-09  Jeff Law  <law@redhat.com>
6119         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6120         unused variable.
6122 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
6124         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6125         prototype.
6127 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6129         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6130         reductions for variable-length vectors.
6132 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
6134         PR other/84889
6135         * common.opt (fdiagnostics-show-line-numbers): New option.
6136         * diagnostic-show-locus.c (class layout): Add fields
6137         "m_show_line_numbers_p" and "m_linenum_width";
6138         (num_digits): New function.
6139         (test_num_digits): New function.
6140         (layout::layout): Initialize new fields.  Update m_x_offset
6141         logic to handle any left margin.
6142         (layout::print_source_line): Print line number when requested.
6143         (layout::start_annotation_line): New member function.
6144         (layout::print_annotation_line): Call it.
6145         (layout::print_leading_fixits): Likewise.
6146         (layout::print_trailing_fixits): Likewise.  Update calls to
6147         move_to_column for new parameter.
6148         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6149         it to potentially call start_annotation_line.
6150         (layout::show_ruler): Call start_annotation_line.
6151         (selftest::test_line_numbers_multiline_range): New selftest.
6152         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6153         and selftest::test_line_numbers_multiline_range.
6154         * diagnostic.c (diagnostic_initialize): Initialize
6155         show_line_numbers_p.
6156         * diagnostic.h (struct diagnostic_context): Add field
6157         "show_line_numbers_p".
6158         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6159         -fno-diagnostics-show-line-numbers.
6160         * dwarf2out.c (gen_producer_string): Add
6161         OPT_fdiagnostics_show_line_numbers to the ignored options.
6162         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6163         one setting" options.
6164         (append_compiler_options): Likewise to the dropped options.
6165         (append_diag_options): Likewise to the passed-on options.
6166         * opts.c (common_handle_option): Handle the new option.
6167         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
6169 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6171         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
6172         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
6173         third argument of type "const signed char" to descriptions of
6174         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
6175         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
6176         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6177         __builtin_bcdsub_ov functions.
6179 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6181         PR tree-optimization/86858
6182         * tree-vect-loop.c (vect_is_simple_reduction): Restore
6183         flow_bb_inside_loop_p calls.
6185 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6187         PR tree-optimization/86871
6188         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6189         instead of gimple_assign_lhs.
6191 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
6193         PR target/86887
6194         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6195         register constraint to operand 0.
6196         (add<mode>3_carryinC): Likewise.
6197         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6199 2018-08-09  Martin Liska  <mliska@suse.cz>
6201         PR c/86895
6202         * common.opt: Remove extra line.
6204 2018-08-09  Martin Liska  <mliska@suse.cz>
6206         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6207         at the end of a line, make first letter capital and end up
6208         a sentence with a dot.
6209         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6210         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6211         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6212         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6213         (PARAM_MAX_ISL_OPERATIONS): Likewise.
6214         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6215         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6216         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6217         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6218         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6219         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6220         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6221         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
6222         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
6223         (PARAM_TREE_REASSOC_WIDTH): Likewise.
6224         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
6225         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
6226         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
6228 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
6230         PR target/84332
6231         * config/s390/s390.c (s390_option_override_internal): Reduce the
6232         stack-clash-protection-probe-interval param if it would be too big
6233         for z900.
6235 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
6237         PR target/46179
6238         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
6239         * config/m68k/m68k.c (handle_move_double): Don't call
6240         m68k_final_prescan_insn.
6241         (m68k_adjust_decorated_operand): Renamed from
6242         m68k_final_prescan_insn, remove first and third operand and
6243         simplify.
6244         (print_operand): Call it.
6245         (print_operand_address): Call it.
6247 2018-08-08  Nathan Sidwell  <nathan@acm.org>
6249         * diagnostic.c (diagnostic_report_current_module): Use
6250         linemap_included_from & linemap_included_from_linemap.
6252 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
6254         * config/aarch64/aarch64-cores.def: Add phecda core.
6255         * config/aarch64/aarch64-tune.md: Regenerate.
6256         * doc/invoke.texi: Add phecda core.
6258 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6260         PR target/85295
6261         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
6262         definitions.
6263         * config/s390/s390.md ("movti"): Add more alternatives for
6264         constant to GPR copies.
6266 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6268         * config/s390/s390.c: Fix whitespace damage throughout the file.
6269         * config/s390/s390.h: Likewise.
6270         * config/s390/tpf.h: Likewise.
6272 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6274         * config/s390/s390.c (s390_loadrelative_operand_p):
6275         Remove TARGET_CPU_ZARCH usages.
6276         (s390_rtx_costs): Likewise.
6277         (s390_legitimate_constant_p): Likewise.
6278         (s390_cannot_force_const_mem): Likewise.
6279         (legitimate_reload_constant_p): Likewise.
6280         (s390_preferred_reload_class): Likewise.
6281         (legitimize_pic_address): Likewise.
6282         (legitimize_tls_address): Likewise.
6283         (s390_split_branches): Removed.
6284         (s390_add_execute): Removed.
6285         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
6286         (s390_mainpool_start): Likewise.
6287         (s390_mainpool_finish): Likewise.
6288         (s390_mainpool_cancel): Removed.
6289         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
6290         (s390_chunkify_cancel): Likewise.
6291         (s390_return_addr_rtx): Likewise.
6292         (s390_register_info): Remove split_branches_pending_p uages.
6293         (s390_optimize_register_info): Likewise.
6294         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
6295         split_branches_pending_p usages.
6296         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
6297         (s390_load_got): Likewise.
6298         (s390_expand_split_stack_prologue): Likewise.
6299         (output_asm_nops): Likewise.
6300         (s390_function_profiler): Likewise.
6301         (s390_emit_call): Likewise.
6302         (s390_conditional_register_usage): Likewise.
6303         (s390_optimize_prologue): Likewise.
6304         (s390_reorg): Remove TARGET_CPU_ZARCH and
6305         split_branches_pending_p usages.
6306         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
6307         usages.
6308         (s390_output_indirect_thunk_function): Likewise.
6309         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
6310         (TARGET_CPU_ZARCH_P): Removed.
6311         (struct machine_function): Remove split_branches_pending_p.
6312         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
6314 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6316         * common/config/s390/s390-common.c (processor_flags_table):
6317         Remove flags.
6318         * config.gcc: Remove with_arch/with_tune support.
6319         * config/s390/2064.md: Remove cpu attribute comparisons.
6320         * config/s390/driver-native.c (s390_host_detect_local_cpu):
6321         Remove MTN.
6322         * config/s390/linux.h (ASM_SPEC):
6323         Remove -march support.
6324         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
6325         Use a table to get an arch level.
6326         * config/s390/s390-opts.h (enum processor_type):
6327         Remove enum values.
6328         * config/s390/s390.c
6329         (processor_table): Remove entries, add arch_level values.
6330         (s390_issue_rate): Remove cases.
6331         (s390_option_override): Adjust
6332         s390_option_override_internal() call.
6333         (s390_option_override_internal): Remove deprecation warning.
6334         (s390_valid_target_attribute_tree): Adjust
6335         s390_option_override_internal() call.
6336         * config/s390/s390.h (struct s390_processor):
6337         Share with s390-c.c, add arch_level field.
6338         * config/s390/s390.md:
6339         Remove occurrences in cpu attribute.
6340         * config/s390/s390.opt: Remove -march/-mtune support.
6341         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
6342         * doc/invoke.texi: Remove deprecation warning.
6344 2018-08-08  Luis Machado  <luis.machado@linaro.org>
6346         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
6347         global.
6348         (qdf24xx_tunings): Set vector cost structure to
6349         qdf24xx_vector_cost.
6351         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
6352         <register_sextend>: Set to 3.
6354 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
6356         PR target/86838
6357         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
6358         * config/aarch64/aarch64-simd.md
6359         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
6360         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
6361         (aarch64_frecpx<mode>): ...this new pattern.
6362         * config/aarch64/aarch64-simd-builtins.def: Remove comment
6363         about aarch64_frecp<FRECP:frecp_suffix><mode>.
6365 2018-08-07  Martin Liska  <mliska@suse.cz>
6367         PR middle-end/83023
6368         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
6369         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
6370         * predict.def (PRED_MALLOC_NONNULL): New predictor.
6371         * doc/extend.texi: Document that malloc attribute adds
6372         hit to compiler.
6374 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
6376         PR target/86785
6377         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6378         Define to speculation_safe_value_not_needed.
6380 2018-08-06  Jeff Law  <law@redhat.com>
6382         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
6383         the vr_values instance to cprop_into_stmt.
6384         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
6385         (cprop_operand): Also query EVRP to determine if OP is a constant.
6387 2018-08-06  Nathan Sidwell  <nathan@acm.org>
6389         * diagnostic.c (diagnostic_report_current_module): Reroll
6390         included-at loop.  Translate text.
6392 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
6394         * function-tests.c (selftest::test_expansion_to_rtl): Call
6395         free_after_compilation.
6397 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6399         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
6401 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
6403         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
6404         loops with memory block operations from getting unrolled.
6406 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
6408         PR target/86807
6409         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6410         Define to speculation_safe_value_not_needed.
6412 2018-08-06  Jeff Law  <law@redhat.com>
6414         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
6415         assert.
6417 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6419         PR target/86662
6420         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
6421         with all enabled __intN types.
6423         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
6425 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6427         * alias.c (record_set): Check for clobber high.
6428         * cfgexpand.c (expand_gimple_stmt): Likewise.
6429         * combine-stack-adj.c (single_set_for_csa): Likewise.
6430         * combine.c (find_single_use_1): Likewise.
6431         (set_nonzero_bits_and_sign_copies): Likewise.
6432         (get_combine_src_dest): Likewise.
6433         (is_parallel_of_n_reg_sets): Likewise.
6434         (try_combine): Likewise.
6435         (record_dead_and_set_regs_1): Likewise.
6436         (reg_dead_at_p_1): Likewise.
6437         (reg_dead_at_p): Likewise.
6438         * dce.c (deletable_insn_p): Likewise.
6439         (mark_nonreg_stores_1): Likewise.
6440         (mark_nonreg_stores_2): Likewise.
6441         * df-scan.c (df_find_hard_reg_defs): Likewise.
6442         (df_uses_record): Likewise.
6443         (df_get_call_refs): Likewise.
6444         * dwarf2out.c (mem_loc_descriptor): Likewise.
6445         * haifa-sched.c (haifa_classify_rtx): Likewise.
6446         * ira-build.c (create_insn_allocnos): Likewise.
6447         * ira-costs.c (scan_one_insn): Likewise.
6448         * ira.c (equiv_init_movable_p): Likewise.
6449         (rtx_moveable_p): Likewise.
6450         (interesting_dest_for_shprep): Likewise.
6451         * jump.c (mark_jump_label_1): Likewise.
6452         * postreload-gcse.c (record_opr_changes): Likewise.
6453         * postreload.c (reload_cse_simplify): Likewise.
6454         (struct reg_use): Add source expr.
6455         (reload_combine): Check for clobber high.
6456         (reload_combine_note_use): Likewise.
6457         (reload_cse_move2add): Likewise.
6458         (move2add_note_store): Likewise.
6459         * print-rtl.c (print_pattern): Likewise.
6460         * recog.c (decode_asm_operands): Likewise.
6461         (store_data_bypass_p): Likewise.
6462         (if_test_bypass_p): Likewise.
6463         * regcprop.c (kill_clobbered_value): Likewise.
6464         (kill_set_value): Likewise.
6465         * reginfo.c (reg_scan_mark_refs): Likewise.
6466         * reload1.c (maybe_fix_stack_asms): Likewise.
6467         (eliminate_regs_1): Likewise.
6468         (elimination_effects): Likewise.
6469         (mark_not_eliminable): Likewise.
6470         (scan_paradoxical_subregs): Likewise.
6471         (forget_old_reloads_1): Likewise.
6472         * reorg.c (find_end_label): Likewise.
6473         (try_merge_delay_insns): Likewise.
6474         (redundant_insn): Likewise.
6475         (own_thread_p): Likewise.
6476         (fill_simple_delay_slots): Likewise.
6477         (fill_slots_from_thread): Likewise.
6478         (dbr_schedule): Likewise.
6479         * resource.c (update_live_status): Likewise.
6480         (mark_referenced_resources): Likewise.
6481         (mark_set_resources): Likewise.
6482         * rtl.c (copy_rtx): Likewise.
6483         * rtlanal.c (reg_referenced_p): Likewise.
6484         (single_set_2): Likewise.
6485         (noop_move_p): Likewise.
6486         (note_stores): Likewise.
6487         * sched-deps.c (sched_analyze_reg): Likewise.
6488         (sched_analyze_insn): Likewise.
6490 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6492         * cse.c (invalidate_reg): New function extracted from...
6493         (invalidate): ...here.
6494         (canonicalize_insn): Check for clobber high.
6495         (invalidate_from_clobbers): invalidate clobber highs.
6496         (invalidate_from_sets_and_clobbers): Likewise.
6497         (count_reg_usage): Check for clobber high.
6498         (insn_live_p): Likewise.
6499         * cselib.c (cselib_expand_value_rtx_1):Likewise.
6500         (cselib_invalidate_regno): Check for clobber in setter.
6501         (cselib_invalidate_rtx): Pass through setter.
6502         (cselib_invalidate_rtx_note_stores):
6503         (cselib_process_insn): Check for clobber high.
6504         * cselib.h (cselib_invalidate_rtx): Add operand.
6506 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6508         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
6509         (mark_not_eliminable): Likewise.
6510         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
6511         * lra-lives.c (process_bb_lives): Check for clobber high.
6512         * lra.c (new_insn_reg): Remember clobber highs.
6513         (collect_non_operand_hard_regs): Check for clobber high.
6514         (lra_set_insn_recog_data): Likewise.
6515         (add_regs_to_insn_regno_info): Likewise.
6516         (lra_update_insn_regno_info): Likewise.
6518 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6520         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
6521         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
6523 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6525         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
6526         (copy_insn_1): Likewise.
6527         (gen_hard_reg_clobber_high): New gen function.
6528         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
6529         * genemit.c (gen_exp): Likewise.
6530         (gen_emit_seq): Pass through info.
6531         (gen_insn): Check for CLOBBER_HIGH.
6532         (gen_expand): Pass through info.
6533         (gen_split): Likewise.
6534         (output_add_clobbers): Likewise.
6535         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
6536         (remove_clobbers): Likewise.
6537         * rtl.h (gen_hard_reg_clobber_high): New declaration.
6539 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6541         * doc/rtl.texi (clobber_high): Add.
6542         (parallel): Add in clobber high
6543         * rtl.c (rtl_check_failed_code3): Add function.
6544         * rtl.def (CLOBBER_HIGH): Add expression.
6545         * rtl.h (RTL_CHECKC3): Add macro.
6546         (rtl_check_failed_code3): Add declaration.
6547         (XC3EXP): Add macro.
6549 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
6551         PR target/86386
6552         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
6553         cfun->machine->max_used_stack_alignment if needed.
6555 2018-08-04  Martin Sebor  <msebor@redhat.com>
6557         PR tree-optimization/86571
6558         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
6559         NaN output to 4.
6561 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
6563         PR target/86799
6564         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6565         Define.
6567 2018-08-03  Jeff Law  <law@redhat.com>
6569         PR target/86795
6570         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6571         Define to speculation_safe_value_not_needed.
6573 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
6575         * doc/gcov.texi (-x): Remove duplicate "to".
6576         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
6577         (-Wif-not-aligned): Remove duplicate "is".
6578         (-flto): Remove duplicate "the".
6579         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
6580         duplicate "v5.00.b".
6581         (MSP430 Options): Remove duplicate "and" from the description
6582         of "-mgprel-sec=regexp".
6583         (x86 Options): Remove duplicate copies of "vmldLog102" and
6584         vmlsLog104 from description of "-mveclibabi=type".
6586 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
6588         * internal-fn.h (first_commutative_argument): Declare.
6589         * internal-fn.c (first_commutative_argument): New function.
6590         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
6591         restrictions for pattern statements.  Use first_commutative_argument
6592         to look for commutative operands in calls to internal functions.
6594 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
6596         * Makefile.in (wide-int-range.o): New.
6597         * tree-vrp.c: Move all the wide_int_* functions to...
6598         * wide-int-range.cc: ...here.
6599         * tree-vrp.h: Move all the wide_int_* prototypes to...
6600         * wide-int-range.h: ...here.
6602 2018-08-03  Tom de Vries  <tdevries@suse.de>
6604         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
6605         UI_NONE.
6606         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
6607         * except.c (output_function_exception_table): Do early exit if
6608         targetm_common.except_unwind_info (&global_options) == UI_NONE.
6610 2018-08-03  Martin Liska  <mliska@suse.cz>
6612         * predict.c (dump_prediction): Change to 2 digits
6613         in fraction part.
6615 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
6617         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
6618         neon_dup_q to...
6619         (falkor_am_1_gtov_gtov): ... a new insn reservation.
6621 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
6623         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
6624         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
6625         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
6626         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
6627         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
6629 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
6631         * diagnostic-show-locus.c (diagnostic_show_locus): Use
6632         pp_take_prefix when saving the existing prefix.
6633         * diagnostic.c (diagnostic_append_note): Likewise.
6634         * langhooks.c (lhd_print_error_function): Likewise.
6635         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
6636         param's type.  Free the existing prefix.
6637         (pp_take_prefix): New function.
6638         (pretty_printer::pretty_printer): Drop the prefix parameter.
6639         Rename the length parameter to match the comment.
6640         (pretty_printer::~pretty_printer): Free the prefix.
6641         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
6642         parameter.
6643         (struct pretty_printer): Drop the "const" from "prefix" field's
6644         type and clarify memory management.
6645         (pp_set_prefix): Drop the "const" from the 2nd param.
6646         (pp_take_prefix): New decl.
6648 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
6650         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
6651         for word_mode_ok here instead of passing as argument.
6652         (expand_block_compare): Change select_block_compare_mode() call.
6653         (expand_strncmp_gpr_sequence): New function.
6654         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
6656 2018-08-02  Jeff Law  <law@redhat.com>
6658         PR target/86790
6659         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6660         Define to speculation_safe_value_not_needed.
6662         PR target/86784
6663         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6664         Define to speculation_safe_value_not_needed.
6666 2018-08-02  Tom de Vries  <tdevries@suse.de>
6668         PR target/86660
6669         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
6670         function.  Return UI_TARGET unconditionally.
6671         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
6672         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
6674 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6676         * genemit.c (print_overload_test): Fix typo.
6678 2018-08-02  Richard Biener  <rguenther@suse.de>
6680         PR tree-optimization/86816
6681         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
6682         which checks for value availability before querying it.
6683         (gvn_uses_equal): Use it.
6684         (same_succ_hash): Likewise.
6685         (gimple_equal_p): Likewise.
6687 2018-08-02  Nick Clifton  <nickc@redhat.com>
6689         PR target/86813
6690         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6691         Define to speculation_safe_value_not_needed.
6693         PR target/86810
6694         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6695         Define to speculation_safe_value_not_needed.
6697         PR target/86810
6698         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6699         Define to speculation_safe_value_not_needed.
6701         PR target/86803
6702         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6703         Define to speculation_safe_value_not_needed.
6705         PR target/86797
6706         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6707         Define to speculation_safe_value_not_needed.
6709         PR target/86791
6710         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6711         Define to speculation_safe_value_not_needed.
6713         PR target/86789
6714         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6715         Define to speculation_safe_value_not_needed.
6717         PR target/86787
6718         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6719         Define to speculation_safe_value_not_needed.
6721         PR target/86782
6722         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6723         speculation_safe_value_not_needed.
6725         PR target/86781
6726         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
6727         to speculation_safe_value_not_needed.
6729 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6731         * doc/md.texi: Expand the documentation of instruction names
6732         to mention port-local uses.  Document '@' in pattern names.
6733         * read-md.h (overloaded_instance, overloaded_name): New structs.
6734         (mapping): Declare.
6735         (md_reader::handle_overloaded_name): New member function.
6736         (md_reader::get_overloads): Likewise.
6737         (md_reader::m_first_overload): New member variable.
6738         (md_reader::m_next_overload_ptr): Likewise.
6739         (md_reader::m_overloads_htab): Likewise.
6740         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
6741         m_next_overload_ptr and m_overloads_htab.
6742         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
6743         (get_mode_token, get_code_token, get_int_token): New functions.
6744         (map_attr_string): Add an optional argument that passes back
6745         the associated iterator.
6746         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
6747         (md_reader::handle_overloaded_name, add_overload_instance): New
6748         functions.
6749         (apply_iterators): Handle '@' names.  Report an error if '@'
6750         is used without iterators.
6751         (initialize_iterators): Initialize the new iterator_group fields.
6752         * genopinit.c (handle_overloaded_code_for)
6753         (handle_overloaded_gen): New functions.
6754         (main): Use them to print declarations of maybe_code_for_* and
6755         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
6756         * genemit.c (print_overload_arguments, print_overload_test)
6757         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
6758         (main): Use it to print definitions of maybe_code_for_* and
6759         maybe_gen_* functions.
6760         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
6761         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
6762         instead of explicit mode checks.
6763         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
6764         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
6765         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
6766         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
6767         (aarch64_expand_compare_and_swap): Likewise
6768         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
6769         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
6770         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
6771         (aarch64_constant_pool_reload_icode): Delete.
6772         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
6773         instead of aarch64_constant_pool_reload_icode.  Use
6774         code_for_aarch64_reload_mov instead of explicit mode checks.
6775         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
6776         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
6777         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
6778         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
6779         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
6780         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
6781         (aarch64_atomic_load_op_code): Delete.
6782         (aarch64_emit_atomic_load_op): Likewise.
6783         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
6784         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
6785         instead of aarch64_emit_atomic_load_op.
6786         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
6787         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
6788         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
6789         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
6790         character before the pattern name.
6791         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
6792         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
6793         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
6794         (aarch64_frecps<mode>): Likewise.
6795         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
6796         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
6797         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
6798         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
6799         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
6801 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6803         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
6804         Allow HFmode constants if TARGET_FP_F16INST.
6806 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
6808         PR target/86014
6809         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
6810         No longer check last store for clobber of address register.
6812 2018-08-02  Martin Liska  <mliska@suse.cz>
6814         PR gcov-profile/86817
6815         * gcov.c (process_all_functions): New function.
6816         (main): Call it.
6817         (process_file): Move functions processing to
6818         process_all_functions.
6820 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
6822         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
6823         "const" to the "gimple *" and "rtx_insn *" parameters.
6824         * dumpfile.h (dump_user_location_t::dump_user_location_t):
6825         Likewise.
6826         (dump_location_t::dump_location_t): Likewise.
6828 2018-08-01  Martin Sebor  <msebor@redhat.com>
6830         PR tree-optimization/86650
6831         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
6832         rather than a "gcall *".  Directly pass the data of interest
6833         to percent_K_format, rather than building a temporary CALL_EXPR
6834         to hold it.
6835         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
6836         (gimple_fold_builtin_strncat): Adjust.
6837         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
6838         gcall* argument with gimple*.
6839         * gimple-ssa-warn-restrict.c (check_call): Same.
6840         (wrestrict_dom_walker::before_dom_children): Same.
6841         (builtin_access::builtin_access): Same.
6842         (check_bounds_or_overlap): Same
6843         (maybe_diag_overlap): Same.
6844         (maybe_diag_offset_bounds): Same.
6845         * tree-diagnostic.c (default_tree_printer): Move usage of
6846         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
6847         to this callsite.
6848         * tree-pretty-print.c (percent_K_format): Add argument.
6849         * tree-pretty-print.h: Add argument.
6850         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
6851         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
6852         (maybe_diag_stxncpy_trunc): Same.
6853         (handle_builtin_stxncpy): Same.
6854         (handle_builtin_strcat): Same.
6856 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6858         * match.pd: Optimise pointer range checks.
6860 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6862         PR tree-optimization/86758
6863         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
6864         to remove pattern statements.
6866 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6868         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
6869         result of dfs_enumerate_from when constructing stmt_vec_infos,
6870         instead of additionally calling get_loop_body.
6872 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6874         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
6875         parameter.
6876         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
6877         When creating an iv, assert that the step is not known to be zero.
6878         (vect_setup_realignment): Update call accordingly.
6879         * tree-vect-stmts.c (vectorizable_store): Likewise.
6880         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
6882 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6884         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
6885         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
6886         (vectorizable_reduction): Likewise.
6887         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6888         (vect_detect_hybrid_slp_stmts): Likewise.
6889         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6891 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
6893         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
6894         (wide_int_set_zero_nonzero_bits): ...this.
6895         (zero_nonzero_bits_from_vr): Rename to...
6896         (vrp_set_zero_nonzero_bits): ...this.
6897         (extract_range_from_multiplicative_op_1): Abstract wide int
6898         code...
6899         (wide_int_range_multiplicative_op): ...here.
6900         (extract_range_from_binary_expr_1): Extract wide int binary
6901         operations into their own functions.
6902         (wide_int_range_lshift): New.
6903         (wide_int_range_can_optimize_bit_op): New.
6904         (wide_int_range_shift_undefined_p): New.
6905         (wide_int_range_bit_xor): New.
6906         (wide_int_range_bit_ior): New.
6907         (wide_int_range_bit_and): New.
6908         (wide_int_range_trunc_mod): New.
6909         (extract_range_into_wide_ints): New.
6910         (vrp_shift_undefined_p): New.
6911         (extract_range_from_multiplicative_op): New.
6912         (vrp_can_optimize_bit_op): New.
6913         * tree-vrp.h (value_range::dump): New.
6914         (wide_int_range_multiplicative_op): New.
6915         (wide_int_range_lshift):New.
6916         (wide_int_range_shift_undefined_p): New.
6917         (wide_int_range_bit_xor): New.
6918         (wide_int_range_bit_ior): New.
6919         (wide_int_range_bit_and): New.
6920         (wide_int_range_trunc_mod): New.
6921         (zero_nonzero_bits_from_bounds): Rename to...
6922         (wide_int_set_zero_nonzero_bits): ...this.
6923         (zero_nonzero_bits_from_vr): Rename to...
6924         (vrp_set_zero_nonzero_bits): ...this.
6925         (range_easy_mask_min_max): Rename to...
6926         (wide_int_range_can_optimize_bit_op): this.
6927         * vr-values.c (simplify_bit_ops_using_ranges): Rename
6928         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6930 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6932         * tree-vectorizer.h (vect_orig_stmt): New function.
6933         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6934         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6935         (vect_create_epilog_for_reduction): Likewise.
6936         (vectorizable_live_operation): Likewise.
6937         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6938         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6939         * tree-vect-stmts.c (vectorizable_call): Likewise.
6940         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6942 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6944         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6945         argument.
6946         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6947         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6948         (vect_transform_loop): Likewise.
6949         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6951 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6953         * tree-vectorizer.h (vect_schedule_slp): Return void.
6954         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6955         (vect_schedule_slp): Likewise.
6957 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6959         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6960         argument.
6961         (vect_transform_loop): Update calls accordingly.  Schedule SLP
6962         instances before the main loop, if any exist.
6964 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6966         PR tree-optimization/86749
6967         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6968         If the lhs is used in a COND_EXPR, check that it is being used
6969         as the "then" or "else" value.
6971 2018-08-01  Tom de Vries  <tdevries@suse.de>
6973         PR target/86800
6974         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6975         speculation_safe_value_not_needed.
6977 2018-08-01  Richard Biener  <rguenther@suse.de>
6979         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6980         as base and offset.
6982 2018-08-01  Martin Liska  <mliska@suse.cz>
6984         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6985         format how successful transformation is dumped.
6986         (gimple_mod_pow2_value_transform): Likewise.
6987         (gimple_mod_subtract_transform): Likewise.
6988         (gimple_stringops_transform): Likewise.
6990 2018-08-01  Martin Liska  <mliska@suse.cz>
6992         PR value-prof/35543
6993         * value-prof.c (interesting_stringop_to_profile_p):
6994         Simplify the code and add BUILT_IN_MEMMOVE.
6995         (gimple_stringops_transform): Likewise.
6997 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
6999         * config/aarch64/aarch64-simd.md
7000         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
7001         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
7002         use GPI iterator instead of SI mode.
7004 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7006         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
7007         rs6000_speculation_barrier.
7008         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
7009         new barrier pattern name.
7011 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7013         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
7014         (speculation_barrier): New insn.
7016 2018-08-01  Richard Biener  <rguenther@suse.de>
7018         PR bootstrap/86724
7019         * graphite.h: Include isl/id.h and isl/space.h to allow build
7020         with ISL 0.20.
7022 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
7024         PR target/86651
7025         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7026         mode for COFF targets.
7027         * defaults.h (TARGET_COFF): Define.
7028         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
7029         TARGET_COFF): Define.
7030         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
7031         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
7032         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
7034 2018-07-31  Alexandre Oliva <oliva@adacore.com>
7036         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
7037         * gimple-streamer-out.c (output_bb): Save it.
7038         * lto-streamer-in.c (input_struct_function_base): Restore
7039         instance discriminator if available.  Create map on demand.
7040         * lto-streamer-out.c (output_struct_function_base): Save it if
7041         available.
7042         * final.c (decl_to_instance_map): Document LTO strategy.
7044 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
7045             Olivier Hainque  <hainque@adacore.com>
7047         * debug.h (decl_to_instance_map_t): New type.
7048         (decl_to_instance_map): Declare.
7049         (maybe_create_decl_to_instance_map): New inline function.
7050         * final.c (bb_discriminator, last_bb_discriminator): New statics,
7051         to track basic block discriminators.
7052         (final_start_function_1): Initialize them.
7053         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
7054         bb_discriminator.
7055         (decl_to_instance_map): New variable.
7056         (map_decl_to_instance, maybe_set_discriminator): New functions.
7057         (notice_source_line): Set discriminator.
7059 2018-07-31  Ian Lance Taylor  <iant@golang.org>
7061         * targhooks.c (default_have_speculation_safe_value): Add
7062         ATTRIBUTE_UNUSED.
7064 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7066         * dump-context.h: Include "pretty-print.h".
7067         (dump_context::refresh_dumps_are_enabled): New decl.
7068         (dump_context::emit_item): New decl.
7069         (class dump_context): Add fields "m_test_pp" and
7070         "m_test_pp_flags".
7071         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7072         (temp_dump_context::get_dumped_text): New decl.
7073         (class temp_dump_context): Add field "m_pp".
7074         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
7075         (dump_context::refresh_dumps_are_enabled): ...and add a test for
7076         m_test_pp.
7077         (set_dump_file): Update for above change.
7078         (set_alt_dump_file): Likewise.
7079         (dump_loc): New overload, taking a pretty_printer *.
7080         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
7081         to any test pretty-printer.
7082         (make_item_for_dump_gimple_stmt): New function, adapted from
7083         optinfo::add_gimple_stmt.
7084         (dump_context::dump_gimple_stmt): Call it, and use the result,
7085         eliminating the direct usage of dump_file and alt_dump_file in
7086         favor of indirectly using them via emit_item.
7087         (make_item_for_dump_gimple_expr): New function, adapted from
7088         optinfo::add_gimple_expr.
7089         (dump_context::dump_gimple_expr): Call it, and use the result,
7090         eliminating the direct usage of dump_file and alt_dump_file in
7091         favor of indirectly using them via emit_item.
7092         (make_item_for_dump_generic_expr): New function, adapted from
7093         optinfo::add_tree.
7094         (dump_context::dump_generic_expr): Call it, and use the result,
7095         eliminating the direct usage of dump_file and alt_dump_file in
7096         favor of indirectly using them via emit_item.
7097         (make_item_for_dump_printf_va): New function, adapted from
7098         optinfo::add_printf_va.
7099         (make_item_for_dump_printf): New function.
7100         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
7101         and use the result, eliminating the direct usage of dump_file and
7102         alt_dump_file in favor of indirectly using them via emit_item.
7103         (make_item_for_dump_dec): New function.
7104         (dump_context::dump_dec): Call it, and use the result,
7105         eliminating the direct usage of dump_file and alt_dump_file in
7106         favor of indirectly using them via emit_item.
7107         (make_item_for_dump_symtab_node): New function, adapted from
7108         optinfo::add_symtab_node.
7109         (dump_context::dump_symtab_node): Call it, and use the result,
7110         eliminating the direct usage of dump_file and alt_dump_file in
7111         favor of indirectly using them via emit_item.
7112         (dump_context::begin_scope): Reimplement, avoiding direct usage
7113         of dump_file and alt_dump_file in favor of indirectly using them
7114         via emit_item.
7115         (dump_context::emit_item): New member function.
7116         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7117         Set up test pretty-printer on the underlying context.  Call
7118         refresh_dumps_are_enabled.
7119         (temp_dump_context::~temp_dump_context): Call
7120         refresh_dumps_are_enabled.
7121         (temp_dump_context::get_dumped_text): New member function.
7122         (selftest::verify_dumped_text): New function.
7123         (ASSERT_DUMPED_TEXT_EQ): New macro.
7124         (selftest::test_capture_of_dump_calls): Run all tests twice, with
7125         and then without optinfo enabled.  Add uses of
7126         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
7127         * dumpfile.h: Update comment for the dump_* API.
7128         * optinfo-emit-json.cc
7129         (selftest::test_building_json_from_dump_calls): Update for new
7130         param for temp_dump_context ctor.
7131         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7132         and "m_owned" field.
7133         (optinfo_item::~optinfo_item): Likewise.
7134         (optinfo::add_item): New member function.
7135         (optinfo::emit): Update comment.
7136         (optinfo::add_string): Delete.
7137         (optinfo::add_printf): Delete.
7138         (optinfo::add_printf_va): Delete.
7139         (optinfo::add_gimple_stmt): Delete.
7140         (optinfo::add_gimple_expr): Delete.
7141         (optinfo::add_tree): Delete.
7142         (optinfo::add_symtab_node): Delete.
7143         (optinfo::add_dec): Delete.
7144         * optinfo.h (class dump_context): New forward decl.
7145         (optinfo::add_item): New decl.
7146         (optinfo::add_string): Delete.
7147         (optinfo::add_printf): Delete.
7148         (optinfo::add_printf_va): Delete.
7149         (optinfo::add_gimple_stmt): Delete.
7150         (optinfo::add_gimple_expr): Delete.
7151         (optinfo::add_tree): Delete.
7152         (optinfo::add_symtab_node): Delete.
7153         (optinfo::add_dec): Delete.
7154         (optinfo::add_poly_int): Delete.
7155         (optinfo_item::optinfo_item): Remove "owned" param.
7156         (class optinfo_item): Remove field "m_owned".
7158 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7160         PR middle-end/86705
7161         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7162         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7163         requested variable alignment.
7164         (expand_one_ssa_partition): Likewise.
7165         (expand_one_var): Likewise.
7167 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7169         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7170         to speculation_safe_value_not_needed.
7172 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7174         * targhooks.h (speculation_safe_value_not_needed): New prototype.
7175         * targhooks.c (speculation_safe_value_not_needed): New function.
7176         * target.def (have_speculation_safe_value): Update documentation.
7177         * doc/tm.texi: Regenerated.
7179 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7181         * config/aarch64/iterators.md (ALLI_TI): New iterator.
7182         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7183         expand.
7184         (despeculate_copy<ALLI:mode>_insn): New insn.
7185         (despeculate_copyti_insn): New insn.
7186         (despeculate_simple<ALLI:mode>): New insn
7187         (despeculate_simpleti): New insn.
7188         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7189         function.
7190         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7191         aarch64_speculation_safe_value.
7192         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7194 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7196         * config/aarch64/aarch64-speculation.cc: New file.
7197         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7198         before pass_reorder_blocks.
7199         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7200         prototype.
7201         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7202         X14 and X15 when tracking speculation.
7203         * config/aarch64/aarch64.md (register name constants): Add
7204         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7205         (unspec): Add UNSPEC_SPECULATION_TRACKER.
7206         (speculation_barrier): New insn attribute.
7207         (cmp<mode>): Allow SP in comparisons.
7208         (speculation_tracker): New insn.
7209         (speculation_barrier): Add speculation_barrier attribute.
7210         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7211         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7212         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7214 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7216         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7217         aarch64_track_speculation is true.
7218         (tb<optab><mode>1): Likewise.
7219         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7220         generate CB[N]Z when tracking speculation.
7221         (aarch64_split_compare_and_swap): Likewise.
7222         (aarch64_split_atomic_op): Likewise.
7224 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7226         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
7228 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7230         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
7231         (speculation_barrier): New insn.
7233 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7235         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
7236         * config/arm/arm.md (speculation_barrier): New expand.
7237         (speculation_barrier_insn): New pattern.
7239 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7241         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
7242         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
7243         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
7244         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
7245         list.
7246         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
7247         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
7248         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
7249         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
7250         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
7251         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
7252         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
7253         * builtins.c (expand_speculation_safe_value): New function.
7254         (expand_builtin): Call it.
7255         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
7256         * doc/extend.texi: Document __builtin_speculation_safe_value.
7257         * doc/md.texi: Document "speculation_barrier" pattern.
7258         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
7259         TARGET_HAVE_SPECULATION_SAFE_VALUE.
7260         * doc/tm.texi: Regenerated.
7261         * target.def (have_speculation_safe_value, speculation_safe_value): New
7262         hooks.
7263         * targhooks.c (default_have_speculation_safe_value): New function.
7264         (default_speculation_safe_value): New function.
7265         * targhooks.h (default_have_speculation_safe_value): Add prototype.
7266         (default_speculation_safe_value): Add prototype.
7268 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7270         * dump-context.h (dump_context::dump_loc): New decl.
7271         * dumpfile.c (dump_context::dump_loc): New member function.
7272         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
7273         and dump_gimple_stmt.
7274         (dump_context::dump_gimple_expr_loc): Likewise, using
7275         dump_gimple_expr.
7276         (dump_context::dump_generic_expr_loc): Likewise, using
7277         dump_generic_expr.
7278         (dump_context::dump_printf_loc_va): Likewise, using
7279         dump_printf_va.
7280         (dump_context::begin_scope): Explicitly using the global function
7281         "dump_loc", rather than the member function.
7283 2018-07-31  Martin Sebor  <msebor@redhat.com>
7285         PR tree-optimization/86741
7286         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
7288 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
7290         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
7291         depend on whether prefetch instructions will be emitted or not.
7292         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
7293         will be emitted or not.
7294         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
7295         (TARGET_SETMEM_PFD): New macros.
7297 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7299         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
7300         (NULL_STMT_VEC_INFO): Delete.
7301         (stmt_vec_info::operator*): Likewise.
7302         (stmt_vec_info::operator gimple *): Likewise.
7303         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
7304         of NULL_STMT_VEC_INFO.
7305         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7306         (vect_reassociating_reduction_p): Likewise.
7307         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7308         (vectorizable_store): Likewise.
7309         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
7310         (vec_info::free_stmt_vec_infos): Likewise.
7312 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7314         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
7315         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
7317 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7319         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
7320         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7321         (vec_info::free_stmt_vec_info): New private member functions.
7322         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
7323         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
7324         * tree-parloops.c (gather_scalar_reductions): Remove calls to
7325         set_stmt_vec_info_vec and free_stmt_vec_infos.
7326         * tree-vect-loop.c (_loop_vec_info): Remove call to
7327         set_stmt_vec_info_vec.
7328         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
7329         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
7330         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
7331         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7332         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
7333         assignments in {vec_info::,}new_stmt_vec_info that are redundant
7334         with the clearing in the xcalloc.
7335         (stmt_vec_info_vec): Delete.
7336         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
7337         (vectorize_loops): Likewise.
7338         (vec_info::~vec_info): Remove argument from call to
7339         free_stmt_vec_infos.
7340         (vec_info::add_stmt): Remove vinfo argument from call to
7341         new_stmt_vec_info.
7343 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7345         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
7346         rather than a gimple stmt.
7347         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
7348         information for pattern statements when passed the original
7349         statement; instead wait to be passed the pattern statement itself.
7350         Don't call set_vinfo_for_stmt here.
7351         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
7352         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
7353         stmt_vec_infos here.
7354         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
7355         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
7356         stmt_vec_infos entry.
7358 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7360         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
7361         * tree-vectorizer.c (vec_info::replace_stmt): New function.
7362         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
7363         * tree-vect-stmts.c (vectorizable_call): Likewise.
7364         (vectorizable_simd_clone_call): Likewise.
7366 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7368         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
7369         * tree-vectorizer.c (vec_info::remove_stmt): New function.
7370         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
7371         * tree-vect-loop.c (vect_transform_loop): Likewise.
7372         * tree-vect-slp.c (vect_schedule_slp): Likewise.
7373         * tree-vect-stmts.c (vect_remove_stores): Likewise.
7375 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7377         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
7378         (vect_dr_stmt): Delete.
7379         * tree-vectorizer.c (vec_info::lookup_dr): New function.
7380         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
7381         of DR_VECT_AUX.
7382         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7383         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
7384         (vect_verify_datarefs_alignment, vect_peeling_supportable)
7385         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7386         (vect_analyze_data_refs): Likewise.
7387         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
7388         argument.
7389         (vect_find_same_alignment_drs): Likewise.
7390         (vect_slp_analyze_node_dependences): Update calls accordingly.
7391         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7392         instead of DR_VECT_AUX.
7393         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
7394         of a vector data references.  Use vec_info::lookup_dr instead of
7395         DR_VECT_AUX.
7396         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
7397         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7398         instead of DR_VECT_AUX.
7400 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7402         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
7403         dr_vec_info.
7404         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7405         accordingly.
7406         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7407         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7408         (vect_gen_prolog_loop_niters): Likewise.
7410 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7412         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
7413         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
7414         (vect_known_alignment_in_bytes, vect_dr_behavior)
7415         (vect_get_scalar_dr_size): Take references as dr_vec_infos
7416         instead of data_references.  Update calls to other routines for
7417         which the same change has been made.
7418         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
7419         dr_vec_infos instead of stmt_vec_infos.
7420         (vect_analyze_data_ref_dependence): Update call accordingly.
7421         (vect_slp_analyze_data_ref_dependence)
7422         (vect_record_base_alignments): Use DR_VECT_AUX.
7423         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
7424         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
7425         (vector_alignment_reachable_p, vect_get_data_access_cost)
7426         (vect_peeling_supportable, vect_analyze_group_access_1)
7427         (vect_analyze_group_access, vect_analyze_data_ref_access)
7428         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
7429         (vect_compile_time_alias, vect_small_gap_p)
7430         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
7431         (vect_supportable_dr_alignment): Take references as dr_vec_infos
7432         instead of data_references.  Update calls to other routines for
7433         which the same change has been made.
7434         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
7435         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7436         (vect_slp_analyze_and_verify_node_alignment)
7437         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7438         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7439         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
7440         above changes.
7441         (_vect_peel_info::dr): Replace with...
7442         (_vect_peel_info::dr_info): ...this new field.
7443         (vect_peeling_hash_get_most_frequent)
7444         (vect_peeling_hash_choose_best_peeling): Update accordingly.
7445         (vect_peeling_hash_get_lowest_cost):
7446         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
7447         routines for which the same change has been made.
7448         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
7449         data_reference.
7450         * tree-vect-loop-manip.c (get_misalign_in_elems)
7451         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
7452         above changes.
7453         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7454         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
7455         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
7456         (get_group_load_store_type, get_negative_load_store_type)
7457         (vect_get_data_ptr_increment, vectorizable_store)
7458         (vectorizable_load): Likewise.
7459         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
7460         Update calls to other routines for which the same change has been made.
7462 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7464         * tree-vectorizer.h (vec_info::move_dr): New member function.
7465         (dataref_aux): Rename to...
7466         (dr_vec_info): ...this and add "dr" and "stmt" fields.
7467         (_stmt_vec_info::dr_aux): Update accordingly.
7468         (_stmt_vec_info::data_ref_info): Delete.
7469         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
7470         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
7471         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
7472         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
7473         of data_ref.
7474         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
7475         (STMT_VINFO_DR_INFO): New macro.
7476         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
7477         (set_dr_misalignment): Update after rename of dataref_aux.
7478         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
7479         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
7480         initialization of STMT_VINFO_DATA_REF.
7481         * tree-vectorizer.c (vec_info::move_dr): New function.
7482         * tree-vect-patterns.c (vect_recog_bool_pattern)
7483         (vect_recog_mask_conversion_pattern)
7484         (vect_recog_gather_scatter_pattern): Use it.
7485         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
7486         the "dr" and "stmt" fields of dr_vec_info instead of
7487         STMT_VINFO_DATA_REF.
7489 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7491         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
7492         (is_pattern_stmt_p): Use it.
7493         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
7494         on pattern statements.
7496 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7498         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
7499         original stmt as a stmt_vec_info rather than a gimple stmt.
7500         (vect_pattern_recog_1): Take the statement directly as a
7501         stmt_vec_info, rather than via a gimple_stmt_iterator.
7502         Update call to vect_mark_pattern_stmts.
7503         (vect_pattern_recog): Update calls accordingly.
7505 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7507         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
7508         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
7509         a vect_def_type for the first argument.
7510         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
7511         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
7512         operand if it isn't defined by a vectorized statement.
7513         (vect_build_gather_load_calls): Remove the mask_dt argument and
7514         update calls to vect_get_vec_def_for_stmt_copy.
7515         (vectorizable_bswap): Likewise the dt argument.
7516         (vectorizable_call): Update calls to vectorizable_bswap and
7517         vect_get_vec_def_for_stmt_copy.
7518         (vectorizable_simd_clone_call, vectorizable_assignment)
7519         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
7520         (vectorizable_comparison): Update calls to
7521         vect_get_vec_def_for_stmt_copy.
7522         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
7523         vect_is_simple_use.
7524         (vect_get_loop_based_defs): Remove dt argument and update call
7525         to vect_get_vec_def_for_stmt_copy.
7526         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
7527         and vect_get_vec_def_for_stmt_copy.
7528         (vectorizable_load): Update calls to vect_build_gather_load_calls
7529         and vect_get_vec_def_for_stmt_copy.
7530         * tree-vect-loop.c (vect_create_epilog_for_reduction)
7531         (vectorizable_reduction, vectorizable_live_operation): Update calls
7532         to vect_get_vec_def_for_stmt_copy.
7534 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7536         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
7537         and gimple stmt arguments with a stmt_vec_info.
7538         (vect_record_base_alignments): Update calls accordingly.
7539         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
7540         and gimple stmt arguments with a stmt_vec_info.
7541         (vect_build_slp_tree_1): Remove vinfo argument and update call
7542         to vect_record_max_nunits.
7543         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
7544         and vect_record_max_nunits.
7546 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7548         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
7549         file and take a stmt_vec_info instead of a gimple stmt.
7550         (supportable_widening_operation, vect_finish_replace_stmt)
7551         (vect_finish_stmt_generation, vect_get_store_rhs)
7552         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
7553         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
7554         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
7555         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
7556         (vect_create_data_ref_ptr, bump_vector_ptr)
7557         (vect_permute_store_chain, vect_setup_realignment)
7558         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
7559         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
7560         (vectorizable_reduction, vectorizable_induction)
7561         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
7562         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
7563         than gimple stmts as arguments.
7564         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
7565         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
7566         (can_group_stmts_p, vect_check_gather_scatter)
7567         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7568         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
7569         (vect_permute_load_chain, vect_shift_permute_load_chain)
7570         (vect_transform_grouped_load)
7571         (vect_record_grouped_load_vectors): Likewise.
7572         * tree-vect-loop.c (vect_fixup_reduc_chain)
7573         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
7574         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
7575         (vectorizable_reduction, vectorizable_induction)
7576         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
7577         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
7578         (vect_get_load_store_mask): Likewise.
7579         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
7580         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
7581         * tree-vect-stmts.c (vect_mark_relevant)
7582         (is_simple_and_all_uses_invariant)
7583         (exist_non_indexing_operands_for_use_p, process_use)
7584         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
7585         (vect_get_vec_def_for_operand, vect_get_vec_defs)
7586         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
7587         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
7588         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
7589         (get_negative_load_store_type, get_load_store_type)
7590         (vect_check_load_store_mask, vect_check_store_rhs)
7591         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
7592         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7593         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7594         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7595         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
7596         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
7597         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
7598         (supportable_widening_operation): Likewise.
7600 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7602         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
7603         a stmt_vec_info instead of a gcall.
7604         (vect_check_gather_scatter): Update call accordingly.
7605         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
7606         of a gphi.
7607         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
7608         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
7609         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
7610         instead of a gimple stmt.
7611         (vect_transform_loop): Update calls accordingly.
7612         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
7613         stmt_vec_infos instead of gimple stmts.
7614         (vect_analyze_slp_instance): Update use accordingly.
7615         * tree-vect-stmts.c (read_vector_array, write_vector_array)
7616         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
7617         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
7618         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
7619         (vect_gen_widened_results_half, vect_get_loop_based_defs)
7620         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
7621         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
7622         down to subroutines.
7624 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7626         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
7627         of the worklist from a vector of gimple stmts to a vector of
7628         stmt_vec_infos.
7629         * tree-vect-stmts.c (vect_mark_relevant, process_use)
7630         (vect_mark_stmts_to_be_vectorized): Likewise
7632 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7634         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
7635         statement before passing it to vect_analyze_stmt.
7636         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
7637         the chain of phi vector definitions.  Track the exit phi via its
7638         stmt_vec_info.
7639         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
7640         STMT_VINFO_REDUC_DEF.
7641         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
7642         stmt_vec_infos to handle the statement chains.
7643         (vect_get_slp_defs): Record the first statement in the node
7644         using a stmt_vec_info.
7645         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
7646         statements here and pass their stmt_vec_info down to subroutines.
7647         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
7648         down to vect_finish_stmt_generation.
7649         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
7650         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
7651         stmt_vec_infos to subroutines.
7652         (vect_remove_stores): Use stmt_vec_infos to handle the statement
7653         chains.
7655 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7657         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
7658         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
7659         (vect_permute_store_chain, vect_setup_realignment)
7660         (vect_permute_load_chain, vect_shift_permute_load_chain)
7661         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
7662         stmts internally, and when passing values to other vectorizer routines.
7663         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7664         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
7665         (vect_analyze_loop_operations, get_initial_def_for_reduction)
7666         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
7667         (vectorizable_reduction, vectorizable_induction)
7668         (vectorizable_live_operation, vect_transform_loop_stmt)
7669         (vect_transform_loop): Likewise.
7670         * tree-vect-patterns.c (vect_reassociating_reduction_p)
7671         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
7672         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
7673         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7674         (vect_slp_analyze_node_operations_1): Likewise.
7675         * tree-vect-stmts.c (vect_mark_relevant, process_use)
7676         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
7677         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
7678         (vect_finish_stmt_generation_1, get_group_load_store_type)
7679         (get_load_store_type, vect_build_gather_load_calls)
7680         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7681         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7682         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7683         (vectorizable_store, vectorizable_load, vectorizable_condition)
7684         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
7685         (supportable_widening_operation): Likewise.
7686         (vect_get_vector_types_for_stmt): Likewise.
7687         * tree-vectorizer.h (vect_dr_behavior): Likewise.
7689 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7691         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7692         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
7693         (vect_permute_store_chain, vect_permute_load_chain)
7694         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
7695         repeated stmt_vec_info lookups.
7696         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7697         (vect_update_ivs_after_vectorizer): Likewise.
7698         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
7699         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7700         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
7701         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7702         (vect_bb_slp_scalar_cost): Likewise.
7703         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
7705 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7707         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
7708         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
7709         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
7710         to get gassigns and gcalls, rather than operating on generc gimple
7711         stmts.
7712         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
7713         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
7714         (vectorizable_load, vect_analyze_stmt): Likewise.
7715         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
7717 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7719         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
7720         return stmt_vec_infos rather than gimple stmts.  Do not accept
7721         null arguments.
7722         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
7723         of a gimple stmt.
7724         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7725         Update use of get_later_stmt.
7726         (vect_get_constant_vectors): Update call accordingly.
7727         (vect_schedule_slp_instance): Likewise
7728         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
7729         (vect_slp_analyze_instance_dependence): Likewise.
7730         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
7732 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7734         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
7735         (stmt_info_for_cost::stmt_info): ...this new field.
7736         (add_stmt_costs): Update accordingly.
7737         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7738         (vect_get_known_peeling_cost): Likewise.
7739         (vect_estimate_min_profitable_iters): Likewise.
7740         * tree-vect-stmts.c (record_stmt_cost): Likewise.
7742 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7744         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
7745         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7746         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7747         accordingly.
7748         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
7750 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7752         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
7753         a gimple stmt to a stmt_vec_info.
7754         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7756 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7758         * tree-vectorizer.h (vec_info::grouped_stores): Change from
7759         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7760         (_loop_vec_info::reduction_chains): Likewise.
7761         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
7762         accordingly.
7763         * tree-vect-slp.c (vect_analyze_slp): Likewise.
7765 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7767         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
7768         a gimple stmt to a stmt_vec_info.
7769         (_stmt_vec_info::next_element): Likewise.
7770         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
7771         (vect_slp_analyze_and_verify_node_alignment)
7772         (vect_analyze_group_access_1, vect_analyze_group_access)
7773         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
7774         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
7775         (vect_supportable_dr_alignment): Update accordingly.
7776         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
7777         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
7778         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
7779         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7780         * tree-vect-slp.c (vect_build_slp_tree_1)
7781         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
7782         (vect_split_slp_store_group, vect_analyze_slp_instance)
7783         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
7784         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
7785         (get_group_load_store_type, get_load_store_type)
7786         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
7787         (vect_transform_stmt, vect_remove_stores): Likewise.
7789 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7791         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
7792         than a gimple stmt.
7793         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7794         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
7795         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
7796         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
7797         (vector_alignment_reachable_p, vect_get_data_access_cost)
7798         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
7799         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
7800         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7801         (vect_analyze_group_access_1, vect_analyze_group_access)
7802         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
7803         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
7804         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
7805         result of vect_dr_stmt and use the stmt_vec_info instead of
7806         the associated gimple stmt.
7807         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7808         (vect_gen_prolog_loop_niters): Likewise.
7809         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7811 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7813         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
7814         to a vec<stmt_vec_info>.
7815         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
7816         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
7817         vec<stmt_vec_info>.
7818         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
7819         to a vec<stmt_vec_info>.
7820         (bst_traits::value_type, bst_traits::value_type): Likewise.
7821         (bst_traits::hash): Update accordingly.
7822         (vect_get_and_check_slp_defs): Change the stmts parameter from
7823         a vec<gimple *> to a vec<stmt_vec_info>.
7824         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
7825         (vect_build_slp_tree): Likewise.
7826         (vect_build_slp_tree_2): Likewise.  Update uses of
7827         SLP_TREE_SCALAR_STMTS.
7828         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
7829         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
7830         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
7831         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
7832         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
7833         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
7834         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
7835         (vect_get_constant_vectors, vect_get_slp_defs)
7836         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
7837         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
7838         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
7839         instead of gimple stmts.
7840         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
7841         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
7842         (vect_slp_analyze_instance_dependence): Update uses of
7843         SLP_TREE_SCALAR_STMTS.
7844         (vect_slp_analyze_and_verify_node_alignment): Likewise.
7845         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
7846         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
7847         (get_initial_defs_for_reduction): Likewise.
7848         (vect_create_epilog_for_reduction): Likewise.
7849         (vectorize_fold_left_reduction): Likewise.
7850         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
7851         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
7852         (can_vectorize_live_stmts): Likewise.
7854 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7856         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
7857         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7858         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
7859         than gimple stmts.
7860         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
7861         of a gimple stmt.
7862         (gather_scalar_reductions): Update after above interface changes.
7863         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
7864         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
7865         than gimple stmts.
7866         (vect_force_simple_reduction): Likewise.
7867         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
7868         LOOP_VINFO_REDUCTIONS.
7869         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7871 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7873         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
7874         a gimple stmt to a stmt_vec_info.
7875         * tree-vect-loop.c (vect_active_double_reduction_p)
7876         (vect_force_simple_reduction, vectorizable_reduction): Update
7877         accordingly.
7879 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7881         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
7882         vec<gimple *> to a vec<stmt_vec_info>.
7883         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
7884         the reduction_phis argument from a vec<gimple *> to a
7885         vec<stmt_vec_info>.
7886         (vectorizable_reduction): Likewise the phis local variable that
7887         is passed to vect_create_epilog_for_reduction.  Update for new type
7888         of SLP_TREE_VEC_STMTS.
7889         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
7890         (vectorizable_live_operation): Likewise.
7891         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
7892         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
7894 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7896         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
7897         a gimple stmt to a stmt_vec_info.
7898         (vectorizable_condition, vectorizable_live_operation)
7899         (vectorizable_reduction, vectorizable_induction): Pass back the
7900         vectorized statement as a stmt_vec_info.
7901         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
7902         use of STMT_VINFO_VEC_STMT.
7903         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
7904         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
7905         as stmt_vec_infos rather than gimple stmts.
7906         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
7907         to a stmt_vec_info.
7908         (vectorizable_live_operation): Likewise.
7909         (vectorizable_reduction, vectorizable_induction): Likewise,
7910         updating use of STMT_VINFO_VEC_STMT.
7911         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
7912         of STMT_VINFO_VEC_STMT.
7913         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
7914         (vectorizable_simd_clone_call, vectorizable_conversion)
7915         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7916         (vectorizable_store, vectorizable_load, vectorizable_condition)
7917         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
7918         from a gimple stmt to a stmt_vec_info.
7919         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
7920         pointer to a stmt_vec_info to the vectorizable_* routines.
7922 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7924         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7925         a gimple stmt to a stmt_vec_info.
7926         (is_pattern_stmt_p): Update accordingly.
7927         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7928         (vect_record_grouped_load_vectors): Likewise.
7929         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7930         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7931         (vect_model_reduction_cost): Likewise.
7932         (vect_create_epilog_for_reduction): Likewise.
7933         (vectorizable_reduction, vectorizable_induction): Likewise.
7934         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7935         Return the stmt_vec_info for the pattern statement.
7936         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7937         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7938         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7939         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7940         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7941         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7942         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7943         (free_stmt_vec_info, vect_is_simple_use): Likewise.
7945 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7947         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7948         (vect_finish_stmt_generation): Likewise.
7949         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7950         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7951         (vect_build_gather_load_calls): Use the return value of the above
7952         functions instead of a separate call to vinfo_for_stmt.  Use narrow
7953         scopes for the input gimple stmt and wider scopes for the associated
7954         stmt_vec_info.  Use vec_info::lookup_def when setting these
7955         stmt_vec_infos from an SSA_NAME definition.
7956         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7957         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7958         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7959         (vectorizable_store, vectorizable_load, vectorizable_condition)
7960         (vectorizable_comparison): Likewise.
7961         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7962         (vectorizable_reduction): Likewise.
7964 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7966         * tree-vectorizer.h (vect_is_simple_use): Add an optional
7967         stmt_vec_info * parameter before the optional gimple **.
7968         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7969         (process_use, vect_get_vec_def_for_operand_1): Update callers.
7970         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7971         * tree-vect-loop.c (vectorizable_reduction): Likewise.
7972         (vectorizable_live_operation): Likewise.
7973         * tree-vect-patterns.c (type_conversion_p): Likewise.
7974         (vect_look_through_possible_promotion): Likewise.
7975         (vect_recog_rotate_pattern): Likewise.
7976         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7978 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7980         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7981         a typedef to a wrapper class.
7982         (NULL_STMT_VEC_INFO): New macro.
7983         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7984         (stmt_vec_info::operator*): New function.
7985         (stmt_vec_info::operator gimple *): Likewise.
7986         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7987         (add_stmt_costs): Likewise.
7988         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7989         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7990         (vect_get_known_peeling_cost): Likewise.
7991         (vect_estimate_min_profitable_iters): Likewise.
7992         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7993         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7994         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7995         (vectorizable_store, free_stmt_vec_infos): Likewise.
7996         (new_stmt_vec_info): Change return type of xcalloc to
7997         _stmt_vec_info *.
7999 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8001         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
8002         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
8003         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
8004         a single_imm_use-based sequence.
8005         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
8007 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8009         * tree-vectorizer.h (vec_info::lookup_def): Declare.
8010         * tree-vectorizer.c (vec_info::lookup_def): New function.
8011         * tree-vect-patterns.c (vect_get_internal_def): Use it.
8012         (vect_widened_op_tree): Likewise.
8013         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8014         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
8015         (vectorizable_reduction): Likewise.
8016         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
8017         of a gimple *.
8018         (vect_is_slp_reduction): Update calls accordingly.  Use
8019         vec_info::lookup_def.
8020         (vect_is_simple_reduction): Likewise
8021         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
8023 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8025         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
8026         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
8027         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
8028         of vinfo_for_stmt.
8029         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
8030         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
8031         (vect_update_vf_for_slp, vect_analyze_loop_operations)
8032         (vect_is_slp_reduction, vectorizable_induction)
8033         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
8034         * tree-vect-patterns.c (vect_init_pattern_stmt):
8035         (vect_determine_min_output_precision_1, vect_determine_precisions)
8036         (vect_pattern_recog): Likewise.
8037         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
8038         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
8039         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
8040         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8041         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
8042         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
8043         info field from a loop to a loop_vec_info.
8045 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8047         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
8048         (vec_info::add_stmt): Declare.
8049         * tree-vectorizer.c (vec_info::add_stmt): New function.
8050         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
8051         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
8052         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8053         (vectorizable_induction): Likewise.
8054         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
8055         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8056         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
8057         (vectorizable_load): Likewise.
8058         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8059         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
8060         (vect_recog_gather_scatter_pattern): Likewise.
8061         (append_pattern_def_seq): Likewise.  Remove a check that is
8062         performed by add_stmt itself.
8064 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8066         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
8067         which make_ssa_name was called with new_stmt before new_stmt
8068         had been created.
8070 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8072         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
8073         split out from...
8074         (vect_is_slp_reduction): ...here...
8075         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
8076         that are already known to be false.
8078 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8080         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
8081         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
8082         STMT_VINFO_NUM_SLP_USES when it's true.
8083         (vect_free_slp_instance): Add a final_p parameter and pass it to
8084         vect_free_slp_tree.
8085         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
8086         (vect_analyze_slp_instance): Likewise.
8087         (vect_slp_analyze_operations): Likewise.
8088         (vect_slp_analyze_bb_1): Likewise.
8089         * tree-vectorizer.c (vec_info): Likewise.
8090         * tree-vect-loop.c (vect_transform_loop): Likewise.
8092 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8094         * tree-vect-loop.c (vectorizable_reduction): Assert that the
8095         function is not called for second and subsequent members of
8096         a reduction group.
8098 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8100         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
8101         cases for nested loops from here to ...
8102         (vect_create_epilog_for_reduction): ...here.  Only call
8103         vect_is_simple_use for inner-loop reductions.
8105 2018-07-31  Martin Liska  <mliska@suse.cz>
8107         PR gcov-profile/85338
8108         PR gcov-profile/85350
8109         PR gcov-profile/85372
8110         * profile.c (struct location_triplet): New.
8111         (struct location_triplet_hash): Likewise.
8112         (output_location): Do not output a BB that
8113         is already recorded for a line.
8114         (branch_prob): Use streamed_locations.
8116 2018-07-31  Martin Liska  <mliska@suse.cz>
8118         PR gcov-profile/85370
8119         * coverage.c (coverage_begin_function): Do not mark target
8120         clones as artificial functions.
8122 2018-07-31  Martin Liska  <mliska@suse.cz>
8124         PR gcov-profile/83813
8125         PR gcov-profile/84758
8126         PR gcov-profile/85217
8127         PR gcov-profile/85332
8128         * profile.c (branch_prob): Do not record GOTO expressions
8129         for GIMPLE statements which locations are already streamed.
8131 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8133         * gcc.c (handle_spec_function): Accept a soft_matched_part
8134         argument, as do_spec_1.  Pass it down to ...
8135         (eval_spec_function): Accept a soft_matched_part argument,
8136         and pass it down to ...
8137         (do_spec_2): Accept a soft_matched_part argument, and pass
8138         it down to do_spec_1.
8139         (do_spec_1): Pass soft_matched_part to handle_spec_function.
8140         (handle_braces): Update call to handle_spec_function.
8141         (driver::set_up_specs): Update calls to do_spec_2.
8142         (compare_debug_dump_opt_spec_function): Likewise.
8143         (compare_debug_self_opt_spec_function): Likewise.
8145 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8147         * common.opt (nolibc): New option.
8148         * doc/invoke.texi (Link Options): Document it.
8149         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8150         * config/alpha/linux.h: Likewise.
8151         * config/arc/elf.h: Likewise.
8152         * config/arm/uclinux-elf.h: Likewise.
8153         * config/arm/unknown-elf.h: Likewise.
8154         * config/avr/avrlibc.h: Likewise.
8155         * config/bfin/bfin.h: Likewise.
8156         * config/bfin/linux.h: Likewise.
8157         * config/bfin/uclinux.h: Likewise.
8158         * config/darwin.h: Likewise.
8159         * config/darwin10.h: Likewise.
8160         * config/darwin12.h: Likewise.
8161         * config/gnu-user.h: Likewise.
8162         * config/lm32/uclinux-elf.h: Likewise.
8163         * config/pa/pa-hpux11.h: Likewise.
8164         * config/pa/pa64-hpux.h: Likewise.
8165         * config/sparc/sparc.h: Likewise.
8167 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8169         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
8170         undefined variables.
8172 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8174         PR target/86640
8175         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8176         instead of GEN_INT.
8178 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8180         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8181         terminated string literal.
8183 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8185         PR rtl-optimization/85160
8186         * combine.c (is_just_move): New function.
8187         (try_combine): Allow combining two instructions into two if neither of
8188         the original instructions was a move.
8190 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
8192         PR target/86673
8193         * doc/extend.texi (Global Register Variables): Discourage use of type
8194         qualifiers.
8195         (Local Register Variables): Likewise.
8197 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
8199         PR tree-optimization/86506
8200         * hwint.h (ceil_log2): Resync with hwint.c implementation.
8202 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
8204         PR target/86547
8205         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8206         hard_regno, make sure no insn between `from` and `to` clobbers it.
8208 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
8209             Tom de Vries  <tdevries@suse.de>
8211         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8212         (PTX_DEFAULT_RUNTIME_DIM): ... this.
8213         (nvptx_goacc_validate_dims): Set default worker and gang dims to
8214         PTX_DEFAULT_RUNTIME_DIM.
8215         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8217 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
8219         * config/pa/pa.c (pa_output_addr_vec): Align address table.
8220         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8221         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
8223 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
8225         * config/rs6000/constraints.md (wG constraint): Delete, no longer
8226         used.
8227         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
8228         predicate to reflect toc fusion has been deleted.
8229         (toc_fusion_mem_raw): Delete, no longer used.
8230         (toc_fusion_mem_wrapped): Likewise.
8231         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
8232         fusion mask bit.
8233         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
8234         Delete, no longer used.
8235         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
8236         meant to be used for toc fusion.
8237         (rs6000_debug_print_mode): Delete toc fusion debugging.
8238         (rs6000_debug_reg_global): Likewise.
8239         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
8240         fusion and secondary reload support that were never used.
8241         (rs6000_option_override_internal): Delete TOC fusion, that was only
8242         partially defined, and it did not work unless you also used the
8243         -mcmodel= switch.
8244         (rs6000_legitimate_address_p): Delete TOC fusion support.
8245         (rs6000_opt_masks): Likewise.
8246         (fusion_wrap_memory_address): Delete function, no longer used.
8247         (fusion_split_address); Delete TOC fusion support.
8248         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
8249         longer used with toc fusion being deleted.
8250         (TARGET_TOC_FUSION_FP): Likewise.
8251         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
8252         UNSPEC.
8253         (toc fusion spliter): Delete TOC fusion support.
8254         (toc_fusionload_<mode>): Likewise.
8255         (toc_fusionload_di): Likewise.
8256         (fusion_gpr_load_<mode>): Delete generator function, this insn no
8257         longer needs to be named.  Rename predicate to delete TOC fusion.
8258         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8259         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8260         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8261         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8262         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
8264 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8266         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
8267         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
8268         __int128 in built-in function prototypes.
8269         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
8270         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
8272 2018-07-27  Martin Sebor  <msebor@redhat.com>
8274         PR tree-optimization/86696
8275         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
8276         types, including enums.
8277         (handle_char_store): Be prepared for the above function to fail.
8279 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
8281         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
8282         when optimization level is lower than 2 or optimize for size.
8284 2018-07-26  Martin Sebor  <msebor@redhat.com>
8286         PR tree-optimization/86043
8287         PR tree-optimization/86042
8288         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
8289         (get_string_cst_length): Rename...
8290         (get_min_string_length): ...to this.  Add argument.
8291         (handle_char_store): Extend to handle multi-character stores by
8292         MEM_REF.
8293         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
8294         * tree.h (initializer_zerop): Add argument.
8296 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
8298         PR middle-end/86660
8299         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
8300         declare target to variables if they have always,{to,from,tofrom} map
8301         kinds.
8303 2018-07-26  Martin Liska  <mliska@suse.cz>
8305         PR lto/86548
8306         * lto-wrapper.c: Add linker_output as prefix
8307         for ltrans_output_file.
8309 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
8311         PR rtl-optimization/85805
8312         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
8313         value for hard registers if that was written in the same mode.
8315 2018-07-26  Martin Liska  <mliska@suse.cz>
8317         PR gcov-profile/86536
8318         * gcov.c (format_gcov): Use printf format %.*f directly
8319         and do not handle special values.
8321 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8323         * common/config/arc/arc-common.c (arc_option_optimization_table):
8324         Update default optimizations for size.
8326 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8328         * config/arc/arc.md (movsf_insn): Add short instruction selection.
8329         * config/arc/constraints.md (CfZ): New constraint.
8330         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
8331         (subsf3_fpu): Likewise.
8332         (cmpsf_fpu): Likewise.
8333         (cmpsf_fpu_uneq): Likewise.
8335 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8337         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
8338         accesses as well.
8339         (arc_is_uncached_mem_p): uncached applies to both the variable and
8340         the pointer.
8342 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8344         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
8345         register names.
8347 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
8349         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
8350         field "m_scopes" from vec to auto_vec.
8352 2018-07-25  Martin Liska  <mliska@suse.cz>
8354         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
8355         return type.
8357 2018-07-25  Richard Biener  <rguenther@suse.de>
8359         PR debug/86654
8360         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
8361         special wrt context_die late.
8362         (gen_subprogram_die): Re-use DIEs in local scope.
8364 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
8366         PR tree-optimization/86644
8367         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
8369 2018-07-25  Martin Liska  <mliska@suse.cz>
8371         PR middle-end/86645
8372         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
8373         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
8375 2018-07-25  Martin Liska  <mliska@suse.cz>
8377         PR sanitizer/79635
8378         * params.def: Explain ASan abbreviation and provide
8379         a documentation link.
8381 2018-07-24  Martin Sebor  <msebor@redhat.com>
8383         PR tree-optimization/86622
8384         PR tree-optimization/86532
8385         * builtins.h (string_length): Declare.
8386         * builtins.c (c_strlen): Correct handling of non-constant offsets.
8387         (check_access): Be prepared for non-constant length ranges.
8388         (string_length): Make extern.
8389         * expr.c (string_constant): Only handle the minor non-constant
8390         array index.  Use string_constant to compute the length of
8391         a generic string constant.
8393 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
8395         PR tree-optimization/86618
8396         * tree-vect-stmts.c (vectorizable_call): Don't take the address
8397         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
8399 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
8401         PR tree-optimization/86636
8402         * json.cc (json::object::set): Fix comment.  Add assertions.
8403         (json::array::append): Move here from json.h.  Add comment and an
8404         assertion.
8405         (json::string::string): Likewise.
8406         * json.h (json::array::append): Move to json.cc.
8407         (json::string::string): Likewise.
8408         * optinfo-emit-json.cc
8409         (optrecord_json_writer::impl_location_to_json): Assert that we
8410         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
8411         wrapper around it.  Expand the location once, rather than three
8412         times.
8413         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
8414         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
8415         wrappers.
8416         (optrecord_json_writer::optinfo_to_json): Likewise, in four
8417         places.  Fix some overlong lines.
8419 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
8421         * config/aarch64/aarch64-simd.md
8422         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
8423         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
8424         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
8425         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
8426         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
8427         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
8428         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
8429         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
8430         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
8432 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
8434         PR middle-end/86627
8435         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
8436         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
8437         and abs_d == d, do the power of two handling if profitable.
8439 2018-07-24  Richard Biener  <rguenther@suse.de>
8441         * match.pd: Add BIT_FIELD_REF canonicalizations.
8443 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8445         PR c/86617
8446         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
8448 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8450         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
8451         terminated STRING_CST object.
8453 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8455         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
8457 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
8459         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
8460         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
8461         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
8462         the elements into a register.
8463         (rs6000_split_v4si_init_di_reg): Delete.
8464         (rs6000_split_v4si_init): Delete.
8465         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
8466         (vsx_init_v4si): Rewrite as a define_expand.
8468 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
8470         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
8471         zero_extend argument from memory): New.
8473 2018-07-22  Martin Sebor  <msebor@redhat.com>
8475         PR bootstrap/86621
8476         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
8477         diagnosing calls with unknown arguments unless -Walloca-larger-than
8478         is restricted to less than PTRDIFF_MAX bytes.
8480 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
8482         * doc/gcov.texi (Invoking Gcov): Editorial changes.
8484 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
8486         * pretty-print.c (text_info::set_location): Remove redundant
8487         "line_table" parameter from call to rich_location::set_range.
8489 2018-07-20  Martin Sebor  <msebor@redhat.com>
8491         PR middle-end/82063
8492         * builtins.c (expand_builtin_alloca): Adjust.
8493         * calls.c (alloc_max_size): Simplify.
8494         * cgraphunit.c (cgraph_node::expand): Adjust.
8495         * common.opt (larger_than_size, warn_frame_larger_than): Remove
8496         variables.
8497         (frame_larger_than_size): Same.
8498         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
8499         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
8500         Initialize.
8501         * doc/invoke.texi (GCC Command Options): Document option arguments.
8502         Explain byte-size arguments and suffixes.
8503         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
8504         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
8505         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
8506         * doc/options.texi (UInteger): Expand.
8507         (Host_Wide_Int, ByteSize): Document new properties.
8508         * final.c (final_start_function_1): Include sizes in an error message.
8509         * function.c (frame_offset_overflow): Same.
8510         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
8511         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
8512         Diagnose unbounded alloca calls only for limits of less than
8513         PTRDIFF_MAX.
8514         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
8515         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
8516         for alloca(0).
8517         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
8518         only for limits of less than PTRDIFF_MAX.
8519         * langhooks-def.h (lhd_handle_option): Change function argument
8520         to HOST_WIDE_INT.
8521         * langhooks.c (lhd_handle_option): Same.
8522         * langhooks.h (handle_option): Same.
8523         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
8524         ByteSize flags.
8525         (var_type, var_type_struct): Same.
8526         (var_set): Handle ByteSize flag.
8527         * optc-gen.awk: Add comments to output to ease debugging.  Make
8528         use of HOST_WIDE_INT where appropriate.
8529         * opts-gen-save.awk:  Use %lx to format unsigned long.
8530         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
8531         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
8532         arguments.  Parse bytes-size suffixes.
8533         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
8534         (enum_value_to_arg): Same.
8535         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
8536         (handle_option): Adjust.
8537         (generate_option): Change function argument to HOST_WIDE_INT.
8538         (cmdline_handle_error): Adjust.
8539         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
8540         (set_option): Change function argument to HOST_WIDE_INT.
8541         (option_enabled): Handle cl_host_wide_int.
8542         (get_option_state): Handle CLVC_SIZE.
8543         (control_warning_option): Same.
8544         * opts.c (common_handle_option): Change function argument to
8545         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
8546         OPT_Wvla_larger_than_.
8547         * opts.h (enum cl_var_type): Add an enumerator.
8548         * stor-layout.c (layout_decl): Print a more meaningful warning.
8549         * toplev.c (output_stack_usage): Adjust.
8551 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
8553         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
8554         call to inline_expand_builtin_string_cmp.
8555         (expand_builtin_strcmp): Likewise.
8556         (expand_builtin_strncmp): Likewise.
8557         (inline_string_cmp): Delete the last parameter, change char_type_node
8558         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
8559         two operands.
8560         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
8561         the inlining expansion on target where the type of the call has same or
8562         narrower precision than unsigned char.
8564 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
8566         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
8567         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
8568         * common.opt (fsave-optimization-record): New option.
8569         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
8570         * doc/invoke.texi (-fsave-optimization-record): New option.
8571         * dumpfile.c: Include "optinfo-emit-json.h".
8572         (struct kv_pair): Move to coretypes.h.
8573         (optgroup_options): Make non-static.
8574         (dump_context::end_scope): Call
8575         optimization_records_maybe_pop_dump_scope.
8576         * dumpfile.h (optgroup_options): New decl.
8577         * json.cc: New file.
8578         * json.h: New file.
8579         * optinfo-emit-json.cc: New file.
8580         * optinfo-emit-json.h: New file.
8581         * optinfo.cc: Include "optinfo-emit-json.h".
8582         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
8583         (optinfo_enabled_p): Check optimization_records_enabled_p.
8584         (optinfo_wants_inlining_info_p): Likewise.
8585         * optinfo.h: Update comment.
8586         * profile-count.c (profile_quality_as_string): New function.
8587         * profile-count.h (profile_quality_as_string): New decl.
8588         (profile_count::quality): New accessor.
8589         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
8590         and optinfo_emit_json_cc_tests.
8591         * selftest.h (selftest::json_cc_tests): New decl.
8592         (selftest::optinfo_emit_json_cc_tests): New decl.
8593         * toplev.c: Include "optinfo-emit-json.h".
8594         (compile_file): Call optimization_records_finish.
8595         (do_compile): Call optimization_records_start.
8596         * tree-ssa-live.c: Include optinfo.h.
8597         (remove_unused_scope_block_p): Retain inlining information if
8598         optinfo_wants_inlining_info_p returns true.
8600 2018-07-20  Richard Biener  <rguenther@suse.de>
8602         PR debug/86585
8603         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
8604         to cover -flto-partition=none.
8606 2018-07-20  Martin Liska  <mliska@suse.cz>
8608         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
8609         (get_decl_source_range): Remove unused function.
8611 2018-07-20  Richard Biener  <rguenther@suse.de>
8613         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
8614         (struct vn_phi_s): Likewise.
8615         (struct vn_reference_s): Likewise.
8616         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
8617         for searching the slot of an entry known to be in the hash itself.
8618         (vn_phi_hasher::equal): Likewise.
8619         (vn_reference_hasher::equal): Likewise.
8620         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
8621         globals.
8622         (optimistic_info, current_info): Remove, keeping only valid_info.
8623         (vn_reference_lookup_1): Remove fallback lookup.
8624         (vn_reference_lookup_2): Likewise.
8625         (vn_nary_op_lookup_1): Likewise.
8626         (vn_phi_lookup): Likewise.
8627         (vn_nary_build_or_lookup_1): Make sure to not chain the built
8628         hash element.
8629         (vn_reference_insert): Adjust, chain the inserted hash element
8630         at last_inserted_ref.
8631         (vn_reference_insert_pieces): Likewise.
8632         (visit_reference_op_call): Likewise.
8633         (vn_nary_op_insert_into): Chain the inserted hash element at
8634         last_inserted_nary.
8635         (vn_nary_op_insert_pieces): Adjust.
8636         (vn_nary_op_insert): Likewise.
8637         (vn_nary_op_insert_stmt): Likewise.
8638         (vn_phi_insert): Adjust, chain the inserted hash element at
8639         last_inserted_phi.
8640         (process_scc): Remove clearing and copying the optimistic
8641         table.  Instead remove elements inserted during an optimistic
8642         iteration from the single table we maintain.
8643         (init_scc_vn): Adjust.
8644         (free_scc_vn): Likewise.
8645         (sccvn_dom_walker::record_cond): Likewise.
8646         (sccvn_dom_walker::after_dom_children): Likewise.
8648 2018-07-19  Martin Sebor  <msebor@redhat.com>
8650         PR tree-optimization/84047
8651         PR tree-optimization/83776
8652         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
8653         (check_array_bounds): Call it.
8655 2018-07-19  Martin Sebor  <msebor@redhat.com>
8657         * align.h (align_flags): Use member initialization.
8659 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
8661         * Makefile.in (OBJS): Add optinfo.o.
8662         * coretypes.h (class symtab_node): New forward decl.
8663         (struct cgraph_node): New forward decl.
8664         (class varpool_node): New forward decl.
8665         * dump-context.h: New file.
8666         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
8667         "tree-pass.h".
8668         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
8669         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
8670         (set_alt_dump_file): Likewise.
8671         (dump_context::~dump_context): New dtor.
8672         (dump_gimple_stmt): Move implementation to...
8673         (dump_context::dump_gimple_stmt): ...this new member function.
8674         Add the stmt to any pending optinfo, creating one if need be.
8675         (dump_gimple_stmt_loc): Move implementation to...
8676         (dump_context::dump_gimple_stmt_loc): ...this new member function.
8677         Start a new optinfo and add the stmt to it.
8678         (dump_gimple_expr): Move implementation to...
8679         (dump_context::dump_gimple_expr): ...this new member function.
8680         Add the stmt to any pending optinfo, creating one if need be.
8681         (dump_gimple_expr_loc): Move implementation to...
8682         (dump_context::dump_gimple_expr_loc): ...this new member function.
8683         Start a new optinfo and add the stmt to it.
8684         (dump_generic_expr): Move implementation to...
8685         (dump_context::dump_generic_expr): ...this new member function.
8686         Add the tree to any pending optinfo, creating one if need be.
8687         (dump_generic_expr_loc): Move implementation to...
8688         (dump_context::dump_generic_expr_loc): ...this new member
8689         function.  Add the tree to any pending optinfo, creating one if
8690         need be.
8691         (dump_printf): Move implementation to...
8692         (dump_context::dump_printf_va): ...this new member function.  Add
8693         the text to any pending optinfo, creating one if need be.
8694         (dump_printf_loc): Move implementation to...
8695         (dump_context::dump_printf_loc_va): ...this new member function.
8696         Start a new optinfo and add the stmt to it.
8697         (dump_dec): Move implementation to...
8698         (dump_context::dump_dec): ...this new member function.  Add the
8699         value to any pending optinfo, creating one if need be.
8700         (dump_context::dump_symtab_node): New member function.
8701         (dump_context::get_scope_depth): New member function.
8702         (dump_context::begin_scope): New member function.
8703         (dump_context::end_scope): New member function.
8704         (dump_context::ensure_pending_optinfo): New member function.
8705         (dump_context::begin_next_optinfo): New member function.
8706         (dump_context::end_any_optinfo): New member function.
8707         (dump_context::s_current): New global.
8708         (dump_context::s_default): New global.
8709         (dump_scope_depth): Delete global.
8710         (dumpfile_ensure_any_optinfo_are_flushed): New function.
8711         (dump_symtab_node): New function.
8712         (get_dump_scope_depth): Reimplement in terms of dump_context.
8713         (dump_begin_scope): Likewise.
8714         (dump_end_scope): Likewise.
8715         (selftest::temp_dump_context::temp_dump_context): New ctor.
8716         (selftest::temp_dump_context::~temp_dump_context): New dtor.
8717         (selftest::verify_item): New function.
8718         (ASSERT_IS_TEXT): New macro.
8719         (ASSERT_IS_TREE): New macro.
8720         (ASSERT_IS_GIMPLE): New macro.
8721         (selftest::test_capture_of_dump_calls): New test.
8722         (selftest::dumpfile_c_tests): Call it.
8723         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
8724         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
8725         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
8726         descriptive comment.
8727         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
8728         (dump_node, dump_bb): Move these unrelated decls.
8729         (class dump_manager): Add leading comment.
8730         * optinfo.cc: New file.
8731         * optinfo.h: New file.
8733 2018-07-19  Michael Collison  <michael.collison@arm.com>
8734             Richard Henderson <rth@redhat.com>
8736         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
8737         (subti): Handle op1 zero.
8738         (subvti4, usub4ti4): New.
8739         (*sub<GPI>3_compare1_imm): New.
8740         (sub<GPI>3_carryinCV): New.
8741         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
8742         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
8744 2018-07-19  Michael Collison  <michael.collison@arm.com>
8745             Richard Henderson <rth@redhat.com>
8747         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
8748         (addti3): Create simpler code if low part is already known to be 0.
8749         (addvti4, uaddvti4): New.
8750         (*add<GPI>3_compareC_cconly_imm): New.
8751         (*add<GPI>3_compareC_cconly): New.
8752         (*add<GPI>3_compareC_imm): New.
8753         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
8754         handle constants within this pattern..
8755         (*add<GPI>3_compareV_cconly_imm): New.
8756         (*add<GPI>3_compareV_cconly): New.
8757         (*add<GPI>3_compareV_imm): New.
8758         (add<GPI>3_compareV): New.
8759         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
8760         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
8761         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
8762         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
8763         with 'comparison' operator.
8764         (*add<GPI>3_compareV_cconly_imm): Ditto.
8765         (*add<GPI>3_compareV_cconly): Ditto.
8766         (*add<GPI>3_compareV_imm): Ditto.
8767         (add<GPI>3_compareV): Ditto.
8768         (add<mode>3_carryinC): Ditto.
8769         (*add<mode>3_carryinC_zero): Ditto.
8770         (*add<mode>3_carryinC): Ditto.
8771         (add<mode>3_carryinV): Ditto.
8772         (*add<mode>3_carryinV_zero): Ditto.
8773         (*add<mode>3_carryinV): Ditto.
8775 2018-07-19  Michael Collison  <michael.collison@arm.com>
8776             Richard Henderson <rth@redhat.com>
8778         * config/aarch64/aarch64-modes.def (CC_V): New.
8779         * config/aarch64/aarch64-protos.h
8780         (aarch64_addti_scratch_regs): Declare
8781         (aarch64_subvti_scratch_regs): Declare.
8782         (aarch64_expand_subvti): Declare.
8783         (aarch64_gen_unlikely_cbranch): Declare
8784         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
8785         for signed overflow using CC_Vmode.
8786         (aarch64_get_condition_code_1): Handle CC_Vmode.
8787         (aarch64_gen_unlikely_cbranch): New function.
8788         (aarch64_addti_scratch_regs): New function.
8789         (aarch64_subvti_scratch_regs): New function.
8790         (aarch64_expand_subvti): New function.
8792 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8794         * config/aarch64/aarch64-option-extensions.def: New entry for profile
8795         extension.
8796         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
8797         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
8798         extension.
8800 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8802         PR target/83009
8803         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8804         address check not strict.
8806 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8808         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
8809         Umq with Umn.
8810         (store_pair_lanes<mode>): Likewise.
8811         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
8812         enum value 'ADDR_QUERY_LDP_STP_N'.
8813         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
8814         (aarch64_print_address_internal): Add declaration.
8815         (aarch64_print_ldpstp_address): Remove.
8816         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
8817         (aarch64_print_operand): Change printing of 'y'.
8818         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
8819         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
8820         'true' rather than '1'.
8821         * config/aarch64/constraints.md (Uml): Likewise.
8822         (Uml): Rename to Umn.
8823         (Umq): Remove.
8825 2018-07-19  Richard Biener  <rguenther@suse.de>
8827         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
8828         a trailing array.
8829         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
8830         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
8831         (vn_reference_hasher): Likewise.
8832         (struct vn_tables_s): Remove obstack and alloc-pool members.
8833         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
8834         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
8835         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
8836         (vn_reference_insert_pieces): Likewise.
8837         (alloc_vn_nary_op_noinit): Adjust.
8838         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
8839         (vn_phi_eq): Adjust.
8840         (shared_lookup_phiargs): Remove.
8841         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
8842         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
8843         (visit_reference_op_call): Likewise.
8844         (copy_nary, copy_phi, copy_reference): Remove.
8845         (process_scc): Rewind the obstack when iterating.  Do not
8846         copy the elements to valid_info but just move them from one
8847         hashtable to the other.
8848         (allocate_vn_table): Adjust.
8849         (free_vn_table): Likewise.
8850         (init_scc_vn): Likewise.
8851         (free_scc_vn): Likewise.
8853 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
8855         PR target/86560
8856         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
8857         indirect_return as function type attribute.
8858         (ix86_attribute_table): Change indirect_return to function
8859         type attribute.
8860         * doc/extend.texi: Update indirect_return attribute.
8862 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
8864         * wide-int.h (widest2_int): New.
8865         * gimple-fold.c (arith_overflowed_p): Use it.
8866         * tree.h (widest2_int_cst): New.
8867         * tree-vrp.c (wide_int_binop_overflow): Rename from
8868         vrp_int_const_binop.
8869         Rewrite to work on trees.
8870         (extract_range_from_multiplicative_op_1): Abstract code to...
8871         (wide_int_range_min_max): ...here.
8872         (wide_int_range_cross_product): ...and here.
8873         (extract_range_from_binary_expr_1): Abstract overflow code to...
8874         (wide_int_range_mult_wrapping): ...here.
8875         * tree-vrp.h (wide_int_range_cross_product): New.
8876         (wide_int_range_mult_wrapping): New.
8878 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
8879             Julia Koval  <julia.koval@intel.com>
8881         * config/i386/x86-tune-costs.h (skylake_memcpy,
8882         skylake_memset): Replace rep_prefix with unrolling for size 512.
8884 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
8886         PR middle-end/86544
8887         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
8888         comparision with EQ_EXPR in last stmt.
8890 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8892         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
8893         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
8894         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
8895         previously known as "PowerPC AltiVec Built-in Functions".  Move
8896         some material to new subsubsections "PowerPC AltiVec Built-in
8897         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
8898         ISA 2.07".
8899         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
8900         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
8901         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
8902         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
8904 2018-07-18  Richard Biener  <rguenther@suse.de>
8906         PR tree-optimization/86557
8907         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
8908         EXACT_DIV_EXPR.
8910 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
8912         * config/s390/s390.c (s390_function_profiler): Generate CFI.
8914 2018-07-17  Jeff Law  <law@redhat.com>
8916         * config/arm/arm.c (get_label_padding): Update for recent
8917         changes to label_to_alignment.
8919         PR tree-optimization/86010
8920         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
8922         * config/mips/mips.c (vr4130_align_insns): Update for recent
8923         changes to label_to_alignment.
8925         * config/frv/frv.c (frv_label_align): Update for recent changes
8926         to label_to_alignment.
8928         * config/nios2/nios2.c (nios2_label_align): Update for recent
8929         changes which dropped ALIGN_LABELS_LOG.
8931 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
8933         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8934         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8936 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
8938         * config/arc/arc.c (arc_label_align): Use align_labels instead of
8939         deprecated align_labels_log.
8941 2018-07-17  Richard Biener  <rguenther@suse.de>
8943         PR lto/86456
8944         * dwarf2out.c (init_sections_and_labels): Always generate
8945         a debug_line_str_section for early LTO debug.
8946         (dwarf2out_finish): Reset debug_line_str_hash output early.
8947         Bump counter for extra dwarf5 .debug_loc labels to not conflict
8948         with fat LTO part.
8949         (dwarf2out_early_finish): Output debug_line_str.
8951 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
8953         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8954         index register on z196 or later.
8956 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
8958         * config/s390/s390.c (s390_default_align): Set default function
8959         alignment to 16.
8960         (s390_override_options_after_change): Call s390_default align.
8961         (s390_option_override_internal): Call s390_default_align.
8962         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8964 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8966         PR middle-end/86542
8967         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8968         to _looptemp_ clauses, other than the first two.
8970 2018-07-17  Martin Liska  <mliska@suse.cz>
8972         * opts.c: Do not enable OPT_falign_* for -Os.
8974 2018-07-17  Martin Liska  <mliska@suse.cz>
8976         * align.h (MAX_CODE_ALIGN): New.
8977         (MAX_CODE_ALIGN_VALUE): New.
8978         * common/config/i386/i386-common.c (ix86_handle_option):
8979         (MAX_CODE_ALIGN): Moved to align.h.
8980         * final.c (MAX_CODE_ALIGN): Likewise.
8981         * opts.c (parse_and_check_align_values):
8982         (MAX_CODE_ALIGN): Likewise.
8983         (MAX_CODE_ALIGN_VALUE): Likewise.
8985 2018-07-17  Martin Liska  <mliska@suse.cz>
8987         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8988         in order to fulfil coding style.
8989         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8990         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8991         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8992         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8993         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8994         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8995         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8996         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8997         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8999 2018-07-17  Martin Liska  <mliska@suse.cz>
9001         * align.h: New file.
9002         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
9003         directly.
9004         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
9005         align_flags of label_to_alignment.
9006         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
9007         align_flags class.
9008         * config/m68k/m68k.c: Do not use removed align_labels_value and
9009         align_loops_value.
9010         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
9011         (LOOP_ALIGN): Likewise.
9012         (LABEL_ALIGN): Likewise.
9013         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9014         Remove not used macro.
9015         (rs6000_loop_align): Change return type to align_flags.
9016         (rs6000_loop_align_max_skip): Remove.
9017         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
9018         Change return type to align_flags.
9019         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9020         Remove not used macro.
9021         (rs6000_loop_align):  Change return type to align_flags.
9022         (rs6000_loop_align_max_skip): Remove.
9023         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
9024         * config/rx/rx-protos.h (rx_align_for_label): Make it
9025         static function.
9026         * config/rx/rx.c (rx_align_for_label): Change return type
9027         to align_flags.
9028         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
9029         macro definitions.
9030         into align_flags class.
9031         (LABEL_ALIGN): Likewise.
9032         (LOOP_ALIGN): Likewise.
9033         * config/s390/s390.c (s390_label_align): Use align_flags
9034         class member.
9035         (s390_asm_output_function_label): Likewise.
9036         * config/sh/sh.c (sh_override_options_after_change):
9037         Use align_flags class directly without macros.
9038         (find_barrier): Likewise.
9039         (barrier_align): Likewise.
9040         (sh_loop_align): Likewise.
9041         * config/spu/spu.c (spu_option_override):
9042         Use align_flags_tuple::get_value instead of removed macros.
9043         (spu_sched_init): Likewise.
9044         * config/spu/spu.h (GTY): Likewise.
9045         * config/visium/visium.c (visium_option_override):
9046         Set "8" as default secondary alignment.
9047         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
9048         in order to guarantee secondary alignment of 8.
9049         * coretypes.h: Include align.h header file.
9050         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
9051         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
9052         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
9053         * doc/tm.texi.in: Likewise.
9054         * final.c (struct label_alignment): Remove not used structure.
9055         (LABEL_ALIGN): Change type to align_flags.
9056         (LOOP_ALIGN): Likewise.
9057         (JUMP_ALIGN): Likewise.
9058         (default_loop_align_max_skip): Remove.
9059         (default_label_align_max_skip): Likewise.
9060         (default_jump_align_max_skip): Likewise.
9061         (default_label_align_after_barrier_max_skip):
9062         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
9063         (LABEL_TO_MAX_SKIP): Remove.
9064         (label_to_alignment): Return align_flags type instead of integer.
9065         (label_to_max_skip): Remove.
9066         (align_fuzz): Use align_flags type.
9067         (compute_alignments): Use align_flags type and use align_flags::max
9068         to combine multiple alignments.
9069         (grow_label_align): Grow vec instead of C array.
9070         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
9071         (shorten_branches):  Use align_flags type and use align_flags::max
9072         to combine multiple alignments.
9073         (final_scan_insn_1): Remove usage of secondary alignment that comes
9074         from label alignment, but instead use proper secondary alignment
9075         which is computed in grow_label_align.
9076         * flags.h (struct align_flags_tuple): Move to align.h.
9077         (struct align_flags): Likewise.
9078         (state_align_loops): Rename to align_loops.
9079         (state_align_jumps): Rename to align_jumps.
9080         (state_align_labels): Rename to align_labels.
9081         (state_align_functions): Rename to align_functions.
9082         (align_loops_log): Remove.
9083         (align_jumps_log): Remove.
9084         (align_labels_log): Remove.
9085         (align_functions_log): Remove.
9086         (align_loops_max_skip): Remove.
9087         (align_jumps_max_skip): Remove.
9088         (align_labels_max_skip): Remove.
9089         (align_functions_max_skip): Remove.
9090         (align_loops_value): Remove.
9091         (align_jumps_value): Remove.
9092         (align_labels_value): Remove.
9093         (align_functions_value): Remove.
9094         * output.h (label_to_alignment): Change return type to align_flags.
9095         (label_to_max_skip): Remove.
9096         * target.def: Remove loop_align_max_skip, label_align_max_skip,
9097         jump_align_max_skip macros.
9098         * targhooks.h (default_loop_align_max_skip): Remove.
9099         (default_label_align_max_skip): Likewise.
9100         (default_jump_align_max_skip): Likewise.
9101         (default_label_align_after_barrier_max_skip): Remove.
9102         * toplev.c (read_log_maxskip): Use ::normalize function.
9103         (parse_N_M): Remove not used argument and also call ::normalize.
9104         (parse_alignment_opts): Do not pass unused arguments.
9105         * varasm.c (assemble_start_function): Use directly align_functions
9106         instead of removed macros.
9107         * system.h: Do not poison removed macros.
9109 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9111         PR middle-end/86539
9112         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
9113         and cond temporaries don't have reference type if iterator has
9114         pointer type.  For init use &for_pre_body instead of pre_p if
9115         for_pre_body is non-empty.
9117 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9119         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9120         double-double modes to SFmode directly directly.
9121         (trunc<mode>sf2_fprs): Delete.
9123 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9125         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9126         for conversions between IFmode and the decimal floating point modes.
9127         (init_float128_ieee): Use the correct names for conversions between
9128         KFmode and the decimal floating point modes.
9130 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9132         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9133         for the conversions between TDmode and IFmode.
9134         (init_float128_ieee): Use more correct names for the conversions
9135         between TDmode and KFmode.
9137 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
9139         PR tree-optimization/86526
9140         * builtins.c (expand_builtin_memcmp): Formatting fixes.
9141         (inline_expand_builtin_string_cmp): Likewise.
9142         (inline_string_cmp): Likewise.  Use c_readstr instead of
9143         builtin_memcpy_read_str.  Add unit_mode temporary.
9145 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9147         PR middle-end/86528
9148         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9149         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9151 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9153         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9154         Alphabetize prototypes of built-in functions, separating out
9155         built-in functions that are listed in this section but should be
9156         described elsewhere.
9158 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
9160         PR target/86511
9161         * expmed.c (emit_store_flag): Do not emit setcc followed by a
9162         conditional move when trapping comparison was split to a
9163         non-trapping one (and vice versa).
9165 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9167         * config/s390/s390.c (s390_function_profiler): Generate nops
9168         instead of profiler call sequences.
9169         * config/s390/s390.opt: Add the new option.
9171 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9173         * config/s390/s390.c (s390_function_profiler): Generate
9174         __mcount_loc section.
9175         * config/s390/s390.opt: Add the new option.
9177 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9179         * common.opt: Add the new warning.
9180         * config/s390/s390.c (s390_function_profiler): Emit "brasl
9181         %r0,__fentry__" when -mfentry is specified.
9182         (s390_option_override_internal): Disallow -mfentry for 31-bit
9183         CPUs.
9184         * config/s390/s390.opt: Add the new option.
9186 2018-07-16  Richard Biener  <rguenther@suse.de>
9188         PR lto/86523
9189         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9190         for function-local FUNCTION_DECL and RESULT_DECL immediately.
9192 2018-07-16  Martin Liska  <mliska@suse.cz>
9194         PR ipa/86529
9195         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9196         to ::get_create.
9198 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9200         * config/arc/arcHS.md: Update ARCHS scheduling rules.
9202 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9204         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9205         for ARCHS4x.
9206         * config/arc/arc-cpus.def (hs4x): New cpu.
9207         (hs4xd): Likewise.
9208         * config/arc/arc-tables.opt: Regenerate.
9209         * config/arc/arc.c (arc_sched_issue_rate): New function.
9210         (TARGET_SCHED_ISSUE_RATE): Define.
9211         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9212         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9213         fpu_cvt.
9214         (attr tune): Add ARCHS4x tune values.
9215         (attr tune_dspmpy): Define.
9216         (*tst): Correct instruction type.
9217         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9218         * config/arc/arcHS4x.md: New file.
9219         * config/arc/fpu.md: Update instruction type attributes.
9220         * config/arc/t-multilib: Regenerate.
9222 2018-07-16  Tom de Vries  <tdevries@suse.de>
9224         PR debug/86455
9225         * var-tracking.c (vt_initialize): Fix pre_dec handling.
9227 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9229         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
9230         early clobber.
9232 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
9234         PR tree-optimization/86514
9235         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
9236         conversion to a boolean type from a type with greater precision.
9238 2018-07-16  Tom de Vries  <tdevries@suse.de>
9240         * var-tracking.c (vt_initialize): Print adjusted insn slim if
9241         dump_flags request TDF_SLIM.
9243 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
9245         * fold-const.c (int_const_binop_1): Abstract...
9246         (wide_int_binop): ...wide int code here.
9247         (poly_int_binop): ...poly int code here.
9248         Abstract the rest of int_const_binop_1 into int_const_binop.
9249         * fold-const.h (wide_int_binop): New.
9250         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
9251         Remove useless PLUS/MINUS_EXPR case.
9252         (zero_nonzero_bits_from_vr): Move wide int code...
9253         (zero_nonzero_bits_from_bounds): ...here.
9254         (extract_range_from_binary_expr_1): Move mask optimization code...
9255         (range_easy_mask_min_max): ...here.
9256         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
9257         (range_easy_mask_min_max): New.
9259 2018-07-15  Jeff Law  <law@redhat.com>
9261         PR target/85993
9262         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
9263         block.
9265 2018-07-14  Jim Wilson  <jimw@sifive.com>
9267         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
9269 2018-07-14  Paul Koning  <ni1d@arrl.net>
9271         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
9273 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
9275         * lto-streamer-out.c (copy_function_or_variable): Dump info about
9276         copying section.
9278 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
9279             Steve Munroe  <munroesj52@gmail.com>
9281         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
9282         (_mm_andnot_si128): Likewise.
9283         (_mm_or_si128): Likewise.
9284         (_mm_xor_si128): Likewise.
9286 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
9288         PR middle-end/78809
9289         * builtins.c (expand_builtin_memcmp): Inline the calls first
9290         when result_eq is false.
9291         (expand_builtin_strcmp): Inline the calls first.
9292         (expand_builtin_strncmp): Likewise.
9293         (inline_string_cmp): New routine. Expand a string compare
9294         call by using a sequence of char comparison.
9295         (inline_expand_builtin_string_cmp): New routine. Inline expansion
9296         a call to str(n)cmp/memcmp.
9297         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
9298         New option.
9299         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
9301 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
9303         * config/arm/driver-arm.c: Include arm-native.h.
9304         (host_detect_local_cpu): Use auto-generated data tables.
9305         (vendors, arm_cpu_table): Delete.  Move part information to ...
9306         * config/arm/arm-cpus.in: ... here.
9307         * config/arm/parsecpu.awk (gen_native): New function.
9308         (vendor, part): New CPU fields.
9309         (END): Add support for building the native CPU detection tables.
9310         * config/arm/t-arm (arm-native.h): Add build rule.
9311         (driver-arm.o): Add dependency on arm-native.h.
9313 2018-07-13  Richard Biener  <rguenther@suse.de>
9315         PR middle-end/85974
9316         * match.pd (addr1 - addr2): Allow either of the operand to
9317         have a conversion.
9319 2018-07-13  Tom de Vries  <tdevries@suse.de>
9321         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
9322         in remap_ssa_name.
9324 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
9326         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
9327         arrays instead of numbered variables.
9329 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
9331         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
9332         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
9334 2018-07-13  Richard Biener  <rguenther@suse.de>
9336         PR debug/86452
9337         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
9338         instead of get_context_die.
9340 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
9341             Richard Biener  <rguenther@suse.de>
9343         PR middle-end/86489
9344         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
9345         that the loop latch destination where phi is defined.
9347 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
9349         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
9350         (riscv_expand_epilogue): Add assertion to check interrupt mode.
9351         (riscv_set_current_function): Extract getting interrupt type to new
9352         function.
9353         (riscv_get_interrupt_type): New function.
9354         (riscv_merge_decl_attributes): New function, checking interrupt type is
9355         same.
9356         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
9358 2018-07-12  Paul Koning  <ni1d@arrl.net>
9360         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
9361         directive.
9363 2018-07-12  Paul Koning  <ni1d@arrl.net>
9365         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
9366         zero reference, add doloop_end instead.
9367         * doc/md.texi (decrement_and_branch_until_zero): Remove.
9368         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
9369         detail for doloop_end.
9371 2018-07-12  Martin Sebor  <msebor@redhat.com>
9373         PR c/86453
9374         * attribs.c (decl_attributes): Reject conflicting attributes before
9375         calling attribute handlers.
9377 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9379         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
9380          parameter.
9381         (gcc::dump_manager::get_dump_file_name): likewise.
9382         (dump_begin): Likewise.
9383         * dumpfile.h (dump_begin): Update prototype.
9384         (gcc::dump_manager::get_dump_file_name,
9385         gcc::dump_manager::get_dump_file_name): Update prototype.
9387 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9389         * internal-fn.h (vectorizable_internal_fn_p): New function.
9390         * tree-vect-slp.c (compatible_calls_p): Likewise.
9391         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
9392         to internal functions.
9393         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
9395 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9397         * fold-const.h (inverse_conditions_p): Declare.
9398         * fold-const.c (inverse_conditions_p): New function.
9399         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
9400         that test the inverse condition of a conditional internal function.
9401         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
9402         * internal-fn.c (internal_fn_mask_index): Handle conditional
9403         internal functions.
9404         (vectorized_internal_fn_supported_p): New function.
9405         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
9406         (any_pred_load_store): Replace with...
9407         (need_to_predicate): ...this new variable.
9408         (redundant_ssa_names): New variable.
9409         (ifcvt_can_use_mask_load_store): Move initial checks to...
9410         (ifcvt_can_predicate): ...this new function.  Handle tree codes
9411         for which a conditional internal function exists.
9412         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
9413         instead of ifcvt_can_use_mask_load_store.  Update after variable
9414         name change.
9415         (predicate_load_or_store): New function, split out from
9416         predicate_mem_writes.
9417         (check_redundant_cond_expr): New function.
9418         (value_available_p): Likewise.
9419         (predicate_rhs_code): Likewise.
9420         (predicate_mem_writes): Rename to...
9421         (predicate_statements): ...this.  Use predicate_load_or_store
9422         and predicate_rhs_code.
9423         (combine_blocks, tree_if_conversion): Update after above name changes.
9424         (ifcvt_local_dce): Handle redundant_ssa_names.
9425         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
9426         general conditional functions.
9427         * tree-vect-stmts.c (vectorizable_call): Likewise.
9429 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9430             Alan Hayward  <alan.hayward@arm.com>
9431             David Sherwood  <david.sherwood@arm.com>
9433         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
9434         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
9435         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
9436         plus and minus and convert them into IFN_COND_FMA-based sequences.
9437         (convert_mult_to_fma): Handle conditional plus and minus.
9439 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9441         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
9442         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
9443         (cond_fnms_optab): New optabs.
9444         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
9445         internal functions.
9446         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
9447         * internal-fn.h (get_conditional_internal_fn): Declare.
9448         (get_unconditional_internal_fn): Likewise.
9449         * internal-fn.c (cond_ternary_direct): New macro.
9450         (expand_cond_ternary_optab_fn): Likewise.
9451         (direct_cond_ternary_optab_supported_p): Likewise.
9452         (FOR_EACH_COND_FN_PAIR): Likewise.
9453         (get_conditional_internal_fn): New function.
9454         (get_unconditional_internal_fn): Likewise.
9455         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
9456         (gimple_match_op::gimple_match_op): Add a new overload for 5
9457         operands.
9458         (gimple_match_op::set_op): Likewise.
9459         (gimple_resimplify5): Declare.
9460         * genmatch.c (decision_tree::gen): Generate simplifications for
9461         5 operands.
9462         * gimple-match-head.c (gimple_simplify): Define an overload for
9463         5 operands.  Handle calls with 5 arguments in the top-level overload.
9464         (convert_conditional_op): Handle conversions from unconditional
9465         internal functions to conditional ones.
9466         (gimple_resimplify5): New function.
9467         (build_call_internal): Pass a fifth operand.
9468         (maybe_push_res_to_seq): Likewise.
9469         (try_conditional_simplification): Try converting conditional
9470         internal functions to unconditional internal functions.
9471         Handle 3-operand unconditional forms.
9472         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
9473         Define ternary equivalents of the current rules for binary conditional
9474         internal functions.
9475         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
9476         ternary operations.
9477         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
9478         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
9479         (optab): Handle them.
9480         (SVE_COND_FP_TERNARY): New int iterator.
9481         (sve_fmla_op, sve_fmad_op): New int attributes.
9482         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
9483         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
9484         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
9486 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9488         * target.def (preferred_else_value): New target hook.
9489         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
9490         * doc/tm.texi: Regenerate.
9491         * targhooks.h (default_preferred_else_value): Declare.
9492         * targhooks.c (default_preferred_else_value): New function.
9493         * internal-fn.h (conditional_internal_fn_code): Declare.
9494         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
9495         (get_conditional_internal_fn): Use it.
9496         (conditional_internal_fn_code): New function.
9497         * gimple-match.h (gimple_match_cond): New struct.
9498         (gimple_match_op): Add a cond member function.
9499         (gimple_match_op::gimple_match_op): Update all forms to take a
9500         gimple_match_cond.
9501         * genmatch.c (expr::gen_transform): Use the same condition as res_op
9502         for the suboperation, but don't specify a particular else_value.
9503         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
9504         (visit_nary_op, visit_reference_op_load): Pass
9505         gimple_match_cond::UNCOND to the gimple_match_op constructor.
9506         * gimple-match-head.c: Include tree-eh.h
9507         (convert_conditional_op): New function.
9508         (maybe_resimplify_conditional_op): Likewise.
9509         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
9510         (gimple_resimplify2): Likewise.
9511         (gimple_resimplify3): Likewise.
9512         (gimple_resimplify4): Likewise.
9513         (maybe_push_res_to_seq): Return null for conditional operations.
9514         (try_conditional_simplification): New function.
9515         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
9516         constructor.
9517         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
9518         IFN_COND_* call.
9519         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
9520         function.
9521         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
9523 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9525         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9526         DECL_FCONTEXT
9527         (hash_tree): Do not hash DECL_FCONTEXT
9528         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
9529         Do not stream DECL_FCONTEXT.
9530         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
9531         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
9533 2018-07-12  Richard Biener  <rguenther@suse.de>
9535         PR debug/86462
9536         * dwarf2out.c (gen_block_die): Only output blocks when they have
9537         at least one !DECL_IGNORED_P variable.
9539 2018-07-12  Richard Biener  <rguenther@suse.de>
9541         PR target/84829
9542         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
9543         Remove -mieee-fp handling.
9545 2018-07-12  Richard Biener  <rguenther@suse.de>
9547         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
9548         left-over from last patch.
9550 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
9552         PR tree-optimization/86492
9553         * gimple-ssa-store-merging.c
9554         (imm_store_chain_info::coalesce_immediate_stores): Call
9555         check_no_overlap even for the merge_overlapping case.  Formatting fix.
9557 2018-07-12  Richard Biener  <rguenther@suse.de>
9559         PR middle-end/86479
9560         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
9561         move possibly trapping operations into the conditional.
9563 2018-07-12  Richard Biener  <rguenther@suse.de>
9565         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
9566         (vn_lookup_simplify_result): Remove recursion limit applied
9567         here.
9568         (vn_nary_build_or_lookup_1): Adjust.
9569         (try_to_simplify): Likewise.
9570         * gimple-match-head.c (gimple_resimplify1): Instead apply one
9571         here.
9572         (gimple_resimplify2): Likewise.
9573         (gimple_resimplify3): Likewise.
9574         (gimple_resimplify4): Likewise.
9576 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
9578         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
9579         Use __mmask64 type instead of __mmask8 for __M argument.
9580         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
9581         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
9582         __U argument.
9583         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
9584         __mmask16 for __M argument.
9585         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
9586         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
9587         to __mmask16 instead of __mmask8.
9588         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
9589         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
9590         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
9591         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
9592         instead of __mmask16 for __U argument.
9593         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
9594         __mmask16 instead of __mmask8 for __U argument.
9595         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
9596         __U argument.
9597         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
9598         __mmask16.
9599         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
9600         argument.
9601         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
9602         __U argument.
9603         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
9604         __mmask16.
9605         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
9606         of __mmask16.
9607         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
9608         __U argument.
9609         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
9610         __U argument.
9611         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
9612         __U argument.
9613         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
9614         __U argument.
9615         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
9616         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
9617         return type as well as __M argument type and all casts from __mmask8
9618         to __mmask32.
9619         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
9620         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
9621         return type as well as __M argument type and all casts from __mmask8
9622         to __mmask16.
9623         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
9624         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
9625         return type as well as __M argument type and all casts from __mmask8
9626         to __mmask32.
9627         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
9628         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
9629         return type as well as __M argument type and all casts from __mmask8
9630         to __mmask16.
9631         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
9632         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
9633         __mmask16.
9635 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
9637         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
9638         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
9639         for __U argument.
9641 2018-07-11  Paul Koning  <ni1d@arrl.net>
9643         * doc/md.texi (define_subst): Document how multiple occurrences of
9644         the same argument in the replacement pattern are handled.
9646 2018-07-11  Paul Koning  <ni1d@arrl.net>
9648         * doc/extend.texi (Common Variable Attributes): Move "mode" into
9649         alphabetical order.
9650         (Common Type Attributes): Add "mode" attribute.
9652 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
9654         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
9655         stream DECL_ORIGINAL_TYPE.
9656         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
9657         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
9658         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
9659         Do not walk original type.
9660         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
9661         external decls.
9662         (write_ts_decl_non_common_tree_pointers): Do not stream
9663         DECL_ORIGINAL_TYPE
9664         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
9665         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
9667 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
9669         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
9670         thread twice from the same starting edge.
9672 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
9674         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
9675         * gimple.c (gimple_call_nonnull_result_p): ...here...
9676         (gimple_call_nonnull_arg): ...and here.
9677         * gimple.h (gimple_call_nonnull_result_p): New.
9678         (gimple_call_nonnull_arg): New.
9680 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
9682         * config/arm/arm-cpus.in: Move information from fpu field of each
9683         cpu definition to the isa field.
9684         * config/arm/parsecpu.awk (fpu): Delete match rule.
9685         (gen_comm_data): Don't add bits from the CPU's FPU entry.
9687 2018-07-11  Richard Biener  <rguenther@suse.de>
9689         PR debug/86457
9690         * dwarf2out.c (init_sections_and_labels): Use
9691         output_asm_line_debug_info consistently.
9692         (dwarf2out_early_finish): Likewise.
9693         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
9694         type units.
9696 2018-07-11  Richard Biener  <rguenther@suse.de>
9698         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
9699         Rework father_bb setting in a way to avoid propagating constants
9700         multiple times on a loop body.
9702 2018-07-10  Mark Wielaard  <mark@klomp.org>
9704         PR debug/86459
9705         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
9707 2018-07-10  Richard Biener  <rguenther@suse.de>
9709         * hash-map.h (hash_map::iterator::operator*): Return
9710         references to key and value.
9712 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
9714         PR c++/86443
9715         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
9716         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
9717         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
9718         (gimplify_omp_for): For composite loops, move outer
9719         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
9720         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
9721         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
9722         TREE_LIST for both the original class iterator and the "last" helper
9723         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
9724         loop, remember has_decl_expr from outer composite loops for the
9725         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
9727 2018-07-09  Martin Sebor  <msebor@redhat.com>
9729         PR middle-end/77357
9730         PR middle-end/86428
9731         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
9732         accessing implicitly initialized array elements.
9733         * expr.c (string_constant): Handle string initializers of
9734         character arrays within aggregates.
9735         * gimple-fold.c (fold_array_ctor_reference): Add argument.
9736         Store element offset.  As a special case, handle zero size.
9737         (fold_nonarray_ctor_reference): Same.
9738         (fold_ctor_reference): Add argument.  Store subobject offset.
9739         * gimple-fold.h (fold_ctor_reference): Add argument.
9741 2018-07-09  Paul Koning  <ni1d@arrl.net>
9743         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
9744         (pdp11_insn_cost): New function.
9745         (pdp11_md_asm_adjust): New function.
9746         (TARGET_INVALID_WITHIN_DOLOOP): Define.
9747         (pdp11_rtx_costs): Update to match machine better.
9748         (output_addr_const_pdp11): Correct format mismatch warnings.
9749         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
9750         * config/pdp11/pdp11.md: General change to add base_cost and/or
9751         length attributes for use by new pdp11_insn_cost function.
9752         (MIN_BRANCH): Correct definition.
9753         (MIN_SOB): Ditto.
9754         (doloop_end): Use standard pattern name for looping pattern.
9755         (doloop_end_nocc): New.
9756         (movsf): Add another constraint alternative.
9757         (zero_extendqihi2): Add constraint alternatives for not in place
9758         extend.
9759         (zero_extendhisi2): Remove.
9760         (shift patterns): Add CC handling variants.
9761         (bswaphi2): New.
9762         (bswapsi2): New.
9763         (rothi3): New.
9764         (define_peephole2): New peephole to recognize mov that sets CC for
9765         subsequent test.
9767 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9769         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
9770         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
9771         wi::add.
9773 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
9775         PR c/86420
9776         * real.c (real_nextafter): Return true if result is denormal.
9778 2018-07-09  Martin Liska  <mliska@suse.cz>
9780         * common.opt: Add back wrongly removed attribute.
9782 2018-07-09  Richard Biener  <rguenther@suse.de>
9784         PR debug/86413
9785         * dwarf2out.c (gen_block_die): For an early generated DIE
9786         always output high/low PC attributes.
9788 2018-07-09  Tom de Vries  <tdevries@suse.de>
9790         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
9791         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
9792         onto VAR_DECL with abstract origin.
9794 2018-07-07  Jim Wilson  <jimw@sifive.com>
9796         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
9798 2018-07-07  Tom de Vries  <tdevries@suse.de>
9800         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
9802 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
9804         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
9805         overflow_type.
9806         (combine_bound): Use wide-int overflow calculation instead of
9807         rolling our own.
9808         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
9809         overflow_type.
9810         * fold-const.c (int_const_binop_2): Same.
9811         (extract_muldiv_1): Same.
9812         (fold_div_compare): Same.
9813         (fold_abs_const): Same.
9814         * match.pd: Same.
9815         * poly-int.h (add): Same.
9816         (sub): Same.
9817         (neg): Same.
9818         (mul): Same.
9819         * predict.c (predict_iv_comparison): Same.
9820         * profile-count.c (slow_safe_scale_64bit): Same.
9821         * simplify-rtx.c (simplify_const_binary_operation): Same.
9822         * tree-chrec.c (tree_fold_binomial): Same.
9823         * tree-data-ref.c (split_constant_offset_1): Same.
9824         * tree-if-conv.c (idx_within_array_bound): Same.
9825         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
9826         * tree-ssa-phiopt.c (minmax_replacement): Same.
9827         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
9828         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
9829         * vr-values.c (vr_values::adjust_range_with_scev): Same.
9830         * wide-int.cc (wi::add_large): Same.
9831         (wi::mul_internal): Same.
9832         (wi::sub_large): Same.
9833         (wi::divmod_internal): Same.
9834         * wide-int.h: Change overflow type to overflow_type for neg, add,
9835         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
9836         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
9837         mul_internal, divmod_internal.
9838         (overflow_type): New enum.
9839         (accumulate_overflow): New.
9841 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
9843         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
9844         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
9846 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
9848         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
9849         argument is checked for zero before entering loop, avoid checking again.
9851 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
9853         * gimplify.h (generic_expr_could_trap_p): Set as global function.
9854         * gimplify.h (generic_expr_could_trap_p): Likwise.
9855         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
9857 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
9859         PR tree-optimization/86401
9860         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
9861         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
9862         (fold_bit_and_mask): ... here.  New helper function for match.pd.
9863         * fold-const.h (fold_bit_and_mask): Declare.
9864         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
9866 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
9868         PR target/86324
9869         * target.def (translate_mode_attribute): New hook.
9870         * targhooks.h (default_translate_mode_attribute): Declare.
9871         * targhooks.c (default_translate_mode_attribute): New function.
9872         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
9873         * doc/tm.texi: Regenerate.
9874         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
9875         (rs6000_translate_mode_attribute): New function.
9877 2018-07-06  Paul Koning  <ni1d@arrl.net>
9879         * doc/md.texi (define_split): Document DONE and FAIL.
9880         (define_peephole2): Ditto.
9882 2018-07-05  Jeff Law  <law@redhat.com>
9884         PR tree-optimization/86010
9885         * tree-ssa-dse.c (compute_trims): More aggressively trim at
9886         both the head and tail of mem* and str* calls.
9888 2018-07-05  Jim Wilson  <jimw@sifive.com>
9890         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
9892 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
9894         * config/aarch64/aarch64-simd.md: correct flags text for
9895         MIN_EXPR replacement.
9897 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
9899         * configure: Regenerated.
9901 2018-07-05  Carl Love  <cel@us.ibm.com>
9903         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
9904         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9905         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
9906         VSX_BUILTIN_DOUBLEL_V4SF.
9908 2018-07-05  Martin Sebor  <msebor@redhat.com>
9910         PR c++/86400
9911         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
9912         than its domain to compute its the upper bound of a char array.
9914 2018-07-05  Nathan Sidwell  <nathan@acm.org>
9916         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
9917         * doc/cpp.texi: Update comment.
9918         * doc/tm.texi: Rebuilt.
9919         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
9920         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
9921         * doc/extend.texi (Backwards Compatibility): Clarify it is system
9922         headers affected by extern "C".
9923         * system.h: Poison NO_IMPLICIT_EXTERN_C.
9924         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9925         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9926         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9927         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9928         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9929         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9930         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9931         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9932         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9933         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9934         config/sparc/sp64-elf.h, config/spu/spu.h,
9935         config/stormy16/stormy16.h, config/v850/v850.h,
9936         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9937         define NO_IMPLICIT_EXTERN_C.
9938         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9940 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9942         PR target/84711
9943         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9944         instead of GET_MODE_SIZE when comparing Units.
9946 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9948         PR target/84711
9949         * rtlanal.c (set_noop_p): Constrain on mode change,
9950         include hard-reg-set.h
9952 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9954         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9956 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
9958         Revert
9959         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9961         PR sanitizer/84250
9962         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9963         libasan.
9964         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9966 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9968         PR sanitizer/84250
9969         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9970         libasan.
9971         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9973 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
9975         PR middle-end/86380
9976         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9978 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
9980         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9981         neg_*_op* variables.
9983 2018-07-04  Martin Liska  <mliska@suse.cz>
9985         * tree-switch-conversion.c: Define
9986         max_ratio_for_speed and max_ratio_for_size constants.
9988 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9989             Martin Liska  <mliska@suse.cz>
9991         PR middle-end/66240
9992         PR target/45996
9993         PR c/84100
9994         * common.opt: Rename align options with 'str_' prefix.
9995         * common/config/i386/i386-common.c (set_malign_value): New
9996         function.
9997         (ix86_handle_option): Use it to set -falign-* options/
9998         * config/aarch64/aarch64-protos.h (struct tune_params): Change
9999         type from int to string.
10000         * config/aarch64/aarch64.c: Update default values from int
10001         to string.
10002         * config/alpha/alpha.c (alpha_override_options_after_change):
10003         Likewise.
10004         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
10005         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10006         max skip conditionally.
10007         * config/i386/freebsd.h (SUBALIGN_LOG): New.
10008         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10009         max skip conditionally.
10010         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10011         max skip conditionally.
10012         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
10013         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10014         max skip conditionally.
10015         * config/i386/i386.c (struct ptt): Change type from int to
10016         string.
10017         (ix86_default_align): Set default values.
10018         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
10019         max skip conditionally.
10020         * config/i386/iamcu.h (SUBALIGN_LOG): New.
10021         (ASM_OUTPUT_MAX_SKIP_ALIGN):
10022         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
10023         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10024         max skip conditionally.
10025         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
10026         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
10027         * config/i386/x86-64.h (SUBALIGN_LOG): New.
10028         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10029         max skip conditionally.
10030         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10031         * config/ia64/ia64.c (ia64_option_override): Set default values
10032         for alignment options.
10033         * config/m68k/m68k.c: Handle new str_align_* options.
10034         * config/mips/mips.c (mips_set_compression_mode): Change
10035         type of constants.
10036         (mips_option_override): Set default values for options.
10037         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
10038         Likewise.
10039         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10040         Likewise.
10041         * config/rx/rx.c (rx_option_override): Likewise.
10042         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
10043         (LABEL_ALIGN): Use align_labels_log.
10044         (LOOP_ALIGN): Use align_loops_align.
10045         * config/s390/s390.c (s390_asm_output_function_label): Use new
10046         macros.
10047         * config/sh/sh.c (sh_override_options_after_change):
10048         Change type of constants.
10049         * config/spu/spu.c (spu_sched_init): Likewise.
10050         * config/sparc/sparc.c (sparc_option_override): Set default
10051         values for options.
10052         * config/visium/visium.c (visium_option_override): Likewise.
10053         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
10054         emit p2align format with last argument if it's not needed.
10055         * doc/invoke.texi: Document extended format of -falign-*.
10056         * final.c: Use align_labels alignment.
10057         * flags.h (struct target_flag_state): Change type to use
10058         align_flags.
10059         (struct align_flags_tuple): New.
10060         (struct align_flags): Likewise.
10061         (align_loops_log): Redefine macro to use new types.
10062         (align_loops_max_skip): Redefine macro to use new types.
10063         (align_jumps_log): Redefine macro to use new types.
10064         (align_jumps_max_skip): Redefine macro to use new types.
10065         (align_labels_log): Redefine macro to use new types.
10066         (align_labels_max_skip): Redefine macro to use new types.
10067         (align_functions_log): Redefine macro to use new types.
10068         (align_loops): Redefine macro to use new types.
10069         (align_jumps): Redefine macro to use new types.
10070         (align_labels): Redefine macro to use new types.
10071         (align_functions): Redefine macro to use new types.
10072         (align_functions_max_skip): Redefine macro to use new types.
10073         (align_loops_value): New macro.
10074         (align_jumps_value): New macro.
10075         (align_labels_value): New macro.
10076         (align_functions_value): New macro.
10077         * function.c (invoke_set_current_function_hook): Propagate
10078         alignment values from flags to global variables default in
10079         topleev.h.
10080         * ipa-icf.c (sem_function::equals_wpa): Use
10081         cl_optimization_option_eq instead of memcmp.
10082         * lto-streamer.h (cl_optimization_stream_out): Support streaming
10083         of string types.
10084         (cl_optimization_stream_in): Likewise.
10085         * optc-save-gen.awk: Support strings in cl_optimization.
10086         * opth-gen.awk: Likewise.
10087         * opts.c (finish_options): Remove error checking of invalid
10088         value ranges.
10089         (MAX_CODE_ALIGN): Remove.
10090         (MAX_CODE_ALIGN_VALUE): Likewise.
10091         (parse_and_check_align_values): New function.
10092         (check_alignment_argument): Likewise.
10093         (common_handle_option): Use check_alignment_argument.
10094         * opts.h (parse_and_check_align_values): Declare.
10095         * toplev.c (init_alignments): Remove.
10096         (read_log_maxskip): New.
10097         (parse_N_M): Likewise.
10098         (parse_alignment_opts): Likewise.
10099         (backend_init_target): Remove usage of init_alignments.
10100         * toplev.h (parse_alignment_opts): Declare.
10101         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
10102         argument.
10103         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
10104         * tree.c (cl_option_hasher::equal): New.
10105         * varasm.c: Use new global macros.
10107 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10109         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
10110         Use a simpler align directive also if MAXSKIP = ALIGN-1.
10111         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10112         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10113         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10114         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10115         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10116         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
10117         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10118         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10119         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10120         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10121         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10123 2018-07-04  Martin Liska  <mliska@suse.cz>
10124             Jonathan Wakely  <jwakely@redhat.com>
10126         * coverage.c: Use correct type.
10127         * doc/invoke.texi: Language correction.
10129 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
10131         PR target/85620
10132         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10133         ENDBRANCH for non-tail call which may return via indirect branch.
10134         * doc/extend.texi: Document indirect_return attribute.
10136 2018-07-03  Martin Sebor  <msebor@redhat.com>
10138         PR tree-optimization/86274
10139         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10140         precondition.
10141         (format_floating): Correct handling of infinities and NaNs.
10143 2018-07-03  Martin Sebor  <msebor@redhat.com>
10145         * print-tree.c (print_real_cst): New function.
10146         (print_node_brief): Call it.
10147         (print_node): Ditto.
10149 2018-07-03  Jeff Law  <law@redhat.com>
10151         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10152         into a single pattern.
10154         * config/h8300/h8300.md (ors code_iterator): New.
10155         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10156         a single pattern and single splitter.
10157         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10158         (iorqi3_1, xorqi3_1): Likewise.
10159         (iorqi3, xorqi3 expanders): Similarly.
10161         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10162         (movmd_internal) into a single pattern using the P mode iterator.
10163         (movmd splitters): Similarly.
10164         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
10165         (movsd splitters): Similarly.
10167         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
10168         ADDB, ADDW and ADDL into a single ADD attribute which selects the
10169         right table based on the size of the operand.
10170         * config/h8300/h8300.md (length_table): Corresponding changes. All
10171         references to "addb", "addw" and "addl" changed to "add".
10172         (btst patterns): Merge two variants into a single pattern.
10173         (tstqi, tsthi): Likewise.
10174         (addhi3_incdec, addsi3_incdec): Likewise.
10175         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
10176         (mulhi3, mulsi3): Likewise.
10177         (udivhi3, udivsi3): Likewise.
10178         (divhi3, divsi3): Likewise.
10179         (andorqi3, andorhi3, andorsi3): Likewise.
10181 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
10183         PR target/85694
10184         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10185         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10187 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10189         PR tree-optimization/85694
10190         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10191         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10192         UNSPEC_URHADD.
10193         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10194         (<u>avg<mode>3_ceil): New patterns.
10196 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
10198         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10199         scan-tree-dump directive.
10200         * gcc.dg/vect/slp-perm-2.c: Likewise.
10201         * gcc.dg/vect/slp-perm-3.c: Likewise.
10202         * gcc.dg/vect/slp-perm-5.c: Likewise.
10203         * gcc.dg/vect/slp-perm-6.c: Likewise.
10204         * gcc.dg/vect/slp-perm-7.c: Likewise.
10205         * gcc.dg/vect/slp-perm-8.c: Likewise.
10207 2018-07-03  Marek Polacek  <polacek@redhat.com>
10209         PR middle-end/86202
10210         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10212 2018-07-03  Richard Biener  <rguenther@suse.de>
10214         PR ipa/86389
10215         * tree-ssa-structalias.c (find_func_clobbers): Properly
10216         handle indirect calls.
10218 2018-07-03  Jeff Law  <law@redhat.com>
10220         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10221         (shifts): New code iterator.
10222         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
10223         expander.  Fix HImode handling on H8/SX.
10224         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
10225         (subqi3, subhi3, subsi3 expanders): Likewise.
10226         (andqi3, andhi3, andsi3 expanders): Likewise.
10227         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
10228         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
10229         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
10230         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
10231         (zero_extendqihi2, zero_extendqisi2): Likewise.
10232         (extendqihi2, extendqisi2): Likewise.
10233         (rotlqi3, rotlhi3, rotlsi3): Likewise.
10234         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
10235         (rotlqi3_1, rotlhi3_1): Likewise.
10236         (logicalhi3_sn, logicalsi3_sn): Likewise.
10237         (logicalhi3, logicalsi3): Likewise.
10239 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10241         * tree-vect-patterns.c (vect_recog_rotate_pattern)
10242         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10243         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
10244         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
10245         type to append_pattern_def_seq instead of creating a stmt_vec_info
10246         directly.
10247         (build_mask_conversion): Likewise.  Remove vinfo argument.
10248         (vect_add_conversion_to_patterm): Likewise, renaming to...
10249         (vect_add_conversion_to_pattern): ...this.
10250         (vect_recog_mask_conversion_pattern): Update call to
10251         build_mask_conversion.  Pass the vector type to
10252         append_pattern_def_seq here too.
10253         (vect_recog_gather_scatter_pattern): Update call to
10254         vect_add_conversion_to_pattern.
10256 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10258         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
10259         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10260         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
10261         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
10262         STMT_VINFO_PATTERN_DEF_SEQ to null here.
10263         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
10264         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10265         append_pattern_def_seq instead of new_pattern_def_seq.
10266         (vect_recog_divmod_pattern): Do both of the above.
10267         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
10268         is null.
10270 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10272         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10273         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
10274         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
10275         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
10276         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
10277         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10278         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
10279         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10280         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10281         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
10282         parameter with a single stmt_vec_info.
10283         (vect_recog_func_ptr): Likewise.
10284         (vect_recog_gather_scatter_pattern): Likewise, folding in...
10285         (vect_try_gather_scatter_pattern): ...this.
10286         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
10287         the stmt_vec_info of the statement to be matched.  Don't clear
10288         STMT_VINFO_RELATED_STMT.
10289         (vect_pattern_recog): Update call accordingly.
10291 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10293         PR tree-optimization/85694
10294         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
10295         (uavgM3_ceil): Document new optabs.
10296         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
10297         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
10298         functions.
10299         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
10300         (savg_ceil_optab): New optabs.
10301         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
10302         (vect_vect_recog_func_ptrs): Add it.
10303         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
10304         constant directly from the associated lhs.
10306 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10308         * tree-vect-patterns.c (vect_split_statement): New function.
10309         (vect_convert_input): Use it to try to split an existing cast.
10311 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10313         * poly-int.h (print_hex): New function.
10314         * dumpfile.h (dump_dec, dump_hex): Declare.
10315         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
10316         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
10317         min_input_precision, operation_precision and operation_sign.
10318         * tree-vect-patterns.c (vect_get_range_info): New function.
10319         (vect_same_loop_or_bb_p, vect_single_imm_use)
10320         (vect_operation_fits_smaller_type): Delete.
10321         (vect_look_through_possible_promotion): Add an optional
10322         single_use_p parameter.
10323         (vect_recog_over_widening_pattern): Rewrite to use new
10324         stmt_vec_info infomration.  Handle one operation at a time.
10325         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
10326         (vect_truncatable_operation_p, vect_set_operation_type)
10327         (vect_set_min_input_precision): New functions.
10328         (vect_determine_min_output_precision_1): Likewise.
10329         (vect_determine_min_output_precision): Likewise.
10330         (vect_determine_precisions_from_range): Likewise.
10331         (vect_determine_precisions_from_users): Likewise.
10332         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
10333         (vect_vect_recog_func_ptrs): Put over_widening first.
10334         Add cast_forwprop.
10335         (vect_pattern_recog): Call vect_determine_precisions.
10337 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10339         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
10340         statements that have been replaced by further pattern statements.
10341         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
10343 2018-07-03  Richard Biener  <rguenther@suse.de>
10345         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10346         always set *dt.  Dump vectype in vectype overload.
10347         * dumpfile.h (dump_gimple_expr): New function.
10348         (dump_gimple_expr_loc): Likewise.
10349         * dumpfile.c (dump_gimple_expr): New function.
10350         (dump_gimple_expr_loc): Likewise.
10352 2018-07-02  Jeff Law  <law@redhat.com>
10354         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
10355         the H8/300, H8/300H and H8/S variants into a single pattern.
10356         (movhi_h8300, movqi_h8300hs): Similarly.
10357         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
10358         (QHI mode iterator): New.
10360         * config/h8300/h8300.md: Remove trailing whitespace.
10362 2018-07-02  Jim Wilson  <jimw@sifive.com>
10364         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
10365         instead of emit_insn for interrupt returns.
10366         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
10367         (riscv_sret, riscv_uret): Likewise.
10369 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10371         * pretty-print.c (selftest::test_pp_format): Move save and restore
10372         of quotes to class auto_fix_quotes, and add an instance.
10373         * selftest.c: Include "intl.h".
10374         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
10375         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
10376         * selftest.h (selftest::auto_fix_quotes): New class.
10378 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
10380         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
10381         (aarch64_sve_prepare_conditional_op): Remove.
10382         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
10383         Allow aarch64_simd_reg_or_zero as select operand; remove
10384         the aarch64_sve_prepare_conditional_op call.
10385         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
10386         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10387         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
10388         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
10389         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
10390         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
10391         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
10392         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
10393         and a splitters to match all of the *_any patterns.
10394         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
10396         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
10397         (SVE_COND_FP_BINARY_REV): Remove.
10398         (sve_int_op_rev, sve_fp_op_rev): New.
10399         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
10400         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
10401         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
10402         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
10403         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
10404         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
10405         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
10406         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
10407         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
10409         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
10410         Remove match_dup 1 from the inner unspec.
10411         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10413         * config/aarch64/aarch64.md (movprfx): New attr.
10414         (length): Default movprfx to 8.
10415         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
10416         (*madd<SVE_I>, *msub<SVE_I): Likewise.
10417         (*<su>mul<SVE_I>3_highpart): Likewise.
10418         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
10419         (*v<ASHIFT><SVE_I>3): Likewise.
10420         (*<su><MAXMIN><SVE_I>3): Likewise.
10421         (*<su><MAXMIN><SVE_F>3): Likewise.
10422         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
10423         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
10424         (*div<SVE_F>4): Likewise.
10426 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
10428         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
10429         in dump string.
10431 2018-07-02  Richard Biener  <rguenther@suse.de>
10433         PR tree-optimization/86363
10434         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
10435         memset argument refers to a non-variable address.
10437 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
10439         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
10440         {PLUS,MINUS}_EXPR code to...
10441         (adjust_symbolic_bound): ...here,
10442         (combine_bound): ...here,
10443         (set_value_range_with_overflow): ...and here.
10445 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
10447         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
10448         code...
10449         (extract_range_from_abs_expr): ...here.
10451 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
10453         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
10454         -fno-omit-frame-pointer when not optimizing.
10456 2018-07-02  Martin Liska  <mliska@suse.cz>
10458         PR ipa/86279
10459         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
10460         (propagate_nothrow): Likewise.
10462 2018-07-02  Martin Liska  <mliska@suse.cz>
10464         PR ipa/86323
10465         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
10467 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10469         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
10470         function in r262149, changing "loc" param from source_location to
10471         const dump_location_t &.
10472         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
10473         declaration, as above.
10475 2018-07-01  Paul Koning  <ni1d@arrl.net>
10477         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10478         -munit-asm, -mgnu-asm, -mdec-asm.
10479         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
10480         (pdp11_output_labelref): New.
10481         (pdp11_output_def): New.
10482         (pdp11_output_addr_vec_elt): New.
10483         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
10484         %# and %@ format codes.
10485         (pdp11_option_override): New.
10486         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
10487         (pdp11_output_ident): New.
10488         (pdp11_asm_named_section): New.
10489         (pdp11_asm_init_sections): New.
10490         (pdp11_file_start): New.
10491         (pdp11_file_end): New.
10492         (output_ascii): Use .ascii/.asciz for -mdec-asm.
10493         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
10494         %o, like %c but octal.
10495         (pdp11_option_override): New.
10496         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
10497         -mdec-asm.
10498         (DATA_SECTION_ASM_OP): Ditto.
10499         (READONLY_DATA_SECTION_ASM_OP): New.
10500         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
10501         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
10502         (ASM_OUTPUT_LABELREF): Ditto.
10503         (ASM_OUTPUT_DEF): Ditto.
10504         (ASM_OUTPUT_EXTERNAL): New.
10505         (ASM_OUTPUT_SOURCE_FILENAME): New.
10506         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
10507         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
10508         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
10509         %# and %@ format codes.
10510         * config/pdp11/pdp11.opt (mgnu-asm): New.
10511         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
10512         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
10513         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
10515 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
10517         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
10518         dereferencing path[] beyond its length.
10519         (debug_path): New.
10520         (debug_all_paths): New.
10521         (rewire_first_differing_edge): New.
10522         (adjust_paths_after_duplication): New.
10523         (duplicate_thread_path): Call adjust_paths_after_duplication.
10524         Add new argument.
10525         (thread_through_all_blocks): Add new argument to
10526         duplicate_thread_path.
10528 2018-06-30  Jim Wilson  <jimw@sifive.com>
10530         * config/riscv/predicates.md (p2m1_shift_operand): New.
10531         (high_mask_shift_operand): New.
10532         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
10533         pattern using p2m1_shift_operand.
10534         (lshsi3_zero_extend_3+2): New combiner pattern using
10535         high_mask_shift_operand.
10537 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10539         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
10540         split out from...
10541         (vect_recog_rotate_pattern): ...here.
10542         (vect_convert_input): Try to insert casts of invariants in the
10543         preheader.
10544         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
10545         preheader to be empty.
10547 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10549         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
10550         vector type.  If given, install it in the new statement's
10551         STMT_VINFO_VECTYPE.
10552         (vect_element_precision): New function.
10553         (vect_unpromoted_value): New struct.
10554         (vect_unpromoted_value::vect_unpromoted_value): New function.
10555         (vect_unpromoted_value::set_op): Likewise.
10556         (vect_look_through_possible_promotion): Likewise.
10557         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
10558         (vect_widened_op_tree, vect_convert_input): Likewise.
10559         (vect_convert_inputs, vect_convert_output): Likewise.
10560         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
10561         to handle the optional cast of the multiplication result and
10562         vect_widened_op_tree to detect the widened multiplication itself.
10563         Do not require the input and output of promotion casts to have
10564         the same sign, but base the signedness of the operation on the
10565         input rather than the result.  If the pattern includes two
10566         promotions, check that those promotions have the same sign.
10567         Do not restrict the MULT_EXPR handling to a double-width result;
10568         handle quadruple-width results and wider.  Use vect_convert_inputs
10569         to convert the inputs to the common type.
10570         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
10571         to handle the optional cast of the ABS result.  Also allow a sign
10572         change or a sign extension between the ABS and MINUS.
10573         Use vect_widened_op_tree to detect the widened subtraction and use
10574         vect_convert_inputs to convert the inputs to the common type.
10575         (vect_handle_widen_op_by_const): Delete.
10576         (vect_recog_widen_op_pattern): New function.
10577         (vect_recog_widen_mult_pattern): Use it.
10578         (vect_recog_widen_shift_pattern): Likewise.
10579         (vect_recog_widen_sum_pattern): Use
10580         vect_look_through_possible_promotion to handle the promoted
10581         PLUS_EXPR operand.
10583 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10585         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
10586         the containing gimple_seq *.
10587         * gimple-iterator.h (gsi_for_stmt): Declare it.
10588         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
10589         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
10590         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10591         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10592         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
10593         checks.
10594         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
10595         split out from...
10596         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
10597         statement being replaced is part of an existing pattern
10598         definition sequence, inserting the new pattern statements before
10599         the original one.
10600         (vect_pattern_recog_1): Don't return a bool.  If the statement
10601         is already part of a pattern, instead apply pattern matching
10602         to the pattern definition statements.  Don't clear the
10603         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
10604         (vect_pattern_recog): Don't break after the first match;
10605         continue processing the pattern definition statements instead.
10606         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
10608 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10610         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10611         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10612         (vect_recog_widen_sum_pattern): Use it.
10614 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10616         * tree-vect-loop.c (vectorizable_reduction): Assert that the
10617         phi is not a pattern statement and has not been replaced by
10618         a pattern statement.
10619         * tree-vect-patterns.c (type_conversion_p): Don't check
10620         STMT_VINFO_IN_PATTERN_P.
10621         (vect_recog_vector_vector_shift_pattern): Likewise.
10622         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
10623         the pattern statement rather than the original statement; check
10624         directly for a WIDEN_MULT_EXPR here.
10625         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
10626         vect_is_simple_use to return the pattern statement rather
10627         than the original statement; use is_pattern_stmt_p to check
10628         for such a pattern statement.
10629         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
10630         to return the pattern statement rather than the original statement;
10631         don't do the same transformation here.
10632         (vect_is_simple_use): If the defining statement has been replaced
10633         by a pattern statement, return the pattern statement instead.
10634         Remove the corresponding (local) transformation from the vectype
10635         overload.
10637 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10639         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
10640         end and default to null.
10641         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10642         (vectorizable_reduction): Update calls accordingly, dropping the
10643         gimple ** argument if the passed-back statement isn't needed.
10644         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
10645         (vect_recog_rotate_pattern): Likewise.
10646         (vect_recog_mask_conversion_pattern): Likewise.
10647         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10648         (vect_mask_constant_operand_p): Likewise.
10649         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
10650         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
10651         (get_group_load_store_type, get_load_store_type): Likewise.
10652         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
10653         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
10654         (vectorizable_conversion, vectorizable_assignment): Likewise.
10655         (vectorizable_shift, vectorizable_operation): Likewise.
10656         (vectorizable_store, vect_is_simple_cond): Likewise.
10657         (vectorizable_condition, vectorizable_comparison): Likewise.
10658         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
10659         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
10660         and move it to the end.  Cope with null def_stmt_outs.
10662 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10664         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
10666 2018-06-29  Jeff Law  <law@redhat.com>
10668         * config/v850/v850.c (v850_legitimate_address_p): Handle large
10669         displacements for TARGET_V850E2V3 and newer.
10670         (TARGET_LRA_P): Remove.  Defaults to LRA now.
10671         * config/v850/v850.md (sign23byte_load): Remove.
10672         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
10673         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
10675 2018-06-29  Martin Liska  <mliska@suse.cz>
10677         PR lto/85759
10678         * coverage.c (coverage_init): Mangle full path name.
10679         * doc/invoke.texi: Document the change.
10680         * gcov-io.c (mangle_path): New.
10681         * gcov-io.h (mangle_path): Likewise.
10682         * gcov.c (mangle_name): Use mangle_path for path mangling.
10684 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10686         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
10687         if starting source register is not even.
10689 2018-06-29  Martin Liska  <mliska@suse.cz>
10691         PR tree-optimization/86263
10692         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
10693         Make edge redirection.
10695 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
10697         * dumpfile.c (dump_loc): Add indentation based on scope depth.
10698         (dump_scope_depth): New variable.
10699         (get_dump_scope_depth): New function.
10700         (dump_begin_scope): New function.
10701         (dump_end_scope): New function.
10702         * dumpfile.h (get_dump_scope_depth): New declaration.
10703         (dump_begin_scope): New declaration.
10704         (dump_end_scope): New declaration.
10705         (class auto_dump_scope): New class.
10706         (AUTO_DUMP_SCOPE): New macro.
10707         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
10708         AUTO_DUMP_SCOPE.
10710 2018-06-29  Richard Biener  <rguenther@suse.de>
10712         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
10713         compute_all_dependences succeeds.
10714         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
10715         exceed --param loop-max-datarefs-for-datadeps.
10717 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
10719         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
10721 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
10723         PR target/86348
10724         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
10725         alternative 0 in preferred_for_speed attribute.
10727 2018-06-28  Paul Koning  <ni1d@arrl.net>
10729         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
10730         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
10731         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
10732         * config/pdp11/pdp11.md: Correct "length" attribute calculation
10733         for shift insn patterns.
10735 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
10737         * cgraph.c (cgraph_node::get_body): Replace assignments to
10738         "dump_file" with calls to set_dump_file.
10739         * dumpfile.c (alt_dump_file): Make static, and group with...
10740         (alt_flags): ...this definition.
10741         (dumps_are_enabled): New variable.
10742         (refresh_dumps_are_enabled): New function.
10743         (set_dump_file): New function.
10744         (set_alt_dump_file): New function.
10745         (gcc::dump_manager::dump_start): Replace assignments to
10746         "dump_file" and "alt_dump_file" with calls to set_dump_file and
10747         set_alt_dump_file.
10748         (gcc::dump_manager::dump_finish): Likewise.
10749         * dumpfile.h (alt_dump_file): Delete decl.
10750         (dumps_are_enabled): New variable decl.
10751         (set_dump_file): New function decl.
10752         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
10753         global.
10754         * tree-nested.c (lower_nested_functions): Replace assignments to
10755         "dump_file" with calls to set_dump_file.
10757 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
10759         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
10760         goto_locus of each outgoing edge of each basic block.
10762 2018-06-28  Richard Biener  <rguenther@suse.de>
10764         * dwarf2out.c (decl_scope_table): Remove.
10765         (push_decl_scope): Likewise.
10766         (pop_decl_scope): Likewise.
10767         (gen_type_die_for_member): Do not call push/pop_decl_scope.
10768         (gen_struct_or_union_type_die): Likewise.
10769         (gen_tagged_type_die): Likewise.
10770         (dwarf2out_init): Do not initialize decl_scope_table.
10771         (dwarf2out_c_finalize): Do not free it.
10773 2018-06-28  Richard Biener  <rguenther@suse.de>
10775         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
10776         deciding whether to not re-use a DIE.
10778 2018-06-28  Richard Biener  <rguenther@suse.de>
10780         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
10781         DW_AT_abstract_origin attribute.
10783 2018-06-28  Martin Liska  <mliska@suse.cz>
10785         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10786         Use newly introduced constants.
10787         * tree-switch-conversion.h (struct jump_table_cluster):
10788         Define max_ratio_for_size and max_ratio_for_speed.
10790 2018-06-28  Martin Liska  <mliska@suse.cz>
10792         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10793         Add new checking assert to catch invalid state.
10794         (jump_table_cluster::can_be_handled): Handle single case
10795         clusters.
10796         (jump_table_cluster::is_beneficial): Bail out for such case.
10797         (bit_test_cluster::find_bit_tests):
10798         Add new checking assert to catch invalid state.
10799         (bit_test_cluster::can_be_handled): Handle single case
10800         clusters.
10801         (bit_test_cluster::is_beneficial): Bail out for such case.
10802         (switch_decision_tree::analyze_switch_statement):
10803         Fix comment.
10805 2018-06-28  Martin Liska  <mliska@suse.cz>
10807         * common.opt: Introduce -completion option.
10808         * gcc.c (driver_handle_option): Handle it.
10809         (driver::main): Print completions if completion
10810         is set.
10811         * opt-suggestions.c (option_proposer::get_completions):
10812         New function.
10813         (option_proposer::suggest_completion): Likewise.
10814         (option_proposer::find_param_completions): Likewise.
10815         (verify_autocompletions): Likewise.
10816         (test_completion_valid_options): Likewise.
10817         (test_completion_valid_params): Likewise.
10818         (in_completion_p): Likewise.
10819         (empty_completion_p): Likewise.
10820         (test_completion_partial_match): Likewise.
10821         (test_completion_garbage): Likewise.
10822         (opt_proposer_c_tests): Likewise.
10823         * opt-suggestions.h: Declare new functions.
10824         * opts.c (common_handle_option): Handle OPT__completion_.
10825         * selftest-run-tests.c (selftest::run_tests): Add
10826         opt_proposer_c_tests.
10827         * selftest.c (assert_str_startswith): New.
10828         * selftest.h (assert_str_startswith): Likewise.
10829         (opt_proposer_c_tests): New.
10830         (ASSERT_STR_STARTSWITH): Likewise.
10832 2018-06-28  Martin Liska  <mliska@suse.cz>
10834         * Makefile.in: Add opt-suggestions.o.
10835         * gcc-main.c: Include opt-suggestions.h.
10836         * gcc.c (driver::driver): Likewise.
10837         (driver::~driver): Remove m_option_suggestions.
10838         (driver::build_option_suggestions): Moved to option_proposer.
10839         (driver::suggest_option): Likewise.
10840         (driver::handle_unrecognized_options): Use option_proposer.
10841         * gcc.h (class driver): Add new memver m_option_proposer.
10842         * opt-suggestions.c: New file.
10843         * opt-suggestions.h: New file.
10845 2018-06-28  Martin Liska  <mliska@suse.cz>
10847         * vec.h (class auto_string_vec): New (moved from auto_argvec).
10848         (auto_string_vec::~auto_string_vec): Likewise.
10850 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
10852         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
10853         prevent_decl_creation_for_types fields up and add reset_location field.
10854         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
10855         statement if id->reset_location is true.
10856         (copy_edges_for_bb): Do not set goto_locus on the new edges if
10857         id->reset_location is true.
10858         (copy_phis_for_bb): Force input_location on the arguments if
10859         id->reset_location is true.
10860         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10861         is set on the function to be inlined.
10863 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
10865         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
10867 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
10869         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
10870         registers for Pmode.
10871         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
10872         hard registers for the clobbered pseudo.
10874 2018-06-27  Paul Koning  <ni1d@arrl.net>
10876         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10877         mutually exclusive options.
10878         * config/pdp11/constraints.md (h): New constraint.
10879         (O): Update definition to match shift code generation.
10880         (D): New constraint.
10881         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
10882         (CCFP): Remove.
10883         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
10884         function.
10885         (output_jump): Change arguments.
10886         (pdp11_fixed_cc_regs): New function.
10887         (pdp11_cc_mode): Ditto.
10888         (pdp11_expand_shift): Ditto.
10889         (pdp11_assemble_shift): Ditto.
10890         (pdp11_small_shift): Ditto.
10891         (pdp11_branch_cost): Remove.
10892         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
10893         from output.
10894         (pdp11_register_move_cost): Update for CC registers.
10895         (pdp11_rtx_costs): Add case for LSHIFTRT.
10896         (pdp11_output_jump): Add CCNZ mode conditional branches.
10897         (notice_update_cc_on_set): Remove.
10898         (pdp11_cc_mode): New function.
10899         (simple_memory_operand): Correct pre/post decrement case.
10900         (no_side_effect_operand): New function.
10901         (pdp11_regno_reg_class): Add CC_REGS class.
10902         (pdp11_fixed_cc_regs): New function.
10903         (pdp11_small_shift): New function.
10904         (pdp11_expand_shift): New function to expand shift insns.
10905         (pdp11_assemble_shift): New function to output shifts.
10906         (pdp11_branch_cost): Remove.
10907         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
10908         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
10909         (WCHAR_TYPE): Ditto.
10910         (PTRDIFF_TYPE): Ditto.
10911         (ADJUST_INSN_LENGTH): New macro.
10912         (FIXED_REGISTERS): Add CC registers.
10913         (CALL_USED_REGISTERS): Ditto.
10914         (reg_class): Ditto.
10915         (REG_CLASS_NAMES): Ditto.
10916         (REG_CLASS_CONTENTS): Ditto.
10917         (SELECT_CC_MODE): Use new function.
10918         (TARGET_FLAGS_REGNUM): New macro.
10919         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
10920         (cc0_reg_rtx): Remove.
10921         (CC_STATUS_MDEP): Remove.
10922         (CC_STATUS_MDEFP_INIT): Remove.
10923         (CC_IN_FPU): Remove.
10924         (NOTICE_UPDATE_CC): Remove.
10925         (REGISTER_NAMES): Add CC registers.
10926         (BRANCH_COST): Change to constant 1.
10927         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10928         handling.
10929         * config/pdp11/pdp11.opt (mbcopy): Remove.
10930         (mbcopy-builtin): Remove.
10931         (mbranch-cheap): Remove.
10932         (mbranch-expensive): Remove.
10933         * config/pdp11/predicates.md (expand_shift_operand): Update to
10934         match shift code generation.
10935         (ccnz_operator): New predicate.
10936         * doc/invoke.texi (PDP-11 Options): Remove deleted options
10937         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10938         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
10939         exclusive options.
10940         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
10941         description of O constraint.
10943 2018-06-27  Jeff Law  <law@redhat.com>
10944             Austin Law  <austinklaw@gmail.com>
10946         * config/v850/v850.md (addsi3_set_flags): New pattern.
10947         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10948         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10949         (zero_extendhisi2_v850_set_flags): Likewise.
10950         (zero_extendqisi2_v850_set_flags): Likewise.
10951         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10952         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10953         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10955         * config/v850/v850-protos.h (notice_update_cc): Remove.
10956         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10957         (v850_print_operand): Handle 'D' and "d".
10958         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10959         Add handling of arithmetic/logical operations compared against zero.
10960         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10961         Do not look at v850_compare_op, instead get mode from last argument.
10962         (v850_gen_compare): Remove
10963         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10964         after reload for prologue insns.
10965         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10966         patterns.
10967         (construct_save_jarl): Likewise.
10968         (TARGET_FLAGS_REGNUM): Define.
10969         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10970         (NOTICE_UPDATE_CC): Remove.
10971         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10972         than cc0.  Conditionalize on reload_completed.
10973         (cmpsi_insn, setfcc_insn): Likewise.
10974         (tst1 splitter): Turn into define_and_split which sets the flags
10975         after reload.
10976         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10977         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10978         (cstoresf4, cstoredf4): Clobber the flags.
10979         (cmpsi, cmpsf, cmpdf): Remove expanders.
10980         (setf_insn): Remove pattern.
10981         (addsi3): Turn into define_and_split which clobbers the flags after
10982         reload and a suitable pattern (addsi3_clobber_flags) for use after
10983         reload.
10984         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10985         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10986         (ashrsi3, ashrsi3_v850e2): Likewise.
10987         (bins): Clobber the flags.
10988         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10989         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10990         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10991         (call_value_internal_short, call_value_internal_long): Likewise.
10992         (callt_save_interrupt, callt_return_interrupt): Likewise.
10993         (save_interrupt, return_interrupt): Likewise.
10994         (callt_save_all_interrupt, save_all_interrupt): Likewise.
10995         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10996         (restore_all_interrupt, _restore_all_interrupt): Likewise.
10997         (All FP comparisons): Only allow after reload has completed.
10998         (trfsr): Likewise.
10999         (divh, divhu): Tweak output template.
11000         (branch_z_normal, branch_z_invert): Remove
11001         (branch_nz_normal, branch_nz_invert): Likewise.
11002         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
11004         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
11005         * config/v850/v850.c (notice_update_cc): Remove.
11006         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
11007         (CC_NO_CARRY): Likewise.
11008         (NOTICE_UPDATE_CC): Define to nothing.
11009         * config/v850/v850.md: Remove block comment on cc0 handling
11010         Remove "cc" attribute from all patterns.  Remove cc_status handling
11011         from all patterns.  Minor formatting fixes.
11013 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11015         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
11016         (cortex-a76.cortex-a55): Likewise.
11017         * config/aarch64/aarch64-tune.md: Regenerate.
11018         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
11019         cortex-a76.cortex-a55.
11021 2018-06-27  Jeff Law  <law@redhat.com>
11023         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
11024         (MULTILIB_DIRNAMES): Similarly.
11026 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
11028         * gimple.h (gimple_return_retbnd): Delete.
11029         (gimple_return_set_retbnd): Likewise.
11030         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11031         gimple_return_set_retbnd.
11032         * gimple-pretty-print.c (dump_gimple_return): Remove call to
11033         gimple_return_retbnd and adjust.
11034         * tree-inline.h (struct copy_body_data): Remove retbnd field.
11035         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
11036         Explicitly return NULL in a couple more cases.  Move assertion
11037         on debug statements and remove unreachable code.
11038         (reset_debug_binding): Do not test id->retbnd.
11039         (expand_call_inline): Do not set it.
11041 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11043         * configure.ac: Add --disable-gcov option.
11044         * configure: Regenerate.
11045         * Makefile.in: Honour @enable_gcov@.
11046         * doc/install.texi: Document --disable-gcov.
11048 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11050         * config/arm/arm-cpus.in (cortex-a76): New entry.
11051         (cortex-a76.cortex-a55): Likewise.
11052         * config/arm/arm-tables.opt: Regenerate.
11053         * config/arm/arm-tune.md: Likewise.
11054         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
11055         * doc/invoke.texi (ARM Options): Document cortex-a76 and
11056         cortex-a76.cortex-a55.
11058 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
11060         PR target/85769
11061         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
11063 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11065         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
11066         comment.
11067         (EPILOGUE_USES): Likewise.
11069 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
11071         * tree-inline.c (remap_location): New function extracted from...
11072         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
11073         (copy_phis_for_bb): ...here.  Call remap_location.
11074         (copy_cfg_body): Adjust call to copy_edges_for_bb.
11076 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
11078         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
11079         unaligned vsx for 16B memset.
11081 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
11083         PR target/86285
11084         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
11085         ieee128_float_type_node to long_double_type_node unless
11086         TARGET_LONG_DOUBLE_128 is set.
11088 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
11090         * cfgloop.c (get_loop_location): Convert return type from
11091         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
11092         by implicit construction from rtx_insn *, and using
11093         dump_user_location_t::from_function_decl for the fallback case.
11094         * cfgloop.h (get_loop_location): Convert return type from
11095         location_t to dump_user_location_t.
11096         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
11097         dump_printf_loc to pass in a dump_location_t rather than a
11098         location_t, via the gimple stmt.
11099         * coverage.c (get_coverage_counts): Update calls to
11100         dump_printf_loc to pass in dump_location_t rather than a
11101         location_t.
11102         * doc/optinfo.texi (Dump types): Convert example of
11103         dump_printf_loc from taking "locus" to taking "insn".  Update
11104         description of the "_loc" calls to cover dump_location_t.
11105         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
11106         "selftest.h".
11107         (dump_user_location_t::dump_user_location_t): New constructors,
11108         from gimple *stmt and rtx_insn *.
11109         (dump_user_location_t::from_function_decl): New function.
11110         (dump_loc): Make static.
11111         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
11112         const dump_location_t &.
11113         (dump_generic_expr_loc): Delete.
11114         (dump_printf_loc): Convert param "loc" from location_t to
11115         const dump_location_t &.
11116         (selftest::test_impl_location): New function.
11117         (selftest::dumpfile_c_tests): New function.
11118         * dumpfile.h: Include "profile-count.h".
11119         (class dump_user_location_t): New class.
11120         (struct dump_impl_location_t): New struct.
11121         (class dump_location_t): New class.
11122         (dump_printf_loc): Convert 2nd param from source_location to
11123         const dump_location_t &.
11124         (dump_generic_expr_loc): Delete.
11125         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11126         const dump_location_t &.
11127         * gimple-fold.c (fold_gimple_assign): Update call to
11128         dump_printf_loc to pass in a dump_location_t rather than a
11129         location_t, via the gimple stmt.
11130         (gimple_fold_call): Likewise.
11131         * gimple-loop-interchange.cc
11132         (loop_cand::analyze_iloop_reduction_var): Update for change to
11133         check_reduction_path.
11134         (tree_loop_interchange::interchange): Update for change to
11135         find_loop_location.
11136         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11137         change in return-type of find_loop_location.
11138         (graphite_regenerate_ast_isl): Likewise.
11139         * graphite-optimize-isl.c (optimize_isl): Likewise.
11140         * graphite.c (graphite_transform_loops): Likewise.
11141         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11142         pass in a dump_location_t rather than a location_t, via the
11143         gimple stmt.
11144         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11145         * ipa.c (walk_polymorphic_call_targets): Likewise.
11146         * loop-unroll.c (report_unroll): Convert "locus" param from
11147         location_t to dump_location_t.
11148         (decide_unrolling): Update for change to get_loop_location's
11149         return type.
11150         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11151         location_t to dump_user_location_t.
11152         (grid_find_single_omp_among_assignments_1): Updates calls to
11153         dump_printf_loc to pass in a dump_location_t rather than a
11154         location_t, via the gimple stmt.
11155         (grid_parallel_clauses_gridifiable): Convert "tloc" from
11156         location_t to dump_location_t.  Updates calls to dump_printf_loc
11157         to pass in a dump_location_t rather than a location_t, via the
11158         gimple stmt.
11159         (grid_inner_loop_gridifiable_p): Likewise.
11160         (grid_dist_follows_simple_pattern): Likewise.
11161         (grid_gfor_follows_tiling_pattern): Likewise.
11162         (grid_target_follows_gridifiable_pattern): Likewise.
11163         (grid_attempt_target_gridification): Convert initialization
11164         of local "grid" from memset to zero-initialization; FIXME: does
11165         this require C++11?  Update call to dump_printf_loc to pass in a
11166         optinfo_location rather than a location_t, via the gimple stmt.
11167         * profile.c (read_profile_edge_counts): Updates call to
11168         dump_printf_loc to pass in a dump_location_t rather than a
11169         location_t
11170         (compute_branch_probabilities): Likewise.
11171         * selftest-run-tests.c (selftest::run_tests): Call
11172         dumpfile_c_tests.
11173         * selftest.h (dumpfile_c_tests): New decl.
11174         * tree-loop-distribution.c (pass_loop_distribution::execute):
11175         Update for change in return type of find_loop_location.
11176         * tree-parloops.c (parallelize_loops): Likewise.
11177         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11178         "locus" from location_t to dump_user_location_t.
11179         (canonicalize_loop_induction_variables): Likewise.
11180         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11181         for change in return type of find_loop_location.
11182         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11183         to dump_printf_loc to pass in a dump_location_t rather than a
11184         location_t, via the stmt.
11185         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11186         Likewise.
11187         * tree-vect-loop-manip.c (find_loop_location): Convert return
11188         type from source_location to dump_user_location_t.
11189         (vect_do_peeling): Update for above change.
11190         (vect_loop_versioning): Update for change in type of
11191         vect_location.
11192         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11193         from location_t to dump_user_location_t.
11194         (vect_estimate_min_profitable_iters): Update for change in type
11195         of vect_location.
11196         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11197         location_t to dump_location_t.
11198         (vect_slp_bb): Update for change in type of vect_location.
11199         * tree-vectorizer.c (vect_location): Convert from source_location
11200         to dump_user_location_t.
11201         (try_vectorize_loop_1): Update for change in vect_location's type.
11202         (vectorize_loops): Likewise.
11203         (increase_alignment): Likewise.
11204         * tree-vectorizer.h (vect_location): Convert from source_location
11205         to dump_user_location_t.
11206         (find_loop_location): Convert return type from source_location to
11207         dump_user_location_t.
11208         (check_reduction_path): Convert 1st param from location_t to
11209         dump_user_location_t.
11210         * value-prof.c (check_counter): Update call to dump_printf_loc to
11211         pass in a dump_user_location_t rather than a location_t; update
11212         call to error_at for change in type of "locus".
11213         (check_ic_target): Update call to dump_printf_loc to
11214         pass in a dump_user_location_t rather than a location_t, via the
11215         call_stmt.
11217 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11219         * config/s390/s390.h (enum processor_flags): Do not use
11220         default tune parameter when -march was specified.
11222 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
11224         PR target/86314
11225         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
11226         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
11227         operands.
11229 2018-06-26  Richard Biener  <rguenther@suse.de>
11231         PR tree-optimization/86287
11232         PR bootstrap/86316
11233         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
11234         (vect_analyze_loop): Initialize n_stmts.
11236 2018-06-26  Richard Biener  <rguenther@suse.de>
11238         PR middle-end/86271
11239         * fold-const.c (fold_convertible_p): Pointer extension
11240         isn't valid.
11242 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
11244         PR debug/86064
11245         * dwarf2out.c (loc_list_has_views): Adjust comments.
11246         (dw_loc_list): Split single cross-partition range with
11247         nonzero locview.
11249 2018-06-25  Jeff Law  <law@redhat.com>
11251         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
11252         on -mbig-switch by default.
11254         * config/v850/predicates.md (const_float_1_operand): Fix match_code
11255         test.
11256         (const_float_0_operand): Remove unused predicate.
11257         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
11258         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
11259         (recipsf2): New expander.  Original pattern now called
11260         (recipsf2_insn).
11261         (recipdf2, recipdf2_insn): Similarly.
11262         (rsqrtsf2, rsqrtsf2_insn): Similarly
11263         (rsqrtdf2, rsqrtdf2_insn): Similarly
11265 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
11267         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
11268         Simplify logic for FreeBSD (twice).
11270 2018-06-25  Martin Sebor  <msebor@redhat.com>
11272         PR tree-optimization/86204
11273         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
11274         a strnlen result if it's less than the length of the string.
11276 2018-06-25  Martin Sebor  <msebor@redhat.com>
11278         PR tree-optimization/85700
11279         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
11280         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
11281         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
11283 2018-06-25  Martin Sebor  <msebor@redhat.com>
11285         * doc/extend.texi (Zero-length arrays): Update and clarify.
11287 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
11289         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
11290         added IEEE/IBM long double multilib support on PowerPC little
11291         endian Linux systems.
11292         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11293         (MULTILIB_DEFAULTS): Likewise.
11294         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11295         Likewise.
11296         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
11297         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
11298         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
11300 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
11302         PR middle-end/86311
11303         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
11304         (REORDER_45): Likewise.
11306 2018-06-25  Jeff Law  <law@redhat.com>
11308         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
11309         dividend to 32 bits.  Adjust length.
11310         (udivmodhi4): Cleanup output template.  Fix length.
11312 2018-06-25  Carl Love  <cel@us.ibm.com>
11314         * config/rs6000/vsx.md: Change word selector to prefered location.
11316 2018-06-25  Richard Biener  <rguenther@suse.de>
11318         PR tree-optimization/86304
11319         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
11320         epilogue-if-converted loops as well.
11322 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
11324         * lto-section-out.c (lto_begin_section): Do not print section
11325         name for noaddr and unnumbered dumps.
11327 2018-06-25  Richard Biener  <rguenther@suse.de>
11329         * tree-vectorizer.h (struct vec_info_shared): New structure
11330         with parts split out from struct vec_info and loop_nest from
11331         struct _loop_vec_info.
11332         (struct vec_info): Adjust accordingly.
11333         (struct _loop_vec_info): Likewise.
11334         (LOOP_VINFO_LOOP_NEST): Adjust.
11335         (LOOP_VINFO_DATAREFS): Likewise.
11336         (LOOP_VINFO_DDRS): Likewise.
11337         (struct _bb_vec_info): Likewise.
11338         (BB_VINFO_DATAREFS): Likewise.
11339         (BB_VINFO_DDRS): Likewise.
11340         (struct _stmt_vec_info): Add dr_aux member.
11341         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
11342         (DR_MISALIGNMENT_UNINITIALIZED): New.
11343         (set_dr_misalignment): Adjust.
11344         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
11345         (vect_analyze_loop): Adjust prototype.
11346         (vect_analyze_loop_form): Likewise.
11347         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11348         Compute dependences lazily.
11349         (vect_record_base_alignments): Use shared datarefs/ddrs.
11350         (vect_verify_datarefs_alignment): Likewise.
11351         (vect_analyze_data_refs_alignment): Likewise.
11352         (vect_analyze_data_ref_accesses): Likewise.
11353         (vect_analyze_data_refs): Likewise.
11354         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
11355         constructor parameter for shared part.
11356         (vect_analyze_loop_form): Pass in shared part and adjust.
11357         (vect_analyze_loop_2): Pass in storage for the number of
11358         stmts.  Move loop nest finding to the caller.  Compute
11359         datarefs lazily.
11360         (vect_analyze_loop): Pass in shared part.
11361         (vect_transform_loop): Verify shared datarefs are unchanged.
11362         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
11363         constructor parameter for shared part.
11364         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
11365         (vect_slp_bb): Verify shared datarefs are unchanged before
11366         transform.
11367         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
11368         change.
11369         (new_stmt_vec_info): Initialize DR_AUX misalignment to
11370         DR_MISALIGNMENT_UNINITIALIZED.
11371         * tree-vectorizer.c (vec_info::vec_info): Add constructor
11372         parameter for shared part.
11373         (vec_info::~vec_info): Adjust.
11374         (vec_info_shared::vec_info_shared): New.
11375         (vec_info_shared::~vec_info_shared): Likewise.
11376         (vec_info_shared::save_datarefs): Likewise.
11377         (vec_info_shared::check_datarefs): Likewise.
11378         (try_vectorize_loop_1): Construct shared part live for analyses
11379         of a single loop for multiple vector sizes.
11380         * tree-parloops.c (gather_scalar_reductions): Adjust.
11382 2018-06-25  Richard Biener  <rguenther@suse.de>
11384         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
11385         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
11386         (vect_analyze_data_refs): Remove similar code from here and
11387         simplify accordingly.
11389 2018-06-25  Richard Biener  <rguenther@suse.de>
11391         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
11392         for reverse storage order accesses rather than asserting
11393         they cannot happen here.
11395 2018-06-25  Tom de Vries  <tdevries@suse.de>
11397         PR debug/86257
11398         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
11399         Use data16 instead of .byte for insn prefix.
11401 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
11403         PR C++/86082
11404         * parser.c (make_char_string_pack): Pass this literal chars
11405         through cpp_interpret_string.
11406         (cp_parser_userdef_numeric_literal): Check the result of
11407         make_char_string_pack.
11409 2018-06-24  Maya Rashish  <coypu@sdf.org>
11411         * ginclude/stddef.h: Simplify conditions around avoiding
11412         re-definition of __size_t.
11414 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11416         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
11417         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
11419 2018-06-22  Maya Rashish  <coypu@sdf.org>
11421         * doc/invoke.texi (mno-fancy-math-387): Update for changes
11422         made to OpenBSD and NetBSD through the years.
11424 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11426         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11427         behavior of vec_pack (vector double, vector double) to match
11428         behavior of vec_float2 (vector double, vector double).
11430 2018-06-22  Olivier Hainque  <hainque@adacore.com>
11432         * gimplify.c (gimplify_function_tree): Prevent creation
11433         of a trampoline for the address of the current function
11434         passed to entry/exit instrumentation hooks.
11436 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
11438         PR target/86222
11439         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
11440         correctly.
11442 2018-06-22  Martin Liska  <mliska@suse.cz>
11444         PR tree-optimization/86263
11445         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11446         Bail out if is_enabled is false.
11447         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
11448         New declaration.
11449         (jump_table_cluster::is_enabled): New function.
11451 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11453         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11454         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
11455         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
11456         (lto_input_ts_binfo_tree_pointers): Likewise.
11457         * tree-streamer-out.c (streamer_write_tree_bitfields,
11458         write_ts_binfo_tree_pointers): Likewise.
11459         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
11461 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11463         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
11465 2018-06-22  Martin Liska  <mliska@suse.cz>
11467         * symbol-summary.h (get): Make it pure and inline move
11468         functionality from ::get function.
11469         (get): Remove and inline into ::get and ::get_create.
11470         (get_create): Move code from ::get function.
11472 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11474         PR target/85994
11475         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
11476         -x assembler-with-cpp.
11478 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11480         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
11481         _FILE_OFFSET_BITS=64 for C++.
11483 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
11485         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
11486         conversion insn that shows up when pr85657-3.c is compiled using
11487         IEEE 128-bit long double.
11488         (neg<mode>2_internal): Use the correct mode to check whether the
11489         mode is IBM extended.
11490         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
11491         multiply and divide external functions from being created more
11492         than once.
11494 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11496         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
11497         functions.
11498         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
11499         the edge can be forwarded.
11500         (cfg_layout_merge_blocks): Likewise.
11502 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11504         * except.c (finish_eh_generation): Commit edge insertions only after
11505         the EH edges have been redirected from post-landing to landing pads.
11507 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11509         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
11510         create_tmp_var_for to create the FRAME decl.
11511         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
11513 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11515         * tree-inline.c (copy_edges_for_bb): Minor tweak.
11516         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
11517         debug statement when resetting its value.
11518         (expand_call_inline): Copy the locus of the call onto the assignment
11519         of the return value, if any.  Use local variable in more cases.
11521 2018-06-21  Martin Liska  <mliska@suse.cz>
11523         * ipa-pure-const.c (propagate_nothrow): Use
11524         funct_state_summaries->get.
11525         (dump_malloc_lattice): Likewise.
11526         (propagate_malloc): Likewise.
11528 2018-06-21  Richard Biener  <rguenther@suse.de>
11530         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
11531         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
11532         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
11533         comment.
11534         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
11535         BLOCK_ABSTRACT_ORIGIN unconditionally.
11537 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
11539         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
11540         deleting it.
11541         * ipa-reference.c (ipa_reference_c_finalize): Delete
11542         ipa_ref_opt_sum_summaries and set it to NULL.
11544 2018-06-21  Tom de Vries  <tdevries@suse.de>
11546         PR tree-optimization/85859
11547         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
11548         test with comment from bb_no_side_effects_p.
11550 2018-06-21  Richard Biener  <rguenther@suse.de>
11552         PR tree-optimization/86232
11553         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
11554         max for constant niter.
11556 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11558         * config/aarch64/aarch64-simd.md
11559         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
11561 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11563         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
11564         Make opernads of the unspec commutative.
11566 2018-06-21  Richard Biener  <rguenther@suse.de>
11568         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
11569         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
11570         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
11571         (vect_analyze_data_ref_dependence): Re-order checks to deal with
11572         NULL DR_STEP.
11573         (vect_record_base_alignments): Do not record base alignment
11574         for gathers or scatters.
11575         (vect_compute_data_ref_alignment): Drop return value that is always
11576         true.  Bail out early for gathers or scatters.
11577         (vect_enhance_data_refs_alignment): Bail out early for gathers
11578         or scatters.
11579         (vect_find_same_alignment_drs): Likewise.
11580         (vect_analyze_data_refs_alignment): Remove dead code.
11581         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11582         (vect_analyze_data_refs): For possible gathers or scatters do
11583         not create an alternate DR, just check their possible validity
11584         and mark them.  Adjust DECL_NONALIASED handling to not rely
11585         on DR_BASE_ADDRESS.
11586         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
11587         update inits of gathers or scatters.
11588         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
11589         Also copy gather/scatter flag to pattern vinfo.
11591 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11593         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11594         behavior of vec_packsu (vector unsigned long long, vector unsigned
11595         long long) to match behavior of vec_packs with same signature.
11597 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
11598             Thomas Schwinge <thomas@codesourcery.com>
11599             Cesar Philippidis  <cesar@codesourcery.com>
11601         * gimplify.c (gimplify_scan_omp_clauses): Add support for
11602         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11603         (gimplify_adjust_omp_clauses): Likewise.
11604         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
11605         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
11606         (gimplify_omp_target_update): Update handling of acc update and
11607         enter/exit data.
11608         * omp-low.c (install_var_field): Remove unused parameter
11609         base_pointers_restrict.
11610         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
11611         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
11612         FINALIZE}
11613         (omp_target_base_pointers_restrict_p): Delete.
11614         (scan_omp_target): Update call to scan_sharing_clauses.
11615         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
11616         FINALIZE}.
11617         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
11618         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11619         (convert_local_omp_clauses): Likewise.
11620         * tree-pretty-print.c (dump_omp_clause): Likewise.
11621         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
11622         FINALIZE}.
11623         (omp_clause_code_name): Likewise.
11625 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
11627         PR debug/86194
11628         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
11629         be narrowed.
11631         PR tree-optimization/86231
11632         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
11633         anti-range don't overwrite *vr0min before using it to compute *vr0max.
11635 2018-06-20  Tom de Vries  <tdevries@suse.de>
11637         PR tree-optimization/86097
11638         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
11639         iv type if signedness of iv type is not the same as that of *nit.
11641 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
11643         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
11644         EDGE_EH edges, verify they are all EDGE_EH.
11646 2018-06-20  Maya Rashish  <coypu@sdf.org>
11648         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
11650 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11652         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
11653         * config/aarch64/aarch64.c (xgene1_tunings): Add
11654         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
11655         (aarch64_mode_valid_for_sched_fusion_p):
11656         Allow 16-byte modes.
11657         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
11658         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
11659         128-bit modes.
11660         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
11661         New pattern.
11662         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
11663         * config/aarch64/iterators.md (VQ2): New mode iterator.
11665 2018-06-20  Martin Liska  <mliska@suse.cz>
11667         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11668         Change default ratio from 10 to 8.
11670 2018-06-20  Martin Liska  <mliska@suse.cz>
11672         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11673         New.
11674         (bit_test_cluster::find_bit_tests): Likewise.
11675         (switch_decision_tree::analyze_switch_statement): Find clusters.
11676         * tree-switch-conversion.h (struct jump_table_cluster): Document
11677         hierarchy.
11679 2018-06-20  Martin Liska  <mliska@suse.cz>
11681         * tree-switch-conversion.c (switch_conversion::collect):
11682         Record m_uniq property.
11683         (switch_conversion::expand): Bail out for special conditions.
11684         (group_cluster::~group_cluster): New.
11685         (group_cluster::group_cluster): Likewise.
11686         (group_cluster::dump): Likewise.
11687         (jump_table_cluster::emit): New.
11688         (switch_decision_tree::fix_phi_operands_for_edges): New.
11689         (struct case_node): Remove struct.
11690         (jump_table_cluster::can_be_handled): New.
11691         (case_values_threshold): Moved to header.
11692         (reset_out_edges_aux): Likewise.
11693         (jump_table_cluster::is_beneficial): New.
11694         (bit_test_cluster::can_be_handled): Likewise.
11695         (add_case_node): Remove.
11696         (bit_test_cluster::is_beneficial): New.
11697         (case_bit_test::cmp): New.
11698         (bit_test_cluster::emit): New.
11699         (expand_switch_as_decision_tree_p): Remove.
11700         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
11701         (fix_phi_operands_for_edge): Likewise.
11702         (switch_decision_tree::analyze_switch_statement): New.
11703         (compute_cases_per_edge): Move ...
11704         (switch_decision_tree::compute_cases_per_edge): ... here.
11705         (try_switch_expansion): Likewise.
11706         (switch_decision_tree::try_switch_expansion): Likewise.
11707         (record_phi_operand_mapping): Likewise.
11708         (switch_decision_tree::record_phi_operand_mapping): Likewise.
11709         (emit_case_decision_tree): Likewise.
11710         (switch_decision_tree::emit): Likewise.
11711         (balance_case_nodes): Likewise.
11712         (switch_decision_tree::balance_case_nodes): Likewise.
11713         (dump_case_nodes): Likewise.
11714         (switch_decision_tree::dump_case_nodes): Likewise.
11715         (emit_jump): Likewise.
11716         (switch_decision_tree::emit_jump): Likewise.
11717         (emit_cmp_and_jump_insns): Likewise.
11718         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
11719         (emit_case_nodes): Likewise.
11720         (switch_decision_tree::emit_case_nodes): Likewise.
11721         (conditional_probability): Remove.
11722         * tree-switch-conversion.h (enum cluster_type): New.
11723         (PRINT_CASE): New.
11724         (struct cluster): Likewise.
11725         (cluster::cluster): Likewise.
11726         (struct simple_cluster): Likewise.
11727         (simple_cluster::simple_cluster): Likewise.
11728         (struct group_cluster): Likewise.
11729         (struct jump_table_cluster): Likewise.
11730         (struct bit_test_cluster): Likewise.
11731         (struct min_cluster_item): Likewise.
11732         (struct case_tree_node): Likewise.
11733         (case_tree_node::case_tree_node): Likewise.
11734         (jump_table_cluster::case_values_threshold): Likewise.
11735         (struct case_bit_test): Likewise.
11736         (struct switch_decision_tree): Likewise.
11737         (struct switch_conversion): Likewise.
11738         (switch_decision_tree::reset_out_edges_aux): Likewise.
11740 2018-06-20  Martin Liska  <mliska@suse.cz>
11742         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
11743         (hoist_edge_and_branch_if_true): Likewise.
11744         (expand_switch_using_bit_tests_p): Likewise.
11745         (struct case_bit_test): Likewise.
11746         (case_bit_test_cmp): Likewise.
11747         (emit_case_bit_tests): Likewise.
11748         (switch_conversion::switch_conversion): New class.
11749         (struct switch_conv_info): Remove old struct.
11750         (collect_switch_conv_info): More to ...
11751         (switch_conversion::collect): ... this.
11752         (check_range): Likewise.
11753         (switch_conversion::check_range): Likewise.
11754         (check_all_empty_except_final): Likewise.
11755         (switch_conversion::check_all_empty_except_final): Likewise.
11756         (check_final_bb): Likewise.
11757         (switch_conversion::check_final_bb): Likewise.
11758         (create_temp_arrays): Likewise.
11759         (switch_conversion::create_temp_arrays): Likewise.
11760         (free_temp_arrays): Likewise.
11761         (gather_default_values): Likewise.
11762         (switch_conversion::gather_default_values): Likewise.
11763         (build_constructors): Likewise.
11764         (switch_conversion::build_constructors): Likewise.
11765         (constructor_contains_same_values_p): Likewise.
11766         (switch_conversion::contains_same_values_p): Likewise.
11767         (array_value_type): Likewise.
11768         (switch_conversion::array_value_type): Likewise.
11769         (build_one_array): Likewise.
11770         (switch_conversion::build_one_array): Likewise.
11771         (build_arrays): Likewise.
11772         (switch_conversion::build_arrays): Likewise.
11773         (gen_def_assigns): Likewise.
11774         (switch_conversion::gen_def_assigns): Likewise.
11775         (prune_bbs): Likewise.
11776         (switch_conversion::prune_bbs): Likewise.
11777         (fix_phi_nodes): Likewise.
11778         (switch_conversion::fix_phi_nodes): Likewise.
11779         (gen_inbound_check): Likewise.
11780         (switch_conversion::gen_inbound_check): Likewise.
11781         (process_switch): Use the newly created class.
11782         (switch_conversion::expand): New.
11783         (switch_conversion::~switch_conversion): New.
11784         * tree-switch-conversion.h: New file.
11786 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11788         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
11789         tree-vect-patterns.c.
11790         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
11791         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
11792         (vect_recog_sad_pattern): Likewise.
11793         (vect_recog_widen_sum_pattern): Likewise.
11794         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
11795         (vect_recog_widen_shift_pattern): Remove the type_in argument.
11796         (vect_recog_rotate_pattern): Likewise.
11797         (vect_recog_mult_pattern): Likewise.
11798         (vect_recog_vector_vector_shift_pattern): Likewise.
11799         (vect_recog_divmod_pattern): Likewise.
11800         (vect_recog_mixed_size_cond_pattern): Likewise.
11801         (vect_recog_bool_pattern): Likewise.
11802         (vect_recog_mask_conversion_pattern): Likewise.
11803         (vect_try_gather_scatter_pattern): Likewise.
11804         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
11805         (vect_recog_over_widening_pattern): Likewise.
11806         (vect_recog_gather_scatter_pattern): Likewise.
11807         (vect_recog_func_ptr): Move from tree-vectorizer.h
11808         (vect_vect_recog_func_ptrs): Move further down the file.
11809         (vect_recog_func): Likewise.  Remove the third argument.
11810         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
11811         (vect_pattern_recog_1): Expect the pattern function to do any
11812         necessary target tests.  Also expect it to provide a vector type.
11813         Remove the type_in handling.
11815 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11817         * tree-vect-patterns.c (vect_pattern_detected): New function.
11818         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
11819         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
11820         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
11821         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
11822         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11823         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11824         (vect_recog_mask_conversion_pattern)
11825         (vect_try_gather_scatter_pattern): Likewise.
11827 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11829         * tree-vect-patterns.c (vect_get_internal_def): New function.
11830         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11831         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
11832         (search_type_for_mask_1): Use it.
11834 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11836         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11837         redundant WIDEN_SUM_EXPR handling.
11838         (vect_recog_sad_pattern): Likewise.
11840 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11842         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11843         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
11844         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
11845         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
11846         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
11848 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11850         * tree-vect-stmts.c (vectorizable_call): Make sure that we
11851         use the stmt_vec_info of the original bb statement for the
11852         new zero assignment, even if the call is part of a pattern.
11854 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11856         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
11857         that the sequence is attached to the original statement rather
11858         than the pattern statement.
11859         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
11860         PATTERN_DEF_SEQ from the original statement rather than
11861         the main pattern statement.
11862         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
11863         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
11864         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
11866 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11868         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
11869         definition statements before the early exit for statements that aren't
11870         live or relevant.
11871         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
11872         split out from...
11873         (vect_transform_loop): ...here.  Process pattern definition
11874         statements without first checking whether the main pattern
11875         statement is live or relevant.
11877 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11879         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
11880         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
11882 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
11884         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
11885         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
11886         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
11887         (expand_block_compare): Change select_block_compare_mode call.
11888         (expand_strncmp_align_check): Use new functions, fix comment.
11889         (emit_final_str_compare_gpr): New function.
11890         (expand_strn_compare): Refactor and clean up code.
11891         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
11893 2018-06-19  Tony Reix  <tony.reix@atos.com>
11894             Damien Bergamini  <damien.bergamini@atos.com>
11895             David Edelsohn  <dje.gcc@gmail.com>
11897         * collect2.c (static_obj): New variable.
11898         (static_libs): New variable.
11899         (is_in_list): Uncomment declaration.
11900         (main): Track AIX libraries linked statically.
11901         (is_in_list): Uncomment definition.
11902         (scan_prog_file): Don't add AIX shared libraries initializer
11903         to constructor list if linking statically.
11905 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
11907         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
11908         constant.
11909         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
11911 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
11913         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
11914         blocks.
11916 2018-06-19  Martin Liska  <mliska@suse.cz>
11918         * config/i386/i386.c (ix86_can_inline_p): Do not use
11919         ipa_fn_summaries::get_create.
11920         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
11921         get.
11922         (devirtualization_time_bonus): Likewise.
11923         (ipcp_propagate_stage): Likewise.
11924         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11925         (edge_set_predicate): Likewise.
11926         (evaluate_conditions_for_known_args): Likewise.
11927         (evaluate_properties_for_edge): Likewise.
11928         (ipa_call_summary::reset): Tranform to ...
11929         (ipa_call_summary::~ipa_call_summary): ... this.
11930         (ipa_fn_summary::reset): Transform to ...
11931         (ipa_fn_summary::~ipa_fn_summary): ... this.
11932         (ipa_fn_summary_t::remove): Rename to ...
11933         (ipa_fn_summary_t::remove_callees): ... this.
11934         (ipa_fn_summary_t::duplicate): Use placement new
11935         instead of memory copy.
11936         (ipa_call_summary_t::duplicate): Likewise.
11937         (ipa_call_summary_t::remove): Remove.
11938         (dump_ipa_call_summary): Change get_create to get.
11939         (ipa_dump_fn_summary): Dump only when summary exists.
11940         (analyze_function_body): Use symbol_summary::get instead
11941         of get_create.
11942         (compute_fn_summary): Likewise.
11943         (estimate_edge_devirt_benefit): Likewise.
11944         (estimate_edge_size_and_time): Likewise.
11945         (inline_update_callee_summaries): Likewise.
11946         (remap_edge_change_prob): Likewise.
11947         (remap_edge_summaries): Likewise.
11948         (ipa_merge_fn_summary_after_inlining): Likewise.
11949         (write_ipa_call_summary): Likewise.
11950         (ipa_fn_summary_write): Likewise.
11951         (ipa_free_fn_summary): Likewise.
11952         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11953         (struct ipa_call_summary): Likewise.
11954         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11955         of get_create.
11956         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11957         (estimate_size_after_inlining): Likewise.
11958         (estimate_growth): Likewise.
11959         (growth_likely_positive): Likewise.
11960         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11961         (inline_call): Likewise.
11962         * ipa-inline.c (caller_growth_limits): Likewise.
11963         (can_inline_edge_p): Likewise.
11964         (can_inline_edge_by_limits_p): Likewise.
11965         (compute_uninlined_call_time): Likewise.
11966         (compute_inlined_call_time): Likewise.
11967         (want_inline_small_function_p): Likewise.
11968         (edge_badness): Likewise.
11969         (update_caller_keys): Likewise.
11970         (update_callee_keys): Likewise.
11971         (inline_small_functions): Likewise.
11972         (inline_to_all_callers_1): Likewise.
11973         (dump_overall_stats): Likewise.
11974         (early_inline_small_functions): Likewise.
11975         (early_inliner): Likewise.
11976         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11977         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11978         * ipa-pure-const.c (malloc_candidate_p): Likewise.
11979         * ipa-split.c (execute_split_functions): Likewise.
11980         * symbol-summary.h: Likewise.
11981         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11983 2018-06-19  Richard Biener  <rguenther@suse.de>
11985         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11986         (vectorize_loops): ... here.  Fix dbgcnt handling.
11987         (try_vectorize_loop): Wrap try_vectorize_loop_1.
11989 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
11991         PR target/86197
11992         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11993         ieee128 argument takes up only one (vector) register, not two (floating
11994         point) registers.
11996 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11998         * gimplify.c (gimplify_init_constructor): Really never clear for an
11999         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
12001 2018-06-19  Richard Biener  <rguenther@suse.de>
12003         PR tree-optimization/86179
12004         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
12005         after failed recognition.
12007 2018-06-18  Martin Sebor  <msebor@redhat.com>
12009         PR middle-end/85602
12010         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
12011         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
12012         Handle integer subtraction.
12013         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
12014         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
12016 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12018         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
12019         param from rtx to rtx_insn *.
12020         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
12021         param.
12022         (frv_ifcvt_modify_insn): Likwise.
12023         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
12024         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
12025         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
12026         as_a <rtx_insn *> cast to local "unprotected_region" once
12027         it's been established that it's not NULL or pc_rtx.
12028         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
12029         param "sethi" from rtx to rtx_insn *.
12030         (nds32_group_float_insns): Likewise for param "insn".
12031         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
12032         param.
12033         (vax_output_int_subtract): Likewise.
12034         * config/vax/vax.c (vax_output_int_add): Likewise for param
12035         "insn".
12036         (vax_output_int_subtract): Likewise.
12037         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
12038         (emit_pattern_after): Likewise for param "after".
12039         (emit_insn_after): Likewise.
12040         (emit_jump_insn_after): Likewise.
12041         (emit_call_insn_after): Likewise.
12042         (emit_debug_insn_after): Likewise.
12043         (emit_pattern_before): Likewise for param "before".
12044         (emit_insn_before): Likewise.
12045         (emit_jump_insn_before): Likewise.
12046         * final.c (get_insn_template): Likewise for param "insn", removing
12047         a cast.
12048         * output.h (get_insn_template): Likewise for 2nd param.
12049         * rtl.h (emit_insn_before): Likewise.
12050         (emit_jump_insn_before): Likewise.
12051         (emit_debug_insn_before_noloc): Likewise.
12052         (emit_insn_after): Likewise.
12053         (emit_jump_insn_after): Likewise.
12054         (emit_call_insn_after): Likewise.
12055         (emit_debug_insn_after): Likewise.
12056         (set_insn_deleted): Likewise for param.
12058 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
12060         PR target/85358
12061         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
12062         floating point modes, so that IFmode is numerically greater than
12063         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
12064         to declare the ordering.  This prevents IFmode from being
12065         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
12066         machine.  Include rs6000-modes.h to share the fractional values
12067         between genmodes* and the rest of the compiler.
12068         (IFmode): Likewise.
12069         (KFmode): Likewise.
12070         (TFmode): Likewise.
12071         * config/rs6000/rs6000-modes.h: New file.
12072         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
12073         meaning of rs6000_long_double_size so that 126..128 selects an
12074         appropriate 128-bit floating point type.
12075         (rs6000_option_override_internal): Likewise.
12076         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
12077         (TARGET_LONG_DOUBLE_128): Change the meaning of
12078         rs6000_long_double_size so that 126..128 selects an appropriate
12079         128-bit floating point type.
12080         (LONG_DOUBLE_TYPE_SIZE): Update comment.
12081         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
12082         source and destination to match the standard usage.
12083         (truncifkf2): Likewise.
12084         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
12085         ISA 2.07 to use an explicit clobber, instead of passing in a
12086         temporary.
12087         (copysign<mode>3_soft): Likewise.
12089 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12091         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12092         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
12093         (vect_slp_analyze_instance_dependence): Likewise.
12094         (vect_enhance_data_refs_alignment): Likewise.
12095         (vect_analyze_data_refs_alignment): Likewise.
12096         (vect_slp_analyze_and_verify_instance_alignment
12097         (vect_analyze_data_ref_accesses): Likewise.
12098         (vect_prune_runtime_alias_test_list): Likewise.
12099         (vect_analyze_data_refs): Likewise.
12100         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
12101         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12102         (vect_analyze_scalar_cycles_1): Likewise.
12103         (vect_get_loop_niters): Likewise.
12104         (vect_analyze_loop_form_1): Likewise.
12105         (vect_update_vf_for_slp): Likewise.
12106         (vect_analyze_loop_operations): Likewise.
12107         (vect_analyze_loop): Likewise.
12108         (vectorizable_induction): Likewise.
12109         (vect_transform_loop): Likewise.
12110         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
12111         * tree-vect-slp.c (vect_analyze_slp): Likewise.
12112         (vect_make_slp_decision): Likewise.
12113         (vect_detect_hybrid_slp): Likewise.
12114         (vect_slp_analyze_operations): Likewise.
12115         (vect_slp_bb): Likewise.
12116         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
12117         (vectorizable_bswap): Likewise.
12118         (vectorizable_call): Likewise.
12119         (vectorizable_simd_clone_call): Likewise.
12120         (vectorizable_conversion): Likewise.
12121         (vectorizable_assignment): Likewise.
12122         (vectorizable_shift): Likewise.
12123         (vectorizable_operation): Likewise.
12124         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12126 2018-06-18  Martin Sebor  <msebor@redhat.com>
12128         PR tree-optimization/81384
12129         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12130         * builtins.c (expand_builtin_strnlen): New function.
12131         (expand_builtin): Call it.
12132         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12133         * builtins.def (BUILT_IN_STRNLEN): New.
12134         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12135         Warn for bounds in excess of maximum object size.
12136         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12137         single-value ranges.  Handle strnlen.
12138         (handle_builtin_strlen): Handle strnlen.
12139         (strlen_check_and_optimize_stmt): Same.
12140         * doc/extend.texi (Other Builtins): Document strnlen.
12142 2018-06-18  Maya Rashish  <coypu@sdf.org>
12144         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12145         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12146         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12148         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12149         here to ...
12150         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12152 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12154         * tree.c (escaped_string::escape): Replace cast to char * by
12155         const_cast<char *> (unescaped).
12157 2018-06-18  Nick Clifton  <nickc@redhat.com>
12159         PR 84195
12160         * tree.c (escaped_string): New class.  Converts an unescaped
12161         string into its escaped equivalent.
12162         (warn_deprecated_use): Use the new class to convert the
12163         deprecation message, if present.
12164         (test_escaped_strings): New self test.
12165         (test_c_tests): Add test_escaped_strings.
12166         * doc/extend.texi (deprecated): Add a note that the
12167         deprecation message is affected by the -fmessage-length
12168         option, and that control characters will be escaped.
12169         (#pragma GCC error): Document this pragma.
12170         (#pragma GCC warning): Likewise.
12171         * doc/invoke.texi (-fmessage-length): Document this option's
12172         effect on the #warning and #error preprocessor directives and
12173         the deprecated attribute.
12175 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
12177         * tree.c (decl_value_expr_lookup): Revert latest change.
12178         (decl_value_expr_insert): Likewise.
12180 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
12182         * gimplify.c (nonlocal_vlas): Delete.
12183         (nonlocal_vla_vars): Likewise.
12184         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12185         referenced VLAs.
12186         (gimplify_body): Do not create and destroy nonlocal_vlas.
12187         * tree-nested.c: Include diagnostic.h.
12188         (use_pointer_in_frame): Tweak.
12189         (lookup_field_for_decl): Add assertion and declare the transformation.
12190         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12191         internal error when the reference is in a wrong context.  Do not
12192         create a debug decl by default.
12193         (note_nonlocal_block_vlas): Delete.
12194         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12195         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
12196         create a debug decl by default.
12197         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12198         call is in a wrong context.
12199         (fixup_vla_decls): New function.
12200         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
12201         debug variables were created.
12202         * tree.c (decl_value_expr_lookup): Add checking assertion.
12203         (decl_value_expr_insert): Likewise.
12205 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12207         PR middle-end/82479
12208         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12209         * tree-scalar-evolution.c (interpret_expr): Likewise.
12210         (expression_expensive_p): Likewise.
12211         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12212         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12213         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12214         (ssa_defined_by_minus_one_stmt_p): New.
12216 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12218         PR middle-end/64946
12219         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12220         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12221         * dojump.c (do_jump): Likewise.
12222         * expr.c (expand_expr_real_2): Check operand type's sign.
12223         * fold-const.c (const_unop): Handle ABSU_EXPR.
12224         (fold_abs_const): Likewise.
12225         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
12226         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12227         (strip_sign_op_1): Likesise.
12228         * match.pd: Add new pattern to generate ABSU_EXPR.
12229         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
12230         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12231         * tree-eh.c (operation_could_trap_helper_p): Likewise.
12232         * tree-inline.c (estimate_operator_cost): Likewise.
12233         * tree-pretty-print.c (dump_generic_node): Likewise.
12234         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
12235         * tree.def (ABSU_EXPR): New.
12237 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
12239         PR middle-end/86095
12240         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
12241         documented as preserved for backward compatibility only.
12242         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
12244         PR rtl-optimization/86108
12245         * bb-reorder.c (create_forwarder_block): Renamed to ...
12246         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
12247         jump from new landing pad to the second part.
12248         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
12249         Adjust callers.
12251 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
12253         PR middle-end/85878
12254         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
12255         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
12256         Only call store_expr for halves if the mode is the same.
12258         PR middle-end/86123
12259         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
12260         Fix up comment formatting.
12262 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12264         * typed-splay-tree.h (typed_splay_tree::remove): New function.
12265         (typed_splay_tree::closure,
12266         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
12267         (typed_splay_tree::typed_splay_tree,
12268         typed_splay_tree::operator =): Declared private.
12269         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
12270         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
12271         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
12272         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
12273         typed_splay_tree::splay_tree_splay,
12274         typed_splay_tree::splay_tree_foreach_helper,
12275         typed_splay_tree::splay_tree_insert,
12276         typed_splay_tree::splay_tree_remove,
12277         typed_splay_tree::splay_tree_lookup,
12278         typed_splay_tree::splay_tree_predecessor,
12279         typed_splay_tree::splay_tree_successor,
12280         typed_splay_tree::splay_tree_min,
12281         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
12282         (typed_splay_tree::root, typed_splay_tree::comp,
12283         typed_splay_tree::delete_key,
12284         typed_splay_tree::delete_value): New data members.
12285         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
12286         typed_splay_tree::remove.
12288 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
12290         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
12291         -mginv and -mno-ginv to the assembler.
12292         * config/mips/mips.opt (-mcrc): New option.
12293         (-mginv): Likewise.
12294         * doc/invoke.text (-mcrc): Document.
12295         (-mginv): Likewise.
12297 2018-06-15  Nick Clifton  <nickc@redhat.com>
12299         PR 84195
12300         * tree.c (escaped_string): New class.  Converts an unescaped
12301         string into its escaped equivalent.
12302         (warn_deprecated_use): Use the new class to convert the
12303         deprecation message, if present.
12304         (test_escaped_strings): New self test.
12305         (test_c_tests): Add test_escaped_strings.
12306         * doc/extend.texi (deprecated): Add a note that the
12307         deprecation message is affected by the -fmessage-length
12308         option, and that control characters will be escaped.
12309         (#pragma GCC error): Document this pragma.
12310         (#pragma GCC warning): Likewise.
12311         * doc/invoke.texi (-fmessage-length): Document this option's
12312         effect on the #warning and #error preprocessor directives and
12313         the deprecated attribute.
12315 2018-06-15  Richard Biener  <rguenther@suse.de>
12317         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
12318         here, also noting vector size used.
12319         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
12320         size used in MSG_OPTIMIZED_LOCATIONS dump.
12321         (pass_slp_vectorize::execute): Adjust.
12323 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
12325         PR target/85968
12326         * config/arc/arc.c (arc_return_address_register): Fix
12327         if-condition.
12329 2018-06-15  Richard Biener  <rguenther@suse.de>
12331         PR middle-end/86159
12332         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
12333         leave useless conversion stripping to force_gimple_operand_gsi.
12334         (gimplify_build2): Likewise.
12335         (gimplify_build1): Likewise.
12337 2018-06-15  Richard Biener  <rguenther@suse.de>
12339         PR middle-end/86076
12340         * tree-cfg.c (move_stmt_op): unshare invariant addresses
12341         before adjusting their block.
12343 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12345         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
12346         multilibs for *-*-rtems*.
12347         * config/riscv/t-rtems: New file.
12349 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
12351         PR middle-end/86122
12352         * match.pd ((A +- CST1) +- CST2): Punt if last resort
12353         unsigned_type_for returns NULL.
12355         PR target/85945
12356         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
12357         subregs of multi-word pseudos unless the float mode has word size.
12359 2018-06-14  Richard Biener  <rguenther@suse.de>
12361         PR middle-end/86139
12362         * tree-vect-generic.c (build_word_mode_vector_type): Remove
12363         duplicate and harmful type_hash_canon.
12364         * tree.c (type_hash_canon): Assert we didn't find ourselves.
12366 2018-06-14  Richard Biener  <rguenther@suse.de>
12368         PR ipa/86124
12369         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
12370         NULL cgraph_node.
12372 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12374         * config/rtems.h (STDINT_LONG32): Define.
12376 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
12377             Prachi Godbole  <prachi.godbole@imgtec.com>
12379         * config/mips/mips-cpus.def: Define P6600.
12380         * config/mips/mips-tables.opt: Regenerate.
12381         * config/mips/mips.c (mips_ucbranch_type): New enum.
12382         (mips_rtx_cost_data): Add support for P6600.
12383         (mips_issue_rate): Likewise.
12384         (mips_multipass_dfa_lookahead): Likewise.
12385         (mips_avoid_hazard): Likewise.
12386         (mips_reorg_process_insns): Likewise.
12387         (mips_classify_branch_p6600): New function.
12388         * config/mips/mips.h (TUNE_P6600): New define.
12389         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
12390         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
12391         * config/mips/mips.md: Include p6600.md.
12392         (processor): Add p6600.
12393         * config/mips/p6600.md: New file.
12394         * doc/invoke.texi: Add p6600 to supported architectures.
12396 2018-06-13  Martin Sebor  <msebor@redhat.com>
12398         PR tree-optimization/86114
12399         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
12400         of integer types.
12401         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
12403 2018-06-13  Richard Biener  <rguenther@suse.de>
12405         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
12406         Properly set vector type of the intermediate stmt.
12407         * tree-vect-stmts.c (vectorizable_operation): The destination
12408         var always has vectype_out type.
12410 2018-06-13  Jeff Law  <law@redhat.com>
12412         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
12413         integer 0 for argument to print_rtl_with_bb.
12414         (rl78_reorg): Likewise.
12416 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
12418         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
12419         from rtx to rtx_insn *.
12420         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
12421         "label".
12422         (add_sched_insns_for_speculation): Likewise for local "target",
12423         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12424         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
12425         from rtx_insn ** to rtx_code_label **.
12426         (reorg_emit_nops): Likewise.
12427         (c6x_reorg): Likewise for local "call_labels".
12428         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
12429         rtx to rtx_insn *.
12430         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
12431         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
12432         the loops over LABEL_REFS.
12433         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
12434         braf_label.
12435         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12436         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
12437         (split_branches): Strengthen local "olabel" from rtx to
12438         rtx_insn *, adding a safe_as_a cast.
12439         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
12440         to "rtx_insn *".
12441         (add_insn_after): Likewise for first two params.
12442         (add_insn_before): Likewise.
12443         (remove_insn): Likewise for param.
12444         (emit_pattern_before_noloc): Likewise for second and third params.
12445         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
12446         (emit_call_insn_before_noloc): Likewise.
12447         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
12448         to "rtx_insn *".
12449         (emit_barrier_before): Likewise.
12450         (emit_label_before): Strengthen "label" param from "rtx" to
12451         "rtx_code_label *".  Strengthen "before" param from "rtx" to
12452         "rtx_insn *".
12453         (emit_insn_after_1): Strengthen "after" param from "rtx" to
12454         "rtx_insn *".
12455         (emit_pattern_after_noloc): Likewise.
12456         (emit_insn_after_noloc): Likewise.
12457         (emit_jump_insn_after_noloc): Likewise.
12458         (emit_call_insn_after_noloc): Likewise.
12459         (emit_debug_insn_after_noloc): Likewise.
12460         (emit_barrier_after): Likewise.
12461         (emit_label_after): Likewise for both params.
12462         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
12463         "loc" param from "int" to "location_t".
12464         (emit_insn_after_setloc): Likewise.
12465         (emit_jump_insn_after_setloc): Likewise.
12466         (emit_call_insn_after_setloc): Likewise.
12467         (emit_debug_insn_after_setloc): Likewise.
12468         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
12469         "loc" param from "int" to "location_t".
12470         (emit_pattern_before): Convert NULL_RTX to NULL.
12471         (emit_insn_before_setloc): Convert "loc" param from "int" to
12472         "location_t".
12473         (emit_jump_insn_before_setloc): Likewise.
12474         (emit_call_insn_before_setloc): Likewise.
12475         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
12476         rtx_insn *.  Convert "loc" param from "int" to "location_t".
12477         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
12478         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
12479         Convert 3rd param from "int" to "location_t".
12480         (emit_barrier_before, emit_barrier_after, next_real_insn):
12481         Strengthen param from rtx to rtx_insn *.
12482         (emit_label_before): Strengthen 1st param from "rtx" to
12483         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
12484         "rtx_insn *".
12485         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
12486         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
12487         Strengthen 2nd param from "rtx" to "rtx_insn *".
12488         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
12489         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
12490         Likewise. Convert 3rd param from "int" to "location_t".
12491         (emit_label_after): Strengthen 1st param from "rtx" to
12492         "rtx_code_label *".
12493         (next_real_insn, remove_insn): Strengthen param from "rtx" to
12494         "rtx_insn *".
12495         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
12496         from "rtx" to "rtx_insn *".
12498 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
12500         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
12501         bodies streamed in with -Q.
12502         * dumpfile.c (dump_files): Add lto-stream-out dump file.
12503         * dumpfile.h (tree_dump_index): Add lto_stream_out.
12504         * gimple-streamer-out.c: Include gimple-pretty-print.h
12505         (output_bb): Dump stmts streamed.
12506         * lto-section-out.c: Include print-tree.h
12507         (lto_begin_section): Dump sections created.
12508         (lto_output_decl_index): Dump decl encoded.
12509         * lto-streamer-out.c: Include print-tree.h
12510         (create_output_block): Dump output block created.
12511         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
12512         (output_function): Dump function output.
12513         (output_constructor): Dump constructor streamed.
12514         (write_global_stream): Output indexes encoded.
12515         (produce_asm_for_decls): Dump streams encoded.
12516         * lto-streamer.c (streamer_dump_file): New global var.
12517         * lto-streamer.h (streamer_dump_file): Declare.
12518         * passes.c (ipa_write_summaries): Initialize streamer dump.
12519         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
12520         in.
12522 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
12524         PR target/86048
12525         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
12526         offsets for register save directives.  Emit a second batch of save
12527         directives, if need be, when the function accesses prior frames.
12529 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12531         * config/arc/fpu.md (fmasf4): Force operand to register.
12532         (fnmasf4): Likewise.
12534 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12536         * config/arc/arc-protos.h (arc_pad_return): Remove.
12537         * config/arc/arc.c (machine_function): Remove force_short_suffix
12538         and size_reason.
12539         (arc_print_operand): Adjust printing of '&'.
12540         (arc_verify_short): Remove conditional printing of short suffix.
12541         (arc_final_prescan_insn): Remove reference to size_reason.
12542         (pad_return): New function.
12543         (arc_reorg): Call pad_return.
12544         (arc_pad_return): Remove.
12545         (arc_init_machine_status): Remove reference to force_short_suffix.
12546         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
12547         (attr length): When attribute iscompact is true force to 2
12548         regardless; in the case of maybe check if we want to force the
12549         instruction to have 4 bytes length.
12550         (nopv): Change it to generate 4 byte long nop as well.
12551         (blockage): New pattern.
12552         (simple_return): Remove call to arc_pad_return.
12553         (p_return_i): Likewise.
12555 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12557         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
12559 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12561         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
12562         ARC cores.
12564 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12566         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
12567         for ARC700 and ARCv2.
12569 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
12571         PR target/86076
12572         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
12573         operands[2] instead of operands[1].
12576 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
12578         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
12579         case, check whether the outer register overlaps an unallocatable
12580         register, not just whether it fits the required class.
12582 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
12584         * poly-int.h (can_div_trunc_p): Add new overload in which all values
12585         are poly_ints.
12586         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
12587         (memrefs_conflict_p): Likewise.
12588         (init_alias_analysis): Likewise.
12589         * cfgexpand.c (expand_debug_expr): Likewise.
12590         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
12591         * cse.c (fold_rtx): Likewise.
12592         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
12593         * expr.c (emit_block_move_hints): Likewise.
12594         (clear_storage_hints, push_block, emit_push_insn): Likewise.
12595         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
12596         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
12597         (emit_group_store): Likewise.
12598         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
12599         to read the PRE/POST_MODIFY increment.
12600         * calls.c (store_one_arg): Use strip_offset.
12601         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
12602         poly_int_rtx_p.
12603         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
12604         by a VEC_SELECT.
12605         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
12606         (simplify_binary_operation_1): Extend CONST_INT handling to
12607         poly_int_rtx_p.
12608         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
12609         than a HOST_WIDE_INT.
12610         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
12611         poly_int64.
12612         (adjust_mems, add_stores): Update accodingly.
12613         (vt_canonicalize_addr): Track polynomial offsets.
12614         (emit_note_insn_var_location): Likewise.
12615         (vt_add_function_parameter): Likewise.
12616         (vt_initialize): Likewise.
12618 2018-06-12  Jeff Law  <law@redhat.com>
12620         * config.gcc (alpha*-*-freebsd*): Remove.
12621         * config/alpha/freebsd.h: Remove.
12623 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
12625         PR other/69968
12626         * spellcheck-tree.c (levenshtein_distance): Rename to...
12627         (get_edit_distance): ...this, and update for underlying renaming.
12628         * spellcheck-tree.h (levenshtein_distance): Rename to...
12629         (get_edit_distance): ...this.
12630         * spellcheck.c (levenshtein_distance): Rename to...
12631         (get_edit_distance): ...this.  Convert from Levenshtein distance
12632         to Damerau-Levenshtein distance by supporting transpositions of
12633         adjacent characters.  Rename "v1" to "v_next" and "v0" to
12634         "v_one_ago".
12635         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
12636         (selftest::test_edit_distance_unit_test_oneway): ...this, and
12637         update for underlying renaming.
12638         (selftest::levenshtein_distance_unit_test): Rename to...
12639         (selftest::test_get_edit_distance_unit): ...this, and update for
12640         underlying renaming.
12641         (selftest::test_find_closest_string): Add example from PR 69968
12642         where transposition helps
12643         (selftest::test_metric_conditions): Update for renaming.
12644         (selftest::test_metric_conditions): Likewise.
12645         (selftest::spellcheck_c_tests): Likewise.
12646         * spellcheck.h (levenshtein_distance): Rename both overloads to...
12647         (get_edit_distance): ...this.
12648         (best_match::consider): Update for renaming.
12650 2018-06-12  Martin Sebor  <msebor@redhat.com>
12652         PR tree-optimization/85259
12653         * builtins.c (compute_objsize): Handle constant offsets.
12654         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
12655         true iff a warning has been issued.
12656         * gimple.h (gimple_nonartificial_location): New function.
12657         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
12658         gimple_nonartificial_location and handle -Wno-system-headers.
12659         (handle_builtin_stxncpy): Same.
12661 2018-06-12  Martin Sebor  <msebor@redhat.com>
12663         PR c/85931
12664         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
12666 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
12668         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
12669         BUILTIN_VEC_XST entries for pointer to double and long long.
12671 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
12673         PR target/85990
12674         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12675         Update comments.
12676         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12677         Likewise.
12679 2018-06-12  Martin Liska  <mliska@suse.cz>
12681         * doc/options.texi: Document IntegerRange.
12683 2018-06-12  Martin Liska  <mliska@suse.cz>
12685         * config/i386/i386.opt: Make MPX-related options as Deprecated.
12686         * opt-functions.awk: Handle Deprecated flag.
12687         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
12688         and report error.
12689         (read_cmdline_option): Report warning for a deprecated option.
12690         * opts.h (struct cl_option): Add new field cl_deprecated.
12691         (CL_ERR_DEPRECATED): New.
12693 2018-06-12  Martin Liska  <mliska@suse.cz>
12695         * doc/options.texi: Document Deprecated option flag.
12697 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12699         * config/arc/arc-arch.h (arc_extras): New enum.
12700         (arc_cpu_t):Add field extra.
12701         (arc_cpu_types): Consider the extras.
12702         * config/arc/arc-cpus.def: Add extras info.
12703         * config/arc/arc-opts.h (processor_type): Consider extra field.
12704         * config/arc/arc.c (arc_override_options): Handle extra field.
12706 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12708         * config/arc/arc-arch.h: Update ARC_OPTX macro.
12709         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
12710         field.
12711         * config/arc/arc.c (arc_init): Update pic warning.
12712         (irq_range): Update irq range parsing warnings.
12713         (arc_override_options): Update various warning messages.
12714         (arc_handle_aux_attribute): Likewise.
12716 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
12718         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
12720 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12722         * doc/sourcebuild.texi: Document usage of line number 0 in verify
12723         compiler messages directives.
12725 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
12727         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
12728         * config/mips/mips-tables.opt: Regenerate.
12729         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
12730         mips64r6.
12731         * doc/invoke.texi: Document -march=i6500.
12733 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
12735         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
12736         (i6400_gpmul): Add cpu_unit.
12737         (i6400_gpdiv): Likewise.
12738         (i6400_msa_add_d): Update reservations.
12739         (i6400_msa_int_add) Likewise.
12740         (i6400_msa_short_logic3) Likewise.
12741         (i6400_msa_short_logic2) Likewise.
12742         (i6400_msa_short_logic) Likewise.
12743         (i6400_msa_move) Likewise.
12744         (i6400_msa_cmp) Likewise.
12745         (i6400_msa_short_float2) Likewise.
12746         (i6400_msa_div_d) Likewise.
12747         (i6400_msa_long_logic1) Likewise.
12748         (i6400_msa_long_logic2) Likewise.
12749         (i6400_msa_mult) Likewise.
12750         (i6400_msa_long_float2) Likewise.
12751         (i6400_msa_long_float4) Likewise.
12752         (i6400_msa_long_float5) Likewise.
12753         (i6400_msa_long_float8) Likewise.
12754         (i6400_fpu_fadd): Include frint type.
12755         (i6400_fpu_store): New define_insn_reservation.
12756         (i6400_fpu_load): Likewise.
12757         (i6400_fpu_move): Likewise.
12758         (i6400_fpu_fcmp): Likewise.
12759         (i6400_fpu_fmadd): Likewise.
12760         (i6400_int_mult): Include imul3nc type and update reservation.
12761         (i6400_int_div): Include idiv3 type and update reservation.
12762         (i6400_int_load): Update to check type not move_type.
12763         (i6400_int_store): Likewise.
12764         (i6400_int_prefetch): Set zero latency.
12766 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
12768         * gcc.c: Document new %@{...} sequence.
12769         (LINK_COMMAND_SPEC): Use it for the -L switches.
12770         (cpp_unique_options): Use it for the -I switches.
12771         (at_file_argbuf): New global variable.
12772         (in_at_file): Likewise.
12773         (alloc_args): Create at_file_argbuf.
12774         (clear_args): Truncate at_file_argbuf.
12775         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
12776         (open_at_file): New function.
12777         (close_at_file): Likewise.
12778         (create_at_file): Delete.
12779         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
12780         <'o'>: Likewise.
12781         <'@'>: New case.
12782         (validate_switches_from_spec): Deal with %@{...} sequence.
12783         (validate_switches): Likewise.
12784         (driver::finalize): Call clear_args.
12786 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
12788         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
12790 2018-06-11  Martin Sebor  <msebor@redhat.com>
12792         * doc/invoke.texi (-Wall): List -Wc++17-compat.
12793         (Wno-class-memaccess): Add @opindex.
12794         (Wno-templates, Wno-multiple-inheritance): Same.
12795         (Wno-virtual-inheritance, Wno-namespaces): Same.
12796         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
12797         (Wno-format-overflow, Wno-format-truncation): Same.
12798         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
12799         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
12800         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
12801         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
12802         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
12803         (Wno-misspelled-isr): Same.
12805 2018-06-11  Martin Sebor  <msebor@redhat.com>
12807         * PR tree-optimization/86083
12808         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
12810 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
12812         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
12814 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
12816         PR target/85755
12817         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
12818         on the correct operand.
12819         (*movdi_internal64): Ditto.
12821 2018-06-11  Martin Liska  <mliska@suse.cz>
12823         PR tree-optimization/86089
12824         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
12826 2018-06-11  Julia Koval  <julia.koval@intel.com>
12828         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
12829         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
12830         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
12832 2018-06-11  Olivier Hainque  <hainque@adacore.com>
12834         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
12835         for Ada with strict dwarf2.
12837 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
12839         PR target/85755
12840         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
12841         addresses.
12843 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
12845         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
12847 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
12849         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
12850         TARGET_ELF.
12852 2018-06-08  Martin Liska  <mliska@suse.cz>
12854         * tree-cfg.h (debug_function): Fix argument type to match
12855         implementation.
12857 2018-06-08  Martin Liska  <mliska@suse.cz>
12859         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
12860         Remove usage of MPX-related (and removed) fields.
12861         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
12863 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
12865         * cfg.c (debug): Use TDF_NONE rather than 0.
12866         * cfghooks.c (debug): Likewise.
12867         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
12868         (struct dump_option_value_info): Convert to...
12869         (struct kv_pair): ...this template type.
12870         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
12871         rather than 0.
12872         (optinfo_verbosity_options): Likewise.
12873         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
12874         OPTGROUP_NONE.
12875         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
12876         than int for "optgroup_flags" param.
12877         (dump_generic_expr_loc): Use dump_flags_t rather than int for
12878         "dump_kind" param.
12879         (dump_dec): Likewise.
12880         (dump_finish): Use TDF_NONE rather than 0.
12881         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
12882         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
12883         than 0.  Update for change to option_ptr.
12884         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
12885         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
12886         0.  Update for changes to optinfo_verbosity_options and
12887         optgroup_options.
12888         (opt_info_switch_p): Convert optgroup_flags from int to
12889         optgroup_flags_t.
12890         (dump_basic_block): Use dump_flags_t rather than int
12891         for "dump_kind" param.
12892         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
12893         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
12894         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
12895         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
12896         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
12897         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
12898         TDF_NONE): Convert from macros to...
12899         (enum dump_flag): ...this new enum.
12900         (dump_flags_t): Update to use enum.
12901         (operator|, operator&, operator~, operator|=, operator&=):
12902         Implement for dump_flags_t.
12903         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
12904         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
12905         Convert from macros to...
12906         (enum optgroup_flag): ...this new enum.
12907         (optgroup_flags_t): New typedef.
12908         (operator|, operator|=): Implement for optgroup_flags_t.
12909         (struct dump_file_info): Convert field "alt_flags" to
12910         dump_flags_t.  Convert field "optgroup_flags" to
12911         optgroup_flags_t.
12912         (dump_basic_block): Use dump_flags_t rather than int for param.
12913         (dump_generic_expr_loc): Likewise.
12914         (dump_dec): Likewise.
12915         (dump_register): Convert param "optgroup_flags" to
12916         optgroup_flags_t.
12917         (opt_info_enable_passes): Likewise.
12918         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
12919         than 0.
12920         * gimple-pretty-print.c (debug): Likewise.
12921         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
12922         (merged_store_group::apply_stores): Likewise.
12923         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12924         * gimple.c (verify_gimple_pp): Likewise.
12925         * graphite-poly.c (print_pbb_body): Likewise.
12926         * passes.c (pass_manager::register_one_dump_file): Convert
12927         local "optgroup_flags" to optgroup_flags_t.
12928         * print-tree.c (print_node): Use TDF_NONE rather than 0.
12929         (debug): Likewise.
12930         (debug_body): Likewise.
12931         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12932         to optgroup_flags_t.
12933         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12934         than 0.
12935         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12936         (convert_mult_to_fma): Likewise.
12937         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12938         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12939         * tree-vect-data-refs.c (dump_lower_bound): Convert param
12940         "dump_kind" to dump_flags_t.
12942 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
12944         * config/rs6000/rs6000.c (min, max): Delete.
12946 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
12948         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12949         -mabi=spe and -mabi=no-spe.
12951 2018-06-08  Martin Liska  <mliska@suse.cz>
12953         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12954         where we expect an existing summary.
12956 2018-06-08  Martin Liska  <mliska@suse.cz>
12958         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12959         * ipa-inline.h (estimate_edge_growth): Likewise.
12961 2018-06-08  Martin Liska  <mliska@suse.cz>
12963         * cgraph.c (function_version_hasher::hash): Use
12964         cgraph_node::get_uid ().
12965         (function_version_hasher::equal):
12966         * cgraph.h (cgraph_node::get_uid): New method.
12967         * ipa-inline.c (update_caller_keys): Use
12968         cgraph_node::get_uid ().
12969         (update_callee_keys): Likewise.
12970         * ipa-utils.c (searchc): Likewise.
12971         (ipa_reduced_postorder): Likewise.
12972         * lto-cgraph.c (input_node): Likewise.
12973         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12974         * symbol-summary.h (symtab_insertion): Likewise.
12975         (symtab_removal): Likewise.
12976         (symtab_duplication): Likewise.
12977         * tree-pretty-print.c (dump_function_header): Likewise.
12978         * tree-sra.c (convert_callers_for_node): Likewise.
12980 2018-06-08  Martin Liska  <mliska@suse.cz>
12982         * cgraph.c (symbol_table::create_edge): Always assign a new
12983         unique number.
12984         (symbol_table::free_edge): Do not recycle numbers.
12985         * cgraph.h (cgraph_edge::get): New method.
12986         * symbol-summary.h (symtab_removal): Use it.
12987         (symtab_duplication): Likewise.
12988         (call_summary::hashable_uid): Remove.
12990 2018-06-08  Martin Liska  <mliska@suse.cz>
12992         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12993         (initialize_growth_caches): Remove.
12994         (free_growth_caches): Likewise.
12995         (do_estimate_edge_time): Use edge_growth_cache.
12996         (do_estimate_edge_size): Likewise.
12997         (do_estimate_edge_hints): Likewise.
12998         * ipa-inline.c (reset_edge_caches): Likewise.
12999         (recursive_inlining): Likewise.
13000         (inline_small_functions): Likewise.
13001         * ipa-inline.h (initialize_growth_caches): Remove.
13002         (estimate_edge_size): Likewise.
13003         (estimate_edge_time): Likewise.
13004         (estimate_edge_hints): Likewise.
13005         (reset_edge_growth_cache): Likewise.
13006         * symbol-summary.h (call_summary::remove): New method.
13008 2018-06-08  Martin Liska  <mliska@suse.cz>
13010         * ipa-cp.c (class edge_clone_summary): New summary.
13011         (grow_edge_clone_vectors): Remove.
13012         (ipcp_edge_duplication_hook): Remove.
13013         (class edge_clone_summary_t): New call_summary class.
13014         (ipcp_edge_removal_hook): Remove.
13015         (edge_clone_summary_t::duplicate): New function.
13016         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
13017         (create_specialized_node): Likewise.
13018         (ipcp_driver): Initialize edge_clone_summaries and do not
13019         register hooks.
13021 2018-06-08  Martin Liska  <mliska@suse.cz>
13023         * symbol-summary.h (get): New function.
13024         (call_summary::m_initialize_when_cloning): New class member.
13026 2018-06-08  Martin Liska  <mliska@suse.cz>
13028         * cgraph.c (cgraph_node::remove): Do not recycle uid.
13029         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
13030         (symbol_table::allocate_cgraph_symbol): Do not set uid.
13031         * passes.c (uid_hash_t): Record removed_nodes by their uids.
13032         (remove_cgraph_node_from_order): Use the removed_nodes set.
13033         (do_per_function_toporder): Likwise.
13034         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
13035         instead of summary_uid.
13036         (symtab_removal): Likewise.
13037         (symtab_duplication): Likewise.
13039 2018-06-08  Martin Liska  <mliska@suse.cz>
13041         * ipa-cp.c (ipcp_store_bits_results): Use
13042         ipcp_transformation_sum.
13043         (ipcp_store_vr_results): Likewise.
13044         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
13045         to ...
13046         (ipcp_transformation_initialize): ... this.
13047         (ipa_set_node_agg_value_chain):
13048         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
13049         (write_ipcp_transformation_info): Likewise.
13050         (read_ipcp_transformation_info): Likewise.
13051         (ipcp_update_bits): Likewise.
13052         (ipcp_update_vr): Likewise.
13053         (ipcp_transform_function): Likewise.
13054         * ipa-prop.h: Rename ipcp_transformation_summary to
13055         ipcp_transformation.
13056         (class ipcp_transformation_t): New function summary.
13057         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
13058         (ipa_get_agg_replacements_for_node): Likewise.
13060 2018-06-08  Martin Liska  <mliska@suse.cz>
13062         * ipa-pure-const.c (struct funct_state_d): Do it class instead
13063         of struct.
13064         (class funct_state_summary_t): New function_summary class.
13065         (has_function_state): Remove.
13066         (get_function_state): Likewise.
13067         (set_function_state): Likewise.
13068         (add_new_function): Likewise.
13069         (funct_state_summary_t::insert): New function.
13070         (duplicate_node_data): Remove.
13071         (remove_node_data): Remove.
13072         (funct_state_summary_t::duplicate): New function.
13073         (register_hooks): Create new funct_state_summaries.
13074         (pure_const_generate_summary): Use it.
13075         (pure_const_write_summary): Likewise.
13076         (pure_const_read_summary): Likewise.
13077         (propagate_pure_const): Likewise.
13078         (propagate_nothrow): Likewise.
13079         (dump_malloc_lattice): Likewise.
13080         (propagate_malloc): Likewise.
13081         (execute): Do not register hooks, just remove summary
13082         instead.
13083         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
13084         constructor.
13086 2018-06-08  Martin Liska  <mliska@suse.cz>
13088         * ipa-reference.c (remove_node_data): Remove.
13089         (duplicate_node_data): Likewise.
13090         (class ipa_ref_var_info_summary_t): New class.
13091         (class ipa_ref_opt_summary_t): Likewise.
13092         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
13093         (get_reference_optimization_summary): Use
13094         ipa_ref_opt_sum_summaries.
13095         (set_reference_vars_info): Remove.
13096         (set_reference_optimization_summary): Likewise.
13097         (ipa_init): Create summaries.
13098         (init_function_info): Use function summary.
13099         (ipa_ref_opt_summary_t::duplicate): New function.
13100         (ipa_ref_opt_summary_t::remove): New function.
13101         (get_read_write_all_from_node): Fix GNU coding style.
13102         (propagate): Use function summary.
13103         (write_node_summary_p): Fix GNU coding style.
13104         (stream_out_bitmap): Likewise.
13105         (ipa_reference_read_optimization_summary): Use function summary.
13106         (ipa_reference_c_finalize): Do not release hooks.
13108 2018-06-08  Martin Liska  <mliska@suse.cz>
13110         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
13111         (analyze_function_body): Extract multiple calls of get_create.
13112         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13113         * ipa-inline.c (recursive_inlining): Use ::get method.
13114         * ipa-inline.h (estimate_edge_growth): Likewise.
13116 2018-06-08  Martin Liska  <mliska@suse.cz>
13118         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13119         HSA_INVALID.
13120         (hsa_function_summary::hsa_function_summary): Use the new enum
13121         value.
13122         (hsa_gpu_implementation_p): Use hsa_summaries::get.
13123         * hsa-gen.c (hsa_get_host_function): Likewise.
13124         (get_brig_function_name): Likewise.
13125         * ipa-hsa.c (process_hsa_functions): Likewise.
13126         (ipa_hsa_write_summary): Likewise.
13127         * symbol-summary.h (symtab_duplication): Use ::get function/
13128         (get): New function.
13130 2018-06-08  Martin Liska  <mliska@suse.cz>
13132         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13133         of get.
13134         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13135         (hsa_register_kernel): Likewise.
13136         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13137         * hsa-gen.c (hsa_get_host_function): Likewise.
13138         (get_brig_function_name): Likewise.
13139         (generate_hsa): Likewise.
13140         (pass_gen_hsail::execute): Likewise.
13141         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13142         (devirtualization_time_bonus): Likewise.
13143         (ipcp_propagate_stage): Likewise.
13144         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13145         (edge_set_predicate): Likewise.
13146         (evaluate_conditions_for_known_args): Likewise.
13147         (evaluate_properties_for_edge): Likewise.
13148         (ipa_fn_summary::reset): Likewise.
13149         (ipa_fn_summary_t::duplicate): Likewise.
13150         (dump_ipa_call_summary): Likewise.
13151         (ipa_dump_fn_summary): Likewise.
13152         (analyze_function_body): Likewise.
13153         (compute_fn_summary): Likewise.
13154         (estimate_edge_devirt_benefit): Likewise.
13155         (estimate_edge_size_and_time): Likewise.
13156         (estimate_calls_size_and_time): Likewise.
13157         (estimate_node_size_and_time): Likewise.
13158         (inline_update_callee_summaries): Likewise.
13159         (remap_edge_change_prob): Likewise.
13160         (remap_edge_summaries): Likewise.
13161         (ipa_merge_fn_summary_after_inlining): Likewise.
13162         (ipa_update_overall_fn_summary): Likewise.
13163         (read_ipa_call_summary): Likewise.
13164         (inline_read_section): Likewise.
13165         (write_ipa_call_summary): Likewise.
13166         (ipa_fn_summary_write): Likewise.
13167         (ipa_free_fn_summary): Likewise.
13168         * ipa-hsa.c (process_hsa_functions): Likewise.
13169         (ipa_hsa_write_summary): Likewise.
13170         (ipa_hsa_read_section): Likewise.
13171         * ipa-icf.c (sem_function::merge): Likewise.
13172         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13173         (do_estimate_edge_time): Likewise.
13174         (estimate_size_after_inlining): Likewise.
13175         (estimate_growth): Likewise.
13176         (growth_likely_positive): Likewise.
13177         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13178         (inline_call): Likewise.
13179         * ipa-inline.c (caller_growth_limits): Likewise.
13180         (can_inline_edge_p): Likewise.
13181         (can_inline_edge_by_limits_p): Likewise.
13182         (compute_uninlined_call_time): Likewise.
13183         (compute_inlined_call_time): Likewise.
13184         (want_inline_small_function_p): Likewise.
13185         (edge_badness): Likewise.
13186         (update_caller_keys): Likewise.
13187         (update_callee_keys): Likewise.
13188         (recursive_inlining): Likewise.
13189         (inline_small_functions): Likewise.
13190         (inline_to_all_callers_1): Likewise.
13191         (dump_overall_stats): Likewise.
13192         (early_inline_small_functions): Likewise.
13193         (early_inliner): Likewise.
13194         * ipa-inline.h (estimate_edge_growth): Likewise.
13195         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13196         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13197         * ipa-prop.h (IPA_NODE_REF): Likewise.
13198         (IPA_EDGE_REF): Likewise.
13199         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13200         (propagate_malloc): Likewise.
13201         * ipa-split.c (execute_split_functions): Likewise.
13202         * symbol-summary.h: Rename get to get_create.
13203         (get): Likewise.
13204         (get_create): Likewise.
13205         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13207 2018-06-08  Martin Liska  <mliska@suse.cz>
13209         * symbol-summary.h (release): Move definition out of class
13210         declaration.
13211         (symtab_removal): Likewise.
13212         (symtab_duplication): Likewise.
13214 2018-06-08  Martin Liska  <mliska@suse.cz>
13216         * symbol-summary.h (function_summary): Move constructor
13217         implementation out of class declaration.
13218         (release): Likewise.
13219         (symtab_insertion): Likewise.
13220         (symtab_removal): Likewise.
13221         (symtab_duplication): Likewise.
13222         (get): Likewise.
13224 2018-06-08  Martin Liska  <mliska@suse.cz>
13226         * Makefile.in: Remove support for MPX (macros, related functions,
13227         fields in cgraph_node, ...).
13228         * builtin-types.def (BT_BND): Likewise.
13229         (BT_FN_BND_CONST_PTR): Likewise.
13230         (BT_FN_CONST_PTR_BND): Likewise.
13231         (BT_FN_VOID_PTR_BND): Likewise.
13232         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
13233         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
13234         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
13235         (expand_builtin_mempcpy_with_bounds): Likewise.
13236         (expand_builtin_memset_with_bounds): Likewise.
13237         (expand_builtin_memset_args): Likewise.
13238         (std_expand_builtin_va_start): Likewise.
13239         (expand_builtin): Likewise.
13240         (expand_builtin_with_bounds): Likewise.
13241         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
13242         (DEF_LIB_BUILTIN_CHKP): Likewise.
13243         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
13244         (DEF_CHKP_BUILTIN): Likewise.
13245         (BUILT_IN_MEMCPY): Likewise.
13246         (BUILT_IN_MEMMOVE): Likewise.
13247         (BUILT_IN_MEMPCPY): Likewise.
13248         (BUILT_IN_MEMSET): Likewise.
13249         (BUILT_IN_STPCPY): Likewise.
13250         (BUILT_IN_STRCAT): Likewise.
13251         (BUILT_IN_STRCHR): Likewise.
13252         (BUILT_IN_STRCPY): Likewise.
13253         (BUILT_IN_STRLEN): Likewise.
13254         (BUILT_IN_MEMCPY_CHK): Likewise.
13255         (BUILT_IN_MEMMOVE_CHK): Likewise.
13256         (BUILT_IN_MEMPCPY_CHK): Likewise.
13257         (BUILT_IN_MEMSET_CHK): Likewise.
13258         (BUILT_IN_STPCPY_CHK): Likewise.
13259         (BUILT_IN_STRCAT_CHK): Likewise.
13260         (BUILT_IN_STRCPY_CHK): Likewise.
13261         * calls.c (store_bounds): Likewise.
13262         (emit_call_1): Likewise.
13263         (special_function_p): Likewise.
13264         (maybe_warn_nonstring_arg): Likewise.
13265         (initialize_argument_information): Likewise.
13266         (finalize_must_preallocate): Likewise.
13267         (compute_argument_addresses): Likewise.
13268         (expand_call): Likewise.
13269         * cfgexpand.c (expand_call_stmt): Likewise.
13270         (expand_return): Likewise.
13271         (expand_gimple_stmt_1): Likewise.
13272         (pass_expand::execute): Likewise.
13273         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
13274         (cgraph_node::remove): Likewise.
13275         (cgraph_node::dump): Likewise.
13276         (cgraph_node::verify_node): Likewise.
13277         * cgraph.h (chkp_function_instrumented_p): Likewise.
13278         (symtab_node::get_alias_target): Likewise.
13279         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
13280         (cgraph_local_p): Likewise.
13281         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
13282         (cgraph_edge::rebuild_references): Likewise.
13283         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
13284         (walk_polymorphic_call_targets): Likewise.
13285         (cgraph_node::expand_thunk): Likewise.
13286         (symbol_table::output_weakrefs): Likewise.
13287         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
13288         (ix86_handle_option): Likewise.
13289         * config/i386/constraints.md: Likewise.
13290         * config/i386/i386-builtin-types.def (BND): Likewise.
13291         (VOID): Likewise.
13292         (PVOID): Likewise.
13293         (ULONG): Likewise.
13294         * config/i386/i386-builtin.def (BDESC_END): Likewise.
13295         (BDESC_FIRST): Likewise.
13296         (BDESC): Likewise.
13297         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
13298         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
13299         * config/i386/i386.c (enum reg_class): Likewise.
13300         (ix86_target_string): Likewise.
13301         (ix86_option_override_internal): Likewise.
13302         (ix86_conditional_register_usage): Likewise.
13303         (ix86_valid_target_attribute_inner_p): Likewise.
13304         (ix86_set_indirect_branch_type): Likewise.
13305         (ix86_set_current_function): Likewise.
13306         (ix86_function_arg_regno_p): Likewise.
13307         (init_cumulative_args): Likewise.
13308         (ix86_function_arg_advance): Likewise.
13309         (ix86_function_arg): Likewise.
13310         (ix86_pass_by_reference): Likewise.
13311         (ix86_function_value_regno_p): Likewise.
13312         (ix86_function_value_1): Likewise.
13313         (ix86_function_value_bounds): Likewise.
13314         (ix86_return_in_memory): Likewise.
13315         (ix86_setup_incoming_vararg_bounds): Likewise.
13316         (ix86_va_start): Likewise.
13317         (indirect_thunk_need_prefix): Likewise.
13318         (print_reg): Likewise.
13319         (ix86_print_operand): Likewise.
13320         (ix86_expand_call): Likewise.
13321         (ix86_output_function_return): Likewise.
13322         (reg_encoded_number): Likewise.
13323         (BDESC_VERIFYS): Likewise.
13324         (ix86_init_mpx_builtins): Likewise.
13325         (ix86_init_builtins): Likewise.
13326         (ix86_emit_cmove): Likewise.
13327         (ix86_emit_move_max): Likewise.
13328         (ix86_expand_builtin): Likewise.
13329         (ix86_builtin_mpx_function): Likewise.
13330         (ix86_get_arg_address_for_bt): Likewise.
13331         (ix86_load_bounds): Likewise.
13332         (ix86_store_bounds): Likewise.
13333         (ix86_load_returned_bounds): Likewise.
13334         (ix86_store_returned_bounds): Likewise.
13335         (ix86_class_likely_spilled_p): Likewise.
13336         (ix86_hard_regno_mode_ok): Likewise.
13337         (x86_order_regs_for_local_alloc): Likewise.
13338         (ix86_mitigate_rop): Likewise.
13339         (ix86_bnd_prefixed_insn_p): Likewise.
13340         (ix86_mpx_bound_mode): Likewise.
13341         (ix86_make_bounds_constant): Likewise.
13342         (ix86_initialize_bounds): Likewise.
13343         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
13344         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
13345         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
13346         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
13347         (TARGET_CHKP_BOUND_MODE): Likewise.
13348         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
13349         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
13350         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
13351         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
13352         * config/i386/i386.h (TARGET_MPX): Likewise.
13353         (TARGET_MPX_P): Likewise.
13354         (VALID_BND_REG_MODE): Likewise.
13355         (FIRST_BND_REG): Likewise.
13356         (LAST_BND_REG): Likewise.
13357         (enum reg_class): Likewise.
13358         (BND_REG_P): Likewise.
13359         (BND_REGNO_P): Likewise.
13360         (BNDmode): Likewise.
13361         (ADJUST_INSN_LENGTH): Likewise.
13362         * config/i386/i386.md: Likewise.
13363         * config/i386/i386.opt: Likewise.
13364         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
13365         (defined): Likewise.
13366         (LINK_MPX): Likewise.
13367         (MPX_SPEC): Likewise.
13368         (LIBMPX_SPEC): Likewise.
13369         (LIBMPXWRAPPERS_SPEC): Likewise.
13370         (CHKP_SPEC): Likewise.
13371         * config/i386/predicates.md: Likewise.
13372         * dbxout.c (dbxout_type): Likewise.
13373         * doc/extend.texi: Likewise.
13374         * doc/invoke.texi: Likewise.
13375         * doc/md.texi: Likewise.
13376         * doc/tm.texi: Likewise.
13377         * doc/tm.texi.in: Likewise.
13378         * dwarf2out.c (is_base_type): Likewise.
13379         (gen_formal_types_die): Likewise.
13380         (gen_subprogram_die): Likewise.
13381         (gen_type_die_with_usage): Likewise.
13382         (gen_decl_die): Likewise.
13383         (dwarf2out_late_global_decl): Likewise.
13384         * expr.c (expand_assignment): Likewise.
13385         (emit_storent_insn): Likewise.
13386         (store_expr_with_bounds): Likewise.
13387         (store_expr): Likewise.
13388         (expand_expr_real_1): Likewise.
13389         * expr.h (store_expr_with_bounds): Likewise.
13390         * function.c (use_register_for_decl): Likewise.
13391         (struct bounds_parm_data): Likewise.
13392         (assign_parms_augmented_arg_list): Likewise.
13393         (assign_parm_find_entry_rtl): Likewise.
13394         (assign_parm_is_stack_parm): Likewise.
13395         (assign_parm_load_bounds): Likewise.
13396         (assign_bounds): Likewise.
13397         (assign_parms): Likewise.
13398         (expand_function_start): Likewise.
13399         * gcc.c (CHKP_SPEC): Likewise.
13400         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
13401         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
13402         (wrestrict_dom_walker::check_call): Likewise.
13403         * gimple.c (gimple_build_call_from_tree): Likewise.
13404         * gimple.h (enum gf_mask): Likewise.
13405         (gimple_call_with_bounds_p): Likewise.
13406         (gimple_call_set_with_bounds): Likewise.
13407         * gimplify.c (gimplify_init_constructor): Likewise.
13408         * ipa-cp.c (initialize_node_lattices): Likewise.
13409         (propagate_constants_across_call): Likewise.
13410         (find_more_scalar_values_for_callers_subset): Likewise.
13411         * ipa-hsa.c (process_hsa_functions): Likewise.
13412         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
13413         * ipa-icf.c (sem_function::merge): Likewise.
13414         * ipa-inline.c (early_inliner): Likewise.
13415         * ipa-pure-const.c (warn_function_noreturn): Likewise.
13416         (warn_function_cold): Likewise.
13417         (propagate_pure_const): Likewise.
13418         * ipa-ref.h (enum GTY): Likewise.
13419         * ipa-split.c (find_retbnd): Likewise.
13420         (consider_split): Likewise.
13421         (split_function): Likewise.
13422         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
13423         * ipa.c (walk_polymorphic_call_targets): Likewise.
13424         (symbol_table::remove_unreachable_nodes): Likewise.
13425         (process_references): Likewise.
13426         (cgraph_build_static_cdtor_1): Likewise.
13427         * lto-cgraph.c (lto_output_node): Likewise.
13428         (output_refs): Likewise.
13429         (compute_ltrans_boundary): Likewise.
13430         (input_overwrite_node): Likewise.
13431         (input_node): Likewise.
13432         (input_cgraph_1): Likewise.
13433         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
13434         * passes.c (pass_manager::execute_early_local_passes): Likewise.
13435         (class pass_chkp_instrumentation_passes): Likewise.
13436         (make_pass_chkp_instrumentation_passes): Likewise.
13437         * passes.def: Likewise.
13438         * rtl.h (struct GTY): Likewise.
13439         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
13440         * stor-layout.c (layout_type): Likewise.
13441         * symtab.c: Likewise.
13442         * target.def: Likewise.
13443         * targhooks.c (default_chkp_bound_type): Likewise.
13444         (default_chkp_bound_mode): Likewise.
13445         (default_builtin_chkp_function): Likewise.
13446         (default_chkp_function_value_bounds): Likewise.
13447         (default_chkp_make_bounds_constant): Likewise.
13448         (default_chkp_initialize_bounds): Likewise.
13449         * targhooks.h (default_chkp_bound_type): Likewise.
13450         (default_chkp_bound_mode): Likewise.
13451         (default_builtin_chkp_function): Likewise.
13452         (default_chkp_function_value_bounds): Likewise.
13453         (default_chkp_make_bounds_constant): Likewise.
13454         (default_chkp_initialize_bounds): Likewise.
13455         * toplev.c (compile_file): Likewise.
13456         (process_options): Likewise.
13457         * tree-core.h (DEF_BUILTIN): Likewise.
13458         (DEF_BUILTIN_CHKP): Likewise.
13459         * tree-inline.c (declare_return_variable): Likewise.
13460         (remap_gimple_stmt): Likewise.
13461         (copy_bb): Likewise.
13462         (initialize_inlined_parameters): Likewise.
13463         (expand_call_inline): Likewise.
13464         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
13465         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
13466         (make_pass_ipa_chkp_produce_thunks): Likewise.
13467         (make_pass_chkp): Likewise.
13468         (make_pass_chkp_opt): Likewise.
13469         (make_pass_chkp_instrumentation_passes): Likewise.
13470         * tree-pretty-print.c (dump_generic_node): Likewise.
13471         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
13472         * tree-ssa-dce.c (propagate_necessity): Likewise.
13473         (eliminate_unnecessary_stmts): Likewise.
13474         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
13475         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
13476         * tree-ssa-sccvn.h: Likewise.
13477         * tree-ssa-strlen.c (get_string_length): Likewise.
13478         (valid_builtin_call): Likewise.
13479         (adjust_last_stmt): Likewise.
13480         (handle_builtin_strchr): Likewise.
13481         (handle_builtin_strcpy): Likewise.
13482         (handle_builtin_stxncpy): Likewise.
13483         (handle_builtin_memcpy): Likewise.
13484         (handle_builtin_strcat): Likewise.
13485         (strlen_check_and_optimize_stmt): Likewise.
13486         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
13487         * tree-streamer-in.c: Likewise.
13488         * tree-streamer.c (record_common_node): Likewise.
13489         * tree.c (tree_code_size): Likewise.
13490         (wide_int_to_tree_1): Likewise.
13491         (type_contains_placeholder_1): Likewise.
13492         (build_common_tree_nodes): Likewise.
13493         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
13494         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
13495         (POINTER_BOUNDS_P): Likewise.
13496         (BOUNDED_TYPE_P): Likewise.
13497         (BOUNDED_P): Likewise.
13498         (CALL_WITH_BOUNDS_P): Likewise.
13499         (pointer_bounds_type_node): Likewise.
13500         * value-prof.c (gimple_ic): Likewise.
13501         * var-tracking.c (vt_add_function_parameters): Likewise.
13502         * varasm.c (make_decl_rtl): Likewise.
13503         (assemble_start_function): Likewise.
13504         (output_constant): Likewise.
13505         (maybe_assemble_visibility): Likewise.
13506         * varpool.c (ctor_for_folding): Likewise.
13507         * chkp-builtins.def: Remove.
13508         * ipa-chkp.c: Remove.
13509         * ipa-chkp.h: Remove.
13510         * rtl-chkp.c: Remove.
13511         * rtl-chkp.h: Remove.
13512         * tree-chkp-opt.c: Remove.
13513         * tree-chkp.c: Remove.
13514         * tree-chkp.h: Remove.
13516 2018-06-07  Carl Love  <cel@us.ibm.com>
13518         * config/rs6000/vsx.md (vextract_fp_from_shorth,
13519         vextract_fp_from_shortl): Add BE support.
13521 2018-06-07  Paul Koning  <ni1d@arrl.net>
13523         * compare-elim.c (try_merge_compare): Don't merge compare if
13524         address contains a side effect.
13525         (try_eliminate_compare): Likewise.
13527 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
13529         * config.gcc: Support "tremont".
13530         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
13531         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13532         PROCESSOR_TREMONT.
13533         * config/i386/i386.c (m_TREMONT): Define.
13534         (processor_target_table): Add "tremont".
13535         (PTA_TREMONT): Define.
13536         (ix86_lea_outperforms): Add TARGET_TREMONT.
13537         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
13538         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
13539         and M_INTEL_GOLDMONT_PLUS.
13540         (fold_builtin_cpu): Add "tremont".
13541         (ix86_add_stmt_cost): Add TARGET_TREMONT.
13542         (ix86_option_override_internal): Add "tremont".
13543         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
13544         (processor_type): Add PROCESSOR_TREMONT.
13545         * config/i386/x86-tune.def: Add m_TREMONT.
13546         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
13548 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13550         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
13551         symbol defined for msp430i* devices to be lower case.
13553 2018-06-07  Richard Biener  <rguenther@suse.de>
13555         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
13556         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
13557         Properly wrap signed arithmetic if overflow wraps.
13559 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
13561         PR tree-optimization/69615
13562         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
13563         of a cast from a same precision integral SSA_NAME in a bb dominated
13564         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
13565         cast to utype if rhs2 has already a compatible type.
13567 2018-06-07  Richard Biener  <rguenther@suse.de>
13569         PR tree-optimization/85935
13570         * graphite-scop-detection.c (find_params_in_bb): Analyze
13571         condition operands with respect to the correct loop.  Assert
13572         the analysis doesn't fail.
13574 2018-06-04  Carl Love  <cel@us.ibm.com>
13576         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
13577         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
13578         as it is slightly cheaper.
13579         (first_match_or_eos_index_<mode>):
13580         Calculate index using natural element order.
13581         (first_match_index_<mode>):
13582         Calculate index using natural element order.
13583         (first_match_or_eos_index_<mode>):
13584         Calculate index using natural order.
13585         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
13586         for BE and LE modes.
13587         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
13588         P9V_BUILTIN_VCLZLSBB_V16QI.
13589         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
13590         specific.
13592 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13594         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
13595         indentation and line wrap for many prototypes.  Add missing
13596         @smallexample directives around block of prototypes for vec_xl and
13597         vec_xst.
13599 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
13601         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
13602         track if we pass or return IEEE 128-bit floating point.
13603         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
13604         C++ mangling that is compatible with GCC 8.1.
13605         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
13606         (init_cumulative_args): Note if we pass or return IEEE 128-bit
13607         floating point types.
13608         (rs6000_function_arg_advance_1): Likewise.
13609         (rs6000_mangle_type): Optionally generate mangled names that match
13610         what GCC 8.1 generated for IEEE 128-bit floating point types.
13611         (rs6000_globalize_decl_name): If we have an external function that
13612         passes or returns IEEE 128-bit types, generate a weak reference
13613         from the mangled name used in GCC 8.1 to the current mangled
13614         name.
13615         (rs6000_init_builtins): Make __ibm128 use the long double type if
13616         long double is IBM extended double.  Make __float128 use the long
13617         double type if long double is IEEE 128-bit.
13619         PR target/85657
13620         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
13621         macro for __ibm128 built-in functions.
13622         (PACK_IF): Add __ibm128 pack/unpack functions.
13623         (UNPACK_IF): Likewise.
13624         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
13625         enable long double built-in functions if long double is IEEE
13626         128-bit floating point.
13627         (rs6000_invalid_builtin): Update long double built-in function
13628         error message.
13629         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
13630         functions, adjust the built-in function to use the long double
13631         built-in function if __ibm128 and long double are the same type.
13632         * doc/extend.texi (PowerPC builtins): Update documention for
13633         __builtin_{,un}pack_longdouble.  Add documentation for
13634         __builtin_{,un}pack_ibm128.
13636 2018-06-06  Jim Wilson  <jimw@sifive.com>
13638         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
13639         (struct machine_function): New field interrupt_mode.
13640         (riscv_handle_type_attribute): New function.  Add forward declaration.
13641         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
13642         (riscv_expand_epilogue): Check interrupt_mode field.
13643         (riscv_set_current_function): Check interrupt attribute args and
13644         set interrupt_mode field.
13645         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
13646         (riscv_sret, riscv_uret): New.
13647         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
13648         new arguments to interrupt attribute.
13650 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
13652         PR target/63177
13653         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
13654         Don't handle -mcpu=power8 if -mpower9-vector is also used.
13656 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13658         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
13659         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
13660         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13661         several redundant entries.
13663 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
13665         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
13666         type from "rtx" to "rtx_insn *".
13667         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
13668         for local "call_insn", removing cast.
13669         (ix86_expand_call): Likewise, introducing a "call_insn" local.
13671 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
13673         PR tree-optimization/86066
13674         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
13675         for BIT_INSERT_EXPR stores.
13677 2018-06-06  Richard Biener  <rguenther@suse.de>
13679         PR tree-optimization/86062
13680         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13681         component refs ontop
13682         of to be offsetted base.
13684 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13686         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
13687         to be static and remove check on interrupt attribute name.
13689 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13691         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13692         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
13694 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
13696         PR target/79924
13697         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
13698         second argument.
13699         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
13700         Remove second argument, change how error is called.
13701         (aarch64_layout_arg): Remove second argument from
13702         aarch64_err_no_fpadvsimd call.
13703         (aarch64_init_cumulative_args): Ditto.
13704         (aarch64_gimplify_va_arg_expr): Ditto.
13705         * config/aarch64/aarch64.md (mov<mode>): Ditto.
13707 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
13709         * config/i386/i386.md (simple_return_indirect_internal): New expander.
13710         (*simple_return_indirect_internal<mode>): Rename from
13711         simple_return_indirect_internal.  Use W mode iterator.
13712         (rstorssp): New expander.
13713         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
13714         (clrssbsy): New expander.
13715         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
13717 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13719         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
13720         __typeof__.
13721         (cmse_check_pointed_object): Likewise.
13723 2018-06-05  Martin Liska  <mliska@suse.cz>
13725         PR gcov-profile/47618
13726         * doc/invoke.texi: Document how -fprofile-dir format
13727         is extended.
13729 2018-06-05  Richard Biener  <rguenther@suse.de>
13731         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
13732         removal pretend DOM info isn't available so we do not update
13733         it and only remove edges, not dominated blocks.  Actually free
13734         DOM info in case we removed something.  Remove unreachable blocks.
13735         (mfb_keep_latches): Work with either DOM info or marked backedges.
13736         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
13737         first.  Mark backedges if DOM info isn't available.
13738         (Re-)compute DOM info after cleanup_control_flow_pre.
13740 2018-06-05  Richard Biener  <rguenther@suse.de>
13742         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
13743         (locus_discrim_hasher::hash): Adjust.
13744         (locus_discrim_hasher::equal): Likewise.
13745         (next_discriminator_for_locus): Work on line directly.
13746         (same_line_p): Pass in expanded locus1 as well.
13747         (assign_discriminators): Avoid redundant location expansions.
13749 2018-06-05  Richard Biener  <rguenther@suse.de>
13751         PR tree-optimization/86046
13752         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
13753         if required after clearing TREE_ADDRESSABLE.
13755 2018-06-05  Richard Biener  <rguenther@suse.de>
13757         PR tree-optimization/86047
13758         * tree-ssa-loop.c (for_each_index): Glob handling of all
13759         decls and constants and really handle all of them.
13761 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13763         PR target/81497
13764         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
13765         qualifier_void_pointer and qualifier_const_void_pointer.
13766         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
13767         (arm_init_builtins): Handle the above.
13768         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
13769         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
13770         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
13771         void intrinsics.
13773 2018-06-05  Martin Liska  <mliska@suse.cz>
13775         * auto-profile.c (read_autofdo_file): Do not use
13776         gcov_ctr_summary struct.
13777         (afdo_callsite_hot_enough_for_early_inline): Likewise.
13778         * coverage.c (struct counts_entry): Likewise.
13779         (read_counts_file): Read just single summary entry.
13780         (get_coverage_counts): Use gcov_summary struct.
13781         * coverage.h (get_coverage_counts): Likewise.
13782         * gcov-dump.c (dump_working_sets): Likewise.
13783         (tag_summary): Dump just single summary.
13784         * gcov-io.c (gcov_write_summary): Write just histogram
13785         summary.
13786         (gcov_read_summary): Read just single summary.
13787         (compute_working_sets): Use gcov_summary struct.
13788         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
13789         of GCOV_COUNTERS_SUMMABLE.
13790         (GCOV_COUNTERS_SUMMABLE): Remove.
13791         (GCOV_FIRST_VALUE_COUNTER): Replace with
13792         GCOV_COUNTER_V_INTERVAL.
13793         (struct gcov_ctr_summary): Remove.
13794         (struct gcov_summary): Directly use fields of former
13795         gcov_ctr_summary.
13796         (compute_working_sets): Use gcov_summary struct.
13797         * gcov.c (read_count_file): Do not use ctrs fields.
13798         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
13799         struct.
13800         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
13801         struct.
13802         * profile.c: Likewise.
13803         * profile.h: Likewise.
13805 2018-06-05  Martin Liska  <mliska@suse.cz>
13807         PR gcov-profile/84846
13808         * gcov.c (output_lines): Print working directory only
13809         in intermediate format.
13811 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
13813         * config/s390/s390-builtin-types.def: Add void function type.
13814         * config/s390/s390-builtins.def: Use the function type for the
13815         tbeginc builtin.
13817 2018-06-04  Jim Wilson  <jimw@sifive.com>
13819         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
13820         to int.
13821         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
13822         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
13823         handle EH_RETURN_DATA_REGNO registers properly.
13824         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
13825         (riscv_expand_epilogue): Update comment.  Change argument name and
13826         type.  Update code to use new name and type.  Pass new args to
13827         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
13828         EXCEPTION_RETURN.
13829         * config/riscv/riscv.md (NORMAL_RETURN): New.
13830         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
13831         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
13832         (eh_return): Call gen_eh_return_internal and emit barrier.
13833         (eh_return_internal): Call riscv_expand_epilogue.
13835 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
13837         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
13838         bit_insertion field and declare can_be_merged_into method.
13839         (merged_store_group::can_be_merged_into): New method.
13840         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
13841         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
13842         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
13844 2018-06-04  Richard Biener  <rguenther@suse.de>
13846         PR tree-optimization/85955
13847         * builtins.c (fold_builtin_sincos): Convert pointers to
13848         destination to appropriate type before dereferencing.
13850 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
13852         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
13854 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
13856         * expr.c (expand_expr_real_1): Force the operand into memory if
13857         its TYPE_MODE is BLKmode and if there is no integer mode for
13858         the number of bits being extracted.
13860 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
13862         PR target/85832
13863         PR target/86036
13864         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13865         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
13867 2018-06-04  Richard Biener  <rguenther@suse.de>
13869         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
13870         (cleanup_tree_cfg_noloop): ... single caller.  Do
13871         start_recording_case_labels later.
13873 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
13875         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
13876         to _IMMINTRIN_H_INCLUDED.
13877         * config/i386/pconfigintrin.h: Ditto.
13878         * config/i386/waitpkgintrin.h: Ditto.
13879         * config/i386/immintrin.h: Add includes for sgxintrin.h,
13880         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
13881         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
13882         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
13883         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
13884         waitpkgintrin.h and cldemoteintrin.h.
13886 2018-06-04  Richard Biener  <rguenther@suse.de>
13888         PR tree-optimization/86038
13889         * tracer.c (find_best_successor): Check probability for
13890         being initialized, bail out if not.
13892 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
13894         PR target/86003
13895         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
13896         of bits to ignore when comparing architectures.
13898 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
13900         PR tree-optimization/69615
13901         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
13902         maximum or minimum of the type, try to merge it also as if
13903         range1 is + [-, x - 1] or + [x + 1, -].
13905         PR c++/86025
13906         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
13908 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
13910         PR tree-optimization/86034
13911         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
13912         the unsigned bitfield type in a bit insertion sequence if it does not
13913         have a larger precision than the bitfield size.
13914         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
13916 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
13918         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
13920 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
13922         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
13923         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13924         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13925         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13927 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
13929         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13930         Disable -fdelete-null-pointer-checks for ELF toolchain.
13932 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13933             Kito Cheng  <kito.cheng@gmail.com>
13935         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13936         (nds32le-*-*, nds32be-*-*): Integrate checking process.
13937         (nds32*-*-*): Add glibc and uclibc conditions.
13938         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13939         (TARGET_EXCEPT_UNWIND_INFO): Define.
13940         * config/nds32/elf.h: New file.
13941         * config/nds32/linux.h: New file.
13942         * config/nds32/nds32-elf.opt: New file.
13943         * config/nds32/nds32-linux.opt: New file.
13944         * config/nds32/nds32-fp-as-gp.c
13945         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13946         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13947         TARGET_LINUX_ABI.
13948         (nds32_asm_file_end): Ditto.
13949         (nds32_print_operand): Ditto.
13950         (nds32_insert_attributes): Ditto.
13951         (nds32_init_libfuncs): New function.
13952         (TARGET_HAVE_TLS): Define.
13953         (TARGET_INIT_LIBFUNCS): Define.
13954         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13955         spec content.
13956         (TARGET_ELF): Apply different mcmodel setting.
13957         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13958         been migrated into elf.h and linux.h files.
13959         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13960         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13961         (mcmodel): The content has been migrated into nds32-elf.opt and
13962         nds32-linux.opt files.
13963         * config/nds32/t-elf: New file.
13964         * config/nds32/t-linux: New file.
13966 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13967             Shiva Chen  <shiva0217@gmail.com>
13969         * config/nds32/constants.md (unspec_volatile_element): Add
13970         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13971         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13972         optimization.
13973         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13974         (make_pass_nds32_fp_as_gp): Declare.
13975         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13976         optmization pass.
13977         (nds32_asm_function_end_prologue): Remove unused asm output.
13978         (nds32_asm_function_begin_epilogue): Remove unused asm output.
13979         (nds32_asm_file_start): Output necessary fp_as_gp information.
13980         (nds32_option_override): Adjust register usage.
13981         (nds32_expand_prologue): Consider fp_as_gp situation.
13982         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13983         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13984         (epilogue): Ditto.
13985         (return): Ditto.
13986         (simple_return): Ditto.
13987         (omit_fp_begin): Output special directive for fp_as_gp.
13988         (omit_fp_end): Output special directive for fp_as_gp.
13989         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13990         mforbid-fp-as-gp): New options.
13992 2018-06-01  Mark Wielaard  <mark@klomp.org>
13994         * dwarf2out.c (dwarf2out_finish): Remove generation of
13995         DW_AT_loclists_base.
13997 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
13999         * gimple-ssa-store-merging.c: Include gimple-fold.h.
14000         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
14001         (struct merged_store_group): Add bit_insertion field.
14002         (dump_char_array): Use standard hexadecimal format.
14003         (merged_store_group::merged_store_group): Set bit_insertion to false.
14004         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
14005         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
14006         also print the mask in the dump file.
14007         (pass_store_merging::gate): Minor tweak.
14008         (imm_store_chain_info::coalesce_immediate): Fix wrong association
14009         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
14010         stores with INTEGER_CST stores.
14011         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
14012         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
14013         and use it throughout.  Generate bit insertion sequences if need be.
14014         (pass_store_merging::process_store): Remove redundant condition.
14015         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
14017 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
14019         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
14020         the 128-bit floating point types.  Fix function comment.
14022 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14024         * config/aarch64/aarch64-simd.md
14025         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
14026         mnemonics.
14027         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
14028         mnemonics.
14030 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14032         PR tree-optimization/85989
14033         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
14034         variable.
14035         (backprop::intersect_uses): Check it when deciding whether this
14036         is a backedge reference.
14037         (backprop::process_block): Add each phi to m_visited_phis
14038         after visiting it, then clear it at the end.
14040 2018-06-01  Richard Biener  <rguenther@suse.de>
14042         * tree-vectorizer.h (vect_dr_stmt): New function.
14043         (vect_get_load_cost): Adjust.
14044         (vect_get_store_cost): Likewise.
14045         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
14046         Use vect_dr_stmt instead of DR_SMTT.
14047         (vect_record_base_alignments): Likewise.
14048         (vect_calculate_target_alignment): Likewise.
14049         (vect_compute_data_ref_alignment): Likewise and make static.
14050         (vect_update_misalignment_for_peel): Likewise.
14051         (vect_verify_datarefs_alignment): Likewise.
14052         (vector_alignment_reachable_p): Likewise.
14053         (vect_get_data_access_cost): Likewise.  Pass down
14054         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
14055         (vect_get_peeling_costs_all_drs): Likewise.
14056         (vect_peeling_hash_get_lowest_cost): Likewise.
14057         (vect_enhance_data_refs_alignment): Likewise.
14058         (vect_find_same_alignment_drs): Likewise.
14059         (vect_analyze_data_refs_alignment): Likewise.
14060         (vect_analyze_group_access_1): Likewise.
14061         (vect_analyze_group_access): Likewise.
14062         (vect_analyze_data_ref_access): Likewise.
14063         (vect_analyze_data_ref_accesses): Likewise.
14064         (vect_vfa_segment_size): Likewise.
14065         (vect_small_gap_p): Likewise.
14066         (vectorizable_with_step_bound_p): Likewise.
14067         (vect_prune_runtime_alias_test_list): Likewise.
14068         (vect_analyze_data_refs): Likewise.
14069         (vect_supportable_dr_alignment): Likewise.
14070         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14071         (vect_gen_prolog_loop_niters): Likewise.
14072         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14073         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
14074         modify DR_STMT.
14075         (vect_recog_mask_conversion_pattern): Likewise.
14076         (vect_try_gather_scatter_pattern): Likewise.
14077         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
14078         to vect_get_store_cost.
14079         (vect_get_store_cost): Get stmt_info instead of DR.
14080         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
14081         (vect_get_load_cost): Get stmt_info instead of DR.
14083 2018-06-01  Richard Biener  <rguenther@suse.de>
14085         PR middle-end/86017
14086         * gimple-fold.c (var_decl_component_p): Also allow offsetted
14087         vars wrapped in MEM_REFs.
14089 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14091         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14092         Fix subreg tests so that we only return a choice between
14093         GENERAL_REGS and FP_REGS if the original classes included both.
14095 2018-06-01  Richard Biener  <rguenther@suse.de>
14097         PR ipa/85960
14098         * tree-ssa-structalias.c (get_function_part_constraint):
14099         Handle NULL fi->decl.
14100         (find_func_aliases_for_call): Properly handle indirect
14101         fi from direct call.
14102         (find_func_clobbers): Likewise.
14103         (ipa_pta_execute): Likewise.
14104         (create_variable_info_for): For functions that are ifunc_resolver
14105         resolve to a varinfo that contains the result of the resolver call.
14106         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
14107         aliases.
14109 2018-05-31  Michael Collison  <michael.collison@arm.com>
14111         * config/aarch64/aarch64.md:
14112         (*fix_to_zero_extenddfdi2): New pattern.
14113         * gcc.target/aarch64/fix_extend1.c: New testcase.
14115 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
14117         PR middle-end/78809
14118         PR middle-end/83026
14119         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14120         and BUILT_IN_STRNCMP_EQ.
14121         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14122         BUILT_IN_STRNCMP_EQ.
14123         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14124         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14125         (gimple_fold_builtin): Likewise.
14126         * tree-ssa-strlen.c (compute_string_length): New function.
14127         (determine_min_obsize): New function.
14128         (handle_builtin_string_cmp): New function to handle calls to
14129         string compare functions.
14130         (strlen_optimize_stmt): Add handling to builtin string compare
14131         calls.
14132         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14133         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14134         * tree.c (build_common_builtin_nodes): Add new defines of
14135         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14137 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
14139         PR target/85984
14140         * bb-reorder.c (pass_partition_blocks::gate): Return false for
14141         functions with naked attribute.
14143 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
14145         * config/i386/sse.md (avx_vec_concat<mode>):
14146         Substitute concat_tg_mode mode attribute with xtg_mode.
14147         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14148         (concat_tg_mode): Remove mode attribute.
14150 2018-05-31  Martin Sebor  <msebor@redhat.com>
14152         PR c/82063
14153         * calls.c (alloc_max_size): Correct a logic error/typo.
14154         Treat excessive arguments as infinite.  Warn for invalid arguments.
14155         * doc/invoke.texi (-Walloc-size-larger-than): Update.
14157 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
14159         PR target/85829
14160         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14161         and movx for Haswell.
14163 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
14164             Cesar Philippidis  <cesar@codesourcery.com>
14166         PR middle-end/85879
14167         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
14168         when emitting error on private/firstprivate reductions.
14169         * omp-low.c (lower_omp_target): Avoid reference-type processing
14170         on pointers for firstprivate clause.
14172 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14174         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14175         (st1x2): Likewise.
14176         (st1x3): Likewise.
14177         * config/aarch64/aarch64-simd.md
14178         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14179         (aarch64_ld1_x3_<mode>): Likewise
14180         (aarch64_st1x2<VALLDIF:mode>): Likewise
14181         (aarch64_st1_x2_<mode>): Likewise
14182         (aarch64_st1x3<VALLDIF:mode>): Likewise
14183         (aarch64_st1_x3_<mode>): Likewise
14184         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14185         (vld1_s8_x3): Likewise.
14186         (vld1_u16_x3): Likewise.
14187         (vld1_s16_x3): Likewise.
14188         (vld1_u32_x3): Likewise.
14189         (vld1_s32_x3): Likewise.
14190         (vld1_u64_x3): Likewise.
14191         (vld1_s64_x3): Likewise.
14192         (vld1_f16_x3): Likewise.
14193         (vld1_f32_x3): Likewise.
14194         (vld1_f64_x3): Likewise.
14195         (vld1_p8_x3): Likewise.
14196         (vld1_p16_x3): Likewise.
14197         (vld1_p64_x3): Likewise.
14198         (vld1q_u8_x3): Likewise.
14199         (vld1q_s8_x3): Likewise.
14200         (vld1q_u16_x3): Likewise.
14201         (vld1q_s16_x3): Likewise.
14202         (vld1q_u32_x3): Likewise.
14203         (vld1q_s32_x3): Likewise.
14204         (vld1q_u64_x3): Likewise.
14205         (vld1q_s64_x3): Likewise.
14206         (vld1q_f16_x3): Likewise.
14207         (vld1q_f32_x3): Likewise.
14208         (vld1q_f64_x3): Likewise.
14209         (vld1q_p8_x3): Likewise.
14210         (vld1q_p16_x3): Likewise.
14211         (vld1q_p64_x3): Likewise.
14212         (vst1_s64_x2): Likewise.
14213         (vst1_u64_x2): Likewise.
14214         (vst1_f64_x2): Likewise.
14215         (vst1_s8_x2): Likewise.
14216         (vst1_p8_x2): Likewise.
14217         (vst1_s16_x2): Likewise.
14218         (vst1_p16_x2): Likewise.
14219         (vst1_s32_x2): Likewise.
14220         (vst1_u8_x2): Likewise.
14221         (vst1_u16_x2): Likewise.
14222         (vst1_u32_x2): Likewise.
14223         (vst1_f16_x2): Likewise.
14224         (vst1_f32_x2): Likewise.
14225         (vst1_p64_x2): Likewise.
14226         (vst1q_s8_x2): Likewise.
14227         (vst1q_p8_x2): Likewise.
14228         (vst1q_s16_x2): Likewise.
14229         (vst1q_p16_x2): Likewise.
14230         (vst1q_s32_x2): Likewise.
14231         (vst1q_s64_x2): Likewise.
14232         (vst1q_u8_x2): Likewise.
14233         (vst1q_u16_x2): Likewise.
14234         (vst1q_u32_x2): Likewise.
14235         (vst1q_u64_x2): Likewise.
14236         (vst1q_f16_x2): Likewise.
14237         (vst1q_f32_x2): Likewise.
14238         (vst1q_f64_x2): Likewise.
14239         (vst1q_p64_x2): Likewise.
14240         (vst1_s64_x3): Likewise.
14241         (vst1_u64_x3): Likewise.
14242         (vst1_f64_x3): Likewise.
14243         (vst1_s8_x3): Likewise.
14244         (vst1_p8_x3): Likewise.
14245         (vst1_s16_x3): Likewise.
14246         (vst1_p16_x3): Likewise.
14247         (vst1_s32_x3): Likewise.
14248         (vst1_u8_x3): Likewise.
14249         (vst1_u16_x3): Likewise.
14250         (vst1_u32_x3): Likewise.
14251         (vst1_f16_x3): Likewise.
14252         (vst1_f32_x3): Likewise.
14253         (vst1_p64_x3): Likewise.
14254         (vst1q_s8_x3): Likewise.
14255         (vst1q_p8_x3): Likewise.
14256         (vst1q_s16_x3): Likewise.
14257         (vst1q_p16_x3): Likewise.
14258         (vst1q_s32_x3): Likewise.
14259         (vst1q_s64_x3): Likewise.
14260         (vst1q_u8_x3): Likewise.
14261         (vst1q_u16_x3): Likewise.
14262         (vst1q_u32_x3): Likewise.
14263         (vst1q_u64_x3): Likewise.
14264         (vst1q_f16_x3): Likewise.
14265         (vst1q_f32_x3): Likewise.
14266         (vst1q_f64_x3): Likewise.
14267         (vst1q_p64_x3): Likewise.
14269 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14271         * config/msp430/msp430.c (msp430_output_labelref): Prepend
14272         user_label_prefix to name.
14274         * tree-core.h: Update comment about the format of NAME string
14275         passed to handler in attribute_spec.
14277         * config/msp430/msp430.md: Remove erroneous subreg expression from
14278         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
14279         zero_extend{q,h}isi2.
14281 2018-05-30  Borislav Petkov  <bp@suse.de>
14283         * doc/extend.texi: Document some architecture specific
14284         constraints and sort entries.
14286 2018-05-30  Martin Sebor  <msebor@redhat.com>
14288         PR middle-end/85369
14289         * builtins.c (expand_builtin_stpcpy_1): New function.
14290         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
14291         only if the former succeeds.
14293 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14295         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
14296         in saphira.
14298 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14300         * doc/invoke.texi (-flinker-output): Document
14302 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14304         * passes.c (ipa_write_summaries): Only modify statements if body
14305         is in memory.
14306         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
14307         incrementally linking.
14308         (ipa_passes): Likewise.
14309         * lto-cgraph.c (lto_output_node): When incrementally linking do not
14310         pass down resolution info.
14311         * common.opt (flag_incremental_link): Update info.
14312         * gcc.c (plugin specs): Turn flinker-output=* to
14313         -plugin-opt=-linker-output-known
14314         * toplev.c (compile_file): Also cut compilation when doing incremental
14315         link.
14316         * flag-types. (enum lto_partition_model): Add
14317         LTO_LINKER_OUTPUT_NOLTOREL.
14318         (invoke.texi): Add -flinker-output docs.
14319         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
14320         link same way as WPA; do not stream in dead initializers.
14322         * dwarf2out.c (dwarf2out_die_ref_for_decl,
14323         darf2out_register_external_decl): Support incremental link.
14325 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14327         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
14329 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14331         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
14332         it down to simple_object_copy_lto_debug_sections.
14333         (run_gcc): Determine incremental LTO link time and configure
14334         lto1 into non-wpa mode, disable renaming of debug sections.
14336 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14338         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14339         descriptions of various incorrectly documented functions.
14341 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14343         Revert:
14344         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14345         address check not strict.
14347 2018-05-30  Richard Biener  <rguenther@suse.de>
14349         PR tree-optimization/85964
14350         * tracer.c (better_p): Drop initialized count check, we only
14351         call the function with initialized counts now.
14352         (find_best_successor): Do find a best edge if one
14353         has uninitialized count.
14354         (find_best_predecessor): Likewise.  Do BB frequency check only
14355         if count is initialized.
14357 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
14359         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14360         (aarch64_ldrstr_offset_compare): New.
14361         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
14362         load/store orderings.
14363         (aarch64_gen_adjusted_ldpstp): Likewise.
14365 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
14367         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14368         Check for subset of GENERAL_REGS and FP_REGS.
14369         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14370         r=w alternative.
14372 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
14374         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
14375         and wi::to_poly_offset.  Add the current offset and then check
14376         whether the sum fits, rather than using an unchecked addition of
14377         a checked term.  Check for a shwi rather than a uhwi.
14378         * expr.c (get_bit_range): Use tree_to_poly_uint64.
14379         (store_constructor): Use poly_int_tree_p.
14380         (expand_expr_real_1): Likewise.
14381         * function.c (assign_temp): Likewise.
14382         * fold-const.c (const_binop): Use poly_int_tree_p and
14383         wi::to_poly_offset.
14384         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
14385         division.
14386         * ipa-icf-gimple.c (func_checker::compare_operand): Use
14387         to_poly_offset for MEM offsets.
14388         * ipa-icf.c (sem_variable::equals): Likewise.
14389         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
14390         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
14391         wi::to_poly_offset for BIT_FIELD_REF offsets.
14392         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
14393         wi::to_poly_offset.
14394         * var-tracking.c (emit_note_insn_var_location): Use
14395         tree_to_poly_uint64.
14397 2018-05-29  Jim Wilson  <jimw@sifive.com>
14399         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
14401 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
14403         PR target/85950
14404         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
14405         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
14406         sequence.
14407         (sse4_1_round<mode>2): Use nonimmediate_operand
14408         for operand 1 predicate.
14410 2018-05-29  Martin Sebor  <msebor@redhat.com>
14411             Richard Biener  <rguenther@suse.de>
14413         PR testsuite/85888
14414         * calls.c (get_size_range): Call determine_value_range instead
14415         of get_value_range..
14416         * tree-vrp.h (determine_value_range): Declared new function.
14417         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
14419 2018-05-29  Richard Biener  <rguenther@suse.de>
14421         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
14422         sure to use non-pattern stmts for get_earlier_stmt arguments.
14423         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
14424         called on pattern stmts.
14425         (get_later_stmt): Likewise.
14427 2018-05-29  Martin Liska  <mliska@suse.cz>
14429         PR gcov-profile/85759
14430         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
14431         env variables.
14433 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
14435         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
14436         VEC_UNPACK_*_EXPR.
14437         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
14438         VEC_PACK_*_EXPR.
14440         PR target/85918
14441         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14442         VEC_PACK_FLOAT_EXPR): New tree codes.
14443         * tree-pretty-print.c (op_code_prio): Handle
14444         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14445         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14446         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14447         * tree-inline.c (estimate_operator_cost): Likewise.
14448         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
14449         * fold-const.c (const_binop): Likewise.
14450         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
14451         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14452         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14453         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
14454         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14455         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14456         * expr.c (expand_expr_real_2): Likewise.
14457         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
14458         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
14459         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
14460         optabs.
14461         * optabs.c (expand_widen_pattern_expr): For
14462         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
14463         sign from result type rather than operand's type.
14464         (expand_binop_directly): For vec_packu_float_optab and
14465         vec_packs_float_optab allow result type to be different from operand's
14466         type.
14467         * optabs-tree.c (optab_for_tree_code): Handle
14468         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14469         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
14470         * tree-vect-generic.c (expand_vector_operations_1):  Handle
14471         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14472         VEC_PACK_FLOAT_EXPR.
14473         * tree-vect-stmts.c (supportable_widening_operation): Handle
14474         FIX_TRUNC_EXPR.
14475         (supportable_narrowing_operation): Handle FLOAT_EXPR.
14476         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
14477         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
14478         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
14479         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
14480         mode attributes.
14481         (vec_pack<floatprefix>_float_<mode>): New expander.
14482         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
14483         attributes.
14484         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
14485         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
14486         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
14487         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
14488         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
14489         Document.
14490         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
14491         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
14492         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14493         VEC_PACK_FLOAT_EXPR): Document.
14495 2018-05-29  Richard Biener  <rguenther@suse.de>
14497         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
14498         member.
14499         (stmt_vec_info_vec): Make pointer.
14500         (init_stmt_vec_info_vec): Remove.
14501         (free_stmt_vec_info_vec): Likewise.
14502         (set_stmt_vec_info_vec): New function.
14503         (free_stmt_vec_infos): Likewise.
14504         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
14505         (set_vinfo_for_stmt): Likewise.
14506         (get_earlier_stmt): Likewise.
14507         (get_later_stmt): Likewise.
14508         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
14509         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
14510         (vec_info::~vec_info): Free stmt_vec_infos.
14511         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
14512         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
14513         (pass_slp_vectorize::execute): Likewise.
14514         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
14515         (free_stmt_vec_info_vec): Likewise.
14516         (set_stmt_vec_info_vec): New function.
14517         (free_stmt_vec_infos): Likewise.
14518         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
14519         the global stmt_vec_info_vec.
14520         * tree-parloops.c (gather_scalar_reductions): Use
14521         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
14522         vector.
14524 2018-05-29  Richard Biener  <rguenther@suse.de>
14526         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
14528 2018-05-29  Martin Liska  <mliska@suse.cz>
14529             David Malcolm  <dmalcolm@redhat.com>
14531         * vec.c (test_reverse): New.
14532         (vec_c_tests): Add new test.
14533         * vec.h (vl_ptr>::reverse): New function.
14535 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
14537         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
14539         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
14540         and later.
14542 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14544         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
14546 2018-05-28  Richard Biener  <rguenther@suse.de>
14548         PR tree-optimization/85933
14549         * tree-vect-data-refs.c (vect_record_base_alignments): Only
14550         look at stmts marked as vectorizable.
14552 2018-05-28  Richard Biener  <rguenther@suse.de>
14554         PR tree-optimization/85934
14555         * tree-vect-generic.c (expand_vector_operations_1): Hoist
14556         vector boolean check before scalar optimization.
14558 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
14560         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
14561         for armv5te.
14563 2018-05-28  Mark Wielaard  <mark@klomp.org>
14565         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
14566         if it is an expression containing a minus sign.
14568 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
14570         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
14572 2018-05-27  Paul Koning  <ni1d@arrl.net>
14574         * config/pdp11/pdp11.md (truncsihi2): Remove.
14576 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
14577             Chung-Ju Wu  <jasonwucj@gmail.com>
14579         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
14580         implementation.
14581         (unaligned_store_dw): Ditto.
14582         * config/nds32/nds32-memory-manipulation.c
14583         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
14584         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
14585         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
14586         (emit_setmem_word_loop): Rename to ...
14587         (emit_setmem_doubleword_loop): ... this.
14588         (nds32_gen_dup_4_byte_to_word_value): New function.
14589         (nds32_gen_dup_8_byte_to_double_word_value): New function.
14590         (nds32_expand_setmem_loop): Refine implementation.
14591         (nds32_expand_setmem_loop_v3m): Ditto.
14592         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
14593         pattern.
14595 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
14597         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
14599 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
14601         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
14602         (nds32_init_machine_status): Initialize machine->attr_naked_p and
14603         machine->attr_no_prologue_p.
14604         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
14605         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
14606         (nds32_expand_epilogue): Consider attr_naked_p.
14607         (nds32_expand_epilogue_v3pop): Likewise.
14608         (nds32_can_use_return_insn): Likewise.
14609         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
14610         attr_no_prologue_p fields.
14611         * config/nds32/nds32.opt (mret-in-naked-func): New option.
14613 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
14615         PR target/85918
14616         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
14617         attributes.
14618         * config/i386/sse.md
14619         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
14620         Rename to ...
14621         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
14622         ... this.
14623         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
14624         Rename to ...
14625         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
14626         ... this.
14627         (*<floatsuffix>floatv2div2sf2): Rename to ...
14628         (*float<floatunssuffix>v2div2sf2): ... this.
14629         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
14630         (float<floatunssuffix>v2div2sf2_mask): ... this.
14631         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
14632         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
14633         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
14634         to ...
14635         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
14636         ... this.
14637         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14638         Rename to ...
14639         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14640         ... this.
14641         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14642         Rename to ...
14643         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14644         ... this.
14645         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
14646         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
14647         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
14648         gen_ufix_truncv8dfv8si2.
14649         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
14650         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
14651         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
14652         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
14653         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
14654         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
14655         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
14656         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
14658 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
14660         PR target/85900
14661         PR target/85345
14662         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
14664 2018-05-25  Jim Wilson  <jimw@sifive.com>
14666         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
14667         * config/riscv/riscv.c (struct machine_function): Add
14668         interrupt_handler_p and attribute_checked_p fields.
14669         (riscv_attribute_table): Add interrupt.
14670         (riscv_interrupt_type_p): New.
14671         (riscv_save_reg_p): Save extra regs for interrupt handler.
14672         (riscv_use_save_libcall): Return false  for interrupt handler.
14673         (riscv_first_stack_step): Add forward declaration.
14674         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
14675         for interrupt handler with large frame.  Use it for saved reg list.
14676         (riscv_expand_prologue): Move flag_stack_usage_info support to
14677         eliminate duplication.
14678         (riscv_expand_epilogue): Generate mret for interrupt handler.
14679         (riscv_epilogue_uses): New.
14680         (riscv_can_use_return_insn): Return false for interrupt handler.
14681         (riscv_function_ok_for_sibcall): Likewise.
14682         (riscv_set_current_function): Add interrupt handler support.
14683         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
14684         * config/riscv/riscv.md (UNSPECV_MRET): New.
14685         (GP_REGNUM): New.
14686         (riscv_frflags, riscv_fsflags): Use tab after opcode.
14687         (riscv_mret): New.
14688         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
14690 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
14692         PR tree-optimization/85712
14693         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
14694         this candidate has already been replaced in-situ by a copy.
14696 2018-05-25  Jason Merrill  <jason@redhat.com>
14698         PR c++/80485 - inline function non-zero address.
14699         * symtab.c (nonzero_address): Check DECL_COMDAT.
14701 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
14703         PR target/83628
14704         * config/alpha/alpha.md (ashlsi3): New insn pattern.
14705         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
14706         extension of SImode operation.  Use const123_operand predicate.
14707         (*saddsi_1): Remove.
14708         (*saddl_se_1): Ditto.
14709         (*ssubsi_1): Ditto.
14710         (*ssubl_se_1): Ditto.
14711         * config/alpha/predicates.md (const123_operand): New predicate.
14712         * config/alpha/constraints.md (P): Use IN_RANGE.
14714 2018-05-25  Richard Biener  <rguenther@suse.de>
14716         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
14717         defaulted to true.
14718         (ref_maybe_used_by_stmt_p): Likewise.
14719         (stmt_may_clobber_ref_p): Likewise.
14720         (stmt_may_clobber_ref_p_1): Likewise.
14721         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
14722         and pass it along.
14723         (ref_maybe_used_by_stmt_p): Likewise.
14724         (stmt_may_clobber_ref_p): Likewise.
14725         (stmt_may_clobber_ref_p_1): Likewise.
14726         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
14727         the alias oracle to disambiguate DRs with stmts DR analysis
14728         couldn't handle.
14729         (vect_analyze_data_refs): Do not give up on not analyzable
14730         DRs for BB vectorization.  Remove code truncating the dataref
14731         vector.
14733 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
14735         PR target/85832
14736         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14737         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
14738         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
14740 2018-05-25  Richard Biener  <rguenther@suse.de>
14742         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
14743         function, combining stmt data ref gathering and fatal analysis
14744         parts.
14745         (vect_analyze_data_refs): Remove now redudnant code and simplify.
14746         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
14747         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
14748         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
14749         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
14751 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14753         PR tree-optimization/85720
14754         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
14755         SCC if all partitions are builtins.
14756         (version_loop_by_alias_check): New parameter.  Generate cancelable
14757         runtime alias check if all partitions are builtins.
14758         (distribute_loop): Update call to above function.
14760 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14762         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
14763         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
14764         (parm_default_def_partition_arg): Ditto.
14765         (set_parm_default_def_partition): Ditto.
14766         (get_parm_default_def_partitions): Ditto and make it static.
14767         (get_undefined_value_partitions): Ditto and make it static.
14768         (remove_ssa_form): Refactor call to init_var_map here.
14769         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
14770         computation for loop region.
14771         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
14772         (register_default_def): Delete.
14773         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
14774         (parm_default_def_partition_arg): Ditto.
14775         (set_parm_default_def_partition): Ditto.
14776         (get_parm_default_def_partitions): Ditto and make it static.
14777         (get_undefined_value_partitions): Ditto and make it static.
14778         (coalesce_with_default, coalesce_with_default): Update comment.
14779         (create_coalesce_list_for_region): New func factored out from
14780         create_outofssa_var_map.
14781         (populate_coalesce_list_for_outofssa): New func factored out from
14782         create_outofssa_var_map and coalesce_ssa_name.
14783         (create_outofssa_var_map): Delete.
14784         (coalesce_ssa_name): Refactor to support live range computation.
14785         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
14786         (get_parm_default_def_partitions): Delete.
14787         (get_undefined_value_partitions): Ditto.
14788         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
14789         computation for loop region.
14790         (new_tree_live_info, loe_visit_block): Ditto.
14791         (live_worklist, set_var_live_on_entry): Ditto.
14792         (calculate_live_on_exit, verify_live_on_entry): Ditto.
14793         * tree-ssa-live.h (struct _var_map): New fields.
14794         (init_var_map): Change decl.
14795         (region_contains_p): New.
14797 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14799         * tree-ssa-live.h (live_merge_and_clear): Delete.
14801 2018-05-25  Richard Biener  <rguenther@suse.de>
14803         PR c++/85912
14804         * tree-dump.c (dequeue_and_dump): Remove access to removed
14805         operand 2 of a SWITCH_EXPR.
14807 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14809         * doc/sourcebuild.texi (vect_double_cond_arith): Include
14810         multiplication and division.
14811         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
14812         (cond_udiv@var{m}, cond_umod@var{m}): Document.
14813         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
14814         (cond_udiv_optab, cond_umod_optab): New optabs.
14815         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
14816         (IFN_COND_RDIV): New internal functions.
14817         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
14818         TRUNC_MOD_EXPR and RDIV_EXPR.
14819         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
14820         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
14821         New unspecs.
14822         (SVE_INT_BINARY): Include mult.
14823         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
14824         (optab, sve_int_op): Handle mult.
14825         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
14826         UNSPEC_COND_DIV.
14827         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
14828         for SVE_INT_BINARY_SD.
14830 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14832         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
14833         (optab, sve_int_op): Handle div and udiv.
14834         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
14835         for SVE_INT_BINARY_SD.
14836         (*<optab><mode>3): New insn for the same.
14838 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14840         * tree-vect-patterns.c: Include predict.h.
14841         (vect_recog_divmod_pattern): Restrict check for division support
14842         to when optimizing for size.
14844 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14846         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
14847         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
14848         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
14849         (gimple_match_op::set_op): Likewise.
14850         (gimple_resimplify4): Declare.
14851         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
14852         (expr::gen_transform): Likewise.
14853         (decision_tree::gen): Generate a simplification routine for 4 operands.
14854         * gimple-match-head.c (gimple_simplify): Add an overload for
14855         4 operands.  In the top-level function, handle up to 4 call
14856         arguments and call gimple_resimplify4.
14857         (gimple_resimplify4): New function.
14858         (build_call_internal): Pass a fourth operand.
14859         (maybe_push_to_seq): Likewise.
14860         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
14861         Fold VEC_COND_EXPRs of an operation and a default value into
14862         an IFN_COND_* function if possible.
14863         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
14864         New unspecs.
14865         (SVE_COND_FP_BINARY): Include them.
14866         (optab, sve_fp_op): Handle them.
14867         (SVE_INT_BINARY_REV): New code iterator.
14868         (SVE_COND_FP_BINARY_REV): New int iterator.
14869         (commutative): New int attribute.
14870         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
14871         Declare.
14872         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
14873         function.
14874         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
14875         (*cond_<optab><mode>): New patterns for reversed operands.
14877 2018-05-25  Richard Biener  <rguenther@suse.de>
14879         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
14880         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
14881         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
14882         (STMT_VINFO_GROUPED_ACCESS): Adjust.
14883         * tree-vect-data-refs.c (everywhere): Adjust users.
14884         * tree-vect-loop.c (everywhere): Likewise.
14885         * tree-vect-slp.c (everywhere): Likewise.
14886         * tree-vect-stmts.c (everywhere): Likewise.
14887         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14889 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14891         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
14892         Rename to...
14893         (gcc_cv_as_section_exclude): ... this.
14894         Try Solaris as #exclude syntax.
14895         * configure: Regenerate.
14896         * config.in: Regenerate.
14897         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
14898         SECTION_EXCLUDE.
14899         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
14900         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
14902         * varasm.c (default_elf_asm_named_section): Don't check if
14903         HAVE_GAS_SECTION_EXCLUDE is defined.
14905 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14907         * doc/md.texi: Update the documentation of the cond_* optabs
14908         to mention the new final operand.  Fix GET_MODE_NUNITS call.
14909         Describe the scalar case too.
14910         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
14911         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
14912         instead of 2.
14913         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
14914         (get_conditional_internal_fn): Update comment.
14915         * tree-vect-loop.c (vectorizable_reduction): Pass the original
14916         accumulator value as a final argument to conditional functions.
14917         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
14918         a define_expand and add an "else" operand.  Assert for now that
14919         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
14920         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
14921         (*cond_<optab><mode>): New patterns.
14922         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
14923         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14924         (UNSPEC_COND_EOR): Delete.
14925         (optab): Remove associated mappings.
14926         (SVE_INT_BINARY): New code iterator.
14927         (sve_int_op): Remove int attribute and add "minus" to the code
14928         attribute.
14929         (SVE_COND_INT_OP): Delete.
14930         (SVE_COND_FP_OP): Rename to...
14931         (SVE_COND_FP_BINARY): ...this.
14933 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14935         * optabs.c (can_reuse_operands_p): New function.
14936         (maybe_legitimize_operands): Try to reuse the results for
14937         earlier operands.
14939 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14941         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14942         Add {q} suffix to insn mnemonic.
14944 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14946         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14947         (msp430_warn_func_return): New.
14949 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
14951         * fold-const.c (tree_nonzero_bits): New function.
14952         * fold-const.h (tree_nonzero_bits): Likewise.
14953         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14954         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14956 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
14958         PR target/85900
14959         PR target/85345
14960         * varasm.c (assemble_alias): Check ifunc_resolver only on
14961         FUNCTION_DECL.
14963 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14965         PR target/85903
14966         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14967         when memory input operand is handled.
14969 2018-05-24  Luis Machado  <luis.machado@linaro.org>
14971         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14972         global.
14973         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14975 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14977         * match.pd: Delay FMA folds until after vectorization.
14979 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14981         PR target/83009
14982         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14983         address check not strict.
14985 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14987         * gimple-match.h (gimple_match_op): New class.
14988         (mprts_hook): Replace parameters with a gimple_match_op *.
14989         (maybe_build_generic_op): Likewise.
14990         (gimple_simplified_result_is_gimple_val): Replace parameters with
14991         a const gimple_match_op *.
14992         (gimple_simplify): Replace code_helper * and tree * parameters with
14993         a gimple_match_op * parameter.
14994         (gimple_resimplify1): Replace code_helper *, tree and tree *
14995         parameters with a gimple_match_op * parameter.
14996         (gimple_resimplify2): Likewise.
14997         (gimple_resimplify3): Likewise.
14998         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14999         parameters with a gimple_match_op * parameter.
15000         * gimple-match-head.c (gimple_simplify): Change prototypes of
15001         auto-generated functions to take a gimple_match_op * instead of
15002         separate code_helper * and tree * parameters.  Make the same
15003         change in the top-level overload and update calls to the
15004         gimple_resimplify routines.  Update calls to the auto-generated
15005         functions and to maybe_push_res_to_seq in the publicly-facing
15006         operation-specific gimple_simplify overloads.
15007         (gimple_match_op::MAX_NUM_OPS): Define.
15008         (gimple_resimplify1): Replace rcode and ops with a single res_op
15009         parameter.  Update call to gimple_simplify.
15010         (gimple_resimplify2): Likewise.
15011         (gimple_resimplify3): Likewise.
15012         (mprts_hook): Replace parameters with a gimple_match_op *.
15013         (maybe_build_generic_op): Likewise.
15014         (build_call_internal): Replace type, nargs and ops with
15015         a gimple_match_op *.
15016         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
15017         with a single gimple_match_op *.  Update calls to mprts_hook,
15018         build_call_internal and gimple_simplified_result_is_gimple_val.
15019         Factor out code that is common to the tree_code and combined_fn cases.
15020         * genmatch.c (expr::gen_transform): Replace tem_code and
15021         tem_ops with a gimple_match_op called tem_op.  Update calls
15022         to the gimple_resimplify functions and maybe_push_res_to_seq.
15023         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
15024         res_ops.  Update call to the gimple_resimplify functions.
15025         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
15026         (decision_tree::gen): Make the functions take a gimple_match_op *
15027         called res_op instead of separate res_code and res_ops parameters.
15028         Update call accordingly.
15029         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
15030         and ops with a single res_op parameter.  Update calls to
15031         maybe_build_generic_op and maybe_push_res_to_seq.
15032         (fold_stmt_1): Update calls to gimple_simplify and
15033         replace_stmt_with_simplification.
15034         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
15035         and gimple_simplified_result_is_gimple_val.
15036         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
15037         gimple_simplify.
15038         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
15039         with a gimple_match_op *.
15040         (vn_nary_build_or_lookup): Likewise.  Update call to
15041         vn_nary_build_or_lookup_1.
15042         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
15043         gimple_match_op *.  Update calls to the gimple_resimplify routines
15044         and to gimple_simplified_result_is_gimple_val.
15045         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
15046         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
15047         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
15048         (visit_nary_op): Likewise.
15049         (visit_reference_op_load): Likewise.
15051 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15053         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
15054         modifier for printing the step amount.
15056 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
15058         PR target/78849
15059         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
15060         types.
15062 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
15064         * doc/sourcebuild.texi (Endianness): New subsubsection.
15066 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15068         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15069         <prefetch_dynamic_strides>: New const bool field.
15070         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15071         prefetch_dynamic_strides.
15072         (exynosm1_prefetch_tune): Likewise.
15073         (thunderxt88_prefetch_tune): Likewise.
15074         (thunderx_prefetch_tune): Likewise.
15075         (thunderx2t99_prefetch_tune): Likewise.
15076         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
15077         false.
15078         (aarch64_override_options_internal): Update to set
15079         PARAM_PREFETCH_DYNAMIC_STRIDES.
15080         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15081         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15082         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15083         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15084         prefetch-dynamic-strides setting.
15086 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15088         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15089         <minimum_stride>: New const int field.
15090         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15091         minimum_stride field defaulting to -1.
15092         (exynosm1_prefetch_tune): Likewise.
15093         (thunderxt88_prefetch_tune): Likewise.
15094         (thunderx_prefetch_tune): Likewise.
15095         (thunderx2t99_prefetch_tune): Likewise.
15096         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
15097         <default_opt_level>: Set to 3.
15098         (aarch64_override_options_internal): Update to set
15099         PARAM_PREFETCH_MINIMUM_STRIDE.
15100         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15101         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15102         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15103         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15104         stride is constant and is below the minimum stride threshold.
15106 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15108         * config/arm/arm-cpus.in (mode26): Delete.
15109         (armv4): Delete mode26 reference.
15110         * config/arm/arm.c (arm_configure_build_target): Delete use of
15111         isa_bit_mode26.
15113 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15115         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
15116         New insn pattern.
15117         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15118         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
15119         for non-SSE modes.
15120         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
15121         (floatunsdidf2): Ditto.
15123 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15125         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15126         (fixuns_trunc<mode>si2_avx512f): Ditto.
15127         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15128         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15129         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15131 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15133         PR rtl-optimization/79985
15134         * df-scan.c (df_insn_refs_collect): Remove special case for
15135         global registers and asm statements.
15137 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15139         * extend.texi (Global Register Variables): Rewrite the bullet list.
15140         Note that the register is available for allocation. Note that access
15141         via inline asm must use constraints. Add note about async-signal
15142         handlers. Remove paragraph about automagic register selection.
15144 2018-05-23  Richard Biener  <rguenther@suse.de>
15146         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15147         of fixed offset from memset VN.
15149 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
15151         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15152         first_interp field.
15153         (alloc_cand_and_find_basis): Initialize first_interp field.
15154         (slsr_process_mul): Modify first_interp field.
15155         (slsr_process_add): Likewise.
15156         (slsr_process_cast): Modify first_interp field for each new
15157         interpretation.
15158         (slsr_process_copy): Likewise.
15159         (dump_candidate): Dump first_interp field.
15160         (replace_mult_candidate): Process all interpretations, not just
15161         subsequent ones.
15162         (replace_rhs_if_not_dup): Likewise.
15163         (replace_one_candidate): Likewise.
15165 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
15167         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
15168         Add new boolean.
15169         (aarch64_needs_frame_chain): New function.
15170         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
15172 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
15174         PR target/84882
15175         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
15176         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15177         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15178         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15179         as true for strict-align.
15180         (aarch64_can_inline_p): Perform checks even when callee has no
15181         attributes to check for strict alignment.
15182         * doc/extend.texi (AArch64 Function Attributes): Document
15183         no-strict-align.
15184         * doc/invoke.texi: (AArch64 Options): Likewise.
15186 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
15188         PR tree-optimization/85853
15189         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15190         the handling of the root of the node to...
15191         (vect_slp_analyze_node_operations_1): ...this new function,
15192         and run the whole thing with the child nodes' def types
15193         set according to their SLP node's def type.
15195 2018-05-23  Richard Biener  <rguenther@suse.de>
15197         PR middle-end/85874
15198         * tree-data-ref.c (create_runtime_alias_checks): Defer
15199         and ignore overflow warnings.
15201 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
15203         PR tree-optimization/85822
15204         * tree-vrp.c (is_masked_range_test): Fix handling of negative
15205         constants.
15207 2018-05-23  Richard Biener  <rguenther@suse.de>
15209         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15210         memset constants via native_interpret_expr.
15212 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
15214         PR target/85345
15215         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15216         attribute.
15217         (cgraph_node::create_alias): Likewise.
15218         (cgraph_node::get_availability): Check ifunc_resolver instead
15219         of looking up ifunc attribute.
15220         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15221         * varasm.c (do_assemble_alias): Likewise.
15222         (assemble_alias): Likewise.
15223         (default_binds_local_p_3): Likewise.
15224         * cgraph.h (cgraph_node): Add ifunc_resolver.
15225         (cgraph_node::only_called_directly_or_aliased_p): Return false
15226         for IFUNC resolver.
15227         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
15228         attribute.
15229         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
15230         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
15231         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
15232         instead of looking up ifunc attribute.
15234 2018-05-22  Luis Machado  <luis.machado@linaro.org>
15236         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
15238 2018-05-22  Martin Sebor  <msebor@redhat.com>
15240         PR middle-end/85359
15241         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
15242         only when expasion succeeds.
15243         (expand_builtin_strcmp): Same.
15244         (expand_builtin_strncmp): Same.
15246 2018-05-22  Martin Sebor  <msebor@redhat.com>
15248         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
15250 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15251             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15253         * config/aarch64/aarch64-ldpstp.md: Replace uses of
15254         aarch64_mem_pair_operand with memory_operand and delete operand swapping
15255         code.
15256         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15257         Add check for legitimate_address.
15258         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
15259         (aarch64_swap_ldrstr_operands): New.
15260         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
15261         Define prototype.
15263 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15264             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15266         * config/aarch64/aarch64.md: New patterns to generate stp
15267         and ldp.
15268         (store_pair_sw, store_pair_dw): New patterns to generate stp for
15269         single words and double words.
15270         (load_pair_sw, load_pair_dw): Likewise.
15271         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
15272         Delete.
15273         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
15274         Delete.
15275         * config/aarch64/aarch64-ldpstp.md: Modify peephole
15276         for different mode ldpstp and add peephole for merged zero stores.
15277         Likewise for loads.
15278         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15279         Add size check.
15280         (aarch64_gen_store_pair): Rename calls to match new patterns.
15281         (aarch64_gen_load_pair): Rename calls to match new patterns.
15282         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
15283         (load_pair<DREG:mode><DREG2:mode>): ... This.
15284         (store_pair<mode>): Rename to...
15285         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
15286         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
15287         New mode iterators.
15288         (V_INT_EQUIV): Handle SImode.
15289         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
15290         New predicate.
15292 2018-05-22  Martin Sebor  <msebor@redhat.com>
15294         PR c/85623
15295         * calls.c (maybe_warn_nonstring_arg): Use string length to set
15296         or ajust the presumed bound on an operation to avoid unnecessary
15297         warnings.
15299 2018-05-22  Martin Sebor  <msebor@redhat.com>
15301         PR tree-optimization/85826
15302         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
15303         assuming that a DECL necesarily has a constant size.
15305 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
15307         PR middle-end/85862
15308         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
15310 2018-05-22  Richard Biener  <rguenther@suse.de>
15312         PR tree-optimization/85834
15313         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
15314         non-constant and non-zero memset arguments.
15316 2018-05-22  Martin Liska  <mliska@suse.cz>
15318         PR ipa/85607
15319         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
15321 2018-05-22  Richard Biener  <rguenther@suse.de>
15323         PR tree-optimization/85863
15324         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
15325         comparisons when vectype is specified.
15326         (vectorizable_condition): Do not specify vectype for
15327         vect_is_simple_cond when SLP vectorizing.
15329 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
15331         PR target/85657
15332         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
15333         define __ibm128 as long double.
15334         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
15335         as a distinct type when IEEE 128-bit support is enabled.
15336         (init_float128_ieee): Fix up conversions between IFmode and IEEE
15337         128-bit types to use the correct functions.
15338         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
15339         convert between 128-bit floating point types that have different
15340         modes but the same representation, instead of using gen_lowpart to
15341         makean alias.
15342         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
15343         KFmode.
15344         (IFKF_reg): New attributes to give the register constraints for
15345         IFmode and KFmode.
15346         (extend<mode>tf2_internal): New insns to mark an explicit
15347         conversion between 128-bit floating point types that have a
15348         different mode but share the same representation.
15350 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
15352         PR tree-optimization/85814
15353         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
15354         a null return from get_strinfo when unsharing the next
15355         strinfo in the chain.
15357 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
15359         PR gcc/84923
15360         * varasm.c (weak_finish): Clean up weak_decls.
15362 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15364         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
15365         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
15366         UNSPEC_UADALP values.
15367         * config/aarch64/iterators.md (ABAL): New int iterator.
15368         (ABDL2): Likewise.
15369         (ADALP): Likewise.
15370         (sur): Add mappings for the above.
15371         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
15372         New define_insn.
15373         (aarch64_<sur>abal<mode>_4): Likewise.
15374         (aarch64_<sur>adalp<mode>_3): Likewise.
15375         (<sur>sadv16qi): New define_expand.
15377 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
15379         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
15380         (*movdf_internal): Ditto.
15381         (*rcpsf2_sse): Ditto.
15382         (*rsqrtsf2_sse): Ditto.
15383         (*sqrt<mode>2_sse): Ditto.
15385 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
15387         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
15388         eor3q<mode>4.
15389         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
15390         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15391         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
15392         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
15393         vbcaxq_s64): New.
15394         * config/aarch64/arm_neon.h: Likewise.
15395         * config/aarch64/iterators.md (VQ_I): New.
15397 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
15399         * config.gcc: Add arc/t-multilib-linux to tmake_file for
15400         arc*-*-linux*.
15401         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
15402         MULTILIB_DIRNAMES
15404 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15406         * config/nds32/constraints.md (S): New constraint.
15407         * config/nds32/nds32.md (call_internal): Use constraint S.
15408         (call_value_internal): Likewise.
15409         (sibcall_internal): Likewise.
15410         (sibcall_value_internal): Likewise.
15412 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15413             Chung-Ju Wu  <jasonwucj@gmail.com>
15415         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15416         into consideration.
15418 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15419             Chung-Ju Wu  <jasonwucj@gmail.com>
15421         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
15422         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
15423         (nds32_rtx_costs_impl): Simplify.
15424         (nds32_address_cost_impl): Simplify.
15425         (nds32_init_rtx_costs): New function.
15426         (nds32_rtx_costs_speed_prefer): Likewise.
15427         (nds32_rtx_costs_size_prefer): Likewise.
15428         (nds32_address_cost_speed_prefer): Likewise.
15429         (nds32_address_cost_speed_fwprop): Likewise.
15430         (nds32_address_cost_size_prefer): Likewise.
15431         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
15432         * config/nds32/nds32.c (nds32_option_override): Use
15433         nds32_init_rtx_costs function.
15435 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15437         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
15438         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
15439         (TARGET_PIPELINE_N8): Likewise.
15440         (TARGET_PIPELINE_N10): Likewise.
15441         (TARGET_PIPELINE_N13): Likewise.
15442         (TARGET_PIPELINE_GRAYWOLF): Likewise.
15444 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
15446         * config/nds32/nds32-fpu.md: Update copyright year.
15448 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15450         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
15452 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15454         * config/nds32/nds32.c
15455         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
15456         * config/nds32/nds32.opt (minline-asm-r15): New option.
15458 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15460         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
15461         MASK_HW_ABS.
15462         * config/nds32/nds32.md (abssi2): New pattern.
15464 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
15466         * config/i386/i386.md (rex64namesuffix): New mode attribute.
15467         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
15468         Merge insn pattern from sse_cvtsi2ss<round_name> and
15469         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
15470         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
15471         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
15472         using SWI48 mode iterator.
15473         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
15474         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
15475         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
15476         pattern from sse_cvttss2si<round_saeonly_name>
15477         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
15478         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
15479         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
15480         using SWI48 mode iterator.
15481         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
15482         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
15483         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
15484         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
15485         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
15486         using SWI48 mode iterator.
15487         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
15488         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
15489         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
15490         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
15491         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
15492         SWI48 mode iterator.
15493         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
15494         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
15495         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
15496         pattern from sse_cvttsd2si<round_saeonly_name>
15497         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
15499 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15501         * config/nds32/nds32-md-auxiliary.c
15502         (nds32_valid_smw_lwm_base_p): Refine.
15503         (nds32_output_smw_single_word): Refine.
15504         (nds32_output_smw_double_word): New.
15505         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
15507 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15509         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
15510         (nds32_output_stack_pop): Refine.
15511         (nds32_expand_unaligned_load): Refine.
15512         (nds32_expand_unaligned_store): Refine.
15514 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
15515             Chung-Ju Wu  <jasonwucj@gmail.com>
15517         * config/nds32/constants.md: Add TP_REGNUM constant.
15518         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
15519         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
15520         UNSPEC_ADD32.
15521         * config/nds32/nds32-doubleword.md: Consider flag_pic.
15522         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
15523         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
15524         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
15525         and PIC code generation.
15526         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
15527         code generation.
15528         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
15529         optimization.
15530         * config/nds32/nds32.md: Support TLS and PIC.
15531         * config/nds32/nds32.c: Support TLS and PIC.
15532         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
15533         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
15534         predicate.
15536 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15538         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
15539         mode with E_ prefix.
15541 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
15542             Chung-Ju Wu  <jasonwucj@gmail.com>
15544         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
15545         * config/nds32/nds32-md-auxiliary.c
15546         (symbolic_reference_mentioned_p): New.
15547         (nds32_legitimize_ict_address): New.
15548         (nds32_expand_ict_move): New.
15549         (nds32_indirect_call_referenced_p): New.
15550         (nds32_symbol_binds_local_p): Delete.
15551         (nds32_long_call_p): Modify.
15552         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
15553         * config/nds32/nds32-protos.h
15554         (symbolic_reference_mentioned_p): Declare.
15555         (nds32_legitimize_ict_address): Declare.
15556         (nds32_expand_ict_move): Declare.
15557         (nds32_indirect_call_referenced_p): Declare.
15558         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
15559         (nds32_relax_group): Use nds32_ict_const_p as condition.
15560         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
15561         (nds32_asm_file_start): Output ict_model directive in asm code.
15562         (nds32_legitimate_address_p): Consider indirect call.
15563         (nds32_print_operand): Consider indirect call.
15564         (nds32_print_operand_address): Consider indirect call.
15565         (nds32_insert_attributes): Handle "indirect_call" attribute.
15566         (TARGET_LEGITIMATE_ADDRESS_P): Define.
15567         (TARGET_LEGITIMATE_CONSTANT_P): Define.
15568         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
15569         (TARGET_DELEGITIMIZE_ADDRESS): Define.
15570         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
15571         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
15572         (TARGET_ICT_MODEL_SMALL): Define.
15573         (TARGET_ICT_MODEL_LARGE): Define.
15574         * config/nds32/nds32.md (movsi): Consider ict model.
15575         (call, call_value): Consider ict model.
15576         (sibcall, sibcall_value): Consider ict model.
15577         * config/nds32/nds32.opt (mict-model): New option.
15578         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
15579         model.
15581 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
15582             Monk Chiang  <sh.chiang04@gmail.com>
15583             Jim Wilson <jimw@sifive.com>
15585         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
15586         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
15587         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
15588         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
15589         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
15590         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
15591         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
15592         compute save_libcall_adjustment properly.
15593         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
15594         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
15595         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
15596         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
15597         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
15598         (ABI_SPEC): Handle mabi=ilp32e.
15599         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
15600         (RVE): Add RVE mask.
15601         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
15602         <-march>: Add rv32e as an example.
15604 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
15606         PR c++/82899
15607         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
15608         (intra_create_variable_infos): Handle C++ constructors.
15610 2018-05-18  Martin Liska  <mliska@suse.cz>
15612         * passes.def: Remove a redundant pass.
15614 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
15616         PR bootstrap/85838
15617         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
15619 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15621         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
15622         (ARMv4): Update.
15623         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
15624         (ARMv6m): Update.
15625         (armv2, armv2a, armv3, armv3m): Delete architectures.
15626         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
15627         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
15628         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
15629         Delete cpus.
15630         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
15631         (*mulsidi3adddi): Likewise.
15632         (mulsidi3): Likewise.
15633         (*mulsidi3_nov6): Likewise.
15634         (umulsidi3): Likewise.
15635         (umulsidi3_nov6): Likewise.
15636         (umaddsidi4): Likewise.
15637         (*umulsidi3adddi): Likewise.
15638         (smulsi3_highpart): Likewise.
15639         (*smulsi3_highpart_nov6): Likewise.
15640         (umulsi3_highpart): Likewise.
15641         (*umulsi3_highpart_nov6): Likewise.
15642         * config/arm/arm.h (arm_arch3m): Delete.
15643         * config/arm/arm.c (arm_arch3m): Delete.
15644         (arm_option_override_internal): Update armv3-related comment.
15645         (arm_configure_build_target): Delete use of isa_bit_mode32.
15646         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
15647         (arm_rtx_costs_internal): Delete check of arm_arch3m.
15648         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
15649         (mulsa3): Likewise.
15650         (mulusa3): Likewise.
15651         * config/arm/arm-protos.h (arm_arch3m): Delete.
15652         * config/arm/arm-tables.opt: Regenerate.
15653         * config/arm/arm-tune.md: Likewise.
15654         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
15655         deleted architectures.
15657 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15659         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
15660         (armv5t, armv5te): New features.
15661         (ARMv5, ARMv5e): Delete fgroups.
15662         (ARMv5t, ARMv5te): Adjust for above changes.
15663         (ARMv6m): Likewise.
15664         (armv5, armv5e): Delete arches.
15665         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
15666         arm_arch5.
15667         (*call_reg_arm): Likewise.
15668         (*call_value_reg_armv5): Likewise.
15669         (*call_value_reg_arm): Likewise.
15670         (*call_symbol): Likewise.
15671         (*call_value_symbol): Likewise.
15672         (*sibcall_insn): Likewise.
15673         (*sibcall_value_insn): Likewise.
15674         (clzsi2): Likewise.
15675         (prefetch): Likewise.
15676         (define_split and define_peephole2 dependent on arm_arch5):
15677         Likewise.
15678         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
15679         arm_arch5e.
15680         (TARGET_ARM_QBIT): Likewise.
15681         (TARGET_DSP_MULTIPLY): Likewise.
15682         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
15683         (arm_arch5, arm_arch5e): Delete.
15684         (arm_arch5t, arm_arch5te): Declare.
15685         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
15686         (arm_arch5t): Declare.
15687         (arm_option_reconfigure_globals): Update for the above.
15688         (arm_options_perform_arch_sanity_checks): Update comment, replace
15689         use of arm_arch5 with arm_arch5t.
15690         (use_return_insn): Likewise.
15691         (arm_emit_call_insn): Likewise.
15692         (output_return_instruction): Likewise.
15693         (arm_final_prescan_insn): Likewise.
15694         (arm_coproc_builtin_available): Likewise.
15695         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
15696         arm_arch5e with arm_arch5t and arm_arch5te.
15697         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
15698         (arm_arch5t, arm_arch5te): Declare.
15699         * config/arm/arm-tables.opt: Regenerate.
15700         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
15701         * config/arm/t-multilib: Likewise.
15702         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
15703         instead of arm_arch5.
15704         (*call_reg_thumb1): Likewise.
15705         (*call_value_reg_thumb1_v5): Likewise.
15706         (*call_value_reg_thumb1): Likewise.
15707         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
15708         unreachable path.
15709         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
15711 2018-05-18  Martin Liska  <mliska@suse.cz>
15713         PR gcov-profile/84846
15714         * doc/gcov.texi: Document -t option of gcov tool.
15716 2018-05-18  Martin Liska  <mliska@suse.cz>
15718         PR gcov-profile/84846
15719         * gcov.c (print_usage): Add new -t option.
15720         (process_args): Handle the option.
15721         (generate_results): Use stdout as output when requested by
15722         the option.
15724 2018-05-18  Martin Liska  <mliska@suse.cz>
15726         PR gcov-profile/84846
15727         * coverage.c (coverage_init): Write PWD to .gcno file.
15728         * doc/gcov.texi: Document how working directory is printed.
15729         * gcov-dump.c (dump_gcov_file): Print PWD.
15730         * gcov.c (output_intermediate_file): Likewise.
15731         (read_graph_file): Read PWD string.
15732         (output_lines): Print PWD.
15734 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15736         PR middle-end/85817
15737         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
15738         for retval and return false if all args to phi are zero.
15740 2018-05-18  Richard Biener  <rguenther@suse.de>
15742         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
15743         method.
15744         (evrp_dom_walker::before_dom_children): Call it.
15746 2018-05-18  Richard Biener  <rguenther@suse.de>
15748         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
15749         results when processing array refs with variable index.
15751 2018-05-18  Toon Moene  <toon@moene.org>
15753         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
15754         directly after that of -floop-interchange. Indicate that both
15755         options are enabled by default when specifying -O3.
15757 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15759         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
15760         iterator.  Delete separate integer-mode vec_set<mode> expander.
15761         (aarch64_simd_vec_setv2di): Delete.
15762         (vec_setv2di): Delete.
15763         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
15764         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
15765         the "w, r" alternative.
15767 2018-05-18  Martin Liska  <mliska@suse.cz>
15769         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
15770         * tree-pass.h (make_pass_lower_switch_O0): New function.
15771         * tree-switch-conversion.c (node_has_low_bound): Remove.
15772         (node_has_high_bound): Likewise.
15773         (node_is_bounded): Likewise.
15774         (class pass_lower_switch): Make it a template type and create
15775         two instances.
15776         (pass_lower_switch::execute): Add template argument.
15777         (make_pass_lower_switch): New function.
15778         (make_pass_lower_switch_O0): New function.
15779         (do_jump_if_equal): Remove.
15780         (emit_case_nodes): Simplify to just handle all 3 cases and leave
15781         all the hard work to tree optimization passes.
15783 2018-05-18  Martin Liska  <mliska@suse.cz>
15785         * dbgcnt.c (limit_low): Renamed from limit.
15786         (limit_high): New variable.
15787         (dbg_cnt_is_enabled): Check for upper limit.
15788         (dbg_cnt): Adjust dumping.
15789         (dbg_cnt_set_limit_by_index): Add new argument for high
15790         value.
15791         (dbg_cnt_set_limit_by_name): Likewise.
15792         (dbg_cnt_process_single_pair): Parse new format.
15793         (dbg_cnt_process_opt): Use strtok.
15794         (dbg_cnt_list_all_counters): Remove 'value' and add
15795         'limit_high'.
15796         * doc/invoke.texi: Document changes.
15798 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
15800         * doc/sourcebuild.texi (scalar_all_fma): Document.
15801         * tree.def (FMA_EXPR): Delete.
15802         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
15803         * internal-fn.c (ternary_direct): New macro.
15804         (expand_ternary_optab_fn): Likewise.
15805         (direct_ternary_optab_supported_p): Likewise.
15806         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
15807         * builtins.c (fold_builtin_fma): Delete.
15808         (fold_builtin_3): Don't call it.
15809         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
15810         * expr.c (expand_expr_real_2): Likewise.
15811         * fold-const.c (operand_equal_p): Likewise.
15812         (fold_ternary_loc): Likewise.
15813         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
15814         * gimple.c (DEFTREECODE): Likewise.
15815         * gimplify.c (gimplify_expr): Likewise.
15816         * optabs-tree.c (optab_for_tree_code): Likewise.
15817         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
15818         * tree-eh.c (operation_could_trap_p): Likewise.
15819         (stmt_could_throw_1_p): Likewise.
15820         * tree-inline.c (estimate_operator_cost): Likewise.
15821         * tree-pretty-print.c (dump_generic_node): Likewise.
15822         (op_code_prio): Likewise.
15823         * tree-ssa-loop-im.c (stmt_cost): Likewise.
15824         * tree-ssa-operands.c (get_expr_operands): Likewise.
15825         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
15826         * fold-const-call.h (fold_fma): Delete.
15827         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
15828         CFN_FNMA and CFN_FNMS.
15829         (fold_fma): Delete.
15830         * genmatch.c (combined_fn): New enum.
15831         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
15832         (commutative_op): New function.
15833         (commutate): Use it.  Handle more than 2 operands.
15834         (dt_operand::gen_gimple_expr): Use commutative_op.
15835         (parser::parse_expr): Allow :c to be used with non-binary
15836         operators if the commutative operand is known.
15837         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
15838         CFN_FMS, CFN_FNMA and CFN_FNMS.
15839         (backprop::process_assign_use): Remove FMA_EXPR handling.
15840         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
15841         (gen_hsa_fma): New function.
15842         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
15843         IFN_FNMA and IFN_FNMS.
15844         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
15845         * gimple-fold.h (follow_all_ssa_edges): Declare.
15846         * gimple-fold.c (follow_all_ssa_edges): New function.
15847         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
15848         gimple_build interface and use follow_all_ssa_edges to fold the result.
15849         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
15850         instead of checking for optabs directly.
15851         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
15852         rather than FMA_EXPRs.
15853         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
15854         call to IFN_FMA instead of an FMA_EXPR.
15856 2018-05-17  Jim Wilson  <jimw@sifive.com>
15858         * expr.c (do_tablejump): When converting index to Pmode, if we have a
15859         sign extended promoted subreg, and the range does not have the sign bit
15860         set, then do a sign extend.
15862         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
15863         test, check for sign extended subreg and/or constant operands, and
15864         do a sign extend in that case.
15866 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
15868         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
15869         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
15870         Add untyped.
15871         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
15872         Change logics_shift_reg to logics_shift_imm.
15873         (thunderx2t99_fp_loadpair_basic): Delete.
15874         (thunderx2t99_fp_storepair_basic): Delete.
15875         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
15876         (thunderx2t99_asimd_polynomial): Delete.
15877         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
15878         and neon_fp_mul_d_scalar_q.
15879         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15880         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
15881         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
15882         (thunderx2t99_asimd_lut): Add missing tbl types.
15883         (thunderx2t99_asimd_ext): Delete.
15884         (thunderx2t99_asimd_load1_1_mult): Delete.
15885         (thunderx2t99_asimd_load1_2_mult): Delete.
15886         (thunderx2t99_asimd_load1_ldp): New.
15887         (thunderx2t99_asimd_load1): New.
15888         (thunderx2t99_asimd_load2): Add missing *load2* types.
15889         (thunderx2t99_asimd_load3): New.
15890         (thunderx2t99_asimd_load4): New.
15891         (thunderx2t99_asimd_store1_1_mult): Delete.
15892         (thunderx2t99_asimd_store1_2_mult): Delete.
15893         (thunderx2t99_asimd_store2_mult): Delete.
15894         (thunderx2t99_asimd_store2_onelane): Delete.
15895         (thunderx2t99_asimd_store_stp): New.
15896         (thunderx2t99_asimd_store1): New.
15897         (thunderx2t99_asimd_store2): New.
15898         (thunderx2t99_asimd_store3): New.
15899         (thunderx2t99_asimd_store4): New.
15901 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
15903         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
15904         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
15906 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
15907             Segher Boessenkool  <segher@kernel.crashing.org>
15909         PR target/85698
15910         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
15911         operand.
15913 2018-05-17  Richard Biener  <rguenther@suse.de>
15915         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
15916         for pruning loop and prune defs feeding only already visited PHIs.
15918 2018-05-17  Richard Biener  <rguenther@suse.de>
15920         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
15922 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
15923             Richard Biener  <rguenther@suse.de>
15925         PR tree-optimization/85793
15926         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15927         for VMAT_ELEMENTWISE.
15929 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
15931         * internal-fn.h (lookup_internal_fn): Declare
15932         * internal-fn.c (lookup_internal_fn): New function.
15933         * gimple.c (gimple_build_call_from_tree): Handle calls to
15934         internal functions.
15935         * gimple-pretty-print.c (dump_gimple_call): Print "." before
15936         internal function names.
15937         * tree-pretty-print.c (dump_generic_node): Likewise.
15938         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15940 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
15942         * gimple-fold.h (gimple_build): Make the function forms take
15943         combined_fn rather than built_in_function.
15944         (gimple_simplify): Likewise.
15945         * gimple-match-head.c (gimple_simplify): Likewise.
15946         * gimple-fold.c (gimple_build): Likewise.
15947         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15948         rather than gimple_build_call_internal.
15949         (get_initial_defs_for_reduction): Likewise.
15950         (vect_create_epilog_for_reduction): Likewise.
15951         (vectorizable_live_operation): Likewise.
15953 2018-05-17  Martin Liska  <mliska@suse.cz>
15955         * gimple-ssa-sprintf.c (format_directive): Do not use
15956         space in between 'G_' and '('.
15958 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
15960         PR target/85323
15961         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15962         even if the mask is not all ones.
15964         PR target/85323
15965         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15966         vector.
15967         (ix86_gimple_fold_builtin): Likewise.
15969         PR target/85323
15970         * config/i386/i386.c: Include tree-vector-builder.h.
15971         (ix86_vector_shift_count): New function.
15972         (ix86_fold_builtin): Fold shift builtins by scalar count.
15973         (ix86_gimple_fold_builtin): Likewise.
15975         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15976         _mm512_setzero): New intrinsics.
15978 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
15979             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15981         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15982         code generation for cases where splatting a value is not useful.
15983         * simplify-rtx.c (simplify_ternary_operation): Simplify
15984         vec_merge across a vec_duplicate and a paradoxical subreg forming
15985         a vector mode to a vec_concat.
15987 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
15989         * config.gcc: Support "goldmont-plus".
15990         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15991         "goldmont-plus".
15992         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15993         PROCESSOR_GOLDMONT_PLUS.
15994         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15995         (processor_target_table): Add "goldmont-plus".
15996         (PTA_GOLDMONT_PLUS): Define.
15997         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15998         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15999         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
16000         (fold_builtin_cpu): Add "goldmont-plus".
16001         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
16002         (ix86_option_override_internal): Add "goldmont-plus".
16003         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
16004         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
16005         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
16006         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
16008 2018-05-17  Richard Biener  <rguenther@suse.de>
16010         PR tree-optimization/85757
16011         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
16012         remove defs that only feed that PHI from further processing.
16014 2018-05-16  Jim Wilson  <jimw@sifive.com>
16016         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
16017         asterisk to name.
16018         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
16020 2018-05-16  Mark Wielaard  <mark@klomp.org>
16022         * dwarf2out.c (count_index_strings): New function.
16023         (output_indirect_strings): Call count_index_strings and generate
16024         header for dwarf_version >= 5.
16026 2018-05-16  Mark Wielaard  <mark@klomp.org>
16028         * dwarf2out.c (dwarf_FORM): New function.
16029         (set_indirect_string): Use dwarf_FORM.
16030         (reset_indirect_string): Likewise.
16031         (size_of_die): Likewise.
16032         (value_format): Likewise.
16033         (output_die): Likewise.
16034         (add_skeleton_AT_string): Likewise.
16035         (output_macinfo_op): Likewise.
16036         (index_string): Likewise.
16037         (output_index_string_offset): Likewise.
16038         (output_index_string): Likewise.
16039         (count_index_strings): Likewise.
16041 2018-05-16  Carl Love  <cel@us.ibm.com>
16043         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16044         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
16046 2018-05-16  Martin Jambor  <mjambor@suse.cz>
16048         * ipa-prop.c (ipa_free_all_edge_args): Remove.
16049         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
16051 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
16053         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
16054         (fnma<mode>4): Likewise.
16055         (fms<mode>4): Likewise.
16056         (fnms<mode>4): Likewise.
16057         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
16058         (aarch64_fnma<mode>4): Likewise.
16059         (aarch64_fms<mode>4): Likewise.
16060         (aarch64_fnms<mode>4): Likewise.
16061         (aarch64_fnmadd<mode>4): Likewise.
16063 2018-05-16  Jason Merrill  <jason@redhat.com>
16065         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
16067 2018-05-16  Richard Biener  <rguenther@suse.de>
16069         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
16070         (dump_stmt_cost): Declare.
16071         (add_stmt_cost): Dump cost we add.
16072         (add_stmt_costs): New function.
16073         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
16074         No longer exported.
16075         (vect_analyze_stmt): Adjust prototype.
16076         (vectorizable_condition): Likewise.
16077         (vectorizable_live_operation): Likewise.
16078         (vectorizable_reduction): Likewise.
16079         (vectorizable_induction): Likewise.
16080         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
16081         cost vector to pass to vectorizable_ and record afterwards.
16082         (vect_model_reduction_cost): Take cost vector argument and adjust.
16083         (vect_model_induction_cost): Likewise.
16084         (vectorizable_reduction): Likewise.
16085         (vectorizable_induction): Likewise.
16086         (vectorizable_live_operation): Likewise.
16087         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
16088         SLP_TREE_NUMBER_OF_VEC_STMTS.
16089         (vect_analyze_slp_cost_1): Remove.
16090         (vect_analyze_slp_cost): Likewise.
16091         (vect_slp_analyze_node_operations): Take visited args and
16092         a target cost vector.  Avoid processing already visited stmt sets.
16093         (vect_slp_analyze_operations): Use a local cost vector to gather
16094         costs and register those of non-discarded instances.
16095         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
16096         (vect_schedule_slp_instance): Remove copying of
16097         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
16098         zero.
16099         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
16100         adding cost.  Record cost entry location.
16101         (vect_prologue_cost_for_slp_op): Function to compute cost of
16102         a constant or invariant generated for SLP vect in the prologue,
16103         split out from vect_analyze_slp_cost_1.
16104         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
16105         (vect_model_promotion_demotion_cost): Likewise.
16106         (vect_model_store_cost): Likewise, make static.
16107         (vect_model_load_cost): Likewise.
16108         (vectorizable_bswap): Add cost vector arg and adjust.
16109         (vectorizable_call): Likewise.
16110         (vectorizable_simd_clone_call): Likewise.
16111         (vectorizable_conversion): Likewise.
16112         (vectorizable_assignment): Likewise.
16113         (vectorizable_shift): Likewise.
16114         (vectorizable_operation): Likewise.
16115         (vectorizable_store): Likewise.
16116         (vectorizable_load): Likewise.
16117         (vectorizable_condition): Likewise.
16118         (vectorizable_comparison): Likewise.
16119         (can_vectorize_live_stmts): Likewise.
16120         (vect_analyze_stmt): Likewise.
16121         (vect_transform_stmt): Adjust calls to vectorizable_*.
16122         * tree-vectorizer.c: Include gimple-pretty-print.h.
16123         (dump_stmt_cost): New function.
16125 2018-05-16  Richard Biener  <rguenther@suse.de>
16127         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16128         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16129         * tree-ssa-dse.c: Include tree-ssa-loop.h.
16130         (check_name): New callback.
16131         (dse_classify_store): Track cycles via a visited bitmap of PHI
16132         defs and simplify handling of in-loop and across loop dead stores
16133         and properly fail for loop-variant refs.  Handle byte-tracking with
16134         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16135         limiting the walk.
16137 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
16139         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16140         (vect_get_mask_type_for_stmt): Likewise.
16141         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16142         split out from...
16143         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
16144         to determine the statement's vector type and the vector type that
16145         should be used for calculating nunits.  Deal with cases in which
16146         the type has to be deferred.
16147         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16148         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16149         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16150         (vect_determine_vf_for_stmt): New functions, split out from...
16151         (vect_determine_vectorization_factor): ...here.
16152         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16153         (vect_get_mask_type_for_stmt): New functions, split out from
16154         vect_determine_vectorization_factor.
16156 2018-05-16  Richard Biener  <rguenther@suse.de>
16158         * tree-cfg.c (verify_gimple_assign_ternary): Properly
16159         verify the [VEC_]COND_EXPR embedded comparison.
16161 2018-05-15  Martin Sebor  <msebor@redhat.com>
16163         PR tree-optimization/85753
16164         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
16165         RECORD_TYPE in addition to ARRAY_TYPE.
16167 2018-05-15  Martin Sebor  <msebor@redhat.com>
16169         PR middle-end/85643
16170         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
16172 2018-05-15  Richard Biener  <rguenther@suse.de>
16174         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
16175         add by_clobber_p one.  Change algorithm to collect all defs
16176         representing uses we need to walk and try reducing them to
16177         a single one before failing.
16178         (dse_dom_walker::dse_optimize_stmt): Adjust.
16180 2018-05-13  Mark Wielaard  <mark@klomp.org>
16182         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16183         (size_of_loc_descr): Likewise.
16184         (output_loc_operands): Likewise.
16185         (output_loc_operands_raw): Likewise.
16186         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16187         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16188         (hash_loc_operands): Likewise.
16189         (compare_loc_operands): Likewise.
16191 2018-05-14  Mark Wielaard  <mark@klomp.org>
16193         * dwarf2out.c (count_index_addrs): New function.
16194         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16196 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16198         PR tree-optimization/83648
16199         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16200         return value as malloc candidate.
16202 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16204         PR ipa/85734
16205         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16206         param as true in call to suggest_attribute.
16208 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
16210         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16211         -mreadonly-in-sdata.
16213 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16215         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16216         New pattern.
16217         (aarch64_crypto_aesd_fused): Likewise.
16219 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
16221         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
16222         (movsi_aarch64): Likewise.
16223         (load_pairsi): Likewise.
16224         (load_pairdi): Likewise.
16225         (store_pairsi): Likewise.
16226         (store_pairdi): Likewise.
16227         (load_pairsf): Likewise.
16228         (load_pairdf): Likewise.
16229         (store_pairsf): Likewise.
16230         (store_pairdf): Likewise.
16231         (zero_extend): Likewise.
16232         (trunc): Swap alternatives.
16233         (fcvt_target): Add '?' to prefer w over r.
16235 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
16237         PR target/85756
16238         * config/i386/i386.md: Disallow non-commutative arithmetics in
16239         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
16240         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
16241         in the peephole2 before it.
16243 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
16245         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
16246         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
16247         (ix86_handle_option): Handle -mcldemote.
16248         * config.gcc: New header.
16249         * config/i386/cldemoteintrin.h: New file.
16250         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
16251         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16252         -mcldemote.
16253         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16254         OPTION_MASK_ISA_CLDEMOTE.
16255         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
16256         (ix86_valid_target_attribute_inner_p): Ditto.
16257         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
16258         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
16259         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
16260         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
16261         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
16262         (cldemote): New.
16263         * config/i386/i386.opt: Add -mcldemote.
16264         * config/i386/x86intrin.h: New header.
16265         * doc/invoke.texi: Add -mcldemote.
16267 2018-05-14  Richard Biener  <rguenther@suse.de>
16269         * doc/match-and-simplify.texi: Adjust :s documentation.
16271 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
16273         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
16274         intended memcpy size.
16275         (REORDER_45): Likewise.
16277 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
16279         * sort.cc: New file.
16280         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
16281         * vec.c (qsort_chk): Use gcc_qsort.
16282         * Makefile.in (OBJS-libcommon): Add sort.o.
16283         (build/sort.o): New target.  Use it...
16284         (BUILD_RTL): ... here, and...
16285         (build/gencfn-macros): ... here, and...
16286         (build/genmatch): ... here.
16288 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16289             Chung-Ju Wu  <jasonwucj@gmail.com>
16291         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
16292         * config/nds32/nds32-graywolf.md: New file.
16293         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
16294         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
16295         pipeline.
16296         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
16297         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
16298         * config/nds32/nds32.md (pipeline_model): Add graywolf.
16299         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
16300         * config/nds32/pipelines.md: Include n15 settings.
16302 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16303             Chung-Ju Wu  <jasonwucj@gmail.com>
16305         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
16306         * config/nds32/nds32-n13.md: New file.
16307         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
16308         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
16309         pipeline.
16310         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
16311         * config/nds32/nds32.md (pipeline_model): Add n13.
16312         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
16313         * config/nds32/pipelines.md: Include n13 settings.
16315 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16316             Chung-Ju Wu  <jasonwucj@gmail.com>
16318         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
16319         * config/nds32/nds32-n10.md: New file.
16320         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
16321         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
16322         pipeline.
16323         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
16324         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
16325         * config/nds32/nds32.md (pipeline_model): Add n10.
16326         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
16327         * config/nds32/pipelines.md: Include n10 settings.
16329 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
16330             Kito Cheng  <kito.cheng@gmail.com>
16331             Chung-Ju Wu  <jasonwucj@gmail.com>
16333         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
16334         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16335         Add enum values for DSP extension instructions.
16336         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
16337         New constraints.
16338         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
16339         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
16340         New code iterators.
16341         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
16342         * config/nds32/nds32-dspext.md: New file for DSP implementation.
16343         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
16344         * config/nds32/nds32-intrinsic.md: Likewise.
16345         * config/nds32/nds32_intrinsic.h: Likewise.
16346         * config/nds32/nds32-md-auxiliary.c: Likewise.
16347         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
16348         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
16349         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
16350         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
16351         * config/nds32/nds32-protos.h: New declarations for DSP extension.
16352         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
16353         TYPE_DMAC in switch statement.
16354         * config/nds32/nds32.c: New checking and implementation for DSP
16355         extension instructions.
16356         * config/nds32/nds32.h: Likewise.
16357         * config/nds32/nds32.md: Likewise.
16358         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
16359         * config/nds32/predicates.md: Implement new predicates for DSP
16360         extension.
16362 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
16364         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
16365         Reformat alternatives and attributes so it is easier to identify
16366         which constraints/attributes go with which instruction.
16367         (mov<mode>_hardfloat32, FMOVE64): Likewise.
16368         (mov<mode>_softfloat32, FMOVE64): Likewise.
16369         (mov<mode>_hardfloat64, FMOVE64): Likewise.
16370         (mov<mode>_softfloat64, FMOVE64): Likewise.
16372 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16374         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16375         subsection.
16376         (Basic PowerPC Built-in Functions): The new name of the
16377         subsection previously known as "PowerPC Built-in Functions".
16378         (Basic PowerPC Built-in Functions Available on all Configurations):
16379         New subsubsection.
16380         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
16381         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
16382         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
16383         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
16385 2018-05-11  Martin Jambor  <mjambor@suse.cz>
16387         PR ipa/85655
16388         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
16389         single const.
16391 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16393         PR target/85733
16394         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
16396 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
16398         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
16399         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
16400         (ix86_handle_option): Handle -mwaitpkg.
16401         * config.gcc: New header.
16402         * config/i386/cpuid.h (bit_WAITPKG): New bit.
16403         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
16404         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
16405         function type.
16406         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16407         OPTION_MASK_ISA_WAITPKG.
16408         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
16409         (ix86_option_override_internal): Add PTA_WAITPKG.
16410         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
16411         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
16412         IX86_BUILTIN_TPAUSE.
16413         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
16414         __builtin_ia32_umwait and __builtin_ia32_tpause.
16415         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
16416         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
16417         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
16418         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
16419         UNSPECV_TPAUSE): New.
16420         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
16421         * config/i386/i386.opt: Add -mwaitpkg.
16422         * config/i386/waitpkgintrin.h: New file.
16423         * config/i386/x86intrin.h: New header.
16424         * doc/invoke.texi: Add -mwaitpkg.
16426 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16428         PR target/85606
16429         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
16430         equivalent.
16431         (cortex-m0): Use armv6s-m isa.
16432         (cortex-m0plus): Likewise.
16433         (cortex-m1): Likewise.
16434         (cortex-m0.small-multiply): Likewise.
16435         (cortex-m0plus.small-multiply): Likewise.
16436         (cortex-m1.small-multiply): Likewise.
16438 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
16439             Jakub Jelinek  <jakub@redhat.com>
16441         PR tree-optimization/85692
16442         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
16443         source permute as well.
16445 2018-05-11  Martin Liska  <mliska@suse.cz>
16447         PR sanitizer/85556
16448         * doc/extend.texi: Document LLVM style format for no_sanitize
16449         attribute.
16451 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
16453         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
16454         mode_supports_vsx_dform_quad to mode_supports_dq_form.
16455         (mode_supports_vsx_dform_quad): Likewise.
16456         (mode_supports_vmx_dform): Move these functions to be next to the
16457         other mode_supports functions.
16458         (mode_supports_dq_form): Likewise.
16459         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
16460         mode_supports_dq_form.
16461         (reg_offset_addressing_ok_p): Likewise.
16462         (offsettable_ok_by_alignment): Likewise.
16463         (rs6000_legitimate_offset_address_p): Likewise.
16464         (legitimate_lo_sum_address_p): Likewise.
16465         (rs6000_legitimize_address): Likewise.
16466         (rs6000_legitimize_reload_address): Likewise.
16467         (rs6000_secondary_reload_inner): Likewise.
16468         (rs6000_preferred_reload_class): Likewise.
16469         (rs6000_output_move_128bit): Likewise.
16471 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
16473         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
16474         Generate SImode target register for null target.
16475         <case IX86_BUILTIN_XGETBV>: Ditto.
16476         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
16477         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
16479 2018-05-10  Carl Love  <cel@us.ibm.com>
16481         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16482         dcbtt and dcbtstt if operands[2] is 0.
16484 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
16486         PR target/85693
16487         * config/i386/sse.md (usadv64qi): New expander.
16489 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
16491         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
16492         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
16493         -maltivec=be support.
16494         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
16495         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
16496         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
16497         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
16498         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
16499         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
16500         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
16501         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
16502         altivec_vsumsws): Adjust.
16503         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
16504         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
16505         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
16506         support.
16507         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
16508         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
16509         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
16510         (altivec_lve<VI_char>x): Delete expand.
16511         (*altivec_lve<VI_char>x_internal): Rename to...
16512         (altivec_lve<VI_char>x): ... this.
16513         (altivec_lvxl_<mode>): Delete expand.
16514         (*altivec_lvxl_<mode>_internal): Rename to ...
16515         (altivec_lvxl_<mode>): ... this.
16516         (altivec_stvxl_<mode>): Delete expand.
16517         (*altivec_stvxl_<mode>_internal): Rename to ...
16518         (altivec_stvxl_<mode>): ... this.
16519         (altivec_stve<VI_char>x): Delete expand.
16520         (*altivec_stve<VI_char>x_internal): Rename to ...
16521         (altivec_stve<VI_char>x): ... this.
16522         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
16523         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
16524         reduc_plus_scal_<mode>): Adjust.
16525         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
16526         comment.
16527         (rs6000_cpu_cpp_builtins): Adjust.
16528         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
16529         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
16530         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
16531         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
16532         -maltivec=be support.
16533         (rs6000_split_vec_extract_var): Adjust.
16534         (rs6000_split_v4si_init): Adjust.
16535         (swap_selector_for_mode): Delete.
16536         (altivec_expand_lvx_be, altivec_expand_stvx_be,
16537         altivec_expand_stvex_be): Delete.
16538         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
16539         -maltivec=be support.
16540         (rs6000_gimple_fold_builtin): Ditto.
16541         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
16542         Adjust.
16543         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
16544         (TARGET_DIRECT_MOVE_64BIT): Adjust.
16545         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
16546         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
16547         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
16548         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
16549         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
16550         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
16551         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
16552         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
16553         anonymous split): Adjust.
16554         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
16555         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
16557 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
16559         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
16560         when --with-gxx-include-dir is also specified.
16561         * configure: Regenerate.
16563 2018-05-09  Jim Wilson  <jimw@sifive.com>
16565         PR target/84797
16566         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
16567         * config/riscv/t-withmultilib: New.
16568         * config/riscv/withmultilib.h: New.
16569         * doc/install.texi: Document RISC-V --with-multilib-list support.
16571 2018-05-09  Richard Biener  <rguenther@suse.de>
16573         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
16574         vector.
16575         (vect_bb_vectorization_profitable_p): Adjust.  Compute
16576         actual scalar cost using the cost vector and the add_stmt_cost
16577         machinery.
16579 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16581         PR rtl-optimization/85645
16582         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
16583         in the REG_CFA_REGISTER note for LR, don't leave it empty.
16585 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16587         PR rtl-optimization/85645
16588         * shrink-wrap.c (spread_components): Return a boolean saying if
16589         anything was changed.
16590         (try_shrink_wrapping_separate): Iterate spread_components until
16591         nothing changes anymore.
16593 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16595         PR rtl-optimization/85645
16596         * regrename.c (build_def_use): Also kill the chains that include the
16597         destination of a REG_CFA_REGISTER note.
16599 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16601         PR rtl-optimization/85645
16602         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
16603         insn that has a REG_CFA_REGISTER note.
16605 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
16607         * cfgexpand.c (expand_clobber): New function.
16608         (expand_gimple_stmt_1): Use it.
16609         * tree-vect-stmts.c (vect_clobber_variable): New function,
16610         split out from...
16611         (vectorizable_simd_clone_call): ...here.
16612         (vectorizable_store): Emit a clobber either side of an
16613         IFN_STORE_LANES sequence.
16614         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
16616 2018-05-09  Tom de Vries  <tom@codesourcery.com>
16618         PR target/85626
16619         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
16620         (define_insn "trap_if_false"): Add exit after trap.
16622 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
16624         PR rtl-optimization/85638
16625         * bb-reorder.c: Include common/common-target.h.
16626         (create_forwarder_block): New function extracted from...
16627         (fix_up_crossing_landing_pad): ...here.  Rename into...
16628         (dw2_fix_up_crossing_landing_pad): ...this.
16629         (sjlj_fix_up_crossing_landing_pad): New function.
16630         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
16631         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
16632         from both partitions and exit the loop after one iteration.
16634 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16636         Revert:
16637         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16638         subsection.
16639         (Basic PowerPC Built-in Functions): The new name of the
16640         subsection previously known as "PowerPC Built-in Functions".
16641         (Basic PowerPC Built-in Functions Available on all Configurations):
16642         New subsubsection.
16643         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16644         subsubsection.
16645         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16646         subsubsection.
16647         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16648         subsubsection.
16649         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16650         subsubsection.
16652 2018-05-08  Jim Wilson  <jimw@sifive.com>
16654         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
16655         (LD_EMUL_SUFFIX): New.
16656         (LINK_SPEC): Use it.
16658 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16660         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16661         subsection.
16662         (Basic PowerPC Built-in Functions): The new name of the
16663         subsection previously known as "PowerPC Built-in Functions".
16664         (Basic PowerPC Built-in Functions Available on all Configurations):
16665         New subsubsection.
16666         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16667         subsubsection.
16668         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16669         subsubsection.
16670         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16671         subsubsection.
16672         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16673         subsubsection.
16675 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
16677         PR target/85683
16678         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
16679         after cmpelim optimization.
16681 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
16683         * config.gcc: Support "goldmont".
16684         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
16685         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16686         PROCESSOR_GOLDMONT.
16687         * config/i386/i386.c (m_GOLDMONT): Define.
16688         (processor_target_table): Add "goldmont".
16689         (PTA_GOLDMONT): Define.
16690         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
16691         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
16692         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
16693         (fold_builtin_cpu): Add "goldmont".
16694         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
16695         (ix86_option_override_internal): Add "goldmont".
16696         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
16697         (processor_type): Add PROCESSOR_GOLDMONT.
16698         * config/i386/i386.md: Add CPU "glm".
16699         * config/i386/glm.md: New file.
16700         * config/i386/x86-tune.def: Add m_GOLDMONT.
16701         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
16703 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
16705         PR target/85572
16706         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
16707         E_V4DImode.
16708         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
16709         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
16710         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
16712         PR target/85317
16713         * config/i386/i386.c (ix86_fold_builtin): Handle
16714         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
16716         PR target/85480
16717         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
16718         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
16720 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
16722         PR target/85658
16723         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
16724         (check_arch): Likewise.
16725         (check_fpu): Return the result rather than printing it.
16726         (end arch): Fix operator precedence.
16727         (end cpu): Likewise.
16728         (END): Print the result from check_fpu.
16730 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
16731             Alan Hayward  <alan.hayward@arm.com>
16732             David Sherwood  <david.sherwood@arm.com>
16734         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
16735         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
16736         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
16737         (*fcmuo<mode>_and): New patterns.
16739 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
16741         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
16742         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
16743         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
16744         (cmp_op, sve_imm_con): New code attributes.
16745         (SVE_COND_INT_CMP, imm_con): Delete.
16746         (cmp_op): Remove above unspecs from int attribute.
16747         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
16748         to...
16749         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
16750         comparison-specific unspecs.
16751         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
16752         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
16753         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
16754         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
16755         (*vec_fcm<cmp_op><mode>): Rename to...
16756         (*fcm<cmp_op><mode>): ...this and adjust likewise.
16757         (*vec_fcmuo<mode>): Rename to...
16758         (*fcmuo<mode>): ...this and adjust likewise.
16759         (*pred_fcm<cmp_op><mode>): New pattern.
16760         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
16761         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
16762         functions.
16763         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
16764         and UNORDERED.
16765         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
16766         (aarch64_emit_sve_predicated_cond): New function.
16767         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
16768         (aarch64_emit_unspec_cond_or): Replace with...
16769         (aarch64_emit_sve_or_conds): ...this new function.  Use
16770         aarch64_emit_sve_ptrue_op for the individual comparisons and
16771         aarch64_emit_binop to OR them together.
16772         (aarch64_emit_inverted_unspec_cond): Replace with...
16773         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
16774         aarch64_emit_sve_ptrue_op for the comparison and
16775         aarch64_emit_unop to invert the result.
16776         (aarch64_expand_sve_vec_cmp_float): Update after the above
16777         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
16779 2018-05-07  Nathan Sidwell  <nathan@acm.org>
16781         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
16782         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
16783         (Backwards Compatibility): Likewise.
16785 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16787         PR bootstrap/85681
16788         Revert:
16789         2018-05-07  Luis Machado  <luis.machado@linaro.org>
16791         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16792         <prefetch_dynamic_strides>: New const bool field.
16793         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16794         prefetch_dynamic_strides.
16795         (exynosm1_prefetch_tune): Likewise.
16796         (thunderxt88_prefetch_tune): Likewise.
16797         (thunderx_prefetch_tune): Likewise.
16798         (thunderx2t99_prefetch_tune): Likewise.
16799         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16800         to false.
16801         (aarch64_override_options_internal): Update to set
16802         PARAM_PREFETCH_DYNAMIC_STRIDES.
16803         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16804         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16805         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16806         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16807         prefetch-dynamic-strides setting.
16809         2018-05-07  Luis Machado  <luis.machado@linaro.org>
16811         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16812         <minimum_stride>: New const int field.
16813         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16814         minimum_stride field.
16815         (exynosm1_prefetch_tune): Likewise.
16816         (thunderxt88_prefetch_tune): Likewise.
16817         (thunderx_prefetch_tune): Likewise.
16818         (thunderx2t99_prefetch_tune): Likewise.
16819         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16820         (aarch64_override_options_internal): Update to set
16821         PARAM_PREFETCH_MINIMUM_STRIDE.
16822         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16823         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16824         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16825         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16826         stride is constant and is below the minimum stride threshold.
16828 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16830         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
16831         to 512.
16833 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16835         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16836         <prefetch_dynamic_strides>: New const bool field.
16837         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16838         prefetch_dynamic_strides.
16839         (exynosm1_prefetch_tune): Likewise.
16840         (thunderxt88_prefetch_tune): Likewise.
16841         (thunderx_prefetch_tune): Likewise.
16842         (thunderx2t99_prefetch_tune): Likewise.
16843         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16844         to false.
16845         (aarch64_override_options_internal): Update to set
16846         PARAM_PREFETCH_DYNAMIC_STRIDES.
16847         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16848         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16849         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16850         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16851         prefetch-dynamic-strides setting.
16853 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16855         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16856         <minimum_stride>: New const int field.
16857         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16858         minimum_stride field.
16859         (exynosm1_prefetch_tune): Likewise.
16860         (thunderxt88_prefetch_tune): Likewise.
16861         (thunderx_prefetch_tune): Likewise.
16862         (thunderx2t99_prefetch_tune): Likewise.
16863         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16864         (aarch64_override_options_internal): Update to set
16865         PARAM_PREFETCH_MINIMUM_STRIDE.
16866         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16867         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16868         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16869         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16870         stride is constant and is below the minimum stride threshold.
16872 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
16874         PR c++/85659
16875         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
16876         the type is addressable.  Don't force op into register if it has
16877         BLKmode.
16879 2018-05-05  Roland McGrath  <mcgrathr@google.com>
16881         PR other/77609
16882         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
16883         any section for which we don't know a specific type it should have,
16884         regardless of name.  Previously this was done only for the exact
16885         names ".init_array", ".fini_array", and ".preinit_array".
16886         (default_elf_asm_named_section): Add comment about
16887         relationship with default_section_type_flags and SECTION_NOTYPE.
16888         (get_section): Don't consider it a type conflict if one side has
16889         SECTION_NOTYPE and the other doesn't, as long as neither has the
16890         SECTION_BSS et al used in the default_section_type_flags logic.
16892 2018-05-05  Tom de Vries  <tom@codesourcery.com>
16894         PR target/85653
16895         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
16896         (workaround_barsyncs): New function.
16897         (nvptx_reorg): Use workaround_barsyncs.
16898         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16899         (define_expand "nvptx_membar_cta"): New define_expand.
16900         (define_insn "*nvptx_membar_cta"): New insn.
16902 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
16904         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
16905         To improve optimization opportunities.
16906         * builtin-types.def: The new needed builtin types for the above.
16908 2018-05-04  Richard Biener  <rguenther@suse.de>
16910         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
16911         * gimple-ssa-store-merging.c
16912         (imm_store_chain_info::output_merged_store): Remove redundant create,
16913         release split_store vector contents on failure.
16914         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
16915         scalar stmt vector on cache hit.
16917 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
16919         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
16920         Xilinx FP support.
16921         * config.gcc (powerpc-xilinx-eabi*): Remove.
16922         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16923         support.
16924         (fusion_addis_mem_combo_load): Ditto.
16925         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16926         FP support.
16927         (rs6000_cpu_cpp_builtins): Ditto.
16928         * config/rs6000/rs6000-linux.c
16929         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16930         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16931         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16932         support.
16933         (rs6000_setup_reg_addr_masks): Ditto.
16934         (rs6000_init_hard_regno_mode_ok): Ditto.
16935         (rs6000_option_override_internal): Ditto.
16936         (legitimate_lo_sum_address_p): Ditto.
16937         (rs6000_legitimize_address): Ditto.
16938         (rs6000_legitimize_reload_address): Ditto.
16939         (rs6000_legitimate_address_p): Ditto.
16940         (abi_v4_pass_in_fpr): Ditto.
16941         (setup_incoming_varargs): Ditto.
16942         (rs6000_gimplify_va_arg): Ditto.
16943         (rs6000_split_multireg_move): Ditto.
16944         (rs6000_savres_strategy): Ditto.
16945         (rs6000_emit_prologue_components): Ditto.
16946         (rs6000_emit_epilogue_components): Ditto.
16947         (rs6000_emit_prologue): Ditto.
16948         (rs6000_emit_epilogue): Ditto.
16949         (rs6000_elf_file_end): Ditto.
16950         (rs6000_function_value): Ditto.
16951         (rs6000_libcall_value): Ditto.
16952         * config/rs6000/rs6000.h: Ditto.
16953         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16954         (TARGET_MINMAX): ... this.  New.
16955         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16956         * config/rs6000/rs6000.md: Remove Xilinx FP support.
16957         (*movsi_internal1_single): Delete.
16958         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16959         mfpu=, mxilinx-fpu): Delete.
16960         * config/rs6000/singlefp.h: Delete.
16961         * config/rs6000/sysv4.h: Remove Xilinx FP support.
16962         * config/rs6000/t-rs6000: Ditto.
16963         * config/rs6000/t-xilinx: Delete.
16964         * config/rs6000/titan.md: Adjust for fp_type removal.
16965         * config/rs6000/vsx.md: Remove Xilinx FP support.
16966         (VStype_simple): Delete.
16967         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16968         * config/rs6000/xfpu.h: Delete.
16969         * config/rs6000/xfpu.md: Delete.
16970         * config/rs6000/xilinx.h: Delete.
16971         * config/rs6000/xilinx.opt: Delete.
16972         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16973         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16975 2018-05-04  Tom de Vries  <tom@codesourcery.com>
16977         PR libgomp/85639
16978         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16979         if ignore == 0.
16981 2018-05-04  Richard Biener  <rguenther@suse.de>
16983         PR middle-end/85627
16984         * tree-complex.c (update_complex_assignment): We are always in SSA form.
16985         (expand_complex_div_wide): Likewise.
16986         (expand_complex_operations_1): Likewise.
16987         (expand_complex_libcall): Preserve EH info of the original stmt.
16988         (tree_lower_complex): Handle removed blocks.
16989         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16990         on complex multiplication and division libcall builtins.
16992 2018-05-04  Richard Biener  <rguenther@suse.de>
16994         PR middle-end/85574
16995         * fold-const.c (negate_expr_p): Restrict negation of operand
16996         zero of a division to when we know that can happen without
16997         overflow.
16998         (fold_negate_expr_1): Likewise.
17000 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
17002         PR libstdc++/85466
17003         * real.h (real_nextafter): Declare.
17004         * real.c (real_nextafter): New function.
17005         * fold-const-call.c (fold_const_nextafter): New function.
17006         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
17007         CASE_CFN_NEXTTOWARD.
17008         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
17009         even when arg1_mode is different from arg0_mode.
17011 2018-05-03  Nathan Sidwell  <nathan@acm.org>
17013         * doc/extend.texi (Deprecated Features): Remove
17014         -ffriend-injection.
17015         (Backwards Compatibility): Likewise.
17016         * doc/invoke.texi (C++ Language Options): Likewise.
17017         (C++ Dialect Options): Likewise.
17019 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
17021         PR target/85530
17022         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
17023         _mm512_mask_mullox_epi64): New intrinsics.
17025 2018-05-03  Tom de Vries  <tom@codesourcery.com>
17027         PR testsuite/85106
17028         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17029         dump files): Add offload-tree.
17031 2018-05-03  Richard Biener  <rguenther@suse.de>
17033         PR tree-optimization/85615
17034         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17035         to loops not nested in BBs loop father to avoid creating multi-entry
17036         loops.
17038 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17040         PR tree-optimization/70291
17041         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
17042         arguments.  Change return type to tree.  Emit libcall as a new
17043         statement rather than replacing existing one when inplace_p is true.
17044         (expand_complex_multiplication_components): New function.
17045         (expand_complex_multiplication): Expand floating-point complex
17046         multiplication using the above.
17047         (expand_complex_division): Rename inner_type parameter to type.
17048         Update expand_complex_libcall call-site.
17049         (expand_complex_operations_1): Update expand_complex_multiplication
17050         and expand_complex_division call-sites.
17052 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17054         PR target/85582
17055         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17056         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17057         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
17058         the highest significant bit of the shift count mask is clear.  In
17059         check whether and[sq]i3 is needed verify that all significant bits
17060         of the shift count other than the highest are set.
17062 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17064         PR libgomp/82428
17065         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
17066         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
17067         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
17068         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
17069         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
17070         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
17071         __builtin_goacc_parlevel_size.
17073 2018-05-02  Richard Biener  <rguenther@suse.de>
17075         PR tree-optimization/85597
17076         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
17077         do not use split vect_get_vec_defs call but call vect_get_slp_defs
17078         directly.
17080 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17082         PR testsuite/85106
17083         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17084         dump files): Add ltrans-tree.
17086 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17088         PR testsuite/85106
17089         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17090         dump files): Add wpa-ipa.
17092 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
17094         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
17095         powerpc*-*-linux*paired* target.
17096         * config/rs6000/750cl.h: Delete.
17097         * config/rs6000/paired.h: Delete.
17098         * config/rs6000/paired.md: Delete.
17099         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
17100         float support.
17101         * config/rs6000/rs6000-builtin.def: Remove paired float support.
17102         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
17103         comment.  Remove paired float support.
17104         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
17105         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
17106         VECTOR_PAIRED.
17107         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
17108         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
17109         declarations.
17110         * config/rs6000/rs6000.c: Remove paired float support.
17111         (paired_expand_vector_init, paired_expand_vector_move,
17112         paired_emit_vector_compare, paired_emit_vector_cond_expr,
17113         (paired_expand_lv_builtin, paired_expand_stv_builtin,
17114         paired_expand_builtin, paired_expand_predicate_builtin,
17115         paired_init_builtins): Delete.
17116         * config/rs6000/rs6000.h: Remove paired float support.
17117         * config/rs6000/rs6000.md: Remove paired float support.
17118         (move_from_CR_ov_bit): Delete.
17119         * config/rs6000/rs6000.opt (mpaired): Delete.
17120         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17121         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17123 2018-05-02  Richard Biener  <rguenther@suse.de>
17125         PR middle-end/85567
17126         * gimplify.c (gimplify_save_expr): When in SSA form allow
17127         SAVE_EXPRs to compute to SSA vars.
17129 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17131         PR target/85582
17132         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17133         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17134         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17135         clobber operands[2], instead use a new pseudo.  Formatting fixes.
17137 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
17139         PR tree-optimization/85586
17140         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17141         exit early for statements in the same group if the accesses are
17142         not strided.
17144 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17146         PR lto/85451
17147         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17148         error message.
17150 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
17152         PR tree-optimization/85143
17153         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17155 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17157         PR lto/85451
17158         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17159         not found" error message.
17161 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17163         PR other/83786
17164         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
17165         * vec.c (test_ordered_remove_if): New function.
17166         (vec_c_tests): Call test_ordered_remove_if.
17167         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
17168         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
17169         * tree-vect-patterns.c (vect_pattern_recog_1): Use
17170         VEC_ORDERED_REMOVE_IF.
17172 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17174         PR tree-optimization/82665
17175         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
17176         pointer subtraction where arguments come from a memchr call.
17178 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
17180         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17181         --push-state --as-needed and --pop-state instead of --as-needed and
17182         --no-as-needed if ld supports it.
17183         * configure: Regenerated.
17185         PR web/85578
17186         * doc/install.texi2html: Replace _002d with - and _002a with * in
17187         generated html files using sed.
17189 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17191         PR c++/85523
17192         * gcc-rich-location.c (blank_line_before_p): New function.
17193         (use_new_line): New function.
17194         (gcc_rich_location::add_fixit_insert_formatted): New function.
17195         * gcc-rich-location.h
17196         (gcc_rich_location::add_fixit_insert_formatted): New function.
17198 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17200         * selftest.c (assert_streq): Rename "expected" and "actual" to
17201         "val1" and "val2".  Extend NULL-handling to cover both inputs
17202         symmetrically, while still requiring both to be non-NULL for a pass.
17203         * selftest.h (assert_streq): Rename "expected" and "actual" to
17204         "val1" and "val2".
17205         (ASSERT_EQ): Likewise.
17206         (ASSERT_EQ_AT): Likewise.
17207         (ASSERT_KNOWN_EQ): Likewise.
17208         (ASSERT_KNOWN_EQ_AT): Likewise.
17209         (ASSERT_NE): Likewise.
17210         (ASSERT_MAYBE_NE): Likewise.
17211         (ASSERT_MAYBE_NE_AT): Likewise.
17212         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
17213         the assertion to pass.
17214         (ASSERT_STREQ_AT): Likewise.
17216 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17218         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17219         interaction with -pie.
17221 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17223         * selftest.h: Fix alphabetization of per-source-file selftest
17224         declarations.
17226 2018-04-30  Jason Merrill  <jason@redhat.com>
17228         PR c++/61982 - dead stores to destroyed objects.
17229         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
17230         of clobber.
17232 2018-04-30  Jason Merrill  <jason@redhat.com>
17234         * tree.c (build_clobber): New.
17235         * tree.h: Declare it.
17236         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
17238 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17240         * diagnostic-show-locus.c (layout::layout): Update for
17241         location_get_source_line returning a char_span.
17242         (struct char_span): Move to input.h.
17243         (struct correction): Update for fields in char_span becoming
17244         private.
17245         (struct source_line): Update for location_get_source_line
17246         returning a char_span.
17247         (layout::print_line): Likewise.
17248         * edit-context.c (edited_file::print_content): Likewise.
17249         (edited_file::print_diff_hunk): Likewise.
17250         (edited_file::print_run_of_changed_lines): Likewise.
17251         (edited_file::get_num_lines): Likewise.
17252         (edited_line::edited_line): Likewise.
17253         * final.c (asm_show_source): Likewise.
17254         * input.c (location_get_source_line): Convert return type
17255         from const char * to char_span, losing the final "line_len"
17256         param.
17257         (dump_location_info): Update for the above.
17258         (get_substring_ranges_for_loc): Likewise.  Use a char_span
17259         when handling the literal within the line.
17260         (test_reading_source_line): Update for location_get_source_line
17261         returning a char_span.
17262         * input.h (class char_span): Move here from
17263         diagnostic-show-locus.c, converting from a struct to a class.
17264         Make data members private.
17265         (char_span::operator bool): New.
17266         (char_span::length): New.
17267         (char_span::get_buffer): New.
17268         (char_span::operator[]): New.
17269         (char_span::subspan): Make const.
17270         (char_span::xstrdup): New.
17271         (location_get_source_line): Convert return type from const char *
17272         to char_span, losing the final "line_size" param.
17274 2018-04-30  Jan Hubicka  <jh@suse.cz>
17276         * lto-wrapper.c (ltrans_priorities): New static var.
17277         (cmp_priority): New.
17278         (run_gcc): Read priorities and if doing parallel build order
17279         the Makefile by them.
17281 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17283         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
17285 2018-04-30  Richard Biener  <rguenther@suse.de>
17287         * tree-cfg.c (verify_address): Remove base argument, add
17288         flag whether to check TREE_ADDRESSABLE and do that.
17289         (verify_expr): Remove.
17290         (verify_types_in_gimple_reference): Add pieces from verify_expr.
17291         (verify_gimple_assign_single): Likewise.
17292         (verify_gimple_switch): Likewise.
17293         (verify_expr_location_1): Dereference tp once.  Add (disabled)
17294         piece from verify_expr.
17295         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
17297 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17299         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17301 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17303         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
17304         (small_data_pattern): Likewise.
17305         (arc_rewrite_small_data): Likewise.
17306         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
17307         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
17308         (get_symbol_alignment): New function.
17309         (legitimate_small_data_address_p): Likewise.
17310         (legitimate_scaled_address): Update, call
17311         legitimate_small_data_address_p.
17312         (output_sdata): New static variable.
17313         (arc_print_operand): Update how we handle small data operands.
17314         (arc_print_operand_address): Likewise.
17315         (arc_legitimate_address_p): Update, use
17316         legitimate_small_data_address_p.
17317         (arc_rewrite_small_data_p): Remove.
17318         (arc_rewrite_small_data_1): Likewise.
17319         (arc_rewrite_small_data): Likewise.
17320         (small_data_pattern): Likewise.
17321         (compact_sda_memory_operand): Update to use
17322         legitimate_small_data_address_p and get_symbol_alignment.
17323         (prepare_move_operands): Don't rewite sdata pattern.
17324         (prepare_extend_operands): Remove.
17325         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
17326         pattern.
17327         (zero_extendqisi2): Likewise.
17328         (zero_extendhisi2): Likewise.
17329         (extendqihi2): Likewise.
17330         (extendqisi2): Likewise.
17331         (extendhisi2): Likewise.
17332         (addsi3): Likewise.
17333         (subsi3): Likewise.
17334         (andsi3): Likewise.
17335         * config/arc/constraints.md (Usd): Change it to memory constraint.
17337 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17339         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
17340         as source of std instructions.
17341         * config/arc/arc.md (movsi_insn): Update pattern predicate to
17342         allow 6-bit constants as source for store instructions.
17343         (movdi_insn): Update instruction pattern to allow 6-bit constants
17344         as source for store instructions.
17346 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17348         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
17350 2018-04-30  Nathan Sidwell  <nathan@acm.org>
17351             Sandra Loosemore <sandra@codesourcery.com>
17353         * dumpfile.c (dump_open): Allow '-' for stdout.
17354         * doc/invoke.texi (Developer Options): Document dump filename
17355         determination early.  Document stdin/stdout selection.
17357 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
17359         Microblaze Target: PIC data text relative
17361         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
17362         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
17363         Add declaration.
17364         * config/microblaze/microblaze.h (microblaze_constant_address_p):
17365         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
17366         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
17367         New addressing mode for data-text relative position indepenedent code.
17368         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
17369         'ADDRESS_SYMBOLIC_TXT_REL'.
17370         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
17371         (microblaze_legitimate_pic_operand): Exclude function calls from
17372         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
17373         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
17374         addresses cases.
17375         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17376         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17377         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
17378         for 'address + offset'.
17379         (microblaze_expand_prologue): Add new function prologue call for
17380         'r20' assignation.
17381         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
17382         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
17383         table in case of TARGET_PIC_DATA_TEXT_REL.
17384         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
17385         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
17386         Add new macros 'UNSPEC_TEXT',
17387         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
17388         + exclude function calls from 'UNSPEC_PLT' in case of data text
17389         relative mode.
17390         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17391         new target hook for generating address diff vector tables in case of
17392         flag_pic.
17393         * doc/tm.texi : Regenerate.
17394         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
17395         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
17396         of addr diff vector generation.
17397         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17398         target hook definition.
17399         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
17400         Add default function for generate_pic_addr_diff_vec -> flag_pic.
17401         * doc/invoke.texi (Add new pic option): Add new microblaze pic
17402         option for data text relative.
17404 2018-04-30  Richard Biener  <rguenther@suse.de>
17406         * tree-chrec.h (evolution_function_is_constant_p): Remove
17407         redundant check.
17408         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
17410 2018-04-30  Richard Biener  <rguenther@suse.de>
17412         PR bootstrap/85571
17413         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
17415 2018-04-30  Richard Biener  <rguenther@suse.de>
17417         PR tree-optimization/28364
17418         PR tree-optimization/85275
17419         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17420         copying first exit test.
17422 2018-04-28  Mark Wielaard  <mark@klomp.org>
17424         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
17425         dwarf_version >= 5.
17426         (dwarf_AT): Handle DW_AT_addr_base.
17427         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
17429 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
17431         PR target/84431
17432         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
17433         (*ashl<dwi>3_doubleword_mask_1): Ditto.
17434         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
17435         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
17437 2018-04-28  Richard Biener  <rguenther@suse.de>
17439         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
17440         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
17441         to reflect use.  Only add interesting stmts.
17443 2018-04-27  Martin Jambor  <mjambor@suse.cz>
17445         PR ipa/85549
17446         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
17447         the jump function allows for passing through aggregate values.
17449 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
17451         * input.h (in_system_header_at): Convert from macro to inline
17452         function.
17453         (from_macro_expansion_at): Likewise.
17454         (from_macro_definition_at): Likewise.
17456 2018-04-27  Jeff Law  <law@redhat.com>
17458         * config.gcc: Mark tile* targets as deprecated/obsolete.
17460 2018-04-27  Richard Biener  <rguenther@suse.de>
17462         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
17463         fix for ILP32.
17465 2018-04-27  Richard Biener  <rguenther@suse.de>
17467         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
17469 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
17471         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
17472         with Yd constraint. Set "preferred_for_speed" attribute from
17473         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
17474         with Yd constraint.
17475         (*movdi_internal): Ditto.
17476         (movti_interunit splitters): Remove
17477         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
17478         (movdi_interunit splitters): Ditto.
17479         * config/i386/constraints.md (Ye): Remove.
17480         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
17482 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17484         PR target/85512
17485         * config/aarch64/constraints.md (Usg): Limit to 31.
17486         (Usj): Limit to 63.
17488 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
17490         PR tree-optimization/85529
17491         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
17492         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
17493         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
17494         zero extension or masking of the MSB bit.
17495         (optimize_range_tests): Add FIRST_BB argument, pass it through
17496         to optimize_range_tests_var_bound.
17497         (maybe_optimize_range_tests, reassociate_bb): Adjust
17498         optimize_range_tests callers.
17500 2018-04-26  Richard Biener  <rguenther@suse.de>
17501             Jakub Jelinek  <jakub@redhat.com>
17503         * cgraph.h (symbol_table): Just declare debug method here.
17504         * symtab.c (symbol_table::debug): Define.
17506 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
17508         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
17510 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
17512         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
17513         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
17514         (*movdi_internal): Substitute Yi and Yj constraint with x
17515         and Ym and Yn constraint with y constraint.  Update "isa"
17516         attribute and set "preferred_for_speed" attribute from
17517         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
17518         (*movsi_internal): Ditto.
17519         (*movdf_internal): Ditto.
17520         (*movsf_internal): Ditto.
17521         (*zero_extendsidi2): Ditto.
17522         * config/i386/sse.md (vec_set<mode>_0): Ditto.
17523         (sse2_loadld): Ditto.
17524         (*vec_extract<ssevecmodelower>_0): Ditto.
17525         (*vec_extractv4si_0_zext_sse4): Ditto.
17526         (vec_concatv2di): Ditto.
17527         (*vec_dup<mode>): Ditto.
17528         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
17529         * config/i386/constraints.md (Yi): Remove.
17530         (Yj): Remove.
17531         (Ym): Remove.
17532         (Yn): Remove.
17534 2018-04-26  Nathan Sidwell  <nathan@acm.org>
17536         * dumpfile.c (dump_open): New.
17537         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
17538         (dump_finish): Detect stdio/stderr by value not name.
17540 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
17542         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
17544 2018-04-26  Tom de Vries  <tom@codesourcery.com>
17546         PR target/84952
17547         * config/nvptx/nvptx.c (verify_neutering_jumps)
17548         (verify_neutering_labels): New function
17549         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
17551 2018-04-26  Tom de Vries  <tom@codesourcery.com>
17553         PR target/84025
17554         * config/nvptx/nvptx.c (needs_neutering_p): New function.
17555         (nvptx_single): Use needs_neutering_p to skip over insns that do not
17556         need neutering.
17558 2018-04-26  Richard Biener <rguenther@suse.de>
17559             Tom de Vries  <tom@codesourcery.com>
17561         PR lto/85422
17562         * lto-streamer-out.c (output_function): Fixup loops if required to match
17563         discovery done in the reader.
17565 2018-04-26  Richard Biener  <rguenther@suse.de>
17567         PR tree-optimization/85116
17568         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
17569         have a loop exit from the single latch predecessor.  Remove
17570         case of header with just condition.
17571         (ch_base::copy_headers): Exclude infinite loops from any
17572         processing.
17573         (pass_ch::execute): Record exits.
17575 2018-04-26  Richard Biener  <rguenther@suse.de>
17577         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
17578         prologue cost vector and pass it to vect_get_load_cost.
17579         (vect_get_peeling_costs_all_drs): Likewise.
17580         (vect_peeling_hash_get_lowest_cost): Likewise.
17581         (vect_enhance_data_refs_alignment): Likewise.
17583 2018-04-26  Richard Biener  <rguenther@suse.de>
17585         PR middle-end/85450
17586         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
17587         checking of integer<->pointer conversions.
17588         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
17589         sign-/zero-extending pointer types.
17590         (expand_omp_for_static_chunk): Likewise.
17592 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
17593             Jean Lee  <xiaoyur347@gmail.com>
17595         * config/mips/mips.c (mips_asan_shadow_offset): New function.
17596         (TARGET_ASAN_SHADOW_OFFSET): Define.
17597         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
17598         true for -fsanitize=address.
17600 2018-04-25  Mark Wielaard  <mark@klomp.org>
17602         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
17603         shorter ones.
17605 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
17607         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
17608         than "alu", remove explicit "memory" and "imm_disp" attributes.
17609         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
17611         PR middle-end/85414
17612         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
17613         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
17614         gen_lowpart_no_emit.
17616 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
17618         PR target/85473
17619         * config/i386/i386.c (ix86_expand_builtin): Change memory
17620         operand to XI, extend p0 to Pmode.
17621         * config/i386/i386.md: Change unspec volatile and operand
17622         1 mode to XI, change operand 0 mode to P.
17624 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17626         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
17627         GET_MODE_MASK before any checking.
17628         (nds32_can_use_bset_p): Likewise.
17629         (nds32_can_use_btgl_p): Likewise.
17631 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17633         * config/nds32/nds32-doubleword.md: New define_split pattern for
17634         illegal register number.
17636 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17638         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
17640 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17642         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
17644 2018-04-25  Richard Biener  <rguenther@suse.de>
17646         * lto-streamer.h (LTO_major_version): Bump to 8.
17648 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
17650         * BASE-VER: Set to 9.0.0.
17652 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
17654         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
17655         in __abskf2 and __powikf2.
17657 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17659         PR target/85512
17660         * config/aarch64/constraints.md (Usg, Usj): New constraints.
17661         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
17662         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
17663         Use the above on operand 2.  Reindent.
17664         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
17666 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
17668         PR target/85485
17669         * common/config/i386/i386-common.c (ix86_handle_option): Don't
17670         handle OPT_mcet.
17671         * config/i386/i386.opt (mcet): Removed.
17672         * doc/install.texi: Remove -mcet documentation.
17673         * doc/invoke.texi: Likewise.
17675 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
17677         PR target/85485
17678         * doc/install.texi: Remove -mcet from bootstrap-cet.
17680 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17682         PR target/85511
17683         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
17684         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
17685         if TARGET_64BIT.
17687         PR target/85503
17688         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
17689         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
17690         containing a CONST_VECTOR.
17692 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
17694         * doc/install.texi: Update newlib dependency for nvptx.
17696 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17698         PR target/85508
17699         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
17700         instead of INTVAL when shifting x left.
17702 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
17704         PR tree-optimization/85478
17705         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
17706         vect_grouped_store_supported for single element vectors.
17708 2018-04-24  Richard Biener  <rguenther@suse.de>
17710         PR target/85491
17711         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
17712         load cost increase to the case of non-constant step.
17714 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17716         PR target/84828
17717         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
17718         destination if any_malformed_asm.
17720 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
17722         PR middle-end/85496
17723         * expr.c (store_field): In the bitfield case, if the value comes from
17724         a function call and is returned in registers by means of a PARALLEL,
17725         do not change the mode of the temporary unless BLKmode and VOIDmode.
17727 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
17729         PR rtl-optimization/85423
17730         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
17731         dependencies to debug insns when the previous insn is non-debug.
17733 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
17735         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
17736         enums into a single definition.
17737         (fls): Fix predicates and printing.
17738         (seti): Likewise.
17740 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
17742         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
17743         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
17744         and short u6 immediate.
17745         (check_if_valid_sleep_operand): Remove.
17746         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
17748 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17750         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
17751         flag_always_save_lp condition.
17752         * config/nds32/nds32.opt (malways-save-lp): New option.
17754 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17756         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
17757         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
17758         * config/nds32/nds32.h
17759         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
17760         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
17762 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17764         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
17765         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
17767 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17768             Chung-Ju Wu  <jasonwucj@gmail.com>
17770         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
17771         Declare.
17772         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
17773         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
17775 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17777         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
17779 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17781         * config/nds32/nds32-protos.h (nds32_data_alignment,
17782         nds32_local_alignment): Declare.
17783         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
17784         nds32_local_alignment): New functions.
17785         (TARGET_CONSTANT_ALIGNMENT): Define.
17786         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
17788 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17790         * config/nds32/nds32.c
17791         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
17792         (TARGET_MODES_TIEABLE_P): Likewise.
17794 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17796         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
17797         level Ofast and Og.
17799 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
17800             Chung-Ju Wu  <jasonwucj@gmail.com>
17802         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17803         for unaligned access.
17804         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17805         unaligned access.
17806         * config/nds32/nds32-intrinsic.md: Likewise.
17807         * config/nds32/nds32_intrinsic.h: Likewise.
17808         * config/nds32/nds32.h (nds32_builtins): Likewise.
17809         * config/nds32/nds32.opt (munaligned-access): New option.
17810         * config/nds32/nds32.c (nds32_asm_file_start): Display
17811         flag_unaligned_access status.
17813 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
17815         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
17816         -mno-relax is present.
17817         * config/riscv/linux.h (LINK_SPEC): Ditto.
17819 2018-04-20  Martin Sebor  <msebor@redhat.com>
17821         PR c/85365
17822         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
17823         for null pointers.
17824         (gimple_fold_builtin_stxcpy_chk): Same.
17825         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
17827 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
17829         PR target/85456
17830         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
17831         __powikf2 when long double is IEEE 128-bit.
17833 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
17835         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
17836         step to make sure stack always aligned.
17838 2018-04-20  Carl Love  <cel@us.ibm.com>
17840         PR target/83402
17841         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
17842         size check for arg0.
17844 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
17845             Tom de Vries  <tom@codesourcery.com>
17847         PR target/85445
17848         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17849         Emit insns for calls too.
17850         (nvptx_find_par): Always look for worker-level predecessor insn.
17851         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
17852         calls.
17853         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
17854         (nvptx_process_pars): Propagate frames for calls.
17856 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
17858         PR target/85469
17859         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
17860         Removed.
17861         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
17862         (ix86_handle_option): Don't handle OPT_mibt.
17863         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
17864         __SHSTK__.
17865         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
17866         has_ibt and ibt.
17867         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
17868         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
17869         (ix86_target_macros): Define __CET__ with flag_cf_protection
17870         for -fcf-protection.
17871         * config/i386/i386.c (isa2_opts): Remove -mibt.
17872         * config/i386/i386.h (TARGET_IBT): Removed.
17873         (TARGET_IBT_P): Likewise.
17874         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
17875         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
17876         * config/i386/i386.opt (mcet): Update help message.
17877         (mshstk): Likewise.
17878         (mibt): Removed.
17879         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
17880         -mcet as an alias for -mshstk.
17882 2018-04-20  Richard Biener <rguenther@suse.de>
17884         PR middle-end/85475
17885         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
17886         complexity by forcing a single use of the multiply operand.
17888 2018-04-20  Martin Jambor  <mjambor@suse.cz>
17890         ipa/85449
17891         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
17892         recursion dependency to only apply to non-clones.
17894 2018-04-20  Martin Jambor  <mjambor@suse.cz>
17896         ipa/85447
17897         * ipa-cp.c (create_specialized_node): Check that clones of
17898         self-recursive edges exist during IPA-CP.
17900 2018-04-19  Toon Moene  <toon@moene.org>
17902         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
17903         by -O3.
17905 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
17907         PR tree-optimization/85467
17908         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
17909         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
17910         VECTOR_CST element to type.
17912 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17914         PR target/85397
17915         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
17916         * config/i386/i386.md (builtin_setjmp_setup): Removed.
17917         (builtin_longjmp): Likewise.
17918         (save_stack_nonlocal): New pattern.
17919         (restore_stack_nonlocal): Likewise.
17921 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17923         PR target/85404
17924         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17925         Replace ASM_OUTPUT_LABEL with fprintf.
17927 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17929         PR target/85417
17930         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17931         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17932         * config/i386/i386-c.c (ix86_target_macros_internal): Also
17933         define __IBT__ and __SHSTK__ for -fcf-protection.
17934         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17935         TARGET_IBT.
17936         (ix86_trampoline_init): Likewise.
17937         (x86_output_mi_thunk): Likewise.
17938         (ix86_notrack_prefixed_insn_p): Likewise.
17939         (ix86_option_override_internal): Don't disallow -fcf-protection.
17940         * config/i386/i386.md (rdssp<mode>): Also enable for
17941         -fcf-protection.
17942         (incssp<mode>): Likewise.
17943         (nop_endbr): Likewise.
17944         * config/i386/i386.opt (mcet): Change help message to built-in
17945         functions only.
17946         (mibt): Likewise.
17947         (mshstk): Likewise.
17948         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17949         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
17950         enable CET built-in functions.
17952 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
17954         * common/config/i386/i386-common.c
17955         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17956         OPTION_MASK_ISA_MOVDIRI_UNSET,
17957         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17958         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17959         * config.gcc (movdirintrin.h): New header.
17960         * config/i386/cpuid.h (bit_MOVDIRI,
17961         bit_MOVDIR64B): New bits.
17962         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17963         and -mmvodir64b.
17964         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17965         (VOID, PVOID, PCVOID)): New function types.
17966         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17967         __builtin_ia32_directstoreu_u64,
17968         __builtin_ia32_movdir64b): New builtins.
17969         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17970         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17971         and -mmovdiri.
17972         (ix86_valid_target_attribute_inner_p): Ditto.
17973         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17974         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17975         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17976         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17977         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17978         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17979         (movdiri<mode>, movdir64b_<mode>): New.
17980         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17981         * config/i386/immintrin.h: Include movdirintrin.h.
17982         * config/i386/movdirintrin.h: New file.
17983         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17985 2018-04-19  Richard Biener  <rguenther@suse.de>
17987         PR middle-end/85455
17988         * cfg.c (clear_bb_flags): When loop state says we have
17989         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17991 2018-04-19  Richard Biener  <rguenther@suse.de>
17993         PR tree-optimization/84737
17994         * tree-vect-data-refs.c (vect_copy_ref_info): New function
17995         copying restrict info.
17996         (vect_setup_realignment): Use it.
17997         * tree-vectorizer.h (vect_copy_ref_info): Declare.
17998         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17999         the first DR to all generated stores.
18000         (vectorizable_load): Likewise for loads.
18002 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18004         PR tree-optimization/85446
18005         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
18006         the integral and pointer types to have the same precision.
18008         * doc/install.texi: Document --disable-cet being the default and
18009         --enable-cet=auto.
18011 2018-04-18  Martin Liska  <mliska@suse.cz>
18013         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
18014         style.
18016 2018-04-18  Martin Liska  <mliska@suse.cz>
18018         Revert
18019         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18021         PR ipa/83983
18022         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18023         arguments if they are comparable.
18025 2018-04-18  Martin Liska  <mliska@suse.cz>
18027         Revert
18028         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
18030         PR lto/84805
18031         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18032         incomplete types.
18034 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
18036         PR target/85388
18037         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
18038         ENDBR after calling __morestack.
18040 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
18042         PR jit/85384
18043         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
18044         by using gcc_base_ver to generate a gcc_driver_version, and use
18045         it when generating GCC_DRIVER_NAME.
18046         * configure: Regenerate.
18048 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
18050         PR target/81084
18051         * config.gcc: Obsolete powerpc*-*-*spe*.
18053 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18055         PR debug/84637
18056         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
18057         (stabstr_D): Change type of unum from unsigned int to
18058         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
18059         type.
18061 2018-04-17  Jim Wilson  <jimw@sifive.com>
18063         PR 84856
18064         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
18065         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
18066         Set arg_pointer_offset after using pretend_args_size.
18068 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18070         PR rtl-optimization/85431
18071         * dse.c (record_store): Ignore zero width stores.
18073         PR sanitizer/85230
18074         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
18075         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
18076         __builtin_stack_restore rather than after it.
18077         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18078         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
18079         argument instead of virtual_dynamic_stack_rtx.
18081 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18083         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
18084         New prototype.
18085         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
18086         Add note to error message to explain internal mapping of overloaded
18087         built-in function name to non-overloaded built-in function name.
18088         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
18089         function.
18091 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
18093         PR target/85424
18094         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
18095         where the inputs overlap with the output.
18097 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18099         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
18100         (=v, v) alternative and explicit "memory" attribute.
18101         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
18102         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18103         attributes.
18104         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
18105         "sselog1" type instead of "sselog".
18106         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
18107         "sselog".  Remove explicit "memory" attribute.
18108         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
18109         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18110         attributes.
18111         (vec_extract_hi_v32hi): Merge all alternatives into one, use
18112         "sselog1" type instead of "sselog".  Remove explicit "memory"
18113         attribute.
18114         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
18115         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18116         attribute.
18117         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18118         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18119         attributes.
18120         (vec_extract_hi_v64qi): Merge all alternatives into one, use
18121         "sselog1" type instead of "sselog".  Remove explicit "memory"
18122         attribute.
18123         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18124         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18125         attribute.
18127         PR target/85430
18128         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18130         PR middle-end/85414
18131         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18132         on a SUBREG.
18134 2018-04-17  Martin Jambor  <mjambor@suse.cz>
18136         PR ipa/85421
18137         * ipa-cp.c (create_specialized_node): Call
18138         expand_all_artificial_thunks if necessary.
18140 2018-04-17  Martin Liska  <mliska@suse.cz>
18142         PR lto/85405
18143         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18144         in message, remote space in between '_G' and '('.
18146 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18148         PR target/85281
18149         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18150         avx512f_vmcmp<mode>3<round_saeonly_name>,
18151         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18152         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18153         avx512f_rndscale<mode><round_saeonly_name>,
18154         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18155         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18156         Use %<iptr>2 instead of %2 for -masm=intel.
18157         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18158         avx512f_vcvttss2usi<round_saeonly_name>,
18159         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18160         -masm=intel.
18161         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18162         avx512f_vcvttsd2usi<round_saeonly_name>,
18163         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18164         Use %q1 instead of %1 for -masm=intel.
18165         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
18166         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
18167         of %3 for -masm=intel.
18168         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
18169         -masm=intel.
18170         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
18171         -masm=intel.
18172         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
18173         -masm=intel.
18174         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
18175         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18176         %g1.
18177         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18178         -masm=intel.
18179         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18180         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18181         %g1 and one with %0 and %1.
18182         (avx512er_vmrcp28<mode><round_saeonly_name>,
18183         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18184         %1 for -masm=intel.
18185         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18186         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18187         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18188         of %0 and %{%4%} for -masm=intel.
18189         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18190         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18191         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18192         order of %0 and %{%5%}%{z%} for -masm=intel.
18194 2018-04-17  Jan Hubicka  <jh@suse.cz>
18196         PR lto/85405
18197         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18199 2018-04-17  Martin Liska  <mliska@suse.cz>
18201         PR ipa/85329
18202         * multiple_target.c (create_dispatcher_calls): Set apostrophes
18203         for target_clone error message.  Make default implementation
18204         clone to be a local declaration.
18205         (separate_attrs): Add new argument and check for an empty
18206         string.
18207         (expand_target_clones): Handle it.
18208         (ipa_target_clone): Make redirection just for target_clones
18209         functions.
18211 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
18212             Tom de Vries  <tom@codesourcery.com>
18214         PR middle-end/84955
18215         * omp-expand.c (expand_oacc_for): Add dummy false branch for
18216         tiled basic blocks without omp continue statements.
18218 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
18220         PR target/83660
18221         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
18222         vec_extract expression as having side effects to make sure it gets
18223         a cleanup point.
18225 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
18227         PR target/85403
18228         * config/i386/i386.c (get_builtin_code_for_version): Check
18229         error_mark_node.
18231 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
18233         PR target/84331
18234         * config.gcc: Support "skylake".
18235         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18236         PROCESSOR_SKYLAKE.
18237         * config/i386/i386.c (m_SKYLAKE): Define.
18238         (processor_target_table): Add "skylake".
18239         (ix86_option_override_internal): Add "skylake".
18240         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
18241         PROCESSOR_CANNONLAKE.
18242         (get_builtin_code_for_version): Fix priority for
18243         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
18244         PROCESSOR_SKYLAKE-AVX512.
18245         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
18246         (processor_type): Add PROCESSOR_SKYLAKE.
18248 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
18249             Jason Merrill  <jason@redhat.com>
18251         PR c++/85112
18252         * convert.c (convert_to_integer_1): Use direct recursion for
18253         enumeral types and types with a precision less than the number
18254         of bits in their mode.
18256 2018-04-16  Julia Koval  <julia.koval@intel.com>
18258         PR target/84413
18259         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
18260         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
18262 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
18264         PR target/85293
18265         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
18266         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
18267         and -mno-direct-move.
18269 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
18271         PR target/83402
18272         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
18273         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
18274         Ensure negative shifts result in {0}.
18276 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
18278         PR rtl-optimization/79916
18279         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
18280         regs (if any) to define how to gnerate SD moves when LRA is in
18281         progress.
18283 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
18285         PR rtl-optimization/85393
18286         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
18287         * except.c (expand_dw2_landing_pad_for_region): Make static.
18288         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
18289         a label and unconditional jump to old_bb, rather than
18290         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
18291         basic block.
18293         PR rtl-optimization/85376
18294         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
18295         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
18296         instead of a specific value.
18298 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18299             Bin Cheng  <bin.cheng@arm.com>
18301         PR tree-optimization/82965
18302         PR tree-optimization/83991
18303         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
18304         by_profile_only parameter.
18305         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
18306         information if the loop was predicted to iterate too many times.
18307         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
18309 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18311         PR lto/71991
18312         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
18313         always inline.
18315 2018-04-13  Martin Liska  <mliska@suse.cz>
18316             Jakub Jelinek  <jakub@redhat.com>
18318         PR middle-end/81657
18319         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
18320         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
18321         * builtins.c (expand_builtin_memory_copy_args): Use
18322         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
18323         handle dest_addr == pc_rtx.
18325 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
18327         PR target/85291
18328         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
18329         asked to not generate direct moves.
18330         (fix_trunc<mode>si2_stfiwx): Similar.
18331         (fix_trunc<mode>si2_internal): Similar.
18333 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18335         PR debug/83157
18336         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
18337         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
18338         lookup if dest in some wider mode is known to be const0_rtx and
18339         if so, record permanent equivalence for it to be ZERO_EXTEND of
18340         the narrower mode destination.
18342 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18344         * lto-streamer-out.c (output_function): Revert 259346.
18345         * omp-expand.c (expand_oacc_for): Likewise.
18347 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
18349         PR rtl-optimization/85354
18350         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
18351         * sel-sched.c (sel_global_init): ... here.
18353 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
18355         PR target/85238
18356         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
18357         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
18358         mode for PE-COFF targets.
18359         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
18360         (i386_pe_asm_lto_end): Likewise.
18361         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
18362         (TARGET_ASM_LTO_END): Likewise.
18363         * config/i386/winnt.c (saved_debug_info_level): New static variable.
18364         (i386_pe_asm_lto_start): New function.
18365         (i386_pe_asm_lto_end): Likewise.
18367 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18368             Richard Biener  <rguenther@suse.de>
18370         PR middle-end/84955
18371         * lto-streamer-out.c (output_function): Fix CFG loop state before
18372         streaming out.
18373         * omp-expand.c (expand_oacc_for): Handle calls to internal
18374         functions like regular functions.
18376 2018-04-12  Richard Biener  <rguenther@suse.de>
18378         PR lto/85371
18379         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
18380         for the early LTO debug to properly generate references to it
18381         during DIE emission.  Do not re-use that for the skeleton for
18382         split-dwarf.
18383         (dwarf2out_early_finish): Likewise.
18385 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18387         PR target/85328
18388         * config/i386/sse.md
18389         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
18390         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
18391         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
18392         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
18393         and output is a reg, avoid creating invalid lowpart subreg, but
18394         instead split into a 512-bit move.  Don't split if not AVX512VL,
18395         input is xmm16+ reg and output is a mem.
18396         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
18397         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
18398         xmm16+ reg and output is a mem.
18400 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18402         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
18403         also for flag_dwarf2_cfi_asm.
18405 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18407         PR rtl-optimization/85342
18408         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
18409         a bool scalar var inside of the loop instead.  Don't try to update
18410         recog_data.operand after failed apply_change_group.
18412 2018-04-12  Tom de Vries  <tom@codesourcery.com>
18414         PR target/85296
18415         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
18416         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
18417         array with flexible array member as array without given dimension.
18418         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
18419         argument for undefined param to true.
18421 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
18423         PR target/85321
18424         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18425         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
18426         from PowerPC section.
18427         * config/rs6000/sysv4.opt (mcall-): Improve help text.
18428         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
18429         help text that is too long.
18430         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
18431         help text that is too long.
18432         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
18433         help text that is too long.
18435 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
18437         * config/alpha/alpha.md (stack_probe_internal): Rename
18438         from "probe_stack".  Update all callers.
18440 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18442         PR rtl-optimization/84566
18443         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
18444         sched_macro_fuse_insns.
18446 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18448         PR target/84301
18449         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
18450         (compute_block_dependences): ... from here.
18452 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18454         PR tree-optimization/85331
18455         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
18456         from int to HOST_WIDE_INT.
18458 2018-04-11  Martin Jambor  <mjambor@suse.cz>
18460         PR ipa/84149
18461         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
18462         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
18463         not the same as the source val.
18464         (cgraph_edge_brings_value_p): New parameter.
18465         (gather_edges_for_value): Pass destination value to
18466         cgraph_edge_brings_value_p.
18467         (perhaps_add_new_callers): Likewise.
18468         (get_info_about_necessary_edges): Likewise and exclude values brought
18469         only by self-recursive edges.
18470         (create_specialized_node): Redirect only clones of self-calling edges.
18471         (+self_recursive_pass_through_p): New function.
18472         (find_more_scalar_values_for_callers_subset): Use it.
18473         (find_aggregate_values_for_callers_subset): Likewise.
18474         (known_aggs_to_agg_replacement_list): Removed.
18475         (decide_whether_version_node): Re-calculate known constants for all
18476         remaining context clones.
18478 2018-04-11  Richard Biener  <rguenther@suse.de>
18480         PR lto/85339
18481         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
18482         from early DWARF output.
18483         (dwarf2out_early_finish): Output line info unconditionally into
18484         early DWARF and add reference to it.
18486 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18488         PR target/85281
18489         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
18490         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
18491         other than V2DFmode using iptr mode attribute.
18492         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
18494 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18496         PR rtl-optimization/84659
18497         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
18499 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18501         PR debug/85302
18502         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
18503         SIZEP is NULL.
18504         (output_loc_list): Pass address of a dummy size variable even in the
18505         locview handling loop.
18506         (index_location_lists): Add comment on why skip_loc_list_entry can't
18507         call size_of_locs.
18509 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18511         PR target/85261
18512         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
18513         into register.
18515 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
18517         PR target/85321
18518         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18519         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
18520         and -mstring-compare-inline-limit.
18522 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18524         PR target/85287
18525         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
18526         for stack clash protection in a register whenever we need it to be in
18527         a register.
18529 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18531         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
18532         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
18534 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18536         PR target/85321
18537         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
18538         the help text.
18539         (mlong-double-): Ditto.
18540         * config/rs6000/sysv4.opt (msdata=): Ditto.
18541         (mtls-size=): Ditto.
18543 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18545         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18546         erroneous entries for
18547         "vector int vec_ldl (int, long int *)", and
18548         "vector unsigned int vec_ldl (int, unsigned long int *)".
18549         Add comments and entries for
18550         "vector bool char vec_ldl (int, bool char *)",
18551         "vector bool short vec_ldl (int, bool short *)",
18552         "vector bool int vec_ldl (int, bool int *)",
18553         "vector bool long long vec_ldl (int, bool long long *)",
18554         "vector pixel vec_ldl (int, pixel *)",
18555         "vector long long vec_ldl (int, long long *)",
18556         "vector unsigned long long vec_ldl (int, unsigned long long *)".
18557         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
18558         type tree bool_long_long_type_node and correct definition of
18559         bool_V2DI_type_node to make reference to this new type tree.
18560         (rs6000_mangle_type): Replace erroneous reference to
18561         bool_long_type_node with bool_long_long_type_node.
18562         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
18563         comments to emphasize sign distinctions for char and int types and
18564         replace RS6000_BTI_bool_long constant with
18565         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
18566         use of RS6000_BTI_pixel.
18567         (bool_long_type_node): Remove this macro definition.
18568         (bool_long_long_type_node): New macro definition
18570 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
18572         PR rtl-optimization/85300
18573         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
18574         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
18575         simplify_unary_operation fails.
18577 2018-04-10  Martin Liska  <mliska@suse.cz>
18579         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
18580         cgraph_edge and ipa_ref.
18582 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
18584         PR target/85177
18585         PR target/85255
18586         * config/i386/sse.md
18587         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
18588         computation of the VEC_MERGE selector from mask.
18589         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
18590         Fix decoding of the VEC_MERGE selector into mask.
18592 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
18594         PR tree-optimization/85286
18595         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
18597 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
18599         * final.c (final_1): Set insn_last_address as well as
18600         insn_current_address.
18602 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18604         PR target/85173
18605         * explow.c (emit_stack_probe): Call validize_mem on memory location
18606         before passing it to gen_probe_stack.  Create address operand and
18607         legitimize it for the probe_stack_address case.
18609 2018-04-09  Jan Hubicka  <jh@suse.cz>
18611         PR lto/85078
18612         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
18613         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
18614         * tree.c (free_lang_data_in_type): Fix handling of binfos;
18615         walk basetypes.
18616         (free_lang_data): Rebuild type inheritance graph.
18618 2018-04-09  Martin Sebor  <msebor@redhat.com>
18620         * invoke.texi (-finline-small-functions): Mention other optimization
18621         options.
18622         (-findirect-inlining, -fpartial-inlining): Same.
18623         (-finline-functions-called-once): Same.
18624         (-freorder-blocks-and-partition): Same.
18626 2018-04-09  Jan Hubicka  <jh@suse.cz>
18628         PR rtl/84058
18629         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
18630         jumps; choose last target that matches the criteria (i.e.
18631         no partition changes for non-crossing jumps).
18632         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
18633         support for redirecting crossing jumps to non-crossing.
18635 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
18637         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
18638         also for naked functions.
18640 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
18642         * config/arc/arc.md (add_shift): New pattern.
18643         (add_shift2): Likewise.
18644         (sub_shift): Likewise.
18645         (sub_shift_cmp0_noout): Likewise.
18646         (compare_si_ashiftsi): Likewise.
18647         (xbfu_cmp0_noout): New combine pattern.
18648         (xbfu_cmp0"): Likewise.
18649         (movsi_set_cc_insn): Place the predicable variant first.
18650         (commutative_binary_cmp0_noout): Remove clobber.
18651         (commutative_binary_cmp0): New pattern.
18652         (noncommutative_binary_cmp0): Likewise.
18653         (noncommutative_binary_cmp0_noout): Likewise.
18654         (noncommutative_binary_comparison_result_used): Removed.
18655         (rsub_cmp0): New pattern.
18656         (rsub_cmp0_noout): Likewise.
18657         (extzvsi): Changed, keep only meaningful variants.
18658         (SQH, SEZ): New iterators.
18659         (SQH_postfix): New mode attribute.
18660         (SEZ_prefix): New code attribute.
18661         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
18662         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
18663         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
18664         of numerical value.
18665         (noncommutative_operator): Check the availability of barrel
18666         shifter option.
18668 2018-04-09  Richard Biener  <rguenther@suse.de>
18670         PR tree-optimization/85284
18671         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
18672         Only use the niter constraining form of simple_iv when the exit
18673         is always executed.
18675 2018-04-09  Tom de Vries  <tom@codesourcery.com>
18677         PR target/84041
18678         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18679         (define_expand "*memory_barrier"): New define_expand.
18680         (define_insn "memory_barrier"): New insn.
18682 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18684         PR rtl-optimization/80463
18685         PR rtl-optimization/83972
18686         PR rtl-optimization/83480
18688         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
18689         correct producer for the insn.
18690         (tidy_control_flow): Fixup seqnos in case of debug insns.
18692 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18694         PR rtl-optimization/83913
18696         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
18697         different sched-times when merging exprs.
18699 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18701         PR rtl-optimization/83962
18703         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
18704         tidy_fallthru_edge and tidy_control_flow.
18706 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18708         PR rtl-optimization/83530
18710         * sel-sched.c (force_next_insn): New global variable.
18711         (remove_insn_for_debug): When force_next_insn is true, also leave only
18712         next insn in the ready list.
18713         (sel_sched_region): When the region wasn't scheduled, make another pass
18714         over it with force_next_insn set to 1.
18716 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
18718         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
18719         into tm_file.
18720         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18721         for interrupt control.
18722         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18723         functions for interrupt control.
18724         * config/nds32/nds32-intrinsic.md: Likewise.
18725         * config/nds32/nds32_intrinsic.h: Likewise.
18726         * config/nds32/nds32.h (nds32_builtins): Likewise.
18728 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
18730         * config/nds32/nds32.c (nds32_init_machine_status,
18731         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
18732         strict_aligned_p field.
18733         (nds32_expand_to_rtl_hook): New function.
18734         (TARGET_EXPAND_TO_RTL_HOOK): Define.
18735         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
18737 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18738             Chung-Ju Wu  <jasonwucj@gmail.com>
18740         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
18741         * config/nds32/nds32-n7.md: New file.
18742         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
18743         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
18744         pipeline.
18745         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
18746         * config/nds32/nds32.md (pipeline_model): Add n7.
18747         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
18748         * config/nds32/pipelines.md: Include n7 settings.
18750 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18751             Chung-Ju Wu  <jasonwucj@gmail.com>
18753         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
18754         * config/nds32/nds32-e8.md: New file.
18755         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
18756         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
18757         pipeline.
18758         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
18759         * config/nds32/nds32.md (pipeline_model): Add e8.
18760         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
18761         * config/nds32/pipelines.md: Include e8 settings.
18763 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18764             Chung-Ju Wu  <jasonwucj@gmail.com>
18766         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
18767         * config/nds32/nds32-n8.md: New file.
18768         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
18769         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
18770         pipeline.
18771         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
18772         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
18773         * config/nds32/nds32.md (pipeline_model): Add n8.
18774         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
18775         * config/nds32/pipelines.md: Include n8 settings.
18777 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18778             Chung-Ju Wu  <jasonwucj@gmail.com>
18780         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
18781         * config/nds32/nds32-n9-2r1w.md: New file.
18782         * config/nds32/nds32-n9-3r2w.md: New file.
18783         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
18784         nds32_register_ports): New or modify for cpu n9.
18785         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
18786         pipeline.
18787         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
18788         * config/nds32/nds32-utils.c: New file.
18789         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
18790         TARGET_MUL_SLOW): Define.
18791         * config/nds32/nds32.md (pipeline_model): New attribute.
18792         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
18793         New options that support cpu n9.
18794         * config/nds32/pipelines.md: Include n9 settings.
18795         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
18797 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
18799         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
18800         information if necessary.
18801         (output_cond_branch_compare_zero): Likewise.
18802         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
18803         (nds32_target_alignment): Refine for alignment.
18804         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
18805         (FUNCTION_BOUNDARY): Modify.
18806         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
18807         align case.
18808         * config/nds32/nds32.opt (malways-align, malign-functions): New.
18810 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
18812         * config/nds32/constants.md (unspec_volatile_element): Add values for
18813         TLB operation and data prefetch.
18814         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18815         functions for TLB operation and data prefetch.
18816         * config/nds32/nds32-intrinsic.md: Likewise.
18817         * config/nds32/nds32_intrinsic.h: Likewise.
18818         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
18819         (nds32_print_operand): Likewise.
18820         * config/nds32/nds32.h (nds32_builtins): Likewise.
18822 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
18823         Andrew Pinski <pinsika@gcc.gnu.org>
18825         PR middle-end/82976
18826         * match.pd: Use constant_boolean_node of correct type instead of
18827         boolean_true_node or boolean_false_node for simplifying
18828         pointer comparisons to zero.
18830 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
18832         PR tree-optimization/80021
18833         * tree.c (verify_type_variant): Make error call in verify_variant_match
18834         translatable and remove final full stop.
18836 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18838         * config/nds32/constants.md (unspec_volatile_element): Add
18839         UNSPEC_VOLATILE_EH_RETURN.
18840         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
18841         nds32_output_stack_pop): Support dwarf exception handling process.
18842         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
18843         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
18844         exception handling process.
18845         (nds32_compute_stack_frame): Likewise.
18846         (nds32_return_addr_rtx): Likewise.
18847         (nds32_initial_elimination_offset): Likewise.
18848         (nds32_expand_prologue): Likewise.
18849         (nds32_expand_epilogue): Likewise.
18850         (nds32_dynamic_chain_address): New function.
18851         * config/nds32/nds32.h (machine_function): Add fields for dwarf
18852         exception handling.
18853         (DYNAMIC_CHAIN_ADDRESS): Define.
18854         (EH_RETURN_DATA_REGNO): Define.
18855         (EH_RETURN_STACKADJ_RTX): Define.
18856         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
18857         patterns for dwarf exception handling.
18859 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18861         * config/nds32/nds32.h: Clean up obsolete macros.
18863 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18865         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18866         Add enum values for particular instructions.
18867         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18868         particular intrinsic functions.
18869         * config/nds32/nds32-intrinsic.md: Likewise.
18870         * config/nds32/nds32_intrinsic.h: Likewise.
18871         * config/nds32/nds32.h (nds32_builtins): Likewise.
18872         * config/nds32/nds32.md (type): Add pbsad and pbsada.
18873         (btst, ave): New patterns for particular instructions.
18875 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18877         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18878         Add enum values for atomic load/store and memory sync.
18879         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
18880         and memory sync.
18881         * config/nds32/nds32-intrinsic.md: Likewise.
18882         * config/nds32/nds32_intrinsic.h: Likewise.
18883         * config/nds32/nds32.h (nds32_builtins): Likewise.
18885 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
18887         PR tree-optimization/85257
18888         * fold-const.c (native_encode_vector): If not all elts could fit
18889         and off is -1, return 0 rather than offset.
18890         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
18891         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
18892         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
18893         adjust buffer in native_interpret_expr call.
18895 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18897         * config/nds32/constants.md (unspec_volatile_element): Add cache
18898         control enum values.
18899         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
18900         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
18901         * config/nds32/nds32.c (nds32_cctl_names): New.
18902         (nds32_print_operand): Handle cache control register names.
18903         * config/nds32/nds32.h (nds32_builtins): New enum values.
18904         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
18905         macros.
18906         * config/nds32/nds32.md (type): Add mmu.
18907         * config/nds32/pipelines.md (simple_insn): Add mmu.
18909 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18911         * config/nds32/nds32.md (type): Remove call.
18912         * config/nds32/pipelines.md (simple_insn): Likewise.
18914 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18916         * config/nds32/constants.md (unspec_volatile_element): Add
18917         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
18918         UNSPEC_VOLATILE_FMFCFG.
18919         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
18920         description for fmfcfg and fmfcsr.
18921         (bdesc_1arg): Add fmtcsr.
18922         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
18923         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18924         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18925         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18926         unspec_fmfcfg): New patterns.
18927         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18928         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18929         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18930         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18931         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18932         __nds32__fmfcfg): Define.
18934 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18936         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18937         intrinsic register names.
18938         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18939         intrinsic register enum values and macros.
18941 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18943         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18944         for load/store addressing form.
18945         (nds32_print_operand_address): Likewise.
18947 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
18949         PR target/85196
18950         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18951         based on LABEL_REF.  Remove useless assertion.
18952         (pic_address_needs_scratch): Fix formatting.
18953         (sparc_legitimize_pic_address): Minor tweaks.
18954         (sparc_delegitimize_address): Adjust assertion accordingly.
18955         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18956         into symbolic_operand.
18957         (movsi_high_pic_label_ref): Likewise.
18958         (movsi_lo_sum_pic_label_ref): Likewise.
18959         (movdi_pic_label_ref): Likewise.
18960         (movdi_high_pic_label_ref): Likewise.
18961         (movdi_lo_sum_pic_label_ref): Likewise.
18963 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
18965         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18966         custom LIB_SPEC setup.
18968 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
18969             Kito Cheng  <kito.cheng@gmail.com>
18971         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18972         * config/riscv/freebsd.h: New.
18974 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18976         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18977         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18978         file.
18980 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18981             Kito Cheng  <kito.cheng@gmail.com>
18983         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18984         nds32_output_call, nds32_symbol_binds_local_p): New functions.
18985         * config/nds32/nds32-protos.h (nds32_output_call,
18986         nds32_output_return): Declare.
18987         * config/nds32/nds32.md: Refine all the call and return patterns.
18989 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18991         PR debug/85252
18992         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18993         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18995         PR rtl-optimization/84872
18996         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18997         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18998         EDGE_CROSSING edge.
19000 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
19002         * expr.c (copy_blkmode_to_reg): Revert 254862.
19003         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
19005 2018-04-06  Richard Biener  <rguenther@suse.de>
19007         PR middle-end/85244
19008         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
19009         after seeing a component reference with an adjacent field.  Treat
19010         refs to arrays at struct end of external decls similar to
19011         refs to unconstrained commons.
19013 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19015         PR sanitizer/85213
19016         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
19017         look through SAVE_EXPRs with non-side-effects argument.  Adjust
19018         recursive calls.
19019         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
19020         save_p here.
19022 2018-04-06  Richard Biener  <rguenther@suse.de>
19024         PR middle-end/85180
19025         * alias.c (find_base_term): New wrapper around find_base_term
19026         unwinding CSELIB_VAL_PTR changes.
19027         (find_base_term): Do not restore CSELIB_VAL_PTR during the
19028         recursion.
19030 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19032         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
19033         instructions.
19034         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
19035         constant definitions.
19036         ("nop"): lr 0,0 -> nopr r0
19037         ("nop_lr0", "nop_lr1"): New insn definitions.
19039 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19041         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
19042         NDS32_V3PUSH_AVAILABLE_P macro.
19044 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
19045             Chung-Ju Wu  <jasonwucj@gmail.com>
19047         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
19048         (nds32*-*-*): Add float and fpu_config into supported_defaults.
19049         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
19050         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
19051         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
19052         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
19053         * config/nds32/constraints.md: New constraints and checking for hard
19054         float configuration.
19055         * config/nds32/iterators.md: New mode iterator and attribute for hard
19056         float configuration.
19057         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
19058         patterns.
19059         * config/nds32/nds32-fpu.md: New file.
19060         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
19061         deal with hard float code generation.
19062         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
19063         ARCH_V3S.
19064         (abi_type, float_reg_number): New enum type.
19065         * config/nds32/nds32-predicates.c: New predicates for hard float.
19066         * config/nds32/nds32-protos.h: Declare functions for hard float.
19067         * config/nds32/nds32.c: Implementation for hard float configuration.
19068         * config/nds32/nds32.h: Definitions for hard float configuration.
19069         * config/nds32/nds32.md: Include hard float machine description and
19070         modify patterns for hard float configuration.
19071         * config/nds32/nds32.opt: New options for hard float configuration.
19072         * config/nds32/predicates.md: New predicates for hard float
19073         configuration.
19075 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19077         * common/config/nds32/nds32-common.c
19078         (nds32_option_optimization_table): Enable -mreleax-hint by default.
19080 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
19082         PR middle-end/85195
19083         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
19084         CONSTRUCTOR_ELT (ctor, ...)->value.
19086 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
19088         PR target/85193
19089         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
19091 2018-04-05  Tom de Vries  <tom@codesourcery.com>
19093         PR target/85204
19094         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
19095         cond jump.
19097 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19098             Kito Cheng  <kito.cheng@gmail.com>
19100         * config/nds32/constraints.md (U33): Fine-tune checking condition.
19101         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
19102         * config/nds32/nds32.h (nds32_16bit_address_type): Add
19103         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
19105 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19106             Kito Cheng  <kito.cheng@gmail.com>
19108         * config/nds32/constraints.md (Ufe): New memory constraint.
19109         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
19110         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
19111         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
19112         operands.
19113         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
19114         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
19116 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19118         * config/nds32/nds32.md: Use optimize_size in the condition for
19119         alu-shift instructions.
19121 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19123         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19125 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19127         * config/nds32/nds32.md (negsi2): Refine pattern.
19129 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19130             Chung-Ju Wu  <jasonwucj@gmail.com>
19132         * config/nds32/iterators.md (shift_rotate): New code iterator.
19133         (shift): New code attribute.
19134         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19135         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19136         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19137         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19138         bit-wise operations.
19139         (andsi3, *andsi3): Ditto.
19140         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19141         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19142         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19143         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19144         nds32_ior_operand, nds32_xor_operand): New predicates.
19146 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19148         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19149         (addsi3, subsi3): ... this.
19151 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19153         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19155 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19157         * config/nds32/nds32.md: Adjust indention.
19159 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19161         * config/nds32/nds32.md (feature): New attribute.
19163 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19165         * config/nds32/nds32.md (subtype): New attribute.
19167 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19169         PR target/85203
19170         * config/arm/arm-builtins.c (arm_expand_builtin): Change
19171         expansion to perform a bitwise AND of the argument followed by a
19172         boolean negation of the result.
19174 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
19176         PR rtl-optimization/84878
19177         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19178         the basic block.  Assert the use reference is not artificial and that
19179         it has an associated insn.
19181 2018-04-04  Michael Matz  <matz@suse.de>
19183         * builtins.c (compute_objsize): Pass correct operand
19184         to array_at_struct_end_p.
19186 2018-04-04  Richard Biener  <rguenther@suse.de>
19188         PR lto/85176
19189         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19190         from contexts for DINFO_LEVEL_TERSE and below.
19192 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19194         * config/nds32/nds32-doubleword.md (move_<mode>): Require
19195         resiter_operand condition.
19196         * config/nds32/nds32.md (*move<mode>): Ditto.
19198 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19199             Monk Chiang  <sh.chiang04@gmail.com>
19201         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19203 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19205         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19207 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19208             Kito Cheng  <kito.cheng@gmail.com>
19210         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19211         nds32_cond_code_str, output_cond_branch,
19212         output_cond_branch_compare_zero, nds32_expand_cbranch,
19213         nds32_expand_cstore, nds32_expand_movcc,
19214         nds32_output_cbranchsi4_equality_zero,
19215         nds32_output_cbranchsi4_equality_reg,
19216         nds32_output_cbranchsi4_equality_reg_or_const_int,
19217         nds32_output_cbranchsi4_greater_less_zero: New functions.
19218         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19219         nds32_expand_cstore, nds32_expand_movcc,
19220         nds32_output_cbranchsi4_equality_zero,
19221         nds32_output_cbranchsi4_equality_reg,
19222         nds32_output_cbranchsi4_equality_reg_or_const_int,
19223         nds32_output_cbranchsi4_greater_less_zero): Declare.
19224         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19225         nds32_rimm11s_operand): New predicates.
19226         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
19227         * config/nds32/nds32.md: Rewrite all the branch and conditional move
19228         patterns.
19230 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19232         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
19233         * config/nds32/nds32.md: Ditto.
19234         * config/nds32/pipelines.md: Ditto.
19236 2018-04-04  Richard Biener  <rguenther@suse.de>
19238         PR tree-optimization/85168
19239         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
19240         propagating abnormals.
19242 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19244         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
19246 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19247             Kito Cheng  <kito.cheng@gmail.com>
19249         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
19250         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
19251         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
19252         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
19253         * config/nds32/nds32.md (sibcall_internal): New.
19254         (sibcall_register): Remove.
19255         (sibcall_immediate): Remove.
19256         (sibcall_value_internal): New.
19257         (sibcall_value_register): Remove.
19258         (sibcall_value_immediate): Remove.
19259         * config/nds32/predicates.md (nds32_general_register_operand): New.
19260         (nds32_call_address_operand): New.
19262 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19264         PR rtl-optimization/85167
19265         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
19266         bb_defs if *split_p, instead preinitialize it to NULL.
19268         PR tree-optimization/85156
19269         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
19270         evaluating the argument multiple times.
19272 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
19274         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
19275         than vector.
19276         (_mm_cvtpd_ps): Likewise.
19277         (_mm_cvttpd_epi32): Likewise.
19278         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
19279         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
19280         vector, pixel, and bool following altivec.h include.
19282 2018-04-03  Martin Sebor  <msebor@redhat.com>
19284         * doc/extend.texi (Common Function Attributes): Clarify.
19285         (const attribute): Likewise.
19286         (pure attribute): Likewise.
19288 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19290         PR target/85169
19291         * config/i386/i386.c (ix86_expand_vector_set): Use
19292         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
19294 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
19296         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
19297         instructions when changing rounding bits to preserve precision bits
19298         in the x87 control word.
19300 2018-04-03  Martin Liska  <mliska@suse.cz>
19302         PR tree-optimization/82491
19303         * rtl.h (strip_offset_and_add): Replace += suboffset with
19304         poly_uint64 () + suboffset.
19306 2018-03-29  Martin Liska  <mliska@suse.cz>
19307             Martin Jambor  <mjambor@suse.cz>
19309         PR ipa/84947
19310         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
19311         param_type is not an integral or pointer type.
19313 2018-04-03  Richard Biener  <rguenther@suse.de>
19315         * sese.h (recompute_all_dominators): Remove.
19317 2018-04-02  Martin Sebor  <msebor@redhat.com>
19319         * doc/invoke.texi (-Wrestrict): Fix typos.
19321 2018-04-02  Jim Wilson  <jimw@sifive.com>
19323         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
19324         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
19325         (<optab>di3, <optab>si3_extend): Likewise.
19326         (<optab>si3_mask, <optab>si3_mask_1): New.
19327         (<optab>di3_mask, <optab>di3_mask_1): New.
19328         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
19329         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
19330         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
19332 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
19334         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
19335         example.
19337 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19339         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
19340         (nds32_canonicalize_comparison): New function.
19342 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19343             Kito Cheng  <kito.cheng@gmail.com>
19344             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19346         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
19347         * config/nds32/constants.md (unspec_volatile_element): Add
19348         UNSPEC_VOLATILE_RELAX_GROUP.
19349         * config/nds32/nds32-relax-opt.c: New file.
19350         * config/nds32/nds32-predicates.c
19351         (nds32_symbol_load_store_p): New function.
19352         * config/nds32/nds32-protos.h
19353         (nds32_symbol_load_store_p): Declare function.
19354         (make_pass_nds32_relax_opt): Declare new rtl pass function.
19355         * config/nds32/nds32.c
19356         (nds32_register_pass): New function to register pass.
19357         (nds32_register_passes): New function to register passes.
19358         * config/nds32/nds32.md (relax_group): New pattern.
19359         * config/nds32/nds32.opt (mrelax-hint): New option.
19360         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
19362 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
19364         * config/nds32/t-nds32: Modify files dependency.
19366 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19368         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
19369         (PROFILE_HOOK): Define its implementation.
19371 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19373         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
19374         type and 32-bit size.
19376 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
19378         PR middle-end/85090
19379         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
19380         (V_128_256): New mode iterator.
19381         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
19382         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
19383         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
19384         of V.
19385         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
19386         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
19388 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
19390         PR target/83315
19391         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
19392         NaN inputs correctly.
19394 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
19396         PR target/80546
19397         * config/rs6000/vsx.md (??r): New mode attribute.
19398         (*vsx_mov<mode>_64bit): Use it.
19399         (*vsx_mov<mode>_32bit): Likewise.
19401 2018-03-30  Martin Sebor  <msebor@redhat.com>
19403         PR tree-optimization/84818
19404         * builtins.c (check_access): Use warning_n.
19406 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19408         PR target/83822
19409         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
19410         condition.
19411         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
19412         condition.
19414 2018-03-30  Julia Koval  <julia.koval@intel.com>
19416         PR target/84413
19417         * x86-tune.def (movx, partial_reg_dependency): Enable for
19418         m_SKYLAKE_AVX512.
19420 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
19422         PR inline-asm/84985
19423         * lra-constraints.c (process_alt_operands): Move setting
19424         this_alternative_matches below.
19426 2018-03-29  Martin Liska  <mliska@suse.cz>
19428         PR lto/84995.
19429         * doc/invoke.texi: Document how LTO works with debug info.
19430         Describe auto-load support of binutils.  Mention 'x86-64'
19431         as valid option value of -march option.
19433 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
19435         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
19437         PR c/85094
19438         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
19439         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
19440         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
19441         checking.
19443 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
19445         PR target/84912
19446         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
19447         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
19448         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
19449         for RS6000_BTM_POWERPC64.
19450         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
19451         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
19452         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
19453         definition.
19454         (DIVDE): Use it.
19455         (DIVDEU): Likewise.
19457 2018-03-28  Carl Love  <cel@us.ibm.com>
19459         Revert
19460         2017-09-27  Carl Love  <cel@us.ibm.com>
19462         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
19463         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
19464         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
19465         fctiw instruction.
19467 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19469         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
19470         instead of __vector bool.
19471         (_mm_max_pu8): Likewise.
19472         (_mm_min_pi16): Likewise.
19474 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
19476         PR target/84912
19477         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
19478         (DIVWEUO): Likewise.
19479         (DIVDEO): Likewise.
19480         (DIVDEUO): Likewise.
19481         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
19482         DIVWEUO and DIVDEUO.
19483         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
19484         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
19485         (div_extend): Likewise.
19486         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
19487         builtin function.
19488         (__builtin_divweuo): Likewise.
19489         (__builtin_divdeo): Likewise.
19490         (__builtin_divdeuo): Likewise.
19492 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
19494         PR target/85095
19495         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
19496         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
19498         PR tree-optimization/82004
19499         * gimple-match-head.c (optimize_pow_to_exp): New function.
19500         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
19501         Don't fold to exp if optimize_pow_to_exp is false.
19503 2018-03-28  Martin Liska  <mliska@suse.cz>
19505         PR other/84819
19506         * calls.c (initialize_argument_information): Fix trailing space.
19507         * common.opt: Fix typo and provide better explanation for
19508         -fsanitize-coverage option.
19509         * config/i386/i386.opt: Fix typo.
19511 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
19512             Martin Liska  <mliska@suse.cz>
19514         PR sanitizer/85081
19515         * gimplify.c (asan_poison_variable): Don't do the check for
19516         gimplify_omp_ctxp here.
19517         (gimplify_decl_expr): Do it here.
19518         (gimplify_target_expr): Likewise.
19520 2018-03-28  Martin Liska  <mliska@suse.cz>
19522         PR target/84988
19523         * config/i386/i386.c (ix86_function_arg_advance): Do not call
19524         chkp_type_bounds_count if MPX is not enabled.
19526 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
19528         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
19530 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
19532         PR target/84914
19533         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
19534         function to create the function decl for complex long double
19535         multiply and divide for -mabi=ieeelongdouble.
19536         (init_float128_ieee): Call it.
19538 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
19540         PR target/85044
19541         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
19542         -fcf-protection=branch -mibt.
19543         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
19545 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19547         PR target/81863
19548         * config/arm/arm.c (arm_valid_symbolic_address): Handle
19549         arm_word_relocations.
19551 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
19553         PR target/85056
19554         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
19555         extern array declarations.
19557 2018-03-27  Richard Biener  <rguenther@suse.de>
19559         PR middle-end/84067
19560         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
19561         explicit single_use checks.
19563 2018-03-27  Richard Biener  <rguenther@suse.de>
19565         PR tree-optimization/85082
19566         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
19567         Valueize the VUSE.
19569 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19571         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
19572         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
19573         Turn on fasynchronous-unwind-tables and funwind-tables.
19575 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
19577         PR target/85073
19578         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
19579         (*bmi_blsr_<mode>_ccz): Ditto.
19581 2018-03-26  Tom de Vries  <tom@codesourcery.com>
19583         PR tree-optimization/85063
19584         * omp-general.c (offloading_function_p): New function.  Factor out
19585         of ...
19586         * omp-offload.c (pass_omp_target_link::gate): ... here.
19587         * omp-general.h (offloading_function_p): Declare.
19588         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
19589         with attribute omp declare target for offloading functions.
19591 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
19593         PR tree-optimization/84005
19594         * tree-data-ref.h (get_base_for_alignment): Declare.
19595         * tree-data-ref.c (get_base_for_alignment_1): New function.
19596         (get_base_for_alignment): Likewise.
19597         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
19598         get_base_for_alignment to find a suitable base object, instead
19599         of always using drb->base_address.
19601 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
19603         PR inline-asm/85022
19604         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
19605         known size by default.
19607 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
19609         PR inline-asm/85030
19610         * lra-constraints.c (process_alt_operands): Don't match BLKmode
19611         and non BLKmode operands.
19613 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19615         PR target/85026
19616         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
19617         Clean up attributes.
19619 2018-03-23  Richard Biener  <rguenther@suse.de>
19621         PR debug/85020
19622         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
19623         we are going to emit early debug for LTO.
19625 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
19627         PR inline-asm/85034
19628         * function.c (match_asm_constraints_1): Don't optimize if input
19629         doesn't satisfy general_operand predicate for output's mode.
19631         PR inline-asm/85022
19632         * alias.c (write_dependence_p): Don't require for x_canonicalized
19633         non-VOIDmode if x has VOIDmode.
19635         PR sanitizer/85029
19636         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
19637         just don't try to optimize it rather than assert it never happens.
19639 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19641         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
19642         macro expansions for definition of ST_INTERNAL_<mode> and
19643         LD_INTERNAL_<mode> builtins.
19644         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
19645         Remove prototype.
19646         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
19647         function.
19648         (altivec_expand_st_builtin): Likewise.
19649         (altivec_expand_builtin): Remove calls to deleted functions.
19650         (rs6000_address_for_altivec): Delete this function.
19651         * config/rs6000/vector.md: Remove expands for
19652         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
19654 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
19656         PR target/84826
19657         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
19658         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
19659         re-computing once computed.
19660         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
19661         (arm_init_machine_status): Initialize
19662         machine->static_chain_stack_bytes.
19664 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19666         PR target/84760
19667         * doc/extend.texi: Add four new prototypes for vec_ld.
19668         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
19669         definitions for more logical presentation.
19670         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
19671         entries for V1TI variants of __builtin_altivec_ld builtin.
19672         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
19673         handling of V1TI variant of LVX icode pattern.
19674         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
19675         (rs6000_gimple_fold_builtin): Likewise.
19676         (altivec_init_builtins): Add code to define
19677         __builtin_altivec_lvx_v1ti function.
19679 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
19681         PR inline-asm/84941
19682         * function.c (match_asm_constraints_1): Don't do the optimization
19683         if input isn't a REG, SUBREG, MEM or constant.
19685 2018-03-22  Tom de Vries  <tom@codesourcery.com>
19687         PR tree-optimization/84956
19688         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
19689         bb_has_abnormal_pred.
19691 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
19693         PR sanitizer/85018
19694         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
19695         DECL_INITIAL (decl) to decl at the end.
19696         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
19697         adjust the comment.
19699 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
19701         * doc/extend.texi (__builtin_tgmath): Document when complex
19702         integer types are treated as _Complex _Float64.
19704 2018-03-21  Tom de Vries  <tom@codesourcery.com>
19706         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
19708 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19710         PR tree-optimization/84960
19711         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
19712         if it is ENTRY block, move them into single succ of ENTRY in that case.
19714 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
19716         PR tree-optimization/84811
19717         * poly-int.h (poly_span_traits): Remove the T3 parameter and
19718         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
19719         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
19720         (known_subrange_p): Update accordingly.  Cast each value involved
19721         in the size comparison, rather than casting the result of the
19722         subtraction.
19724 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19726         PR tree-optimization/84982
19727         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
19728         by flipping the least significant bit rather than all bits from
19729         bitpos to bitpos + bitsize - 1.
19731 2018-03-21  Nathan Sidwell  <nathan@acm.org>
19733         * doc/extend.texi (Deprecated Features): Remove mention of
19734         long-deleted deprecations.
19736 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
19738         PR jit/84288
19739         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
19740         * configure: Regenerate.
19742 2018-03-21  Tom de Vries  <tom@codesourcery.com>
19744         PR tree-optimization/83126
19745         * tree-parloops.c (num_phis): New function.
19746         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
19748 2018-03-21  Nathan Sidwell  <nathan@acm.org>
19750         * doc/extend.texi (Deprecated Features): Update deprecated flags,
19751         mention anon-struct/union members and trailing attributes.
19753 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
19755         PR tree-optimization/84969
19756         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
19757         builtin memset partitions if they set different rhs values.
19759 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19761         PR rtl-optimization/84989
19762         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
19763         VEC_DUPLICATE with scalar result mode.
19765 2018-03-21  Martin Liska  <mliska@suse.cz>
19767         PR ipa/84963
19768         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
19769         not intended return statement.
19771 2018-03-21  Martin Liska  <mliska@suse.cz>
19773         PR target/84988
19774         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
19775         (chkp_find_bound_slots_1): Limit number of iterations.
19777 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
19779         PR target/84838
19780         * Minor grammar fixes for x86 options.
19782 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19784         PR debug/84875
19785         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
19786         holding REG_CFA_RESTORE notes, instead turn them into a USE.
19788 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
19790         PR target/83789
19791         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
19792         (altivec_lvx_<mode>_1op): Likewise.
19793         (altivec_stvx_<mode>_2op): Likewise.
19794         (altivec_stvx_<mode>_1op): Likewise.
19795         (altivec_lvx_<VM2:mode>): New define_expand.
19796         (altivec_stvx_<VM2:mode>): Likewise.
19797         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
19798         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19799         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
19800         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19801         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
19802         (rs6000_gen_lvx): Likewise.
19803         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
19804         (altivec_expand_stv_builtin): Likewise.
19805         (altivec_expand_builtin): Likewise.
19806         * config/rs6000/vector.md: Likewise.
19808 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19810         PR target/82518
19811         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
19812         BYTES_BIG_ENDIAN.
19814 2018-03-20  Richard Biener  <rguenther@suse.de>
19816         PR target/84986
19817         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
19818         sign-conversions as zero, fall back to standard scalar_stmt
19819         cost for the rest.
19821 2018-03-20  Martin Liska  <mliska@suse.cz>
19823         PR ipa/84825
19824         * predict.c (rebuild_frequencies): Handle case when we have
19825         PROFILE_ABSENT, but flag_guess_branch_prob is false.
19827 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19829         PR target/84990
19830         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
19831         flag_section_anchors.
19832         * varasm.c (use_blocks_for_decl_p): Remove hack for
19833         dw2_force_const_mem.
19835         PR target/84845
19836         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
19837         to ...
19838         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
19839         be created, use lowpart_subreg of operands[0] rather than operands[0]
19840         itself.
19841         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
19842         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
19843         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
19844         and n constraint instead of aarch64_shift_imm_di and Usd.
19845         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
19846         (*aarch64_<optab>_reg_minus<mode>3): ... this.
19848 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
19850         PR target/82989
19851         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
19852         to favor GPR over NEON registers.
19853         (<shift>di3_neon): Likewise.
19855 2018-03-20  Tom de Vries  <tom@codesourcery.com>
19857         PR target/84952
19858         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
19859         (nvptx_process_pars): Emit bar.sync asap and alap.
19861 2018-03-20  Tom de Vries  <tom@codesourcery.com>
19863         PR target/84954
19864         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
19865         seen_label if seen_label is already set.
19867 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19869         PR target/84945
19870         * config/i386/i386.c (fold_builtin_cpu): For features above 31
19871         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
19872         Use 1U instead of 1.  Formatting fixes.
19874         PR c/84953
19875         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
19876         instead of TREE_TYPE (s1) for the return value.
19878 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
19880         PR tree-optimization/84946
19881         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
19882         bitsize + bitsize in poly_uint64 rather than poly_int64.
19884         PR sanitizer/78651
19885         * dwarf2asm.c: Include fold-const.c.
19886         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
19887         of decl rather than decl itself.
19889         PR rtl-optimization/84643
19890         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
19892 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
19894         PR sanitizer/78651
19895         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
19896         calling assemble_variable.
19898 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
19900         PR target/81647
19901         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
19902         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
19904 2018-03-19  Jim Wilson  <jimw@sifive.com>
19906         PR bootstrap/84856
19907         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
19908         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
19909         (riscv_first_stack_step): Likewise.
19910         (riscv_option_override): Use STACK_BOUNDARY instead of
19911         MIN_STACK_BOUNDARY.
19912         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
19913         MIN_STACK_BOUNDARY.
19914         (BIGGEST_ALIGNMENT): Set to 128.
19915         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19916         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
19917         STACK_BOUNDARY.
19919 2018-03-19  Richard Biener  <rguenther@suse.de>
19921         PR tree-optimization/84933
19922         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
19923         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19925 2018-03-19  Richard Biener  <rguenther@suse.de>
19927         PR tree-optimization/84859
19928         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19929         (cond_if_else_store_replacement): Perform sinking operation on
19930         single-store BBs regardless of MAX_STORES_TO_SINK setting.
19931         Generalize what a BB with a single eligible store is.
19933 2018-03-19  Richard Biener  <rguenther@suse.de>
19935         PR tree-optimization/84929
19936         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19937         chrec_is_positive against non-chrec arg.
19939 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
19941         PR target/84711
19942         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19944 2018-03-18  Martin Liska  <mliska@suse.cz>
19946         PR rtl-optimization/84635
19947         * regrename.c (build_def_use): Use matches_mode only when
19948         matches >= 0.
19950 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
19952         PR tree-optimization/84913
19953         * tree-vect-loop.c (vectorizable_reduction): Don't try to
19954         vectorize chains of COND_EXPRs.
19956 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19958         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19960 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19962         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19964 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19966         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19968 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19969             Kito Cheng  <kito.cheng@gmail.com>
19971         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19972         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19973         (nds32_adjust_reg_alloc_order): New function.
19974         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19976 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
19978         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19979         nds32_print_operand, nds32_print_operand_address): Use
19980         HOST_WIDE_INT_PRINT_DEC instead.
19982 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19984         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19986 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
19988         PR target/84902
19989         * config/i386/i386.c (initial_ix86_tune_features,
19990         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19991         unsigned long long.
19992         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19993         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19994         rather than 1u << ix86_tune.  Formatting fix.
19995         (ix86_option_override_internal): Change ix86_arch_mask from
19996         unsigned int to unsigned HOST_WIDE_INT, initialize to
19997         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19998         (ix86_function_specific_restore): Likewise.
20000 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20002         PR target/84899
20003         * postreload.c (reload_combine_recognize_pattern): Perform
20004         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
20005         truncate_int_for_mode the result for the destination's mode.
20007         PR c/84909
20008         * hsa-gen.c (mem_type_for_type): Fix comment typo.
20009         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
20010         Likewise.
20011         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20012         Likewise.
20014 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
20016         PR target/84876
20017         * lra-assigns.c (lra_split_hard_reg_for): Don't use
20018         regno_allocno_class_array and sorted_pseudos.
20019         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
20020         insns where regno is used.
20022 2018-03-16  Martin Liska  <mliska@suse.cz>
20024         PR ipa/84833
20025         * multiple_target.c (create_dispatcher_calls): Redirect
20026         reference in the symbol table.
20028 2018-03-16  Martin Liska  <mliska@suse.cz>
20030         PR ipa/84722
20031         * multiple_target.c (create_dispatcher_calls): Redirect also
20032         an alias.
20034 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20036         PR c++/79937
20037         PR c++/82410
20038         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
20039         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
20040         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
20042 2018-03-16  Julia Koval  <julia.koval@intel.com>
20044         * doc/invoke.texi (Skylake Server): Add CLWB.
20045         Cannonlake): Remove CLWB.
20047 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20049         PR tree-optimization/84841
20050         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
20051         1 << 3.
20052         (FLOAT_ONE_CONST_TYPE): Define.
20053         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
20054         (sort_by_operand_rank): Put entries with higher constant_type last
20055         rather than first to match comments.
20057 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
20059         * config/nios2/nios2.md (movsi_internal): Fix thinko in
20060         split predicate.
20062 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20064         PR c++/79085
20065         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
20066         check and use address of target always.
20068 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
20070         PR target/84574
20071         * config/i386/i386.c (indirect_thunk_needed): Update comments.
20072         (indirect_thunk_bnd_needed): Likewise.
20073         (indirect_thunks_used): Likewise.
20074         (indirect_thunks_bnd_used): Likewise.
20075         (indirect_return_needed): New.
20076         (indirect_return_bnd_needed): Likewise.
20077         (output_indirect_thunk_function): Add a bool argument for
20078         function return.
20079         (output_indirect_thunk_function): Don't generate alias for
20080         function return thunk.
20081         (ix86_code_end): Call output_indirect_thunk_function to generate
20082         function return thunks.
20083         (ix86_output_function_return): Set indirect_return_bnd_needed
20084         and indirect_return_needed instead of indirect_thunk_bnd_needed
20085         and indirect_thunk_needed.
20087 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
20089         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
20090         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
20091         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
20093 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
20094             Paul Hua <paul.hua.gm@gmail.com>
20096         PR c/84852
20097         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
20099 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
20101         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
20102         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
20103         resp. SFmode cases.
20105 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
20107         PR target/84711
20108         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
20109         instead of GET_MODE_SIZE when comparing Units.
20111 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
20113         PR target/68256
20114         * varasm.c (hash_section): Return an unchangeble hash value
20115         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
20116         Return !aarch64_can_use_per_function_literal_pools_p ().
20118 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20120         PR target/84860
20121         * optabs.c (emit_conditional_move): Pass address of cmode's copy
20122         rather than address of cmode as last argument to prepare_cmp_insn.
20124 2018-03-15  Julia Koval  <julia.koval@intel.com>
20126         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20127         F_AVX512VNNI, F_AVX512BITALG): New.
20129 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
20131         PR target/83451
20132         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
20133         insn for floating-point loads and stores.
20135 2018-03-14  Carl Love  <cel@us.ibm.com>
20137         * config/rs6000/rs6000-c.c: Add macro definitions for
20138         ALTIVEC_BUILTIN_VEC_PERMXOR.
20139         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20140         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20141         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20142         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20143         UNSPEC_VPERMXOR.
20144         * config/doc/extend.texi: Add prototypes for vec_permxor.
20146 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
20148         PR c/84852
20149         * diagnostic-show-locus.c (class layout_point): Convert m_line
20150         from int to linenum_type.
20151         (line_span::comparator): Use linenum "compare" function when
20152         comparing line numbers.
20153         (test_line_span): New function.
20154         (layout_range::contains_point): Convert param "row" from int to
20155         linenum_type.
20156         (layout_range::intersects_line_p): Likewise.
20157         (layout::will_show_line_p): Likewise.
20158         (layout::print_source_line): Likewise.
20159         (layout::should_print_annotation_line_p): Likewise.
20160         (layout::print_annotation_line): Likewise.
20161         (layout::print_leading_fixits): Likewise.
20162         (layout::annotation_line_showed_range_p): Likewise.
20163         (struct line_corrections): Likewise for field m_row.
20164         (line_corrections::line_corrections): Likewise for param "row".
20165         (layout::print_trailing_fixits): Likewise.
20166         (layout::get_state_at_point): Likewise.
20167         (layout::get_x_bound_for_row): Likewise.
20168         (layout::print_line): Likewise.
20169         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
20170         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
20171         * input.c (selftest::test_linenum_comparisons): New function.
20172         (selftest::input_c_tests): Call it.
20173         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
20174         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
20175         * selftest.h (ASSERT_GT): New macro.
20176         (ASSERT_GT_AT): New macro.
20177         (ASSERT_LT): New macro.
20178         (ASSERT_LT_AT): New macro.
20180 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
20182         PR rtl-optimization/84780
20183         * combine.c (distribute_links): Don't make a link based on pc_rtx.
20185 2018-03-14  Martin Liska  <mliska@suse.cz>
20187         * tree.c (record_node_allocation_statistics): Use
20188         get_stats_node_kind.
20189         (get_stats_node_kind): New function extracted from
20190         record_node_allocation_statistics.
20191         (free_node): Use get_stats_node_kind.
20193 2018-03-14  Richard Biener  <rguenther@suse.de>
20195         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20196         that the value-set of ANTIC_IN doesn't grow.
20198         Revert
20199         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20200         member.
20201         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20202         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20204 2018-03-14  Julia Koval  <julia.koval@intel.com>
20206         * config.gcc (icelake-client, icelake-server): New.
20207         (icelake): Remove.
20208         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20209         (initial_ix86_arch_features): Ditto.
20210         (PTA_SKYLAKE): Add SGX.
20211         (PTA_ICELAKE): Remove.
20212         (PTA_ICELAKE_CLIENT): New.
20213         (PTA_ICELAKE_SERVER): New.
20214         (ix86_option_override_internal): Split up icelake on icelake client and
20215         icelake server.
20216         (get_builtin_code_for_version): Ditto.
20217         (fold_builtin_cpu): Ditto.
20218         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20219         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20220         * config/i386/i386.h (processor_type): Ditto.
20221         * doc/invoke.texi: Ditto.
20223 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
20225         PR sanitizer/83392
20226         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
20227         INTEGER_CST offset, add it together with bitpos / 8 and
20228         sign extend based on POINTER_SIZE.
20230         PR target/84844
20231         Revert
20232         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
20234         PR target/78090
20235         * config/i386/constraints.md (Yc): New register constraint.
20236         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
20237         Use Yc constraint for alternative 2 of operand 0.  Remove
20238         preferred_for_speed attribute.
20240 2018-03-14  Richard Biener  <rguenther@suse.de>
20242         PR tree-optimization/84830
20243         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
20244         with the old one to avoid oscillations.
20246 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
20248         PR target/83712
20249         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
20250         pseudos.
20251         (assign_by_spills): Return a flag of reload assignment failure.
20252         Do not process the reload assignment failures.  Do not spill other
20253         reload pseudos if they has the same reg class.  Update n if
20254         necessary.
20255         (lra_assign): Add a return arg.  Set up from the result of
20256         assign_by_spills call.
20257         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20258         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20259         usage_insns if it is not NULL.
20260         (spill_hard_reg_in_range): New function.
20261         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20262         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20263         function prototypes.
20264         (lra_assign): Change prototype.
20265         * lra.c (lra): Add code to deal with fails by splitting hard reg
20266         live ranges.
20268 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
20270         * config/riscv/riscv.opt (mrelax): New option.
20271         * config/riscv/riscv.c (riscv_file_start): Emit ".option
20272         "norelax" when riscv_mrelax is disabled.
20273         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
20275 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20277         PR target/84743
20278         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
20279         reassociation for int modes.
20281 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20283         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
20284         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
20285         for big-endian.
20286         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
20287         * config/aarch64/aarch64-sve.md
20288         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
20289         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
20290         (*extend<mode><Vwide>2): Rename to...
20291         (aarch64_sve_extend<mode><Vwide>2): ...this.
20292         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
20293         renaming the old pattern to...
20294         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
20295         unsigned packs.
20296         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
20297         define_expand, renaming the old pattern to...
20298         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
20299         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
20300         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
20301         account when deciding which SVE instruction the optab should use.
20302         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
20304 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20306         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
20307         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
20308         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
20309         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
20310         (tlsdesc_small_<mode>): Turn a define_expand and use
20311         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
20312         (tlsdesc_small_advsimd_<mode>): ...this.
20313         (tlsdesc_small_sve_<mode>): New pattern.
20315 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20317         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
20318         (UNSPEC_UMUL_HIGHPART): New constants.
20319         (MUL_HIGHPART): New int iteraor.
20320         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
20321         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20322         define_expand.
20323         (*<su>mul<mode>3_highpart): New define_insn.
20325 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20327         PR lto/84805
20328         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20329         incomplete types.
20331 2018-03-13  Martin Liska  <mliska@suse.cz>
20333         PR ipa/84658.
20334         * (sem_item_optimizer::sem_item_optimizer): Initialize new
20335         vector.
20336         (sem_item_optimizer::~sem_item_optimizer): Release it.
20337         (sem_item_optimizer::merge_classes): Register variable aliases.
20338         (sem_item_optimizer::fixup_pt_set): New function.
20339         (sem_item_optimizer::fixup_points_to_sets): Likewise.
20340         * ipa-icf.h: Declare new variables and functions.
20342 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
20344         PR middle-end/84834
20345         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
20346         integer_pow2p@2 and test integer_pow2p in condition.
20347         (A < 0 ? C : 0): Similarly for @1.
20349         PR middle-end/84831
20350         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
20351         characters starting at p contain '\0' character, don't look beyond
20352         that.
20354         PR target/84827
20355         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
20356         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
20358         PR target/84828
20359         * reg-stack.c (change_stack): Change update_end var from int to
20360         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
20361         also call set_block_for_insn on the newly added insns and rescan.
20363         PR target/84786
20364         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
20365         on the last operand.
20367         PR c++/84704
20368         * tree.c (stabilize_reference_1): Return save_expr (e) for
20369         STATEMENT_LIST even if it doesn't have side-effects.
20371 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
20373         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
20375 2018-03-12  Renlin Li  <renlin.li@arm.com>
20377         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
20378         aarch64_output_scalar_simd_mov_immediate.
20380 2018-03-12  Martin Sebor  <msebor@redhat.com>
20382         PR tree-optimization/83456
20383         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
20384         for perfectly overlapping calls to memcpy.
20385         (gimple_fold_builtin_memory_chk): Same.
20386         (gimple_fold_builtin_strcpy): Handle no-warning.
20387         (gimple_fold_builtin_stxcpy_chk): Same.
20388         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
20390 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20392         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
20393         parameter.  Use it for SFmode.
20394         (rs6000_function_arg_advance_1): Adjust.
20395         (rs6000_function_arg): Adjust.
20396         (rs6000_gimplify_va_arg): Pass false for that new parameter.
20398 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20400         PR rtl-optimization/84169
20401         PR rtl-optimization/84780
20402         * combine.c (can_combine_p): Check for a 2-insn combination whether
20403         the destination register is used between the two insns, too.
20405 2018-03-12  Richard Biener  <rguenther@suse.de>
20407         PR tree-optimization/84803
20408         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20409         for refs DR analysis didn't process.
20411 2018-03-12  Richard Biener  <rguenther@suse.de>
20413         PR tree-optimization/84777
20414         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
20415         force-vectorize loops ignore whether we are optimizing for size.
20417 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
20419         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
20420         (TARGET_MD_ASM_ADJUST): Define.
20422 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
20423             Kito Cheng  <kito.cheng@gmail.com>
20424             Chung-Ju Wu  <jasonwucj@gmail.com>
20426         * config/nds32/nds32.c (nds32_compute_stack_frame,
20427         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
20428         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
20429         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
20430         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
20431         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
20432         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
20433         * config/nds32/nds32.md (prologue, epilogue): Use macro
20434         NDS32_V3PUSH_AVAILABLE_P to do checking.
20436 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
20438         PR debug/58150
20439         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
20440         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
20441         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
20442         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
20443         addition of most attributes on !orig_type_die or the attribute not
20444         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
20446 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20447             Chung-Ju Wu  <jasonwucj@gmail.com>
20449         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
20450         __NDS32_VH__ macro.
20451         * config/nds32/nds32.opt (mvh): New option.
20453 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20454             Chung-Ju Wu  <jasonwucj@gmail.com>
20456         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
20457         function.
20458         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
20459         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
20460         definition.
20462 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20463             Chung-Ju Wu  <jasonwucj@gmail.com>
20465         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
20466         function.
20467         * config/nds32/nds32-multiple.md (strlensi): New pattern.
20468         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
20470 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
20471             Kito Cheng  <kito.cheng@gmail.com>
20472             Chung-Ju Wu  <jasonwucj@gmail.com>
20474         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
20475         UNSPEC_FFMISM and UNSPEC_FLMISM.
20476         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
20477         for ffb, ffmism and flmism.
20478         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
20479         (unspec_ffmism): Ditto.
20480         (unspec_flmism): Ditto.
20481         (nds32_expand_builtin_impl): Check if string extension is available.
20482         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
20483         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
20485 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
20487         Reverting patch:
20488         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
20490         PR target/83712
20491         * lra-assigns.c (assign_by_spills): Return a flag of reload
20492         assignment failure.  Do not process the reload assignment
20493         failures.  Do not spill other reload pseudos if they has the same
20494         reg class.
20495         (lra_assign): Add a return arg.  Set up from the result of
20496         assign_by_spills call.
20497         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20498         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20499         usage_insns if it is not NULL.
20500         (spill_hard_reg_in_range): New function.
20501         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20502         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20503         function prototypes.
20504         (lra_assign): Change prototype.
20505         * lra.c (lra): Add code to deal with fails by splitting hard reg
20506         live ranges.
20508 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
20510         PR target/84807
20511         * config/i386/i386.opt: Replace Enforcment with Enforcement.
20513 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
20515         PR debug/84620
20516         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
20517         (dw_val_node): Add val_symbolic_view.
20518         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
20519         (symview_upper_bound): New.
20520         (new_line_info_table): Initialize symviews_since_reset.
20521         (dwarf2out_source_line): Count symviews_since_reset and set
20522         symview_upper_bound.
20523         (dw_val_equal_p): Handle symview.
20524         (add_AT_symview): New.
20525         (print_dw_val): Handle symview.
20526         (attr_checksum, attr_checksum_ordered): Likewise.
20527         (same_dw_val_p, size_of_die): Likewise.
20528         (value_format, output_die): Likewise.
20529         (add_high_low_attributes): Use add_AT_symview for entry_view.
20530         (dwarf2out_finish): Reset symview_upper_bound, clear
20531         zero_view_p.
20533 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
20535         PR target/83969
20536         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
20537         Add strict argument and use it.
20538         (rs6000_split_multireg_move): Update for new strict argument.
20539         (mem_operand_gpr): Disallow all non-offsettable addresses.
20540         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
20542 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
20544         PR target/84772
20545         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
20546         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
20547         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
20549         PR c++/84767
20550         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
20551         decl, use remap_type if we want to use the type.
20553 2018-03-09  Martin Sebor  <msebor@redhat.com>
20555         PR tree-optimization/84526
20556         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20557         Remove dead code.
20558         (builtin_access::generic_overlap): Be prepared to handle non-array
20559         base objects.
20561 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
20563         PR rtl-optimization/84682
20564         * lra-constraints.c (process_address_1): Check is_address flag
20565         for address constraints.
20566         (process_alt_operands): Likewise.
20567         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
20568         preprocess_constraints.
20569         * recog.h (preprocess_constraints): Add oploc parameter.
20570         Adjust callers.
20571         * recog.c (preprocess_constraints): Test address_operand for
20572         CT_ADDRESS constraints.
20574 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
20576         PR target/83712
20577         * lra-assigns.c (assign_by_spills): Return a flag of reload
20578         assignment failure.  Do not process the reload assignment
20579         failures.  Do not spill other reload pseudos if they has the same
20580         reg class.
20581         (lra_assign): Add a return arg.  Set up from the result of
20582         assign_by_spills call.
20583         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20584         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20585         usage_insns if it is not NULL.
20586         (spill_hard_reg_in_range): New function.
20587         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20588         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20589         function prototypes.
20590         (lra_assign): Change prototype.
20591         * lra.c (lra): Add code to deal with fails by splitting hard reg
20592         live ranges.
20594 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20596         PR target/83193
20597         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
20598         Accept complain bool parameter.  Only emit errors if it is true.
20599         (arm_parse_cpu_option_name): Likewise.
20600         (arm_target_thumb_only): Adjust callers of the above.
20601         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
20602         prototype to take a default true bool parameter.
20603         (arm_parse_arch_option_name): Likewise.
20605 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
20606             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
20608         PR jit/64089
20609         PR jit/84288
20610         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
20611         * configure: Regenerate.
20612         * configure.ac ("linker --version-script option"): New.
20613         ("linker soname option"): New.
20615 2018-03-09  Richard Biener  <rguenther@suse.de>
20617         PR tree-optimization/84775
20618         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
20619         immediate uses of predicate stmts and mark them modified.
20621         Revert
20622         PR tree-optimization/84178
20623         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20624         to caller.
20625         (version_loop_for_if_conversion): Delay update_ssa call.
20626         (tree_if_conversion): Delay update_ssa until after predicate
20627         insertion.
20629 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
20631         PR target/84763
20632         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
20633         when the function accesses prior frames.
20635 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
20637         PR debug/84456
20638         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
20639         gen_llsym, otherwise call maybe_gen_llsym.
20641         PR inline-asm/84742
20642         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
20643         has ',' character inside of it.
20645 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20647         PR target/84748
20648         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
20649         as clobbering CC_REGNUM.
20651 2018-03-08  Richard Biener  <rguenther@suse.de>
20653         PR middle-end/84552
20654         * tree-scalar-evolution.c: Include tree-into-ssa.h.
20655         (follow_copies_to_constant): Do not follow SSA names registered
20656         for update.
20658 2018-03-08  Richard Biener  <rguenther@suse.de>
20660         PR tree-optimization/84178
20661         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20662         to caller.
20663         (version_loop_for_if_conversion): Delay update_ssa call.
20664         (tree_if_conversion): Delay update_ssa until after predicate
20665         insertion.
20667 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
20669         PR tree-optimization/84178
20670         * tree-if-conv.c (release_bb_predicate): Remove the
20671         the assertion that the stmts have NULL use_ops.
20672         Discard the statements, asserting that they haven't
20673         yet been added to a BB.
20675 2018-03-08  Richard Biener  <rguenther@suse.de>
20677         PR tree-optimization/84746
20678         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
20679         (phi_translate): Pass in destination ANTIC_OUT set.
20680         (phi_translate_1): Likewise.  For a simplified result lookup
20681         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
20682         (phi_translate_set): Adjust.
20683         (do_pre_regular_insertion): Likewise.
20684         (do_pre_partial_partial_insertion): Likewise.
20686 2018-03-08  Martin Liska  <mliska@suse.cz>
20688         PR gcov-profile/84735
20689         * doc/gcov.texi: Document usage of profile files.
20690         * gcov-io.h: Document changes in the format.
20692 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
20694         PR debug/84404
20695         PR debug/84408
20696         * dwarf2out.c (struct dw_line_info_table): Update comments for
20697         view == -1.
20698         (FORCE_RESET_NEXT_VIEW): New.
20699         (FORCE_RESETTING_VIEW_P): New.
20700         (RESETTING_VIEW_P): Check for -1 too.
20701         (ZERO_VIEW_P): Likewise.
20702         (new_line_info_table): Force-reset next view.
20703         (dwarf2out_begin_function): Likewise.
20704         (dwarf2out_source_line): Simplify zero_view_p initialization.
20705         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
20706         view directly.  Omit view when omitting .loc at line 0.
20708 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
20710         PR tree-optimization/84740
20711         * tree-switch-conversion.c (process_switch): Call build_constructors
20712         only if info.phi_count is non-zero.
20714         PR tree-optimization/84739
20715         * tree-tailcall.c (find_tail_calls): Check call arguments against
20716         DECL_ARGUMENTS (current_function_decl) rather than
20717         DECL_ARGUMENTS (func) when checking for tail recursion.
20719 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
20721         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
20722         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
20723         Volker Reichelt's entry and add entries for people that perform
20724         GCC fuzzy testing and report numerous bugs.
20726 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
20728         PR target/82411
20729         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
20730         readonly data in sdata, if that is disabled.
20731         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
20732         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
20733         -mreadonly-in-sdata option.
20735 2018-03-07  Martin Sebor  <msebor@redhat.com>
20737         PR tree-optimization/84468
20738         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
20739         basic block when looking for nul assignment.
20741 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
20743         PR target/84277
20744         * except.h (output_function_exception_table): Adjust prototype.
20745         * except.c (output_function_exception_table): Remove FNNAME parameter
20746         and add SECTION parameter.  Ouput one part of the table at a time.
20747         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
20748         the first part of the exception table and emit unwind directives.
20749         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
20750         (i386_pe_seh_cold_init): Likewise.
20751         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
20752         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
20753         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
20754         (ix86_output_call_insn): Emit a nop in one more case for SEH.
20755         * config/i386/winnt.c: Include except.h.
20756         (struct seh_frame_state): Add reg_offset, after_prologue and
20757         in_cold_section fields.
20758         (i386_pe_seh_end_prologue): Set seh->after_prologue.
20759         (i386_pe_seh_cold_init): New function.
20760         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
20761         to seh->in_cold_section.
20762         (seh_emit_push): Record the offset of the push.
20763         (seh_emit_save): Record the offet of the save.
20764         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
20765         Test seh->after_prologue to disregard the epilogue.
20766         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
20767         (i386_pe_end_cold_function): New function.
20769 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
20771         PR fortran/84565
20772         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
20773         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
20775         PR c++/84704
20776         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
20777         on tmp_var.
20778         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
20779         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
20781         PR middle-end/84723
20782         * multiple_target.c: Include tree-inline.h and intl.h.
20783         (expand_target_clones): Diagnose and fail if node->definition and
20784         !tree_versionable_function_p (node->decl).
20786 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
20788         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
20789         sprint_ul.
20790         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
20791         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20792         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
20794 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
20796         PR target/84710
20797         * combine.c (try_combine): Use reg_or_subregno instead of handling
20798         just paradoxical SUBREGs and REGs.
20800 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
20802         * config/arc/arc.c (arc_finalize_pic): Remove function.
20803         (arc_must_save_register): We use single base PIC register, remove
20804         checks to save/restore the PIC register.
20805         (arc_expand_prologue): Likewise.
20806         * config/arc/arc-protos.h (arc_set_default_type_attributes):
20807         Remove.
20808         (arc_verify_short): Likewise.
20809         (arc_attr_type): Likewise.
20810         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
20811         (walk_stores): Likewise.
20812         (arc_address_cost): Make it static.
20813         (arc_verify_short): Likewise.
20814         (branch_dest): Likewise.
20815         (arc_attr_type): Likewise.
20816         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
20817         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
20818         (arc_final_prescan_insn): Remove inserting the nops due to
20819         hardware hazards.  It is done in reorg step.
20820         (insn_length_variant_t): Remove.
20821         (insn_length_parameters_t): Likewise.
20822         (arc_insn_length_parameters): Likewise.
20823         (arc_get_insn_variants): Likewise.
20824         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
20826 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
20828         PR inline-asm/84683
20829         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
20830         assertion failure.
20832         PR tree-optimization/84687
20833         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
20834         on new_node->decl.
20835         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
20837 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20839         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
20840         Rename to ppc_speculation_barrier.
20841         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
20842         __builtin_ppc_speculation_barrier.
20844 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
20846         PR target/84700
20847         * combine.c (combine_simplify_rtx): Don't try to simplify if
20848         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
20849         are equal to x.
20851 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
20853         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
20854         to 32 bytes when compiling for POWER9.
20856 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
20858         PR target/84564
20859         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
20860         regparm >= 3 with no arg reg available also for calls with
20861         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
20863         PR target/84524
20864         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
20865         orig,vex.
20866         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
20868 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
20870         PR target/84264
20871         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
20873 2018-03-05  Richard Biener  <rguenther@suse.de>
20875         PR tree-optimization/84486
20876         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
20877         When inserting a __builtin_assume_aligned call set the LHS
20878         SSA name alignment info accordingly.
20880 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
20882         PR tree-optimization/84114
20883         * config/aarch64/aarch64.c (aarch64_reassociation_width)
20884         Avoid reassociation of FLOAT_MODE addition.
20886 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
20888         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
20889         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
20890         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
20891         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
20892         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
20893         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
20894         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
20895         and -mwbnoinvd.
20896         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
20897         __builtin_ia32_wbinvd): New builtins.
20898         (SPECIAL_ARGS2): New.
20899         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
20900         (SPECIAL_ARGS2): New.
20901         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
20902         (ix86_valid_target_attribute_inner_p): Ditto.
20903         (ix86_init_mmx_sse_builtins): Add special_args2.
20904         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
20905         TARGET_WBNOINVD_P): New.
20906         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
20907         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
20908         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
20909         * config/i386/immintrin.h (_wbinvd): New intrinsic.
20910         * config/i386/pconfigintrin.h: New file.
20911         * config/i386/wbnoinvdintrin.h: Ditto.
20912         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
20913         wbnoinvdintrin.h.
20914         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
20916 2018-03-05  Richard Biener  <rguenther@suse.de>
20918         PR tree-optimization/84670
20919         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20920         member.
20921         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20922         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20923         (compute_antic_aux): Only assert the number of values in ANTIC_IN
20924         doesn't grow if all successors (recursively) were visited at least
20925         once.
20927 2018-03-05  Richard Biener  <rguenther@suse.de>
20929         PR tree-optimization/84650
20930         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20931         if executed in the loop pipeline.
20933 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
20935         * doc/configfiles.texi (Configuration Files): Move info about
20936         conditionalizing $target-protos.h to...
20937         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
20938         differs from $target-protos.h.
20940 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
20941             Chung-Ju Wu  <jasonwucj@gmail.com>
20943         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20944         * config/nds32/nds32-multiple.md (setmemsi): Define.
20945         * config/nds32/nds32-memory-manipulation.c
20946         (nds32_gen_dup_4_byte_to_word_value): New.
20947         (emit_setmem_word_loop): New.
20948         (emit_setmem_byte_loop): New.
20949         (nds32_expand_setmem_loop): New.
20950         (nds32_expand_setmem_loop_v3m): New.
20951         (nds32_expand_setmem_unroll): New.
20952         (nds32_expand_setmem): New.
20954 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20955             Chung-Ju Wu  <jasonwucj@gmail.com>
20957         * config/nds32/nds32-memory-manipulation.c
20958         (nds32_emit_load_store): New.
20959         (nds32_emit_post_inc_load_store): New.
20960         (nds32_emit_mem_move): New.
20961         (nds32_emit_mem_move_block): New.
20962         (nds32_expand_movmemsi_loop_unknown_size): New.
20963         (nds32_expand_movmemsi_loop_known_size): New.
20964         (nds32_expand_movmemsi_loop): New.
20965         (nds32_expand_movmemsi_unroll): New.
20966         (nds32_expand_movmemqi): Rename ...
20967         (nds32_expand_movmemsi): ... to this.
20968         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20969         (movmemsi): ... to this.
20970         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20971         (nds32_expand_movmemsi): ... to this.
20973 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20974             Monk Chiang  <sh.chiang04@gmail.com>
20975             Chung-Ju Wu  <jasonwucj@gmail.com>
20977         * config/nds32/nds32-protos.h
20978         (nds32_expand_load_multiple): New arguments.
20979         (nds32_expand_store_multiple): Ditto.
20980         (nds32_valid_multiple_load_store): Rename ...
20981         (nds32_valid_multiple_load_store_p): ... to this.
20982         * config/nds32/nds32-memory-manipulation.c
20983         (nds32_expand_load_multiple): Refine implementation.
20984         (nds32_expand_store_multiple): Ditto.
20985         * config/nds32/nds32-multiple.md
20986         (load_multiple): Update nds32_expand_load_multiple interface.
20987         (store_multiple): Update nds32_expand_store_multiple interface.
20988         * config/nds32/nds32-predicates.c
20989         (nds32_valid_multiple_load_store): Rename ...
20990         (nds32_valid_multiple_load_store_p): ... to this and refine
20991         implementation.
20992         * config/nds32/predicates.md
20993         (nds32_load_multiple_and_update_address_operation): New predicate.
20994         (nds32_store_multiple_and_update_address_operation): New predicate.
20996 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20997             Chung-Ju Wu  <jasonwucj@gmail.com>
20999         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
21000         (combo): New attribute.
21001         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
21003 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21005         * config/nds32/nds32.opt: Change -mcmodel= default value.
21007 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
21008             Monk Chiang  <sh.chiang04@gmail.com>
21009             Chung-Ju Wu  <jasonwucj@gmail.com>
21011         * config/nds32/constants.md (unspec_element): New enum.
21012         * config/nds32/constraints.md (Umw): New constraint.
21013         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
21014         * config/nds32/nds32-intrinsic.md: Likewise.
21015         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
21016         (nds32_valid_smw_lwm_base_p): New.
21017         (nds32_output_smw_single_word): New.
21018         (nds32_output_lmw_single_word): New.
21019         (nds32_expand_unaligned_load): New.
21020         (nds32_expand_unaligned_store): New.
21021         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
21022         (nds32_output_smw_single_word): Declare.
21023         (nds32_output_lmw_single_word): Declare.
21024         (nds32_expand_unaligned_load): Declare.
21025         (nds32_expand_unaligned_store): Declare.
21026         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
21027         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
21028         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
21029         NDS32_BUILTIN_UASTORE_DW.
21030         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
21031         predicate.
21033 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
21034             Kito Cheng  <kito.cheng@gmail.com>
21035             Chung-Ju Wu  <jasonwucj@gmail.com>
21037         * config/nds32/nds32-intrinsic.c
21038         (nds32_expand_builtin_null_ftype_reg): Delete.
21039         (nds32_expand_builtin_reg_ftype_imm): Ditto.
21040         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
21041         (nds32_read_argument): New.
21042         (nds32_legitimize_target): Ditto.
21043         (nds32_legitimize_argument): Ditto.
21044         (nds32_check_constant_argument): Ditto.
21045         (nds32_expand_unop_builtin): Ditto.
21046         (nds32_expand_unopimm_builtin): Ditto.
21047         (nds32_expand_binop_builtin): Ditto.
21048         (nds32_builtin_decl_impl): Ditto.
21049         (builtin_description): Ditto.
21050         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
21051         (nds32_init_builtins_impl): Ditto.
21052         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
21053         (nds32_builtin_decl): New.
21054         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
21055         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
21057 2018-03-02  Jeff Law  <law@redhat.com>
21059         * reorg.c (stop_search_p): Handle DEBUG_INSN.
21060         (redundant_insn, fill_simple_delay_slots): Likewise.
21061         (fill_slots_from_thread): Likewise.
21062         * resource.c (mark_referenced_resources): Likewise.
21063         (mark_set_resources, find_dead_or_set_registers): Likewise.
21065 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21067         * substring-locations.h (format_warning_va): Formatting fix for
21068         ATTRIBUTE_GCC_DIAG.
21069         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
21070         argument.
21071         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
21072         * substring-locations.c: Include intl.h.
21073         (format_warning_va): Turned into small wrapper around
21074         format_warning_n_va, renamed to ...
21075         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
21076         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
21077         use ngettext.
21078         (format_warning_at_substring_n): New function.
21079         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
21080         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
21081         format_warning_at_substring with just a shorter name instead of
21082         const function pointer.
21083         (fmtwarn_n): New function.
21084         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
21085         appropriate, get rid of all the fmtstr temporaries, move conditionals
21086         with G_() wrapped string literals directly into fmtwarn arguments,
21087         cast dir.len to (int), formatting fixes.
21089 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
21091         * doc/invoke.texi: Remove "Cilk Plus" references.
21093 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21094             Richard Biener  <rguenther@suse.de>
21096         PR ipa/84628
21097         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
21098         for error or warning attributes if CALL_FROM_THUNK_P is set.
21099         Formatting fixes.
21101 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21103         PR target/56540
21104         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
21105         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
21107         PR target/56540
21108         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
21109         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
21111         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
21112         instead of -1U in last predictors element's probability member.
21114 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21116         PR ipa/83983
21117         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21118         arguments if they are comparable.
21120 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21122         PR tree-optimization/84634
21123         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21124         masks and masked_loop_p with a single loop_masks, making sure it's
21125         null for bb vectorization.
21127 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21129         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21130         (vect_analyze_data_ref_access): Use loop->safe_len rather than
21131         loop->force_vectorize to check whether there is no alias.
21133 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21135         PR target/84614
21136         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21137         prototypes.
21138         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21139         comments.
21140         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21141         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21142         instead of a loop around prev_real_insn.
21143         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21144         prev_real_insn.
21146         PR inline-asm/84625
21147         * config/i386/i386.c (ix86_print_operand): Use conditional
21148         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21149         zero vector.
21151 2018-03-02  Richard Biener  <rguenther@suse.de>
21153         PR tree-optimization/84427
21154         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21155         (bitmap_set_subtract_values): Rewrite to handle multiple
21156         exprs per value.
21157         (clean): Likewise.
21158         (prune_clobbered_mems): Likewise.
21159         (phi_translate): Take edge instead of pred/phiblock.
21160         (phi_translate_1): Likewise.
21161         (phi_translate_set): Likewise.  Insert all translated
21162         exprs for a value into the set, keeping possibly multiple
21163         expressions per value.
21164         (compute_antic_aux): Adjust for phi_translate changes.
21165         When intersecting union the expressions and prune those
21166         not in the final value set, keeping possibly multiple
21167         expressions per value.  Do not use value-insertion
21168         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
21169         all expressions.  Add verification that the value-sets
21170         only shrink during iteration.
21171         (compute_partial_antic_aux): Adjust for the phi_translate changes.
21172         (do_pre_regular_insertion): Likewise.
21173         (do_pre_partial_partial_insertion): Likewise.
21175 2018-03-02  Richard Biener  <rguenther@suse.de>
21177         PR target/82005
21178         * config/darwin.c (saved_debug_info_level): New static global.
21179         (darwin_asm_lto_start): Disable debug info generation for LTO out.
21180         (darwin_asm_lto_end): Restore debug info generation settings.
21182 2018-03-01  Martin Liska  <mliska@suse.cz>
21184         PR sanitizer/82484
21185         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21186         volatile arguments.
21188 2018-03-01  Richard Biener  <rguenther@suse.de>
21190         PR debug/84645
21191         * dwarf2out.c (gen_variable_die): Properly handle late VLA
21192         type annotation with LTO when debug was disabled at compile-time.
21194 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
21196         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21197         XINT with INTVAL.
21198         (mips_final_postscan_insn): Likewise.
21200 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
21202         PR rtl-optimization/84528
21203         * alias.c (init_alias_target): Add commentary.
21204         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21205         a unique base value if the frame pointer is not eliminated
21206         to the stack pointer.
21208 2018-03-01  Tom de Vries  <tom@codesourcery.com>
21210         PR rtl-optimization/83327
21211         * lra-int.h (hard_regs_spilled_into): Declare.
21212         * lra.c (hard_regs_spilled_into): Define.
21213         (init_reg_info): Init hard_regs_spilled_into.
21214         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21215         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21216         (process_bb_lives): Handle hard_regs_spilled_into.
21217         (lra_create_live_ranges_1): Before doing liveness propagation, clear
21218         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21220 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
21222         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
21223         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
21224         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
21225         * config/rs6000/aix72.h: New file.
21227 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
21229         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
21230         instead of warning_at with conditional singular and plural messages
21231         where possible.
21233         PR target/52991
21234         * stor-layout.c (update_alignment_for_field): For
21235         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
21236         && !DECL_PACKED (field), do the alignment update, just use
21237         only desired_align instead of MAX (type_align, desired_align)
21238         as the alignment.
21239         (place_field): Don't do known_align < desired_align handling
21240         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
21241         is non-NULL, instead do it after rli->prev_field handling and
21242         only if not within a bitfield word.  For DECL_PACKED (field)
21243         use type_align of BITS_PER_UNIT.
21245 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
21247         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
21248         superfluous parentheses and trailing spaces.
21250 2018-02-28  Richard Biener  <rguenther@suse.de>
21252         PR tree-optimization/84584
21253         * graphite-scop-detection.c (scop_detection::add_scop): Discard
21254         SCoPs with fake exit edge.
21256 2018-02-28  Martin Liska  <mliska@suse.cz>
21258         PR testsuite/84597
21259         * timevar.c (timer::print): Fix format to properly print 100%
21260         values.
21262 2018-02-28  Richard Biener  <rguenther@suse.de>
21264         PR middle-end/84607
21265         * genmatch.c (capture_info::walk_match): Do not mark
21266         captured expressions without operands as expr_p given
21267         they act more like predicates and should be subject to
21268         "lost tail" side-effect preserving.
21270 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
21272         PR rtl-optimization/81611
21273         * auto-inc-dec.c (attempt_change): Move dead note from
21274         mem_insn if it's the next use of regno
21275         (find_address): Take address use of reg holding
21276         non-incremented value.  Add parm to limit search to the named
21277         reg only.
21278         (merge_in_block): Attempt to use a mem insn that is the next
21279         use of the original regno.
21281 2018-02-27  Martin Sebor  <msebor@redhat.com>
21283         PR c++/83871
21284         * doc/invoke.texi (-Wmissing-attributes): New option.
21285         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
21287 2018-02-27  Martin Sebor  <msebor@redhat.com>
21289         PR translation/84207
21290         * diagnostic-core.h (warning_n, error_n, inform_n): Change
21291         n argument to unsigned HOST_WIDE_INT.
21292         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
21293         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
21294         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
21295         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
21297 2018-02-27  Richard Biener  <rguenther@suse.de>
21299         PR tree-optimization/84512
21300         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21301         Do not use the estimate returned from record_stmt_cost for
21302         the scalar iteration cost but sum properly using add_stmt_cost.
21304 2018-02-27  Richard Biener  <rguenther@suse.de>
21306         PR tree-optimization/84466
21307         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21308         Adjust last change to less strictly validate use operands.
21310 2018-02-27  Martin Liska  <mliska@suse.cz>
21312         PR gcov-profile/84548
21313         * gcov.c (process_file): Allow partial overlap and consider it
21314         also as group functions.
21315         (output_lines): Properly calculate range of lines for a group.
21317 2018-02-27  Martin Liska  <mliska@suse.cz>
21319         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
21320         'ggc' suffixes.  Change first column width.
21321         (timer::print): Fix formatting of the column.
21323 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
21325         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
21326         preserve inline entry blocks for the sake of debug inline
21327         entry point markers alone.
21328         (remove_unused_locals): Suggest in comments a better place to
21329         force the preservation of inline entry blocks that are
21330         otherwise unused, but do not preserve them.
21332 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21334         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
21336 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21338         PR target/84039
21339         * config/i386/constraints.md (Bs): Replace
21340         ix86_indirect_branch_register with
21341         TARGET_INDIRECT_BRANCH_REGISTER.
21342         (Bw): Likewise.
21343         * config/i386/i386.md (indirect_jump): Likewise.
21344         (tablejump): Likewise.
21345         (*sibcall_memory): Likewise.
21346         (*sibcall_value_memory): Likewise.
21347         Peepholes of indirect call and jump via memory: Likewise.
21348         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
21349         (*sibcall_value_GOT_32): Likewise.
21350         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21351         (GOT_memory_operand): Likewise.
21352         (call_insn_operand): Likewise.
21353         (sibcall_insn_operand): Likewise.
21354         (GOT32_symbol_operand): Likewise.
21355         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
21357 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21359         PR rtl-optimization/83496
21360         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
21361         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
21362         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
21363         redundant insn, if any.
21364         (relax_delay_slots): Likewise.
21365         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
21367 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
21369         PR tree-optimization/83965
21370         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
21371         that grouped statements are part of a reduction chain.  Return
21372         true if the statement is not marked as a reduction itself but
21373         is part of a group.
21374         (vect_recog_dot_prod_pattern): Don't check whether the statement
21375         is part of a group here.
21376         (vect_recog_sad_pattern): Likewise.
21377         (vect_recog_widen_sum_pattern): Likewise.
21379 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21381         PR debug/84545
21382         * final.c (rest_of_clean_state): Also look for calls inside sequences.
21384 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21386         PR target/84530
21387         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
21388         the bool argument.
21389         (ix86_output_indirect_function_return): New prototype.
21390         (ix86_split_simple_return_pop_internal): Likewise.
21391         * config/i386/i386.c (indirect_return_via_cx): New.
21392         (indirect_return_via_cx_bnd): Likewise.
21393         (indirect_thunk_name): Handle return va CX_REG.
21394         (output_indirect_thunk_function): Create alias for
21395         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
21396         (ix86_output_indirect_jmp): Remove the bool argument.
21397         (ix86_output_indirect_function_return): New function.
21398         (ix86_split_simple_return_pop_internal): Likewise.
21399         * config/i386/i386.md (*indirect_jump): Don't pass false
21400         to ix86_output_indirect_jmp.
21401         (*tablejump_1): Likewise.
21402         (simple_return_pop_internal): Change it to define_insn_and_split.
21403         Call ix86_split_simple_return_pop_internal to split it for
21404         -mfunction-return=.
21405         (simple_return_indirect_internal): Call
21406         ix86_output_indirect_function_return instead of
21407         ix86_output_indirect_jmp.
21409 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
21411         PR bootstrap/84405
21412         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
21413         memset and value initialization afterwards.
21415 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
21417         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
21419 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
21421         PR target/84521
21422         * common/config/aarch64/aarch64-common.c
21423         (aarch_option_optimization_table[]): Switch
21424         off fomit-frame-pointer
21426 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
21427             Chung-Ju Wu  <jasonwucj@gmail.com>
21429         * config/nds32/nds32-multiple.md (load_multiple): Disallow
21430         volatile memory.
21431         (store_multiple): Ditto.
21433 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
21435         * config.gcc: Add --with-cpu support for nds32 target.
21436         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
21437         * config/nds32/nds32.opt: Add -mcpu= option.
21439 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
21441         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
21442         isel=yes): Warn for these deprecated options.
21444 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
21446         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
21447         ISA_2_5_MASKS_EMBEDDED.
21449 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
21451         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
21452         p->max as pointers rather than using iterative_hash_expr.
21454 2018-02-23  Carl Love  <cel@us.ibm.com>
21456         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
21457         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
21458         BU_P8V_OVERLOAD_2.
21459         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
21460         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
21461         P8V_BUILTIN_VEC_VUNSIGNED2.
21463 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
21465         PR target/81572
21466         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
21467         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
21468         LRA_UNKNOWN_ALT.
21469         * lra-constraints.c (curr_insn_transform): Set up
21470         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
21471         LRA_UNKNOWN_ALT.
21472         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
21473         * lra-eliminations.c (spill_pseudos): Ditto.
21474         (process_insn_for_elimination): Ditto.
21475         * lra-lives.c (reg_early_clobber_p): Use the new macros.
21476         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
21477         LRA_NON_CLOBBERED_ALT.
21479 2018-02-22  Martin Sebor  <msebor@redhat.com>
21481         PR tree-optimization/84480
21482         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
21483         to maybe_diag_stxncpy_trunc.  Call it.
21484         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
21485         from gimple_fold_builtin_strcpy.  Print inlining stack.
21486         (handle_builtin_stxncpy): Print inlining stack.
21487         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
21489 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
21491         PR target/84176
21492         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
21493         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
21494         and -fcheck-pointer-bounds are used together.
21495         (indirect_thunk_prefix): New enum.
21496         (indirect_thunk_need_prefix): New function.
21497         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
21498         "_nt" instead of "_bnd" for NOTRACK prefix.
21499         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
21500         (output_indirect_thunk_function): Likewise.
21501         (): Likewise.
21502         (ix86_code_end): Update output_indirect_thunk_function calls.
21503         (ix86_output_indirect_branch_via_reg): Replace
21504         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
21505         (ix86_output_indirect_branch_via_push): Likewise.
21506         (ix86_output_function_return): Likewise.
21507         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
21508         incompatible with -fcf-protection=branch and
21509         -fcheck-pointer-bounds.
21511 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
21513         PR target/83335
21514         * config/aarch64/aarch64.c (aarch64_print_address_internal):
21515         Change gcc_assert call to output_operand_lossage.
21517 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
21519         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
21521 2018-02-22  DJ Delorie  <dj@redhat.com>
21522             Sebastian Perta  <sebastian.perta@renesas.com>
21523             Oleg Endo  <olegendo@gcc.gnu.org>
21525         * config/rx/rx.c (rx_rtx_costs): New function.
21526         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
21528 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21530         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
21532 2018-02-22  Martin Liska  <mliska@suse.cz>
21534         PR driver/83193
21535         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
21536         Add "native" as a possible value.
21538 2018-02-22  Martin Liska  <mliska@suse.cz>
21540         PR driver/83193
21541         * config/i386/i386.c (ix86_option_override_internal):
21542         Add "native" as a possible value for -march and -mtune.
21544 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
21546         PR target/84502
21547         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
21548         to all type variants.
21550         PR tree-optimization/84503
21551         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
21552         width as info->bitpos + info->bitsize - start.
21553         (merged_store_group::merge_overlapping): Simplify width computation.
21554         (check_no_overlap): New function.
21555         (imm_store_chain_info::try_coalesce_bswap): Compute expected
21556         start + width and last_order of the group, fail if check_no_overlap
21557         fails.
21558         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
21559         to group if check_no_overlap fails.
21561 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
21563         * config/rs6000/altivec.md: Delete contraint arguments to
21564         define_expand, define_split, and define_peephole2, and in
21565         define_insn_and_split if always unused.
21566         * config/rs6000/darwin.md: Ditto.
21567         * config/rs6000/dfp.md: Ditto.
21568         * config/rs6000/rs6000.md: Ditto.
21569         * config/rs6000/sync.md: Ditto.
21570         * config/rs6000/vector.md: Ditto.
21571         * config/rs6000/vsx.md: Ditto.
21573 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
21575         * config/rs6000/altivec.md: Write output control strings as braced
21576         blocks instead of double-quoted strings.
21577         * config/rs6000/darwin.md: Ditto.
21578         * config/rs6000/rs6000.md: Ditto.
21579         * config/rs6000/vector.md: Ditto.
21580         * config/rs6000/vsx.md: Ditto.
21582 2018-02-21  Jason Merrill  <jason@redhat.com>
21584         PR c++/84314 - ICE with templates and fastcall attribute.
21585         * attribs.c (build_type_attribute_qual_variant): Remove assert.
21587 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
21589         * ipa-cp.c (determine_versionability): Fix comment typos.
21591 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
21593         PR c/84229
21594         * ipa-cp.c (determine_versionability): Do not version functions caling
21595         va_arg_pack.
21597 2018-02-21  Martin Liska  <mliska@suse.cz>
21599         PR driver/83193
21600         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
21601         Add "native" as a possible value.
21602         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
21603         the macro when native cpu detection is available.
21605 2018-02-21  Martin Liska  <mliska@suse.cz>
21607         PR driver/83193
21608         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
21609         Add "native" as a possible value.
21610         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
21611         when native cpu detection is available.
21613 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
21614             Martin Sebor  <msebor@redhat.com>
21616         PR tree-optimization/84478
21617         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
21618         false.
21619         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
21620         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
21621         support which is conservatively correct, for 2 only stay conservative
21622         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
21623         argument to the 2 argument get_range_strlen, adjust 6 arg
21624         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
21625         false.
21626         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
21627         (gimple_fold_builtin_strlen): Pass true as last argument to
21628         get_range_strlen.
21630 2018-02-20  Martin Sebor  <msebor@redhat.com>
21632         PR middle-end/84095
21633         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
21634         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
21635         (builtin_memref::builtin_memref): Factor out parts into
21636         set_base_and_offset and call it.
21638 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
21640         PR middle-end/84406
21641         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
21642         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
21643         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
21644         search at the associated MODE_INT.
21646 2018-02-20  Jeff Law  <law@redhat.com>
21648         PR middle-end/82123
21649         PR tree-optimization/81592
21650         PR middle-end/79257
21651         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
21652         for range data rather than using global data.
21653         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
21654         range data rather than using global data.
21655         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
21656         pass it to children as needed.
21657         (struct directive::fmtresult): Similarly.
21658         (struct directive::set_width): Similarly.
21659         (struct directive::set_precision): Similarly.
21660         (format_integer, format_directive, parse_directive): Similarly.
21661         (format_none): Accept unnamed vr_values parameter.
21662         (format_percent, format_floating, format_character): Similarly.
21663         (format_string, format_plain): Similarly.
21664         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
21665         the EVRP range analyzer for range data rather than using global data.
21666         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
21667         gimple-ssa-evrp-analyze.h
21668         (class sprintf_dom_walker): Add after_dom_children member function.
21669         Add evrp_range_analyzer member.
21670         (sprintf_dom_walker::before_dom_children): Call into the EVRP
21671         range analyzer as needed.
21672         (sprintf_dom_walker::after_dom_children): New member function.
21673         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
21674         if not optimizing.
21675         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
21676         (evrp_range_analyzer::pop_to_marker): Likewise.
21678 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
21680         PR tree-optimization/84419
21681         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
21682         with the required type if its current type is compatible but
21683         different.
21685 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
21687         PR middle-end/82004
21688         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
21689         after vectorization.
21691 2018-02-20  Martin Liska  <mliska@suse.cz>
21693         PR driver/83193
21694         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
21695         possible values if we don't have a hint.
21697 2018-02-20  Martin Liska  <mliska@suse.cz>
21699         PR c/84310
21700         PR target/79747
21701         * final.c (shorten_branches): Build align_tab array with one
21702         more element.
21703         * opts.c (finish_options): Add alignment option limit check.
21704         (MAX_CODE_ALIGN): Likewise.
21705         (MAX_CODE_ALIGN_VALUE): Likewise.
21706         * doc/invoke.texi: Document maximum allowed option value for
21707         all -falign-* options.
21709 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
21711         PR target/84146
21712         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
21713         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
21714         * var-tracking.c (emit_note_insn_var_location): Remove all references
21715         to NOTE_INSN_CALL_ARG_LOCATION.
21716         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
21717         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
21718         Use copy_rtx_if_shared.
21719         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
21720         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
21721         (dwarf2out_var_location): Remove handling of
21722         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
21723         on call_insn.
21724         * final.c (final_scan_insn): Remove all references to
21725         NOTE_INSN_CALL_ARG_LOCATION.
21726         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
21727         before dumping final insns.
21728         * except.c (emit_note_eh_region_end): Remove all references to
21729         NOTE_INSN_CALL_ARG_LOCATION.
21730         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
21731         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
21732         * config/arc/arc.c (hwloop_optimize): Likewise.
21733         * config/arm/arm.c (create_fix_barrier): Likewise.
21734         * config/s390/s390.c (s390_chunkify_start): Likewise.
21735         * config/sh/sh.c (find_barrier): Likewise.
21736         * config/i386/i386.c (rest_of_insert_endbranch,
21737         ix86_seh_fixup_eh_fallthru): Likewise.
21738         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
21739         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
21740         * config/frv/frv.c (frv_function_prologue): Likewise.
21741         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
21742         reg note.
21743         (note_outside_basic_block_p): Remove all references to
21744         NOTE_INSN_CALL_ARG_LOCATION.
21745         * gengtype.c (adjust_field_rtx_def): Likewise.
21746         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
21747         Likewise.
21748         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
21749         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
21751         PR c++/84444
21752         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
21753         is ADDR_EXPR.
21755         PR tree-optimization/84452
21756         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
21757         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
21758         is NULL.
21760 2018-02-19  Martin Liska  <mliska@suse.cz>
21762         PR sanitizer/82183
21763         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
21765 2018-02-19  Martin Liska  <mliska@suse.cz>
21766             Richard Sandiford  <richard.sandiford@linaro.org>
21768         PR tree-optimization/82491
21769         * gimple-fold.c (get_base_constructor): Make earlier bail out
21770         to prevent ubsan.
21772 2018-02-19  Carl Love  <cel@us.ibm.com>
21774         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
21775         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
21776         BU_P8V_OVERLOAD_1.
21777         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
21778         P8V_BUILTIN_VEC_NEG.
21780 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
21782         * config/rl78/rl78.md (movdf): New define expand.
21784 2018-02-19  Martin Liska  <mliska@suse.cz>
21786         PR other/80589
21787         * doc/invoke.texi: Fix typo.
21788         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
21790 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
21792         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
21793         handle rs6000_single_float and rs6000_double_float specially for
21794         e500 family CPUs.
21796 2018-02-16  Jeff Law  <law@redhat.com>
21798         * config/rx/rx.c (add_pop_cfi_notes): New function.;
21799         (pop_regs): Use it.
21801 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
21803         PR ipa/84425
21804         * ipa-inline.c (inline_small_functions): Fix a typo.
21806 2018-02-16  Nathan Sidwell  <nathan@acm.org>
21808         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
21810 2018-02-16  Carl Love  <cel@us.ibm.com>
21812         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
21813         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
21814         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
21815         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
21816         expansion to P8V_BUILTIN_VEC_FLOAT2.
21818 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
21820         PR rtl-optimization/70023
21821         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
21822         src_regno into account.
21824 2018-02-16  Carl Love  <cel@us.ibm.com>
21826         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
21827         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
21828         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
21829         * config/rs6000/rs6000.c: Remove case statements for
21830         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
21831         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
21832         and P9V_BUILTIN_VEC_VINSERT4B.
21833         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
21834         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
21835         * config/rs6000/vsx.md:
21836         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
21837         vec_insert4b.
21839 2018-02-16  Carl Love  <cel@us.ibm.com>
21841         * config/rs6000/altivec.h: Add builtin names vec_extract4b
21842         vec_insert4b.
21843         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
21844         definitions.
21845         * config/rs6000/rs6000-c.c: Add the definitions for
21846         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
21847         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
21848         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
21849         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
21850         definition for insert4b and define insn *insert3b_internal.
21851         * doc/extend.texi: Add documentation for vec_extract4b.
21853 2018-02-16  Nathan Sidwell  <nathan@acm.org>
21855         * doc/extend.texi (Backwards Compatibility): Mention friend
21856         injection.  Note for-scope is deprecated.
21857         * doc/invoke.texi (-ffriend-injection): Deprecate.
21859 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
21861         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
21862         that moved to I2, also allow destinations that are a paradoxical
21863         subreg (instead of a normal reg).
21865 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
21867         PR target/83831
21868         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
21869         to QImode.
21871 2018-02-16  Richard Biener  <rguenther@suse.de>
21873         PR tree-optimization/84037
21874         PR tree-optimization/84016
21875         PR target/82862
21876         * config/i386/i386.c (ix86_builtin_vectorization_cost):
21877         Adjust vec_construct for the fact we need additional higher latency
21878         128bit inserts for AVX256 and AVX512 vector builds.
21879         (ix86_add_stmt_cost): Scale vector construction cost for
21880         elementwise loads.
21882 2018-02-16  Richard Biener  <rguenther@suse.de>
21884         PR tree-optimization/84417
21885         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
21886         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
21887         (non_rewritable_lvalue_p): Likewise, use poly-ints.
21889 2018-02-16  Martin Liska  <mliska@suse.cz>
21891         PR sanitizer/84307
21892         * internal-fn.def (ASAN_CHECK): Set proper flags.
21893         (ASAN_MARK): Likewise.
21895 2018-02-16  Julia Koval  <julia.koval@intel.com>
21897         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
21898         from PTA_CANNONLAKE.
21900 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
21902         PR target/84272
21903         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
21904         Use ++iter rather than iter++ for std::list iterators.
21905         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
21906         defer deleting them until all nodes in the forest are processed.  Do
21907         free even leaf nodes.  Change to_process into auto_vec.
21909         PR bootstrap/84405
21910         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
21911         * vec.h (vec_default_construct): Use memset instead of placement new
21912         if BROKEN_VALUE_INITIALIZATION is defined.
21913         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
21914         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
21915         is defined.
21917         PR rtl-optimization/83723
21918         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
21919         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
21920         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
21921         recursive calls.
21922         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21923         callers.
21924         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21926 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
21928         PR rtl-optimization/81443
21929         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21930         from inner REGs to paradoxical SUBREGs.
21932 2018-02-16  Richard Biener  <rguenther@suse.de>
21934         PR tree-optimization/84399
21935         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21936         For operands we can analyze at their definition make sure we can
21937         analyze them at each use as well.
21939 2018-02-16  Richard Biener  <rguenther@suse.de>
21941         PR tree-optimization/84190
21942         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21943         volatile accesses if the decl isn't volatile.
21945 2018-02-15  Jason Merrill  <jason@redhat.com>
21947         PR c++/84314 - ICE with templates and fastcall attribute.
21948         * attribs.c (build_type_attribute_qual_variant): Don't clobber
21949         TYPE_CANONICAL on an existing type.
21951 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
21953         PR tree-optimization/84383
21954         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21955         dstoff nor call operand_equal_p if dstbase is NULL.
21957         PR tree-optimization/84334
21958         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21959         also a CONSTANT_CLASS_P, punt.
21961 2018-02-14  Jim Wilson  <jimw@sifive.com>
21963         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21964         first SMALL_OPERAND check.  New local min_second_step.  Move assert
21965         to where locals are set.  Add TARGET_RVC support.
21966         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21968 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
21970         * doc/invoke.texi: Correct -Wformat-overflow code sample.
21972 2018-02-14  Martin Sebor  <msebor@redhat.com>
21974         PR tree-optimization/83698
21975         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21976         arrays constrain the offset range to their bounds.
21977         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21978         (builtin_access::overlap): Avoid setting the size of overlap if it's
21979         already been set.
21980         (maybe_diag_overlap): Also consider arrays when deciding what values
21981         of offsets to include in diagnostics.
21983 2018-02-14  Martin Sebor  <msebor@redhat.com>
21985         PR c/84108
21986         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21987         that correspond to the kind of a declaration.
21989 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
21991         PR target/83984
21992         * config/pa/pa.md: Load address of PIC label using the linkage table
21993         if the label is nonlocal.
21995 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21997         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21998         warning message if user requests -maltivec=be.
21999         * doc/invoke.texi: Document deprecation of -maltivec=be.
22001 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
22003         PR target/84220
22004         * config/rs6000/rs6000-c.c: Update definitions for
22005         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
22006         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
22008 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22010         PR target/84239
22011         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
22012         add _get_ssp intrinsics. Remove argument from
22013         __builtin_ia32_rdssp[d|q].
22014         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
22015         * config/i386/i386-builtin.def: Remove argument from
22016         __builtin_ia32_rdssp[d|q].
22017         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
22018         ix86_expand_special_args_builtin for _rdssp[d|q].
22019         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
22020         Clear register before usage.
22021         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
22022         Add documentation for new _get_ssp and _inc_ssp intrinsics.
22024 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
22026         PR tree-optimization/84357
22027         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
22028         operand 1 of an ARRAY_REF too.
22030 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
22032         PR target/83831
22033         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
22034         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22035         declarations.
22036         (set_of_reg): New struct.
22037         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
22038         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
22039         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22040         functions.
22041         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
22042         Split into bitclr, bitset, bitinvert patterns if appropriate.
22043         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
22044         use rx_fuse_in_memory_bitop.
22045         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
22046         to named insn, correct maximum insn length.
22048 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
22050         PR target/79242
22051         * machmode.def: Define a complex mode for PARTIAL_INT.
22052         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
22053         MODE_PARTIAL_INT.
22054         * doc/rtl.texi: Document CSPImode.
22055         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
22056         handling.
22057         (msp430_hard_regno_nregs_with_padding): Likewise.
22059 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
22061         PR target/84279
22062         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
22064 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
22066         PR rtl-optimization/84169
22067         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
22068         we generated a parallel as new i3 and we split that to new i2 and i3
22069         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
22070         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
22071         those to i2, not i1.  Partially rewrite this scan code.
22073 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22075         PR c/82210
22076         * stor-layout.c (place_field): For variable length fields, adjust
22077         offset_align afterwards not just based on the field's alignment,
22078         but also on the size.
22080         PR middle-end/84309
22081         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
22082         of exps and logs in the use_exp2 case.
22084 2018-02-13  Jeff Law  <law@redhat.com>
22086         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
22087         entry for "vector".
22089         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
22090         ARGS as unused.
22092 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
22094         PR debug/84342
22095         PR debug/84319
22096         * common.opt (gas-loc-support, gas-locview-support): New.
22097         (ginline-points, ginternal-reset-location-views): New.
22098         * doc/invoke.texi: Document them.  Use @itemx where intended.
22099         (gvariable-location-views): Adjust.
22100         * target.def (reset_location_view): New.
22101         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
22102         (TARGET_RESET_LOCATION_VIEW): New.
22103         * doc/tm.texi: Rebuilt.
22104         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
22105         (dwarf2out_default_as_locview_support): New.
22106         (output_asm_line_debug_info): Use option variables.
22107         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
22108         (output_loc_list): Likewise.
22109         (add_high_low_attributes): Check option variables.
22110         Don't output entry view attribute in strict mode.
22111         (gen_inlined_subroutine_die): Check option variables.
22112         (dwarf2out_inline_entry): Likewise.
22113         (init_sections_and_labels): Likewise.
22114         (dwarf2out_early_finish): Likewise.
22115         (maybe_reset_location_view): New, from...
22116         (dwarf2out_var_location): ... here.  Call it.
22117         * debug.h (dwarf2out_default_as_loc_support): Declare.
22118         (dwarf2out_default_as_locview_support): Declare.
22119         * hooks.c (hook_int_rtx_insn_0): New.
22120         * hooks.h (hook_int_rtx_insn_0): Declare.
22121         * toplev.c (process_options): Take -gas-loc-support and
22122         -gas-locview-support from dwarf2out.  Enable
22123         -gvariable-location-views by default only with locview
22124         assembler support.  Enable -ginternal-reset-location-views by
22125         default only if the target defines the corresponding hook.
22126         Enable -ginline-points by default if location views are
22127         enabled; force it disabled if statement frontiers are
22128         disabled.
22129         * tree-inline.c (expand_call_inline): Check option variables.
22130         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22132 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
22134         PR tree-optimization/84321
22135         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22136         handling.  Also check whether the anti-range contains any values
22137         that satisfy the mask; switch to a VR_RANGE if not.
22139 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
22141         PR sanitizer/84340
22142         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22144 2018-02-13  Martin Jambor  <mjambor@suse.cz>
22146         PR c++/83990
22147         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22148         of call statements, also set location of a load to a temporary.
22150 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
22152         * config/rl78/rl78.c (add_vector_labels): New function.
22153         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22154         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22155         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22156         which checks that no arguments are passed.
22157         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22158         * doc/extend.texi: Documentation for the new attribute.
22160 2018-02-13  Andreas Schwab  <schwab@suse.de>
22162         * config/riscv/linux.h (CPP_SPEC): Define.
22164 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22166         PR target/84335
22167         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
22168         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
22169         OPTION_MASK_ISA_AES as first argument to def_builtin_const
22170         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
22171         instead of OPTION_MASK_ISA_PCLMUL as first argument to
22172         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
22173         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
22174         temporarily for AES and PCLMUL builtins.
22176         PR tree-optimization/84339
22177         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22178         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22179         Formatting fixes.
22181         PR middle-end/84309
22182         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22183         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22184         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22185         inline function.
22186         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22187         inline function.
22188         * omp-simd-clone.h: New file.
22189         * omp-simd-clone.c: Include omp-simd-clone.h.
22190         (expand_simd_clones): No longer static.
22191         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22192         cgraph.h and omp-simd-clone.h.
22193         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22194         (vect_recog_widen_shift_pattern): Formatting fix.
22195         (vect_pattern_recog_1): Don't check optab for calls.
22197         PR target/84336
22198         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22199         operands[2] into a REG before using gen_lowpart on it.
22201 2018-02-12  Jeff Law  <law@redhat.com>
22203         PR target/83760
22204         * config/sh/sh.c (find_barrier): Consider a sibling call
22205         a barrier as well.
22207         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22208         successfully back substituting a reg.
22210 2018-02-12  Richard Biener  <rguenther@suse.de>
22212         PR tree-optimization/84037
22213         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22214         parameter, move visited init to caller.
22215         (vect_slp_analyze_operations): Separate cost from validity
22216         check, initialize visited once for all instances.
22217         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22218         for all instances.
22219         * tree-vect-stmts.c (vect_model_simple_cost): Make early
22220         out an assert.
22221         (vect_model_promotion_demotion_cost): Likewise.
22222         (vectorizable_bswap): Guard cost modeling with !slp_node
22223         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
22224         SLP stmts.
22225         (vectorizable_call): Likewise.
22226         (vectorizable_conversion): Likewise.
22227         (vectorizable_assignment): Likewise.
22228         (vectorizable_shift): Likewise.
22229         (vectorizable_operation): Likewise.
22230         (vectorizable_store): Likewise.
22231         (vectorizable_load): Likewise.
22232         (vectorizable_condition): Likewise.
22233         (vectorizable_comparison): Likewise.
22235 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
22237         PR sanitizer/84307
22238         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
22239         (ASAN_MARK): Fix fnspec to account for return value, change pointer
22240         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
22242 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22244         PR middle-end/83665
22245         * params.def (inline-min-speedup): Increase from 8 to 15.
22246         (max-inline-insns-auto): Decrease from 40 to 30.
22247         * ipa-split.c (consider_split): Add some buffer for function to
22248         be considered inlining candidate.
22249         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
22250         default values.
22252 2018-02-12  Richard Biener  <rguenther@suse.de>
22254         PR tree-optimization/84037
22255         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
22256         matched stmts if we cannot swap the non-matched ones.
22258 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22260         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
22261         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
22262         _mm_maskz_scalef_round_ss): New intrinsics.
22263         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
22264         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
22265         __builtin_ia32_scalefss_round): Remove.
22266         (__builtin_ia32_scalefsd_mask_round,
22267         __builtin_ia32_scalefss_mask_round): New intrinsics.
22268         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
22269         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
22270         ((match_operand:VF_128 2 "<round_nimm_predicate>"
22271         "<round_constraint>")): Changed to ...
22272         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
22273         "<round_scalar_constraint>")): ... this.
22274         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
22275         %0, %1, %2<round_op3>}"): Changed to ...
22276         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
22277         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
22278         %2<round_scalar_mask_op3>}"): ... this.
22279         * config/i386/subst.md (round_scalar_nimm_predicate): New.
22281 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22283         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
22284         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
22285         (_mm_maskz_sqrt_round_ss): New intrinsics.
22286         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
22287         (__builtin_ia32_sqrtsd_mask_round)
22288         (__builtin_ia32_sqrtss_mask_round): New builtins.
22289         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
22290         (__builtin_ia32_sqrtss_round): Remove.
22291         (__builtin_ia32_sqrtsd_mask_round)
22292         (__builtin_ia32_sqrtss_mask_round): New builtins.
22293         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
22294         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
22295         ((match_operand:VF_128 1 "vector_operand"
22296         "xBm,<round_constraint>")): Changed to ...
22297         ((match_operand:VF_128 1 "vector_operand"
22298         "xBm,<round_scalar_constraint>")): ... this.
22299         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
22300         %0, %2, %<iptr>1<round_op3>}): Changed to ...
22301         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
22302         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
22303         %<iptr>1<round_scalar_mask_op3>}): ... this.
22304         ((set_attr "prefix" "<round_prefix>")): Changed to ...
22305         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
22307 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
22309         PR target/84266
22310         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
22311         Cast vec_cmpeq result to correct type.
22312         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
22313         Cast vec_cmpgt result to correct type.
22315 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
22317         * final.c (final_scan_insn_1): Renamed from...
22318         (final_scan_insn): ... this.  New wrapper, to recover
22319         seen from the outermost call in recursive ones.
22320         * config/sparc/sparc.c (output_return): Drop seen from call.
22321         (output_sibcall): Likewise.
22322         * config/visium/visium.c (output_branch): Likewise.
22324 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
22326         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
22327         function label.
22329 2018-02-10  Alan Modra  <amodra@gmail.com>
22331         PR target/84300
22332         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
22333         Specify LR as an input.
22335 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
22337         PR sanitizer/83987
22338         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
22339         remove_member_access_dummy_vars): New functions.
22340         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
22341         lower_omp_1, execute_lower_omp): Use them.
22343         PR rtl-optimization/84308
22344         * shrink-wrap.c (spread_components): Release todo vector.
22346 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
22348         PR rtl-optimization/57193
22349         * ira-color.c (struct allocno_color_data): Add member
22350         conflict_allocno_hard_prefs.
22351         (update_conflict_allocno_hard_prefs): New.
22352         (bucket_allocno_compare_func): Add a preference based on
22353         conflict_allocno_hard_prefs.
22354         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
22355         (color_allocnos): Remove a dead code.  Initiate
22356         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
22358 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22360         PR target/84226
22361         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
22362         constraint from =wa to wa.  Avoid a subreg on the output operand,
22363         instead use a pseudo and subreg it in a move.
22364         (p9_xxbrd_<mode>): Changed to ...
22365         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
22366         (p9_xxbrd_v2df): New expander.
22367         (p9_xxbrw_<mode>): Changed to ...
22368         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
22369         (p9_xxbrw_v4sf): New expander.
22371 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22373         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
22375 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
22377         PR target/83926
22378         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
22379         multiply in 32-bit mode.
22380         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
22381         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
22382         mode.
22384 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22386         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
22387         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
22388         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
22389         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
22391 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
22393         PR lto/84213
22394         * dwarf2out.c (is_trivial_indirect_ref): New function.
22395         (dwarf2out_late_global_decl): Do not generate a location
22396         attribute for variables that have a non-trivial DECL_VALUE_EXPR
22397         and that are not defined in the current unit.
22399 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22401         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
22402         instead of a libcall for UNORDERED.
22404 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
22406         PR target/82641
22407         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
22408         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
22410 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22412         PR target/PR84295
22413         * config/s390/s390.c (s390_set_current_function): Invoke
22414         s390_indirect_branch_settings also if fndecl didn't change.
22416 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22418         * config/rs6000/rs6000.md (blockage): Set length to zero.
22420 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22422         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
22424 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22426         PR sanitizer/84285
22427         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
22428         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
22429         -static-lib*san.
22431         PR debug/84252
22432         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
22433         PARALLEL incoming that failed vt_get_decl_and_offset check.
22435         PR middle-end/84237
22436         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
22437         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
22438         TREE_READONLY bit.
22439         (get_variable_section): For decls in named .bss* sections pass true as
22440         second argument to bss_initializer_p.
22442 2018-02-09  Marek Polacek  <polacek@redhat.com>
22443             Jakub Jelinek  <jakub@redhat.com>
22445         PR c++/83659
22446         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
22447         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
22448         Sync some changes from cxx_fold_indirect_ref.
22450 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22452         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
22453         markers.
22454         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
22455         (BLOCK_INLINE_ENTRY_LABEL): New.
22456         (dwarf2out_var_location): Disregard inline entry markers.
22457         (inline_entry_data): New struct.
22458         (inline_entry_data_hasher): New hashtable type.
22459         (inline_entry_data_hasher::hash): New.
22460         (inline_entry_data_hasher::equal): New.
22461         (inline_entry_data_table): New variable.
22462         (add_high_low_attributes): Add DW_AT_entry_pc and
22463         DW_AT_GNU_entry_view attributes if a pending entry is found
22464         in inline_entry_data_table.  Add old entry_pc attribute only
22465         if debug nonbinding markers are disabled.
22466         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
22467         markers are enabled.
22468         (block_within_block_p, dwarf2out_inline_entry): New.
22469         (dwarf2out_finish): Check that no entries remained in
22470         inline_entry_data_table.
22471         * final.c (reemit_insn_block_notes): Handle inline entry notes.
22472         (final_scan_insn, notice_source_line): Likewise.
22473         (rest_of_clean_state): Skip inline entry markers.
22474         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
22475         markers.
22476         * gimple.c (gimple_build_debug_inline_entry): New.
22477         * gimple.h (enum gimple_debug_subcode): Add
22478         GIMPLE_DEBUG_INLINE_ENTRY.
22479         (gimple_build_debug_inline_entry): Declare.
22480         (gimple_debug_inline_entry_p): New.
22481         (gimple_debug_nonbind_marker_p): Adjust.
22482         * insn-notes.def (INLINE_ENTRY): New.
22483         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
22484         inline entry marker notes.
22485         (print_insn): Likewise.
22486         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
22487         (INSN_DEBUG_MARKER_KIND): Likewise.
22488         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
22489         * tree-inline.c (expand_call_inline): Build and insert
22490         debug_inline_entry stmt.
22491         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
22492         inline entry blocks early, if nonbind markers are enabled.
22493         (dump_scope_block): Dump fragment info.
22494         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
22495         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
22496         (gimple_build_debug_inline_entry): New.
22497         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
22498         Enable/disable inline entry points too.
22499         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
22500         (DEBUG_INSN): Describe inline entry markers.
22502         * common.opt (gvariable-location-views): New.
22503         (gvariable-location-views=incompat5): New.
22504         * config.in: Rebuilt.
22505         * configure: Rebuilt.
22506         * configure.ac: Test assembler for view support.
22507         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
22508         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
22509         * dwarf2out.c (var_loc_view): New typedef.
22510         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
22511         (dwarf2out_locviews_in_attribute): New.
22512         (dwarf2out_locviews_in_loclist): New.
22513         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
22514         (enum dw_line_info_opcode): Add LI_adv_address.
22515         (struct dw_line_info_table): Add view.
22516         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
22517         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
22518         (zero_view_p): New variable.
22519         (ZERO_VIEW_P): New macro.
22520         (output_asm_line_debug_info): New.
22521         (struct var_loc_node): Add view.
22522         (add_AT_view_list, AT_loc_list): New.
22523         (add_var_loc_to_decl): Add view param.  Test it against last.
22524         (new_loc_list): Add view params.  Record them.
22525         (AT_loc_list_ptr): Handle loc and view lists.
22526         (view_list_to_loc_list_val_node): New.
22527         (print_dw_val): Handle dw_val_class_view_list.
22528         (size_of_die): Likewise.
22529         (value_format): Likewise.
22530         (loc_list_has_views): New.
22531         (gen_llsym): Set vl_symbol too.
22532         (maybe_gen_llsym, skip_loc_list_entry): New.
22533         (dwarf2out_maybe_output_loclist_view_pair): New.
22534         (output_loc_list): Output view list or entries too.
22535         (output_view_list_offset): New.
22536         (output_die): Handle dw_val_class_view_list.
22537         (output_dwarf_version): New.
22538         (output_compilation_unit_header): Use it.
22539         (output_skeleton_debug_sections): Likewise.
22540         (output_rnglists, output_line_info): Likewise.
22541         (output_pubnames, output_aranges): Update version comments.
22542         (output_one_line_info_table): Output view numbers in asm comments.
22543         (dw_loc_list): Determine current endview, pass it to new_loc_list.
22544         Call maybe_gen_llsym.
22545         (loc_list_from_tree_1): Adjust.
22546         (add_AT_location_description): Create view list attribute if
22547         needed, check it's absent otherwise.
22548         (convert_cfa_to_fb_loc_list): Adjust.
22549         (maybe_emit_file): Call output_asm_line_debug_info for test.
22550         (dwarf2out_var_location): Reset views as needed.  Precompute
22551         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
22552         attribute.  Set view.
22553         (new_line_info_table): Reset next view.
22554         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
22555         (dwarf2out_source_line): Likewise.  Output view resets and labels to
22556         the assembler, or select appropriate line info opcodes.
22557         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
22558         (optimize_string_length): Catch it.  Adjust.
22559         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
22560         dw_val_class_view_list, and remove it if no longer needed.
22561         (hash_loc_list): Hash view numbers.
22562         (loc_list_hasher::equal): Compare them.
22563         (optimize_location_lists): Check whether a view list symbol is
22564         needed, and whether the locview attribute is present, and
22565         whether they match.  Remove the locview attribute if no longer
22566         needed.
22567         (index_location_lists): Call skip_loc_list_entry for test.
22568         (dwarf2out_finish): Call output_asm_line_debug_info for test.
22569         Use output_dwarf_version.
22570         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
22571         (struct dw_val_node): Add val_view_list.
22572         * final.c (SEEN_NEXT_VIEW): New.
22573         (set_next_view_needed): New.
22574         (clear_next_view_needed): New.
22575         (maybe_output_next_view): New.
22576         (final_start_function): Rename to...
22577         (final_start_function_1): ... this.  Take pointer to FIRST,
22578         add SEEN parameter.  Emit param bindings in the initial view.
22579         (final_start_function): Reintroduce SEEN-less interface.
22580         (final): Rename to...
22581         (final_1): ... this.  Take SEEN parameter.  Output final pending
22582         next view at the end.
22583         (final): Reintroduce seen-less interface.
22584         (final_scan_insn): Output pending next view before switching
22585         sections or ending a block.  Mark the next view as needed when
22586         outputting variable locations.  Notify debug backend of section
22587         changes, and of location view changes.
22588         (rest_of_handle_final): Adjust.
22589         * toplev.c (process_options): Autodetect value for debug variable
22590         location views option.  Warn on incompat5 without -gdwarf-5.
22591         * doc/invoke.texi (gvariable-location-views): New.
22592         (gvariable-location-views=incompat5): New.
22593         (gno-variable-location-views): New.
22595 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
22597         PR tree-optimization/84136
22598         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
22599         that the result of find_edge is non-NULL.
22601 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
22603         PR target/83008
22604         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
22605         storing integer register in SImode.  Fix cost of 256 and 512
22606         byte aligned SSE register store.
22608 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
22610         * config/i386/i386.c (ix86_multiplication_cost): Fix
22611         multiplication cost for TARGET_AVX512DQ.
22613 2018-02-08  Marek Polacek  <polacek@redhat.com>
22615         PR tree-optimization/84238
22616         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
22617         get_range_strlen.
22619 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22621         PR tree-optimization/84265
22622         * tree-vect-stmts.c (vectorizable_store): Don't treat
22623         VMAT_CONTIGUOUS accesses as grouped.
22624         (vectorizable_load): Likewise.
22626 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22628         PR tree-optimization/81635
22629         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
22630         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
22631         (test_round_for_mask): New functions.
22632         (wide_int_cc_tests): Call test_round_for_mask.
22633         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
22634         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
22635         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
22636         range returned by get_range_info.
22638 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22640         PR ipa/81360
22641         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
22642         * symtab.c: Include builtins.h
22643         (symtab_node::output_to_lto_symbol_table_p): Move here
22644         from lto-streamer-out.c:output_symbol_p.
22645         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
22646         (output_symbol_p): Move all logic to symtab.c
22647         (produce_symtab): Update.
22649 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22651         * config/s390/s390-opts.h (enum indirect_branch): Define.
22652         * config/s390/s390-protos.h (s390_return_addr_from_memory)
22653         (s390_indirect_branch_via_thunk)
22654         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
22655         (enum s390_indirect_branch_type): Define.
22656         * config/s390/s390.c (struct s390_frame_layout, struct
22657         machine_function): Remove.
22658         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
22659         (indirect_branch_table_label_no, indirect_branch_table_name):
22660         Define variables.
22661         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
22662         (enum s390_indirect_branch_option): Define.
22663         (s390_return_addr_from_memory): New function.
22664         (s390_handle_string_attribute): New function.
22665         (s390_attribute_table): Add new attribute handler.
22666         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
22667         (s390_indirect_branch_via_thunk): New function.
22668         (s390_indirect_branch_via_inline_thunk): New function.
22669         (s390_function_ok_for_sibcall): When jumping via thunk disallow
22670         sibling call optimization for non z10 compiles.
22671         (s390_emit_call): Force indirect branch target to be a single
22672         register.  Add r1 clobber for non-z10 compiles.
22673         (s390_emit_epilogue): Emit return jump via return_use expander.
22674         (s390_reorg): Handle JUMP_INSNs as execute targets.
22675         (s390_option_override_internal): Perform validity checks for the
22676         new command line options.
22677         (s390_indirect_branch_attrvalue): New function.
22678         (s390_indirect_branch_settings): New function.
22679         (s390_set_current_function): Invoke s390_indirect_branch_settings.
22680         (s390_output_indirect_thunk_function):  New function.
22681         (s390_code_end): Implement target hook.
22682         (s390_case_values_threshold): Implement target hook.
22683         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
22684         macros.
22685         * config/s390/s390.h (struct s390_frame_layout)
22686         (struct machine_function): Move here from s390.c.
22687         (TARGET_INDIRECT_BRANCH_NOBP_RET)
22688         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
22689         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
22690         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
22691         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
22692         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
22693         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
22694         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
22695         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
22696         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
22697         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
22698         (mnemonic attribute): Add values which aren't recognized
22699         automatically.
22700         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
22701         pattern for branch conversion.  Fix mnemonic attribute.
22702         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
22703         indirect branch via thunk if requested.
22704         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
22705         ("*indirect_jump"): Disable for branch conversion using out of
22706         line thunks.
22707         ("indirect_jump_via_thunk<mode>_z10")
22708         ("indirect_jump_via_thunk<mode>")
22709         ("indirect_jump_via_inlinethunk<mode>_z10")
22710         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
22711         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
22712         ("casesi_jump_via_inlinethunk<mode>_z10")
22713         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
22714         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
22715         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
22716         ("*indirect2_jump"): Disable for branch conversion.
22717         ("casesi_jump"): Turn into expander and expand patterns for branch
22718         conversion.
22719         ("return_use"): New expander.
22720         ("*return"): Emit return via thunk and rename it to ...
22721         ("*return<mode>"): ... this one.
22722         * config/s390/s390.opt: Add new options and and enum for the
22723         option values.
22725 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22727         * lra-constraints.c (match_reload): Unconditionally use
22728         gen_lowpart_SUBREG, rather than selecting between that
22729         and equivalent gen_rtx_SUBREG code.
22731 2018-02-08  Richard Biener  <rguenther@suse.de>
22733         PR tree-optimization/84233
22734         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
22735         changed flag instead of boguously re-using phi_inserted.
22737 2018-02-08  Martin Jambor  <mjambor@suse.cz>
22739         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
22740         static local variables.
22742 2018-02-08  Richard Biener  <rguenther@suse.de>
22744         PR tree-optimization/84278
22745         * tree-vect-stmts.c (vectorizable_store): When looking for
22746         smaller vector types to perform grouped strided loads/stores
22747         make sure the mode is supported by the target.
22748         (vectorizable_load): Likewise.
22750 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
22752         * config/aarch64/aarch64.c (aarch64_components_for_bb):
22753         Increase LDP/STP opportunities by adding adjacent callee-saves.
22755 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
22757         PR rtl-optimization/84068
22758         PR rtl-optimization/83459
22759         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
22761 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
22763         PR tree-optimization/84224
22764         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
22765         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
22766         non-zero arguments.
22768 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
22770         PR target/84113
22771         * config/rs6000/altivec.md (*restore_world): Remove LR use.
22772         * config/rs6000/predicates.md (restore_world_operation): Adjust op
22773         count, remove one USE.
22775 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22777         * doc/install.texi (Configuration): Document the
22778         --with-long-double-format={ibm,ieee} PowerPC configuration
22779         options.
22781         PR target/84154
22782         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
22783         Convert from define_expand to be define_insn_and_split.  Rework
22784         float/double/_Float128 conversions to QI/HI/SImode to work with
22785         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
22786         conversions to QI/HImode types did a store and then a load to
22787         truncate the value.  For conversions to VSX registers, don't split
22788         the insn, instead emit the code directly.  Use the code iterator
22789         any_fix to combine signed and unsigned conversions.
22790         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
22791         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
22792         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22793         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
22794         (fix_<mode>di2_hw): Likewise.
22795         (fixuns_<mode>di2_hw): Likewise.
22796         (fix_<mode>si2_hw): Likewise.
22797         (fixuns_<mode>si2_hw): Likewise.
22798         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
22799         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22800         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
22801         fix<uns>_trunc<SFDF:mode>si2_p8.
22802         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
22803         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
22804         (fix<uns>_<mode>_mem): Likewise.
22805         (fctiw<u>z_<mode>_mem): Likewise.
22806         (fix<uns>_<mode>_mem): Likewise.
22807         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
22808         the register allocator from doing a direct move to the GPRs to do
22809         a store, and instead use the ISA 3.0 store byte/half-word from
22810         vector register instruction.  For IEEE 128-bit floating point,
22811         also optimize stores of 32-bit ints.
22812         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
22814 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
22816         * genextract.c (push_pathstr_operand): New function to support
22817         [a-zA-Z].
22818         (walk_rtx): Call push_pathstr_operand.
22819         (print_path): Support [a-zA-Z].
22821 2018-02-07  Richard Biener  <rguenther@suse.de>
22823         PR tree-optimization/84037
22824         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
22825         (cse_and_gimplify_to_preheader): Declare.
22826         (vect_get_place_in_interleaving_chain): Likewise.
22827         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22828         ivexpr_map.
22829         (_loop_vec_info::~_loop_vec_info): Delete it.
22830         (cse_and_gimplify_to_preheader): New function.
22831         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
22832         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
22833         (vectorizable_load): Likewise.  For grouped stores always base
22834         the IV on the first element.
22835         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
22836         condition before gimplifying.
22838 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
22840         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
22841         *DIV_EXPR and *MOD_EXPR.
22843 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
22845         PR target/84248
22846         * config/i386/i386.c (ix86_option_override_internal): Mask out
22847         the CF_SET bit when checking -fcf-protection.
22849 2018-02-07  Tom de Vries  <tom@codesourcery.com>
22851         PR libgomp/84217
22852         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
22853         enough.
22855 2018-02-07  Richard Biener  <rguenther@suse.de>
22857         PR tree-optimization/84204
22858         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
22859         this place.
22861         PR tree-optimization/84205
22862         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
22863         special-case isl_ast_op_zdiv_r.
22865         PR tree-optimization/84223
22866         * graphite-scop-detection.c (gather_bbs::before_dom_children):
22867         Only add conditions from within the region.
22868         (gather_bbs::after_dom_children): Adjust.
22870 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
22872         PR target/84209
22873         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
22874         * config/avr/avr.md: Only post-reload split REG-REG moves if
22875         either register is GENERAL_REG_P.
22877 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
22879         PR tree-optimization/84235
22880         * tree-ssa-scopedtables.c
22881         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
22882         if the subtraction is performed in floating point type where NaNs are
22883         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
22884         build 1.  Formatting fix.
22886 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
22888         PR target/84146
22889         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
22890         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
22891         and skip it regardless of bb boundaries.  Use CALL_P macro,
22892         don't test INSN_P (insn) together with CALL_P or JUMP_P check
22893         unnecessarily, formatting fix.
22895 2018-02-06  Michael Collison  <michael.collison@arm.com>
22897         * config/arm/thumb2.md:
22898         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
22899         (*thumb_mov_notscc): Ditto.
22901 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
22903         PR target/84154
22904         * config/rs6000/rs6000.md (su code attribute): Use "u" for
22905         unsigned_fix, not "s".
22907 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22909         * configure.ac (gcc_fn_eh_frame_ro): New function.
22910         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
22911         correct .eh_frame permissions.
22912         * configure: Regenerate.
22914 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
22916         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
22917         irrelevant options.
22919 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22921         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22922         Display warning message for -mno-speculate-indirect-jumps.
22924 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
22926         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22927         Undocumented.
22928         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22930 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
22932         PR tree-optimization/84225
22933         * tree-eh.c (find_trapping_overflow): Only call
22934         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22936 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22938         PR target/84145
22939         * config/i386/i386.c: Reimplement the check of possible options
22940         -mibt/-mshstk conbination. Change error messages.
22941         * doc/invoke.texi: Fix a typo: remove extra '='.
22943 2018-02-06  Marek Polacek  <polacek@redhat.com>
22945         PR tree-optimization/84228
22946         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22948 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
22950         PR target/82641
22951         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22952         emitted arch directives.
22953         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22954         __ARM_FEATURE_COPROC before changing architectures.
22956 2018-02-06  Richard Biener  <rguenther@suse.de>
22958         * config/i386/i386.c (print_reg): Fix typo.
22959         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22961 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
22963         * configure: Regenerate.
22965 2018-02-05  Martin Sebor  <msebor@redhat.com>
22967         PR tree-optimization/83369
22968         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22969         inlining context.
22971 2018-02-05  Martin Liska  <mliska@suse.cz>
22973         * doc/invoke.texi: Cherry-pick upstream r323995.
22975 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
22977         * ira.c (ira_init_register_move_cost): Adjust comment.
22979 2018-02-05  Martin Liska  <mliska@suse.cz>
22981         PR gcov-profile/84137
22982         * doc/gcov.texi: Fix typo in documentation.
22984 2018-02-05  Martin Liska  <mliska@suse.cz>
22986         PR gcov-profile/83879
22987         * doc/gcov.texi: Document necessity of --dynamic-list-data when
22988         using dlopen functionality.
22990 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
22992         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22993         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22994         _mm_maskz_range_ss, _mm_mask_range_round_ss,
22995         _mm_maskz_range_round_ss): New intrinsics.
22996         (__builtin_ia32_rangesd128_round)
22997         (__builtin_ia32_rangess128_round): Remove.
22998         (__builtin_ia32_rangesd128_mask_round,
22999         __builtin_ia32_rangess128_mask_round): New builtins.
23000         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
23001         __builtin_ia32_rangess128_round): Remove.
23002         (__builtin_ia32_rangesd128_mask_round,
23003         __builtin_ia32_rangess128_mask_round): New builtins.
23004         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
23005         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
23006         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
23007         "<round_saeonly_constraint>")): Changed to ...
23008         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
23009         "<round_saeonly_scalar_constraint>")): ... this.
23010         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
23011         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
23012         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
23013         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
23014         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
23016 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
23018         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
23019         options.
23020         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
23021         Remove all values except native, 8540 and 8548.
23023 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
23025         * config/i386/i386.c (ix86_output_function_return): Pass
23026         INVALID_REGNUM, instead of -1, as invalid register number to
23027         indirect_thunk_name and output_indirect_thunk.
23029 2018-02-02  Julia Koval  <julia.koval@intel.com>
23031         * config.gcc: Add -march=icelake.
23032         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
23033         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
23034         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
23035         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
23036         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
23037         (processor_target_table): Add icelake.
23038         (ix86_option_override_internal): Handle new PTAs.
23039         (get_builtin_code_for_version): Handle icelake.
23040         (M_INTEL_COREI7_ICELAKE): New.
23041         (fold_builtin_cpu): Handle icelake.
23042         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
23043         * doc/invoke.texi: Add -march=icelake.
23045 2018-02-02  Julia Koval  <julia.koval@intel.com>
23047         * config/i386/i386.c (ix86_option_override_internal): Change flags type
23048         to wide_int_bitmask.
23049         * wide-int-bitmask.h: New.
23051 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23053         PR target/84066
23054         * config/i386/i386.md: Replace Pmode with word_mode in
23055         builtin_setjmp_setup and builtin_longjmp to support x32.
23057 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
23059         PR target/56010
23060         PR target/83743
23061         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
23062         #include "opts.h".
23063         (rs6000_supported_cpu_names): New static variable.
23064         (linux_cpu_translation_table): Likewise.
23065         (elf_platform) <cpu>: Define new static variable and use it.
23066         Translate kernel AT_PLATFORM name to canonical name if needed.
23067         Error if platform name is unknown.
23069 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
23071         PR target/84089
23072         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
23074 2018-02-01  Jeff Law  <law@redhat.com>
23076         PR target/84128
23077         * config/i386/i386.c (release_scratch_register_on_entry): Add new
23078         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
23079         the scratch if RELEASE_VIA_POP is false.
23080         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
23081         If we have to save a temporary register, decrement SIZE appropriately.
23082         Pass new arguments to release_scratch_register_on_entry.
23083         (ix86_adjust_stack_and_probe): Likewise.
23084         (ix86_emit_probe_stack_range): Pass new arguments to
23085         release_scratch_register_on_entry.
23087 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
23089         PR rtl-optimization/84157
23090         * combine.c (change_zero_ext): Use REG_P predicate in
23091         front of HARD_REGISTER_P predicate.
23093 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
23095         * config/avr/avr.c (avr_option_override): Move disabling of
23096         -fdelete-null-pointer-checks to...
23097         * common/config/avr/avr-common.c (avr_option_optimization_table):
23098         ...here.
23100 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23102         PR tree-optimization/81635
23103         * tree-data-ref.c (split_constant_offset_1): For types that
23104         wrap on overflow, try to use range info to prove that wrapping
23105         cannot occur.
23107 2018-02-01  Renlin Li  <renlin.li@arm.com>
23109         PR target/83370
23110         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
23111         TAILCALL_ADDR_REGS.
23112         (aarch64_register_move_cost): Likewise.
23113         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
23114         TAILCALL_ADDR_REGS.
23115         (REG_CLASS_NAMES): Likewise.
23116         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
23117         TAILCALL_ADDR_REGS. Remove IP registers.
23118         * config/aarch64/aarch64.md (Ucs): Update register constraint.
23120 2018-02-01  Richard Biener  <rguenther@suse.de>
23122         * domwalk.h (dom_walker::dom_walker): Add additional constructor
23123         for specifying RPO order and allow NULL for that.
23124         * domwalk.c (dom_walker::dom_walker): Likewise.
23125         (dom_walker::walk): Handle NULL RPO order.
23126         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23127         in RPO order.
23128         (rewrite_update_dom_walker): Likewise.
23129         (mark_def_dom_walker): Likewise.
23131 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23133         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23134         (aarch64_maybe_expand_sve_subreg_move): Declare.
23135         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23136         * config/aarch64/predicates.md (aarch64_any_register_operand): New
23137         predicate.
23138         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23139         that are semantically a reverse operation.
23140         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23141         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23142         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23143         functions.
23144         (aarch64_can_change_mode_class): For big-endian, forbid changes
23145         between two SVE modes if they have different element sizes.
23147 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23149         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23150         the TImode handling for big-endian targets.
23152 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23154         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23155         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
23156         not just bytes.
23157         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23158         Remove BSWAP handing for big-endian targets and use the form of
23159         LD1RQ appropariate for the mode.
23161 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23163         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23164         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
23165         duplicated element.
23167 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23169         PR tearget/83845
23170         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
23171         check for operands that need to go through aarch64_sve_reload_be.
23173 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
23175         PR tree-optimization/81661
23176         PR tree-optimization/84117
23177         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23178         * tree-eh.c: Include gimplify.h.
23179         (find_trapping_overflow, replace_trapping_overflow,
23180         rewrite_to_non_trapping_overflow): New functions.
23181         * tree-vect-loop.c: Include tree-eh.h.
23182         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23183         * tree-data-ref.c: Include tree-eh.h.
23184         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23186 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
23188         PR rtl-optimization/84123
23189         * combine.c (change_zero_ext): Check if hard register satisfies
23190         can_change_dest_mode before calling gen_lowpart_SUBREG.
23192 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
23194         PR target/82444
23195         * ira.c (ira_init_register_move_cost): Remove assert.
23197 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23199         PR rtl-optimization/84071
23200         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23201         * doc/tm.texi: Regenerate.
23203 2018-01-31  Richard Biener  <rguenther@suse.de>
23205         PR tree-optimization/84132
23206         * tree-data-ref.c (analyze_miv_subscript): Properly
23207         check whether evolution_function_is_affine_multivariate_p
23208         before calling gcd_of_steps_may_divide_p.
23210 2018-01-31  Julia Koval  <julia.koval@intel.com>
23212         PR target/83618
23213         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23214         * config/i386/i386.md (rdpid_rex64) New.
23215         (rdpid): Make 32bit only.
23217 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
23219         PR lto/84105
23220         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23221         an IDENTIFIER_NODE for FUNCTION_TYPE's.
23223 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23225         Revert
23226         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23228         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23230 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23232         PR rtl-optimization/84071
23233         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
23234         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
23236 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23238         * config/arc/arc.c (arc_handle_aux_attribute): New function.
23239         (arc_attribute_table): Add 'aux' attribute.
23240         (arc_in_small_data_p): Consider aux like variables.
23241         (arc_is_aux_reg_p): New function.
23242         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
23243         (arc_get_aux_arg): New function.
23244         (prepare_move_operands): Handle aux-register access.
23245         (arc_handle_aux_attribute): New function.
23246         * doc/extend.texi (ARC Variable attributes): Add subsection.
23248 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23250         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
23251         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
23252         (arc_attribute_table): Add 'uncached' attribute.
23253         (arc_print_operand): Print '.di' flag for uncached memory
23254         accesses.
23255         (arc_in_small_data_p): Do not consider for small data the uncached
23256         types.
23257         (arc_is_uncached_mem_p): New function.
23258         * config/arc/predicates.md (compact_store_memory_operand): Check
23259         for uncached memory accesses.
23260         (nonvol_nonimm_operand): Likewise.
23261         * doc/extend.texi (ARC Type Attribute): New subsection.
23263 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
23265         PR c/84100
23266         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
23267         falign-loops=): Add Optimization flag.
23269 2018-01-30  Jeff Law  <law@redhat.com>
23271         PR target/84064
23272         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
23273         INT_REGISTERS_SAVED.  Check it prior to calling
23274         get_scratch_register_on_entry.
23275         (ix86_adjust_stack_and_probe): Similarly.
23276         (ix86_emit_probe_stack_range): Similarly.
23277         (ix86_expand_prologue): Corresponding changes.
23279 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23281         PR target/40411
23282         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
23283         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
23285 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
23287         PR target/84112
23288         * lra-constraints.c (curr_insn_transform): Process AND in the
23289         address.
23291 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
23293         PR rtl-optimization/83986
23294         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
23295         dependence against last_pending_memory_flush in addition to
23296         pending_jump_insns.
23298 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
23300         PR tree-optimization/81611
23301         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
23302         copies.
23304 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23306         PR target/83758
23307         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
23308         a reg rtx.
23310 2018-01-30  Richard Biener  <rguenther@suse.de>
23311             Jakub Jelinek  <jakub@redhat.com>
23313         PR tree-optimization/84111
23314         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
23315         inner loops added during recursion, as they don't have up-to-date
23316         SSA form.
23318 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23320         PR ipa/81360
23321         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
23322         (can_inline_edge_by_limits_p): ... here.
23323         (can_early_inline_edge_p, check_callers,
23324         update_caller_keys, update_callee_keys, recursive_inlining,
23325         add_new_edges_to_heap, speculation_useful_p,
23326         inline_small_functions,
23327         inline_small_functions, flatten_function,
23328         inline_to_all_callers_1): Update.
23330 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23332         * profile-count.c (profile_count::combine_with_ipa_count): Handle
23333         zeros correctly.
23335 2018-01-30  Richard Biener  <rguenther@suse.de>
23337         PR tree-optimization/83008
23338         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
23339         invariant and constant vector uses in stmts when they need
23340         more than one stmt.
23342 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23344         PR bootstrap/84017
23345         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
23346         * configure: Regenerate.
23348 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23350         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
23351         pattern.
23352         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
23353         Use gen_rtx_REG rather than gen_lowpart.
23355 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23357         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
23358         rather than 0 when creating partial subregs.
23360 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23362         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
23363         of usage.
23365 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
23367         PR target/81550
23368         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
23369         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
23370         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
23371         flags.  This restores the settings used before the 2017-07-24.
23372         Turning off pre increment/decrement/modify allows IVOPTS to
23373         optimize DF/SF loops where the index is an int.
23375 2018-01-29  Richard Biener  <rguenther@suse.de>
23376             Kelvin Nilsen  <kelvin@gcc.gnu.org>
23378         PR bootstrap/80867
23379         * tree-vect-stmts.c (vectorizable_call): Don't call
23380         targetm.vectorize_builtin_md_vectorized_function if callee is
23381         NULL.
23383 2018-01-22  Carl Love  <cel@us.ibm.com>
23385         * doc/extend.tex: Fix typo in second arg in
23386         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
23388 2018-01-29  Richard Biener  <rguenther@suse.de>
23390         PR tree-optimization/84086
23391         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
23392         (flush_ssaname_freelist): When SSA names were released reset
23393         the SCEV hash table.
23395 2018-01-29  Richard Biener  <rguenther@suse.de>
23397         PR tree-optimization/84057
23398         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
23399         removed paths when removing edges.
23401 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
23403         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
23404         -mfunction-return=@var{choice}.
23406 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
23408         PR diagnostic/84034
23409         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
23410         Handle CR like TAB.
23411         (layout::print_source_line): Likewise.
23412         (test_get_line_width_without_trailing_whitespace): Add test cases.
23414 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
23416         PR middle-end/84040
23417         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
23418         debug insns.
23420 2018-01-26  Jim Wilson  <jimw@sifive.com>
23422         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
23424         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
23425         specified.
23427 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23429         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
23430         and CMP + SUB-immediate -> SUBS.
23432 2018-01-26  Martin Sebor  <msebor@redhat.com>
23434         PR tree-optimization/83896
23435         * tree-ssa-strlen.c (get_string_len): Rename...
23436         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
23437         Avoid assuming length is constant.
23438         (handle_char_store): Use HOST_WIDE_INT for string length.
23440 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
23442         PR target/81763
23443         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
23444         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
23446 2018-01-26  Richard Biener  <rguenther@suse.de>
23448         PR rtl-optimization/84003
23449         * dse.c (record_store): Only record redundant stores when
23450         the earlier store aliases at least all accesses the later one does.
23452 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
23454         PR rtl-optimization/83985
23455         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
23456         REG_CFA_RESTORE insns.
23457         (delete_unmarked_insns): Don't ignore separate shrink wrapping
23458         REG_CFA_RESTORE insns here.
23460         PR c/83989
23461         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
23462         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
23464 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23466         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
23467         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
23468         (arc_init): Likewise.
23469         (arc_override_options): Likewise.
23470         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
23471         value.
23472         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
23473         support.
23474         * config/arc/arc.h (TARGET_DBNZ): Define.
23475         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
23476         properly set the tune attribute.
23477         (dbnz): Use TARGET_DBNZ guard.
23478         * config/arc/arc.opt (mtune): Add core3 option.
23480 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23482         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
23483         recognize new pic like addresses.
23484         (arc_delegitimize_address): Clean up.
23486 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23488         * config/arc/arc-arches.def: Option mrf16 valid for all
23489         architectures.
23490         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
23491         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
23492         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
23493         * config/arc/arc-tables.opt: Regenerate.
23494         * config/arc/arc.c (arc_conditional_register_usage): Handle
23495         reduced register file case.
23496         (arc_file_start): Set must have build attributes.
23497         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
23498         mrf16 option value.
23499         * config/arc/arc.opt (mrf16): Add new option.
23500         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
23501         * config/arc/genmultilib.awk: Handle new mrf16 option.
23502         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
23503         * config/arc/t-multilib: Regenerate.
23504         * doc/invoke.texi (ARC Options): Document mrf16 option.
23506 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23508         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
23509         * config/arc/arc.c (arc_handle_secure_attribute): New function.
23510         (arc_attribute_table): Add 'secure_call' attribute.
23511         (arc_print_operand): Print secure call operand.
23512         (arc_function_ok_for_sibcall): Don't optimize tail calls when
23513         secure.
23514         (arc_is_secure_call_p): New function.  * config/arc/arc.md
23515         (call_i): Add support for sjli instruction.
23516         (call_value_i): Likewise.
23517         * config/arc/constraints.md (Csc): New constraint.
23519 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23520             John Eric Martin  <John.Martin@emmicro-us.com>
23522         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
23523         * config/arc/arc.c (_arc_jli_section): New struct.
23524         (arc_jli_section): New type.
23525         (rc_jli_sections): New static variable.
23526         (arc_handle_jli_attribute): New function.
23527         (arc_attribute_table): Add jli_always and jli_fixed attribute.
23528         (arc_file_end): New function.
23529         (TARGET_ASM_FILE_END): Define.
23530         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
23531         (arc_add_jli_section): New function.
23532         (jli_call_scan): Likewise.
23533         (arc_reorg): Call jli_call_scan.
23534         (arc_output_addsi): Remove 'S' from printing asm operand.
23535         (arc_is_jli_call_p): New function.
23536         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
23537         operand.
23538         (movhi_insn): Likewise.
23539         (movsi_insn): Likewise.
23540         (movsi_set_cc_insn): Likewise.
23541         (loadqi_update): Likewise.
23542         (load_zeroextendqisi_update): Likewise.
23543         (load_signextendqisi_update): Likewise.
23544         (loadhi_update): Likewise.
23545         (load_zeroextendhisi_update): Likewise.
23546         (load_signextendhisi_update): Likewise.
23547         (loadsi_update): Likewise.
23548         (loadsf_update): Likewise.
23549         (movsicc_insn): Likewise.
23550         (bset_insn): Likewise.
23551         (bxor_insn): Likewise.
23552         (bclr_insn): Likewise.
23553         (bmsk_insn): Likewise.
23554         (bicsi3_insn): Likewise.
23555         (cmpsi_cc_c_insn): Likewise.
23556         (movsi_ne): Likewise.
23557         (movsi_cond_exec): Likewise.
23558         (clrsbsi2): Likewise.
23559         (norm_f): Likewise.
23560         (normw): Likewise.
23561         (swap): Likewise.
23562         (divaw): Likewise.
23563         (flag): Likewise.
23564         (sr): Likewise.
23565         (kflag): Likewise.
23566         (ffs): Likewise.
23567         (ffs_f): Likewise.
23568         (fls): Likewise.
23569         (call_i): Remove 'S' asm letter, add jli instruction.
23570         (call_value_i): Likewise.
23571         * config/arc/arc.op (mjli-always): New option.
23572         * config/arc/constraints.md (Cji): New constraint.
23573         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
23574         operand.
23575         (subsf3_fpx): Likewise.
23576         (mulsf3_fpx): Likewise.
23577         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
23578         asm operand.
23579         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
23580         function attrbutes.
23581         * doc/invoke.texi (ARC): Document mjli-always option.
23583 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
23585         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
23586         avoid addition with 0 and use incw and decw where possible.
23588 2018-01-26  Richard Biener  <rguenther@suse.de>
23590         PR tree-optimization/81082
23591         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
23592         association if it requires casting to unsigned.
23593         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
23594         from fold_plusminus_mult_expr to catch important cases late when
23595         range info is available.
23597 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23599         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
23600         * configure.ac (hidden_linkonce): New test.
23601         * configure: Regenerate.
23602         * config.in: Regenerate.
23604 2018-01-26  Julia Koval  <julia.koval@intel.com>
23606         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
23607         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
23608         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
23609         _mm_mask_bitshuffle_epi64_mask): Fix type.
23610         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
23611         USI_FTYPE_V4DI_V4DI_USI): Remove.
23612         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
23613         __builtin_ia32_vpshufbitqmb256_mask,
23614         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
23615         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
23616         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
23618 2018-01-26  Alan Modra  <amodra@gmail.com>
23620         PR target/84033
23621         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
23622         UNSPEC_VBPERMQ.  Sort other unspecs.
23624 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
23626         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
23628 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
23630         PR middle-end/83055
23631         * predict.c (drop_profile): Do not push/pop cfun; update also
23632         node->count.
23633         (handle_missing_profiles): Fix logic looking for zero profiles.
23635 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
23637         PR middle-end/83977
23638         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
23639         on functions with #pragma omp declare simd or functions with simd
23640         attribute.
23641         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
23642         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
23643         Remove trailing \n from warning_at calls.
23645 2018-01-25  Tom de Vries  <tom@codesourcery.com>
23647         PR target/84028
23648         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23649         for neutered workers.
23651 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
23653         PR target/68467
23654         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
23655         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
23657 2018-01-24  Jeff Law  <law@redhat.com>
23659         PR target/83994
23660         * i386.c (get_probe_interval): Move to earlier point.
23661         (ix86_compute_frame_layout): If -fstack-clash-protection and
23662         the frame is larger than the probe interval, then use pushes
23663         to save registers rather than reg->mem moves.
23664         (ix86_expand_prologue): Remove conditional for int_registers_saved
23665         assertion.
23667 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
23669         PR target/84014
23670         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
23671         min/max for never referenced object.
23673 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
23675         PR middle-end/83977
23676         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
23677         here.
23678         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
23679         attributes from DECL_ATTRIBUTES (decl) without affecting
23680         DECL_ATTRIBUTES (current_function_decl).
23681         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
23682         functions with non-NULL DECL_ABSTRACT_ORIGIN.
23684 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
23686         PR tree-optimization/83979
23687         * fold-const.c (fold_comparison): Use constant_boolean_node
23688         instead of boolean_{true,false}_node.
23690 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
23692         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
23693         with zero counts.
23695 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23697         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23698         Simplify the clause that sets the length attribute.
23699         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23700         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
23701         clause that sets the length attribute.
23702         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23704 2018-01-24  Tom de Vries  <tom@codesourcery.com>
23706         PR target/83589
23707         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
23708         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
23709         Add strict parameter.
23710         (prevent_branch_around_nothing): Insert dummy insn between branch to
23711         label and label with no ptx insn inbetween.
23712         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
23714 2018-01-24  Tom de Vries  <tom@codesourcery.com>
23716         PR target/81352
23717         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23718         for neutered threads in warp.
23719         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
23721 2018-01-24  Richard Biener  <rguenther@suse.de>
23723         PR tree-optimization/83176
23724         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
23725         operands.
23727 2018-01-24  Richard Biener  <rguenther@suse.de>
23729         PR tree-optimization/82819
23730         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
23731         code generating pluses that are no-ops in the target precision.
23733 2018-01-24  Richard Biener  <rguenther@suse.de>
23735         PR middle-end/84000
23736         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
23738 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23740         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
23741         to merge probabilities.
23742         * predict.c (probably_never_executed): Also mark as cold functions
23743         with global 0 profile and guessed local profile.
23744         * profile-count.c (profile_probability::combine_with_count): New
23745         member function.
23746         * profile-count.h (profile_probability::operator*,
23747         profile_probability::operator*=, profile_probability::operator/,
23748         profile_probability::operator/=): Reduce precision to adjusted
23749         and set value to guessed on contradictory divisions.
23750         (profile_probability::combine_with_freq): Remove.
23751         (profile_probability::combine_wiht_count): Declare.
23752         (profile_count::force_nonzero):: Set to adjusted.
23753         (profile_count::probability_in):: Set quality to adjusted.
23754         * tree-ssa-tail-merge.c (replace_block_by): Use
23755         combine_with_count.
23757 2018-01-23  Andrew Waterman  <andrew@sifive.com>
23758             Jim Wilson  <jimw@sifive.com>
23760         * config/riscv/riscv.c (riscv_stack_boundary): New.
23761         (riscv_option_override): Set riscv_stack_boundary.  Handle
23762         riscv_preferred_stack_boundary_arg.
23763         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
23764         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
23765         (STACK_BOUNDARY): Set to riscv_stack_boundary.
23766         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
23767         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
23768         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
23770 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
23772         PR target/83905
23773         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
23774         of struct ix86_frame.
23775         (ix86_expand_epilogue): Likewise.  Add a local variable for
23776         the reg_save_offset field in struct ix86_frame.
23778 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
23780         PR tree-optimization/82604
23781         * tree-loop-distribution.c (enum partition_kind): New enum item
23782         PKIND_PARTIAL_MEMSET.
23783         (partition_builtin_p): Support above new enum item.
23784         (generate_code_for_partition): Ditto.
23785         (compute_access_range): Differentiate cases that equality can be
23786         proven at all loops, the innermost loops or no loops.
23787         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
23788         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
23789         (finalize_partitions, distribute_loop): Don't fuse partition of
23790         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
23791         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
23792         parloop is enabled.
23794 2018-01-23  Martin Liska  <mliska@suse.cz>
23796         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
23797         order to ignore the predictor.
23798         (PRED_POLYMORPHIC_CALL): Likewise.
23799         (PRED_RECURSIVE_CALL): Likewise.
23801 2018-01-23  Martin Liska  <mliska@suse.cz>
23803         * tree-profile.c (tree_profiling): Print function header to
23804         aware reader which function we are working on.
23805         * value-prof.c (gimple_find_values_to_profile): Do not print
23806         not interesting value histograms.
23808 2018-01-23  Martin Liska  <mliska@suse.cz>
23810         * profile-count.h (enum profile_quality): Add
23811         profile_uninitialized as the first value. Do not number values
23812         as they are zero based.
23813         (profile_count::verify): Update sanity check.
23814         (profile_probability::verify): Likewise.
23816 2018-01-23  Nathan Sidwell  <nathan@acm.org>
23818         * doc/invoke.texi (ffor-scope): Deprecate.
23820 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23822         PR tree-optimization/83510
23823         * domwalk.c (set_all_edges_as_executable): New function.
23824         (dom_walker::dom_walker): Convert bool param
23825         "skip_unreachable_blocks" to enum reachability.  Move setup of
23826         edge flags to set_all_edges_as_executable and only do it when
23827         reachability is REACHABLE_BLOCKS.
23828         * domwalk.h (enum dom_walker::reachability): New enum.
23829         (dom_walker::dom_walker): Convert bool param
23830         "skip_unreachable_blocks" to enum reachability.
23831         (set_all_edges_as_executable): New decl.
23832         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
23833         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
23834         "reachability".
23835         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
23836         but converting true to REACHABLE_BLOCKS.
23837         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
23838         * tree-vrp.c
23839         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
23840         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
23841         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
23842         REACHABLE_BLOCKS.
23843         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
23844         if check_all_array_refs will be called.
23846 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23848         * tree.c (selftest::test_location_wrappers): Add more test
23849         coverage.
23851 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23853         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
23854         (selftest::test_bit_in_range): Likewise.
23856 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
23858         PR testsuite/83888
23859         * doc/sourcebuild.texi (vect_float): Say that the selector
23860         only describes the situation when -funsafe-math-optimizations is on.
23861         (vect_float_strict): Document.
23863 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
23865         PR tree-optimization/83965
23866         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
23867         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
23868         instead of checking only for a reduction.
23869         (vect_recog_widen_sum_pattern): Likewise.
23871 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23873         * predict.c (probably_never_executed): Only use precise profile info.
23874         (compute_function_frequency): Skip after inlining hack since we now
23875         have quality checking.
23877 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23879         * profile-count.h (profile_probability::very_unlikely,
23880         profile_probability::unlikely, profile_probability::even): Set
23881         precision to guessed.
23883 2018-01-23  Richard Biener  <rguenther@suse.de>
23885         PR tree-optimization/83963
23886         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
23887         Properly terminate dominator walk when crossing the exit edge not
23888         when visiting its source block.
23890 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
23892         PR c++/83918
23893         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
23894         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
23896 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
23898         PR tree-optimization/83957
23899         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
23900         semicolon after for body surrounded by braces.
23902         PR tree-optimization/83081
23903         * profile-count.h (profile_probability::split): New method.
23904         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
23905         Use profile_probability::split.
23906         (do_compare_rtx_and_jump): Fix adjustment of probabilities
23907         when splitting a single conditional jump into 2.
23909 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
23911         PR tree-optimization/69452
23912         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
23913         decl.
23915 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23917         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
23918         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
23919         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
23921 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23923         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23924         declaration.
23925         * config/rl78/rl78.md (movdi): New define_expand.
23926         * config/rl78/rl78.c (rl78_split_movdi): New function.
23928 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
23930         PR target/83862
23931         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23932         no longer used.
23933         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23934         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23935         128-bit to produce an UNSPEC move to get the double word with the
23936         signbit and then a shift directly to do signbit.
23937         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23938         implementation with a new version that just does either a direct
23939         move or a regular move.  Move memory interface to separate insns.
23940         Move insns so they are next to the expander.
23941         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23942         with signbit move.  Split big and little endian case.
23943         (signbit<mode>2_dm_mem_le): Likewise.
23944         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23945         (signbit<mode>2_dm2): Likewise.
23947 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23949         * config/rl78/rl78.md (anddi3): New define_expand.
23951 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23953         * config/rl78/rl78.md (umindi3): New define_expand.
23955 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23957         * config/rl78/rl78.md (smindi3): New define_expand.
23959 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23961         * config/rl78/rl78.md (smaxdi3): New define_expand.
23963 2018-01-22  Carl Love  <cel@us.ibm.com>
23965         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23966         LVX_V1TI): Add macro expansion.
23967         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23968         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23969         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23970         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23971         Change check to determine if the instruction is a byte reversing
23972         entry.  Fix typo in comment.
23973         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23974         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23975         Add def_builtin calls for new builtins.
23976         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23977         Add define_insn expansion.
23979 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23981         * config/rl78/rl78.md (umaxdi3): New define_expand.
23983 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23985         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23986         for non-QImode registers.
23988 2018-01-22  Richard Biener  <rguenther@suse.de>
23990         PR tree-optimization/83963
23991         * graphite-scop-detection.c (scop_detection::get_sese): Delay
23992         including the loop exit block.
23993         (scop_detection::merge_sese): Likewise.
23994         (scop_detection::add_scop): Do it here instead.
23996 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23998         * doc/sourcebuild.texi (arm_softfloat): Document.
24000 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
24002         PR gcc/77734
24003         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
24004         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
24005         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
24007 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24008             David Edelsohn  <dje.gcc@gmail.com>
24010         PR target/83946
24011         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24012         Change "crset eq" to "crset 2".
24013         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24014         (*call_indirect_aix<mode>_nospec): Likewise.
24015         (*call_value_indirect_aix<mode>_nospec): Likewise.
24016         (*call_indirect_elfv2<mode>_nospec): Likewise.
24017         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
24018         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
24019         change assembly output from . to $.
24020         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24021         (indirect_jump<mode>_nospec): Change assembly output from . to $.
24022         (*tablejump<mode>_internal1_nospec): Likewise.
24024 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
24026         PR target/80870
24027         * config/sh/sh_optimize_sett_clrt.cc:
24028         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
24030 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
24032         PR tree-optimization/83940
24033         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
24034         offset_dt to vect_constant_def rather than vect_unknown_def_type.
24035         (vect_check_load_store_mask): Add a mask_dt_out parameter and
24036         use it to pass back the definition type.
24037         (vect_check_store_rhs): Likewise rhs_dt_out.
24038         (vect_build_gather_load_calls): Add a mask_dt argument and use
24039         it instead of a call to vect_is_simple_use.
24040         (vectorizable_store): Update calls to vect_check_load_store_mask
24041         and vect_check_store_rhs.  Use the dt returned by the latter instead
24042         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
24043         instead of calls to vect_is_simple_use.  Pass the scalar rather
24044         than the vector operand to vect_is_simple_use when handling
24045         second and subsequent copies of an rhs value.
24046         (vectorizable_load): Update calls to vect_check_load_store_mask
24047         and vect_build_gather_load_calls.  Use the cached mask_dt and
24048         gs_info.offset_dt instead of calls to vect_is_simple_use.
24050 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
24052         PR middle-end/83945
24053         * tree-emutls.c: Include gimplify.h.
24054         (lower_emutls_2): New function.
24055         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
24056         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
24057         it before further processing.
24059         PR target/83930
24060         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
24061         UINTVAL (trueop1) instead of INTVAL (op1).
24063 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
24065         PR debug/81570
24066         PR debug/83728
24067         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
24068         INCOMING_FRAME_SP_OFFSET if not defined.
24069         (scan_trace): Add ENTRY argument.  If true and
24070         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
24071         emit a note to adjust the CFA offset.
24072         (create_cfi_notes): Adjust scan_trace callers.
24073         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
24074         INCOMING_FRAME_SP_OFFSET in the CIE.
24075         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
24076         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
24077         Likewise.
24078         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
24079         * doc/tm.texi: Regenerated.
24081 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24083         PR rtl-optimization/83147
24084         * lra-constraints.c (remove_inheritance_pseudos): Use
24085         lra_substitute_pseudo_within_insn.
24087 2018-01-19  Tom de Vries  <tom@codesourcery.com>
24088             Cesar Philippidis  <cesar@codesourcery.com>
24090         PR target/83920
24091         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
24093 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
24095         PR target/83790
24096         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
24097         spaces for function labels.
24099 2018-01-19  Martin Liska  <mliska@suse.cz>
24101         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
24102         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
24103         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
24104         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
24105         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
24106         (PRED_CONST_RETURN): Change from 69 to 65.
24107         (PRED_NULL_RETURN): Change from 91 to 71.
24108         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
24109         (PRED_LOOP_GUARD): Change from 66 to 73.
24111 2018-01-19  Martin Liska  <mliska@suse.cz>
24113         * predict.c (predict_insn_def): Add new assert.
24114         (struct branch_predictor): Change type to signed integer.
24115         (test_prediction_value_range): Amend test to cover
24116         PROB_UNINITIALIZED.
24117         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24118         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24119         (PRED_LOOP_ITERATIONS_MAX): Likewise.
24120         (PRED_LOOP_IV_COMPARE): Likewise.
24121         * predict.h (PROB_UNINITIALIZED): Define new constant.
24123 2018-01-19  Martin Liska  <mliska@suse.cz>
24125         * predict.c (dump_prediction): Add new format for
24126         analyze_brprob.py script which is enabled with -details
24127         suboption.
24128         * profile-count.h (precise_p): New function.
24130 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24132         PR tree-optimization/83922
24133         * tree-vect-loop.c (vect_verify_full_masking): Return false if
24134         there are no statements that need masking.
24135         (vect_active_double_reduction_p): New function.
24136         (vect_analyze_loop_operations): Use it when handling phis that
24137         are not in the loop header.
24139 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24141         PR tree-optimization/83914
24142         * tree-vect-loop.c (vectorizable_induction): Don't convert
24143         init_expr or apply the peeling adjustment for inductions
24144         that are nested within the vectorized loop.
24146 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24148         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24149         instead of NEG.
24151 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24153         PR sanitizer/81715
24154         PR testsuite/83882
24155         * function.h (gimplify_parameters): Add gimple_seq * argument.
24156         * function.c: Include gimple.h and options.h.
24157         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24158         for the added local temporaries if needed.
24159         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24160         if there are any parameter cleanups, wrap whole body into a
24161         try/finally with the cleanups.
24163 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
24165         PR target/82964
24166         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
24167         Use GET_MODE_CLASS for scalar floating point.
24169 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24171         PR ipa/82256
24172         patch by PaX Team
24173         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
24174         Fix call of call_cgraph_insertion_hooks.
24176 2018-01-18  Martin Sebor  <msebor@redhat.com>
24178         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24180 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24182         PR ipa/83619
24183         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24184         frequencies.
24186 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
24188         PR other/70268
24189         * common.opt: (-ffile-prefix-map): New option.
24190         * opts.c (common_handle_option): Defer it.
24191         * opts-global.c (handle_common_deferred_options): Handle it.
24192         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24193         * file-prefix-map.h: New file.
24194         (remap_debug_filename, add_debug_prefix_map): ...here.
24195         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24196         * final.c (debug_prefix_map, add_debug_prefix_map
24197         remap_debug_filename): Move to...
24198         * file-prefix-map.c: New file.
24199         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24200         generalize, get rid of alloca(), use strrchr() instead of strchr().
24201         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24202         Implement in terms of add_prefix_map().
24203         (remap_macro_filename, remap_debug_filename): Implement in term of
24204         remap_filename().
24205         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24206         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24207         * dbxout.c: Include file-prefix-map.h.
24208         * varasm.c: Likewise.
24209         * vmsdbgout.c: Likewise.
24210         * xcoffout.c: Likewise.
24211         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24212         * doc/cppopts.texi (-fmacro-prefix-map): Document.
24213         * doc/invoke.texi (-ffile-prefix-map): Document.
24214         (-fdebug-prefix-map): Update description.
24216 2018-01-18  Martin Liska  <mliska@suse.cz>
24218         * config/i386/i386.c (indirect_thunk_name): Document that also
24219         lfence is emitted.
24220         (output_indirect_thunk): Document why both instructions
24221         (pause and lfence) are generated.
24223 2018-01-18  Richard Biener  <rguenther@suse.de>
24225         PR tree-optimization/83887
24226         * graphite-scop-detection.c
24227         (scop_detection::get_nearest_dom_with_single_entry): Remove.
24228         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
24229         (scop_detection::merge_sese): Re-implement with a flood-fill
24230         algorithm that properly finds a SESE region if it exists.
24232 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24234         PR c/61240
24235         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
24236         pointer_diff optimizations use view_convert instead of convert.
24238 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24240         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24241         Generate different code for -mno-speculate-indirect-jumps.
24242         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24243         (*call_indirect_aix<mode>): Disable for
24244         -mno-speculate-indirect-jumps.
24245         (*call_indirect_aix<mode>_nospec): New define_insn.
24246         (*call_value_indirect_aix<mode>): Disable for
24247         -mno-speculate-indirect-jumps.
24248         (*call_value_indirect_aix<mode>_nospec): New define_insn.
24249         (*sibcall_nonlocal_sysv<mode>): Generate different code for
24250         -mno-speculate-indirect-jumps.
24251         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24253 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
24255         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
24256         long double type, set the flags for noting the default long double
24257         type, even if we don't pass or return a long double type.
24259 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
24261         PR ipa/83051
24262         * ipa-inline.c (flatten_function): Do not overwrite final inlining
24263         failure.
24265 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
24267         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
24268         support for merge[hl].
24269         (fold_mergehl_helper): New helper function.
24270         (tree-vector-builder.h): New #include for tree_vector_builder usage.
24271         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
24272         (altivec_vmrglw_direct): Add xxmrglw insn.
24274 2018-01-17  Andrew Waterman  <andrew@sifive.com>
24276         * config/riscv/riscv.c (riscv_conditional_register_usage): If
24277         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
24279 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
24281         PR lto/83121
24282         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
24283         call the lto_location_cache before reading the
24284         DECL_SOURCE_LOCATION of the types.
24286 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
24287             Richard Sandiford  <richard.sandiford@linaro.org>
24289         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
24290         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
24291         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
24292         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
24293         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
24294         Add declaration.
24295         * config/aarch64/constraints.md (aarch64_movti_operand):
24296         Limit immediates.
24297         * config/aarch64/predicates.md (Uti): Add new constraint.
24299 2018-01-17  Carl Love  <cel@us.ibm.com>
24301         * config/rs6000/vsx.md (define_expand xl_len_r,
24302         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
24303         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
24304         lxvll.
24305         (define_expand, define_insn): Move the shift left from  the
24306         define_insn to the define_expand for lxvl and stxvl instructions.
24307         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
24308         and XL_LEN_R definitions to PURE.
24310 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
24312         * config/i386/i386.c (indirect_thunk_name): Declare regno
24313         as unsigned int.  Compare regno with INVALID_REGNUM.
24314         (output_indirect_thunk): Ditto.
24315         (output_indirect_thunk_function): Ditto.
24316         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
24317         in the call to output_indirect_thunk_function.
24319 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
24321         PR middle-end/83884
24322         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
24323         rather than the size of inner_type to determine the stack slot size
24324         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
24326 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
24328         PR target/83546
24329         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
24330         to PTA_SILVERMONT.
24332 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
24334         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
24335         endian Linux systems to optionally enable multilibs for selecting
24336         the long double type if the user configured an explicit type.
24337         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
24338         have no long double multilibs if not defined.
24339         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
24340         warn if the user used -mabi={ieee,ibm}longdouble and we built
24341         multilibs for long double.
24342         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
24343         appropriate multilib option.
24344         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
24345         multilib options.
24346         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
24347         for building long double multilibs.
24348         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
24350 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
24352         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
24353         copies.
24355         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
24356         64 bits.
24357         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
24358         128 bits.
24360         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
24361         variables.
24363         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
24364         return value.
24366 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24368         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
24369         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
24371 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24373         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
24374         different rtl trees depending on TARGET_64BIT.
24375         (rs6000_gen_lvx): Likewise.
24377 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24379         * config/visium/visium.md (nop): Tweak comment.
24380         (hazard_nop): Likewise.
24382 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24384         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
24385         -mspeculate-indirect-jumps.
24386         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
24387         for -mno-speculate-indirect-jumps.
24388         (*call_indirect_elfv2<mode>_nospec): New define_insn.
24389         (*call_value_indirect_elfv2<mode>): Disable for
24390         -mno-speculate-indirect-jumps.
24391         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
24392         (indirect_jump): Emit different RTL for
24393         -mno-speculate-indirect-jumps.
24394         (*indirect_jump<mode>): Disable for
24395         -mno-speculate-indirect-jumps.
24396         (*indirect_jump<mode>_nospec): New define_insn.
24397         (tablejump): Emit different RTL for
24398         -mno-speculate-indirect-jumps.
24399         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
24400         (tablejumpsi_nospec): New define_expand.
24401         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
24402         (tablejumpdi_nospec): New define_expand.
24403         (*tablejump<mode>_internal1): Disable for
24404         -mno-speculate-indirect-jumps.
24405         (*tablejump<mode>_internal1_nospec): New define_insn.
24406         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
24407         option.
24409 2018-01-16  Artyom Skrobov tyomitch@gmail.com
24411         * caller-save.c (insert_save): Drop unnecessary parameter.  All
24412         callers updated.
24414 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24415             Richard Biener  <rguenth@suse.de>
24417         PR libgomp/83590
24418         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
24419         return early, inline manually is_gimple_sizepos.  Make sure if we
24420         call gimplify_expr we don't end up with a gimple constant.
24421         * tree.c (variably_modified_type_p): Don't return true for
24422         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
24423         * gimplify.h (is_gimple_sizepos): Remove.
24425 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24427         PR tree-optimization/83857
24428         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
24429         vectorizable_live_operation for pure SLP statements.
24430         (vectorizable_live_operation): Handle PHIs.
24432 2018-01-16  Richard Biener  <rguenther@suse.de>
24434         PR tree-optimization/83867
24435         * tree-vect-stmts.c (vect_transform_stmt): Precompute
24436         nested_in_vect_loop_p since the scalar stmt may get invalidated.
24438 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24440         PR c/83844
24441         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
24442         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
24443         If off is not INTEGER_CST, issue a may not be aligned warning
24444         rather than isn't aligned.  Use isn%'t rather than isn't.
24445         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
24446         into MULT_EXPR.
24447         <case MULT_EXPR>: Improve the case when bottom and one of the
24448         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
24449         operand, in that case check if the other operand is multiple of
24450         bottom divided by the INTEGER_CST operand.
24452 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24454         PR target/83858
24455         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
24456         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
24457         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
24458         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
24459         * config/pa/pa.c (pa_function_arg_advance): Likewise.
24460         (pa_function_arg, pa_arg_partial_bytes): Likewise.
24461         (pa_function_arg_size): New function.
24463 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24465         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
24466         in a separate statement.
24468 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24470         PR tree-optimization/83847
24471         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
24472         group gathers and scatters.
24474 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24476         PR rtl-optimization/86620
24477         * params.def (max-sched-ready-insns): Bump minimum value to 1.
24479         PR rtl-optimization/83213
24480         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
24481         to last if both are JUMP_INSNs.
24483         PR tree-optimization/83843
24484         * gimple-ssa-store-merging.c
24485         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
24486         store_immediate_info for bswap/nop orig_stores.
24488 2018-01-15  Andrew Waterman  <andrew@sifive.com>
24490         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
24491         !TARGET_MUL.
24492         <UDIV>: Increase cost if !TARGET_DIV.
24494 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
24496         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
24497         (define_attr "cr_logical_3op"): New.
24498         (cceq_ior_compare): Adjust.
24499         (cceq_ior_compare_complement): Adjust.
24500         (*cceq_rev_compare): Adjust.
24501         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
24502         (is_cracked_insn): Adjust.
24503         (insn_must_be_first_in_group): Adjust.
24504         * config/rs6000/40x.md: Adjust.
24505         * config/rs6000/440.md: Adjust.
24506         * config/rs6000/476.md: Adjust.
24507         * config/rs6000/601.md: Adjust.
24508         * config/rs6000/603.md: Adjust.
24509         * config/rs6000/6xx.md: Adjust.
24510         * config/rs6000/7450.md: Adjust.
24511         * config/rs6000/7xx.md: Adjust.
24512         * config/rs6000/8540.md: Adjust.
24513         * config/rs6000/cell.md: Adjust.
24514         * config/rs6000/e300c2c3.md: Adjust.
24515         * config/rs6000/e500mc.md: Adjust.
24516         * config/rs6000/e500mc64.md: Adjust.
24517         * config/rs6000/e5500.md: Adjust.
24518         * config/rs6000/e6500.md: Adjust.
24519         * config/rs6000/mpc.md: Adjust.
24520         * config/rs6000/power4.md: Adjust.
24521         * config/rs6000/power5.md: Adjust.
24522         * config/rs6000/power6.md: Adjust.
24523         * config/rs6000/power7.md: Adjust.
24524         * config/rs6000/power8.md: Adjust.
24525         * config/rs6000/power9.md: Adjust.
24526         * config/rs6000/rs64.md: Adjust.
24527         * config/rs6000/titan.md: Adjust.
24529 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24531         * config/i386/predicates.md (indirect_branch_operand): Rewrite
24532         ix86_indirect_branch_register logic.
24534 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24536         * config/i386/constraints.md (Bs): Update
24537         ix86_indirect_branch_register check.  Don't check
24538         ix86_indirect_branch_register with GOT_memory_operand.
24539         (Bw): Likewise.
24540         * config/i386/predicates.md (GOT_memory_operand): Don't check
24541         ix86_indirect_branch_register here.
24542         (GOT32_symbol_operand): Likewise.
24544 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24546         * config/i386/predicates.md (constant_call_address_operand):
24547         Rewrite ix86_indirect_branch_register logic.
24548         (sibcall_insn_operand): Likewise.
24550 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24552         * config/i386/constraints.md (Bs): Replace
24553         ix86_indirect_branch_thunk_register with
24554         ix86_indirect_branch_register.
24555         (Bw): Likewise.
24556         * config/i386/i386.md (indirect_jump): Likewise.
24557         (tablejump): Likewise.
24558         (*sibcall_memory): Likewise.
24559         (*sibcall_value_memory): Likewise.
24560         Peepholes of indirect call and jump via memory: Likewise.
24561         * config/i386/i386.opt: Likewise.
24562         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24563         (GOT_memory_operand): Likewise.
24564         (call_insn_operand): Likewise.
24565         (sibcall_insn_operand): Likewise.
24566         (GOT32_symbol_operand): Likewise.
24568 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
24570         PR middle-end/83837
24571         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
24572         type rather than type addr's type points to.
24573         (expand_omp_atomic_mutex): Likewise.
24574         (expand_omp_atomic): Likewise.
24576 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24578         PR target/83839
24579         * config/i386/i386.c (output_indirect_thunk_function): Use
24580         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
24581         for  __x86_return_thunk.
24583 2018-01-15  Richard Biener  <rguenther@suse.de>
24585         PR middle-end/83850
24586         * expmed.c (extract_bit_field_1): Fix typo.
24588 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24590         PR target/83687
24591         * config/arm/iterators.md (VF): New mode iterator.
24592         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
24593         Remove integer-related logic from pattern.
24594         (neon_vabd<mode>_3): Likewise.
24596 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
24598         PR middle-end/82694
24599         * common.opt (fstrict-overflow): No longer an alias.
24600         (fwrapv-pointer): New option.
24601         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
24602         also for pointer types based on flag_wrapv_pointer.
24603         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
24604         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
24605         opts->x_flag_wrapv got set.
24606         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
24607         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
24608         POINTER_TYPE_OVERFLOW_UNDEFINED.
24609         * match.pd: Likewise in address comparison pattern.
24610         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
24612 2018-01-15  Richard Biener  <rguenther@suse.de>
24614         PR lto/83804
24615         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
24616         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
24617         Reset type names to their identifier if their TYPE_DECL doesn't
24618         have linkage (and thus is used for ODR and devirt).
24619         (save_debug_info_for_decl): Remove.
24620         (save_debug_info_for_type): Likewise.
24621         (add_tree_to_fld_list): Adjust.
24622         * tree-pretty-print.c (dump_generic_node): Make dumping of
24623         type names more robust.
24625 2018-01-15  Richard Biener  <rguenther@suse.de>
24627         * BASE-VER: Bump to 8.0.1.
24629 2018-01-14  Martin Sebor  <msebor@redhat.com>
24631         PR other/83508
24632         * builtins.c (check_access): Avoid warning when the no-warning bit
24633         is set.
24635 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
24637         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
24638         * ira-color (allocno_hard_regs_compare): Likewise.
24640 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
24642         PR target/83013
24643         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
24644         Use .pushsection/.popsection.
24646 2018-01-14  Martin Sebor  <msebor@redhat.com>
24648         PR c++/81327
24649         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
24651 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
24653         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
24654         entry from extra_headers.
24655         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
24656         extra_headers, make the list bitwise identical to the i?86-*-* one.
24658 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24660         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
24661         -mcmodel=large with -mindirect-branch=thunk,
24662         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
24663         -mfunction-return=thunk-extern.
24664         * doc/invoke.texi: Document -mcmodel=large is incompatible with
24665         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
24666         -mfunction-return=thunk and -mfunction-return=thunk-extern.
24668 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24670         * config/i386/i386.c (print_reg): Print the name of the full
24671         integer register without '%'.
24672         (ix86_print_operand): Handle 'V'.
24673         * doc/extend.texi: Document 'V' modifier.
24675 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24677         * config/i386/constraints.md (Bs): Disallow memory operand for
24678         -mindirect-branch-register.
24679         (Bw): Likewise.
24680         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24681         (GOT_memory_operand): Likewise.
24682         (call_insn_operand): Likewise.
24683         (sibcall_insn_operand): Likewise.
24684         (GOT32_symbol_operand): Likewise.
24685         * config/i386/i386.md (indirect_jump): Call convert_memory_address
24686         for -mindirect-branch-register.
24687         (tablejump): Likewise.
24688         (*sibcall_memory): Likewise.
24689         (*sibcall_value_memory): Likewise.
24690         Disallow peepholes of indirect call and jump via memory for
24691         -mindirect-branch-register.
24692         (*call_pop): Replace m with Bw.
24693         (*call_value_pop): Likewise.
24694         (*sibcall_pop_memory): Replace m with Bs.
24695         * config/i386/i386.opt (mindirect-branch-register): New option.
24696         * doc/invoke.texi: Document -mindirect-branch-register option.
24698 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24700         * config/i386/i386-protos.h (ix86_output_function_return): New.
24701         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
24702         set function_return_type.
24703         (indirect_thunk_name): Add ret_p to indicate thunk for function
24704         return.
24705         (output_indirect_thunk_function): Pass false to
24706         indirect_thunk_name.
24707         (ix86_output_indirect_branch_via_reg): Likewise.
24708         (ix86_output_indirect_branch_via_push): Likewise.
24709         (output_indirect_thunk_function): Create alias for function
24710         return thunk if regno < 0.
24711         (ix86_output_function_return): New function.
24712         (ix86_handle_fndecl_attribute): Handle function_return.
24713         (ix86_attribute_table): Add function_return.
24714         * config/i386/i386.h (machine_function): Add
24715         function_return_type.
24716         * config/i386/i386.md (simple_return_internal): Use
24717         ix86_output_function_return.
24718         (simple_return_internal_long): Likewise.
24719         * config/i386/i386.opt (mfunction-return=): New option.
24720         (indirect_branch): Mention -mfunction-return=.
24721         * doc/extend.texi: Document function_return function attribute.
24722         * doc/invoke.texi: Document -mfunction-return= option.
24724 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24726         * config/i386/i386-opts.h (indirect_branch): New.
24727         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
24728         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
24729         with local indirect jump when converting indirect call and jump.
24730         (ix86_set_indirect_branch_type): New.
24731         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
24732         (indirectlabelno): New.
24733         (indirect_thunk_needed): Likewise.
24734         (indirect_thunk_bnd_needed): Likewise.
24735         (indirect_thunks_used): Likewise.
24736         (indirect_thunks_bnd_used): Likewise.
24737         (INDIRECT_LABEL): Likewise.
24738         (indirect_thunk_name): Likewise.
24739         (output_indirect_thunk): Likewise.
24740         (output_indirect_thunk_function): Likewise.
24741         (ix86_output_indirect_branch_via_reg): Likewise.
24742         (ix86_output_indirect_branch_via_push): Likewise.
24743         (ix86_output_indirect_branch): Likewise.
24744         (ix86_output_indirect_jmp): Likewise.
24745         (ix86_code_end): Call output_indirect_thunk_function if needed.
24746         (ix86_output_call_insn): Call ix86_output_indirect_branch if
24747         needed.
24748         (ix86_handle_fndecl_attribute): Handle indirect_branch.
24749         (ix86_attribute_table): Add indirect_branch.
24750         * config/i386/i386.h (machine_function): Add indirect_branch_type
24751         and has_local_indirect_jump.
24752         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
24753         to true.
24754         (tablejump): Likewise.
24755         (*indirect_jump): Use ix86_output_indirect_jmp.
24756         (*tablejump_1): Likewise.
24757         (simple_return_indirect_internal): Likewise.
24758         * config/i386/i386.opt (mindirect-branch=): New option.
24759         (indirect_branch): New.
24760         (keep): Likewise.
24761         (thunk): Likewise.
24762         (thunk-inline): Likewise.
24763         (thunk-extern): Likewise.
24764         * doc/extend.texi: Document indirect_branch function attribute.
24765         * doc/invoke.texi: Document -mindirect-branch= option.
24767 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
24769         PR ipa/83051
24770         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
24772 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
24774         * ipa-inline.c (want_inline_small_function_p): Return false if
24775         inlining has already failed with CIF_FINAL_ERROR.
24776         (update_caller_keys): Call want_inline_small_function_p before
24777         can_inline_edge_p.
24778         (update_callee_keys): Likewise.
24780 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24782         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
24783         New function.
24784         (rs6000_quadword_masked_address_p): Likewise.
24785         (quad_aligned_load_p): Likewise.
24786         (quad_aligned_store_p): Likewise.
24787         (const_load_sequence_p): Add comment to describe the outer-most loop.
24788         (mimic_memory_attributes_and_flags): New function.
24789         (rs6000_gen_stvx): Likewise.
24790         (replace_swapped_aligned_store): Likewise.
24791         (rs6000_gen_lvx): Likewise.
24792         (replace_swapped_aligned_load): Likewise.
24793         (replace_swapped_load_constant): Capitalize argument name in
24794         comment describing this function.
24795         (rs6000_analyze_swaps): Add a third pass to search for vector loads
24796         and stores that access quad-word aligned addresses and replace
24797         with stvx or lvx instructions when appropriate.
24798         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
24799         New function prototype.
24800         (rs6000_quadword_masked_address_p): Likewise.
24801         (rs6000_gen_lvx): Likewise.
24802         (rs6000_gen_stvx): Likewise.
24803         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
24804         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
24805         when memory address is aligned.
24806         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
24807         this split to select lvx instruction when memory address is aligned.
24808         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
24809         instruction when memory address is aligned.
24810         (*vsx_le_perm_load_v16qi): Likewise.
24811         (four unnamed splitters): Modify to select the stvx instruction
24812         when memory is aligned.
24814 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
24816         * predict.c (determine_unlikely_bbs): Handle correctly BBs
24817         which appears in the queue multiple times.
24819 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24820             Alan Hayward  <alan.hayward@arm.com>
24821             David Sherwood  <david.sherwood@arm.com>
24823         * tree-vectorizer.h (vec_lower_bound): New structure.
24824         (_loop_vec_info): Add check_nonzero and lower_bounds.
24825         (LOOP_VINFO_CHECK_NONZERO): New macro.
24826         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
24827         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
24828         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
24829         fields.  Make seg_len the distance travelled, not including the
24830         access size.
24831         (dr_direction_indicator): Declare.
24832         (dr_zero_step_indicator): Likewise.
24833         (dr_known_forward_stride_p): Likewise.
24834         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
24835         tree-ssanames.h.
24836         (runtime_alias_check_p): Allow runtime alias checks with
24837         variable strides.
24838         (operator ==): Compare access_size and align.
24839         (prune_runtime_alias_test_list): Rework for new distinction between
24840         the access_size and seg_len.
24841         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
24842         segment lengths.
24843         (get_segment_min_max): New function.
24844         (create_intersect_range_checks): Use it.
24845         (dr_step_indicator): New function.
24846         (dr_direction_indicator): Likewise.
24847         (dr_zero_step_indicator): Likewise.
24848         (dr_known_forward_stride_p): Likewise.
24849         * tree-loop-distribution.c (data_ref_segment_size): Return
24850         DR_STEP * (niters - 1).
24851         (compute_alias_check_pairs): Update call to the dr_with_seg_len
24852         constructor.
24853         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
24854         (vect_preserves_scalar_order_p): New function, split out from...
24855         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
24856         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
24857         (vect_vfa_access_size): New function.
24858         (vect_vfa_align): Likewise.
24859         (vect_compile_time_alias): Take access_size_a and access_b arguments.
24860         (dump_lower_bound): New function.
24861         (vect_check_lower_bound): Likewise.
24862         (vect_small_gap_p): Likewise.
24863         (vectorizable_with_step_bound_p): Likewise.
24864         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
24865         depencies if the vectorization factor is 1.  Convert the checks
24866         for nonzero steps into checks on the bounds of DR_STEP.  Try using
24867         a bunds check for variable steps if the minimum required step is
24868         relatively small. Update calls to the dr_with_seg_len
24869         constructor and to vect_compile_time_alias.
24870         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
24871         function.
24872         (vect_loop_versioning): Call it.
24873         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
24874         when retrying.
24875         (vect_estimate_min_profitable_iters): Account for any bounds checks.
24877 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24878             Alan Hayward  <alan.hayward@arm.com>
24879             David Sherwood  <david.sherwood@arm.com>
24881         * doc/sourcebuild.texi (vect_scatter_store): Document.
24882         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
24883         optabs.
24884         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
24885         Document.
24886         * genopinit.c (main): Add supports_vec_scatter_store and
24887         supports_vec_scatter_store_cached to target_optabs.
24888         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
24889         IFN_MASK_SCATTER_STORE.
24890         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
24891         functions.
24892         * internal-fn.h (internal_store_fn_p): Declare.
24893         (internal_fn_stored_value_index): Likewise.
24894         * internal-fn.c (scatter_store_direct): New macro.
24895         (expand_scatter_store_optab_fn): New function.
24896         (direct_scatter_store_optab_supported_p): New macro.
24897         (internal_store_fn_p): New function.
24898         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
24899         IFN_MASK_SCATTER_STORE.
24900         (internal_fn_mask_index): Likewise.
24901         (internal_fn_stored_value_index): New function.
24902         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
24903         for scatter stores.
24904         * optabs-query.h (supports_vec_scatter_store_p): Declare.
24905         * optabs-query.c (supports_vec_scatter_store_p): New function.
24906         * tree-vectorizer.h (vect_get_store_rhs): Declare.
24907         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
24908         true for scatter stores.
24909         (vect_gather_scatter_fn_p): Handle scatter stores too.
24910         (vect_check_gather_scatter): Consider using scatter stores if
24911         supports_vec_scatter_store_p.
24912         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
24913         scatter stores too.
24914         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24915         internal_fn_stored_value_index.
24916         (check_load_store_masking): Handle scatter stores too.
24917         (vect_get_store_rhs): Make public.
24918         (vectorizable_call): Use internal_store_fn_p.
24919         (vectorizable_store): Handle scatter store internal functions.
24920         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
24921         when deciding whether the end of the group has been reached.
24922         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
24923         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24924         (mask_scatter_store<mode>): New insns.
24926 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24927             Alan Hayward  <alan.hayward@arm.com>
24928             David Sherwood  <david.sherwood@arm.com>
24930         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24931         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24932         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24933         function.
24934         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24935         Use vect_truncate_gather_scatter_offset if we can't treat the
24936         operation as a normal gather load or scatter store.
24937         (get_group_load_store_type): Take the gather_scatter_info
24938         as argument.  Try using a gather load or scatter store for
24939         single-element groups.
24940         (get_load_store_type): Update calls to get_group_load_store_type
24941         and vect_use_strided_gather_scatters_p.
24943 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24944             Alan Hayward  <alan.hayward@arm.com>
24945             David Sherwood  <david.sherwood@arm.com>
24947         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24948         optional tree argument.
24949         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24950         null target hooks.
24951         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24952         but continue to use the current value as a fallback.
24953         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24954         to compare the updates.
24955         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24956         (get_load_store_type): Use it when handling a strided access.
24957         (vect_get_strided_load_store_ops): New function.
24958         (vect_get_data_ptr_increment): Likewise.
24959         (vectorizable_load): Handle strided gather loads.  Always pass
24960         a step to vect_create_data_ref_ptr and bump_vector_ptr.
24962 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24963             Alan Hayward  <alan.hayward@arm.com>
24964             David Sherwood  <david.sherwood@arm.com>
24966         * doc/md.texi (gather_load@var{m}): Document.
24967         (mask_gather_load@var{m}): Likewise.
24968         * genopinit.c (main): Add supports_vec_gather_load and
24969         supports_vec_gather_load_cached to target_optabs.
24970         * optabs-tree.c (init_tree_optimization_optabs): Use
24971         ggc_cleared_alloc to allocate target_optabs.
24972         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24973         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24974         functions.
24975         * internal-fn.h (internal_load_fn_p): Declare.
24976         (internal_gather_scatter_fn_p): Likewise.
24977         (internal_fn_mask_index): Likewise.
24978         (internal_gather_scatter_fn_supported_p): Likewise.
24979         * internal-fn.c (gather_load_direct): New macro.
24980         (expand_gather_load_optab_fn): New function.
24981         (direct_gather_load_optab_supported_p): New macro.
24982         (direct_internal_fn_optab): New function.
24983         (internal_load_fn_p): Likewise.
24984         (internal_gather_scatter_fn_p): Likewise.
24985         (internal_fn_mask_index): Likewise.
24986         (internal_gather_scatter_fn_supported_p): Likewise.
24987         * optabs-query.c (supports_at_least_one_mode_p): New function.
24988         (supports_vec_gather_load_p): Likewise.
24989         * optabs-query.h (supports_vec_gather_load_p): Declare.
24990         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24991         and memory_type field.
24992         (NUM_PATTERNS): Bump to 15.
24993         * tree-vect-data-refs.c: Include internal-fn.h.
24994         (vect_gather_scatter_fn_p): New function.
24995         (vect_describe_gather_scatter_call): Likewise.
24996         (vect_check_gather_scatter): Try using internal functions for
24997         gather loads.  Recognize existing calls to a gather load function.
24998         (vect_analyze_data_refs): Consider using gather loads if
24999         supports_vec_gather_load_p.
25000         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
25001         (vect_get_gather_scatter_offset_type): Likewise.
25002         (vect_convert_mask_for_vectype): Likewise.
25003         (vect_add_conversion_to_patterm): Likewise.
25004         (vect_try_gather_scatter_pattern): Likewise.
25005         (vect_recog_gather_scatter_pattern): New pattern recognizer.
25006         (vect_vect_recog_func_ptrs): Add it.
25007         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25008         internal_fn_mask_index and internal_gather_scatter_fn_p.
25009         (check_load_store_masking): Take the gather_scatter_info as an
25010         argument and handle gather loads.
25011         (vect_get_gather_scatter_ops): New function.
25012         (vectorizable_call): Check internal_load_fn_p.
25013         (vectorizable_load): Likewise.  Handle gather load internal
25014         functions.
25015         (vectorizable_store): Update call to check_load_store_masking.
25016         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
25017         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
25018         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
25019         (aarch64_gather_scale_operand_d): New predicates.
25020         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
25021         (mask_gather_load<mode>): New insns.
25023 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25024             Alan Hayward  <alan.hayward@arm.com>
25025             David Sherwood  <david.sherwood@arm.com>
25027         * optabs.def (fold_left_plus_optab): New optab.
25028         * doc/md.texi (fold_left_plus_@var{m}): Document.
25029         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
25030         * internal-fn.c (fold_left_direct): Define.
25031         (expand_fold_left_optab_fn): Likewise.
25032         (direct_fold_left_optab_supported_p): Likewise.
25033         * fold-const-call.c (fold_const_fold_left): New function.
25034         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
25035         * tree-parloops.c (valid_reduction_p): New function.
25036         (gather_scalar_reductions): Use it.
25037         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
25038         (vect_finish_replace_stmt): Declare.
25039         * tree-vect-loop.c (fold_left_reduction_fn): New function.
25040         (needs_fold_left_reduction_p): New function, split out from...
25041         (vect_is_simple_reduction): ...here.  Accept reductions that
25042         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
25043         (vect_force_simple_reduction): Also store the reduction type in
25044         the assignment's STMT_VINFO_REDUC_TYPE.
25045         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
25046         (merge_with_identity): New function.
25047         (vect_expand_fold_left): Likewise.
25048         (vectorize_fold_left_reduction): Likewise.
25049         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
25050         scalar phi in place for it.  Check for target support and reject
25051         cases that would reassociate the operation.  Defer the transform
25052         phase to vectorize_fold_left_reduction.
25053         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
25054         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
25055         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
25057 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25059         * tree-if-conv.c (predicate_mem_writes): Remove redundant
25060         call to ifc_temp_var.
25062 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25063             Alan Hayward  <alan.hayward@arm.com>
25064             David Sherwood  <david.sherwood@arm.com>
25066         * target.def (legitimize_address_displacement): Take the original
25067         offset as a poly_int.
25068         * targhooks.h (default_legitimize_address_displacement): Update
25069         accordingly.
25070         * targhooks.c (default_legitimize_address_displacement): Likewise.
25071         * doc/tm.texi: Regenerate.
25072         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
25073         as an argument, moving assert of ad->disp == ad->disp_term to...
25074         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
25075         Try calling targetm.legitimize_address_displacement before expanding
25076         the address rather than afterwards, and adjust for the new interface.
25077         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
25078         Match the new hook interface.  Handle SVE addresses.
25079         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
25080         new hook interface.
25082 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25084         * Makefile.in (OBJS): Add early-remat.o.
25085         * target.def (select_early_remat_modes): New hook.
25086         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
25087         * doc/tm.texi: Regenerate.
25088         * targhooks.h (default_select_early_remat_modes): Declare.
25089         * targhooks.c (default_select_early_remat_modes): New function.
25090         * timevar.def (TV_EARLY_REMAT): New timevar.
25091         * passes.def (pass_early_remat): New pass.
25092         * tree-pass.h (make_pass_early_remat): Declare.
25093         * early-remat.c: New file.
25094         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
25095         function.
25096         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
25098 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25099             Alan Hayward  <alan.hayward@arm.com>
25100             David Sherwood  <david.sherwood@arm.com>
25102         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
25103         vfm1 with a bound_epilog parameter.
25104         (vect_do_peeling): Update calls accordingly, and move the prologue
25105         call earlier in the function.  Treat the base bound_epilog as 0 for
25106         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
25107         this base when peeling for gaps.
25108         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
25109         with fully-masked loops.
25110         (vect_estimate_min_profitable_iters): Handle the single peeled
25111         iteration in that case.
25113 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25114             Alan Hayward  <alan.hayward@arm.com>
25115             David Sherwood  <david.sherwood@arm.com>
25117         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25118         single-element interleaving even if the size is not a power of 2.
25119         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25120         accesses for single-element interleaving if the group size is
25121         not a power of 2.
25123 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25124             Alan Hayward  <alan.hayward@arm.com>
25125             David Sherwood  <david.sherwood@arm.com>
25127         * doc/md.texi (fold_extract_last_@var{m}): Document.
25128         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25129         * optabs.def (fold_extract_last_optab): New optab.
25130         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25131         * internal-fn.c (fold_extract_direct): New macro.
25132         (expand_fold_extract_optab_fn): Likewise.
25133         (direct_fold_extract_optab_supported_p): Likewise.
25134         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25135         * tree-vect-loop.c (vect_model_reduction_cost): Handle
25136         EXTRACT_LAST_REDUCTION.
25137         (get_initial_def_for_reduction): Do not create an initial vector
25138         for EXTRACT_LAST_REDUCTION reductions.
25139         (vectorizable_reduction): Leave the scalar phi in place for
25140         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
25141         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
25142         epilogue code for EXTRACT_LAST_REDUCTION and defer the
25143         transform phase to vectorizable_condition.
25144         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25145         split out from...
25146         (vect_finish_stmt_generation): ...here.
25147         (vect_finish_replace_stmt): New function.
25148         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25149         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25150         pattern.
25151         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25153 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25154             Alan Hayward  <alan.hayward@arm.com>
25155             David Sherwood  <david.sherwood@arm.com>
25157         * doc/md.texi (extract_last_@var{m}): Document.
25158         * optabs.def (extract_last_optab): New optab.
25159         * internal-fn.def (EXTRACT_LAST): New internal function.
25160         * internal-fn.c (cond_unary_direct): New macro.
25161         (expand_cond_unary_optab_fn): Likewise.
25162         (direct_cond_unary_optab_supported_p): Likewise.
25163         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25164         loops using EXTRACT_LAST.
25165         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
25166         (extract_last_<mode>): ...this optab.
25167         (vec_extract<mode><Vel>): Update accordingly.
25169 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25170             Alan Hayward  <alan.hayward@arm.com>
25171             David Sherwood  <david.sherwood@arm.com>
25173         * target.def (empty_mask_is_expensive): New hook.
25174         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
25175         * doc/tm.texi: Regenerate.
25176         * targhooks.h (default_empty_mask_is_expensive): Declare.
25177         * targhooks.c (default_empty_mask_is_expensive): New function.
25178         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25179         if the target says that empty masks are expensive.
25180         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25181         New function.
25182         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25184 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25185             Alan Hayward  <alan.hayward@arm.com>
25186             David Sherwood  <david.sherwood@arm.com>
25188         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25189         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25190         (vect_use_loop_mask_for_alignment_p): New function.
25191         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25192         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25193         niters_skip argument.  Make sure that the first niters_skip elements
25194         of the first iteration are inactive.
25195         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25196         Update call to vect_set_loop_masks_directly.
25197         (get_misalign_in_elems): New function, split out from...
25198         (vect_gen_prolog_loop_niters): ...here.
25199         (vect_update_init_of_dr): Take a code argument that specifies whether
25200         the adjustment should be added or subtracted.
25201         (vect_update_init_of_drs): Likewise.
25202         (vect_prepare_for_masked_peels): New function.
25203         (vect_do_peeling): Skip prologue peeling if we're using a mask
25204         instead.  Update call to vect_update_inits_of_drs.
25205         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25206         mask_skip_niters.
25207         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25208         alignment.  Do not include the number of peeled iterations in
25209         the minimum threshold in that case.
25210         (vectorizable_induction): Adjust the start value down by
25211         LOOP_VINFO_MASK_SKIP_NITERS iterations.
25212         (vect_transform_loop): Call vect_prepare_for_masked_peels.
25213         Take the number of skipped iterations into account when calculating
25214         the loop bounds.
25215         * tree-vect-stmts.c (vect_gen_while_not): New function.
25217 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25218             Alan Hayward  <alan.hayward@arm.com>
25219             David Sherwood  <david.sherwood@arm.com>
25221         * doc/sourcebuild.texi (vect_fully_masked): Document.
25222         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
25223         default value to 0.
25224         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
25225         split out from...
25226         (vect_analyze_loop_2): ...here. Don't check the vectorization
25227         factor against the number of loop iterations if the loop is
25228         fully-masked.
25230 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25231             Alan Hayward  <alan.hayward@arm.com>
25232             David Sherwood  <david.sherwood@arm.com>
25234         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
25235         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
25236         (dump_groups): Update accordingly.
25237         (iv_use::mem_type): New member variable.
25238         (address_p): New function.
25239         (record_use): Add a mem_type argument and initialize the new
25240         mem_type field.
25241         (record_group_use): Add a mem_type argument.  Use address_p.
25242         Remove obsolete null checks of base_object.  Update call to record_use.
25243         (find_interesting_uses_op): Update call to record_group_use.
25244         (find_interesting_uses_cond): Likewise.
25245         (find_interesting_uses_address): Likewise.
25246         (get_mem_type_for_internal_fn): New function.
25247         (find_address_like_use): Likewise.
25248         (find_interesting_uses_stmt): Try find_address_like_use before
25249         calling find_interesting_uses_op.
25250         (addr_offset_valid_p): Use the iv mem_type field as the type
25251         of the addressed memory.
25252         (add_autoinc_candidates): Likewise.
25253         (get_address_cost): Likewise.
25254         (split_small_address_groups_p): Use address_p.
25255         (split_address_groups): Likewise.
25256         (add_iv_candidate_for_use): Likewise.
25257         (autoinc_possible_for_pair): Likewise.
25258         (rewrite_groups): Likewise.
25259         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
25260         (determine_group_iv_cost): Update after split of USE_ADDRESS.
25261         (get_alias_ptr_type_for_ptr_address): New function.
25262         (rewrite_use_address): Rewrite address uses in calls that were
25263         identified by find_address_like_use.
25265 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25266             Alan Hayward  <alan.hayward@arm.com>
25267             David Sherwood  <david.sherwood@arm.com>
25269         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
25270         TARGET_MEM_REFs.
25271         * gimple-expr.h (is_gimple_addressable: Likewise.
25272         * gimple-expr.c (is_gimple_address): Likewise.
25273         * internal-fn.c (expand_call_mem_ref): New function.
25274         (expand_mask_load_optab_fn): Use it.
25275         (expand_mask_store_optab_fn): Likewise.
25277 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25278             Alan Hayward  <alan.hayward@arm.com>
25279             David Sherwood  <david.sherwood@arm.com>
25281         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
25282         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
25283         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
25284         (cond_umax@var{mode}): Document.
25285         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
25286         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
25287         (cond_umin_optab, cond_umax_optab): New optabs.
25288         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
25289         (COND_IOR, COND_XOR): New internal functions.
25290         * internal-fn.h (get_conditional_internal_fn): Declare.
25291         * internal-fn.c (cond_binary_direct): New macro.
25292         (expand_cond_binary_optab_fn): Likewise.
25293         (direct_cond_binary_optab_supported_p): Likewise.
25294         (get_conditional_internal_fn): New function.
25295         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
25296         Cope with reduction statements that are vectorized as calls rather
25297         than assignments.
25298         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
25299         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
25300         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
25301         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
25302         (UNSPEC_COND_EOR): New unspecs.
25303         (optab): Add mappings for them.
25304         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
25305         (sve_int_op, sve_fp_op): New int attributes.
25307 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25308             Alan Hayward  <alan.hayward@arm.com>
25309             David Sherwood  <david.sherwood@arm.com>
25311         * optabs.def (while_ult_optab): New optab.
25312         * doc/md.texi (while_ult@var{m}@var{n}): Document.
25313         * internal-fn.def (WHILE_ULT): New internal function.
25314         * internal-fn.h (direct_internal_fn_supported_p): New override
25315         that takes two types as argument.
25316         * internal-fn.c (while_direct): New macro.
25317         (expand_while_optab_fn): New function.
25318         (convert_optab_supported_p): Likewise.
25319         (direct_while_optab_supported_p): New macro.
25320         * wide-int.h (wi::udiv_ceil): New function.
25321         * tree-vectorizer.h (rgroup_masks): New structure.
25322         (vec_loop_masks): New typedef.
25323         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
25324         and fully_masked_p.
25325         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
25326         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
25327         (vect_max_vf): New function.
25328         (slpeel_make_loop_iterate_ntimes): Delete.
25329         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
25330         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
25331         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
25332         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
25333         internal-fn.h, stor-layout.h and optabs-query.h.
25334         (vect_set_loop_mask): New function.
25335         (add_preheader_seq): Likewise.
25336         (add_header_seq): Likewise.
25337         (interleave_supported_p): Likewise.
25338         (vect_maybe_permute_loop_masks): Likewise.
25339         (vect_set_loop_masks_directly): Likewise.
25340         (vect_set_loop_condition_masked): Likewise.
25341         (vect_set_loop_condition_unmasked): New function, split out from
25342         slpeel_make_loop_iterate_ntimes.
25343         (slpeel_make_loop_iterate_ntimes): Rename to..
25344         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
25345         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
25346         (vect_do_peeling): Update call accordingly.
25347         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
25348         loops.
25349         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25350         mask_compare_type, can_fully_mask_p and fully_masked_p.
25351         (release_vec_loop_masks): New function.
25352         (_loop_vec_info): Use it to free the loop masks.
25353         (can_produce_all_loop_masks_p): New function.
25354         (vect_get_max_nscalars_per_iter): Likewise.
25355         (vect_verify_full_masking): Likewise.
25356         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
25357         retries, and free the mask rgroups before retrying.  Check loop-wide
25358         reasons for disallowing fully-masked loops.  Make the final decision
25359         about whether use a fully-masked loop or not.
25360         (vect_estimate_min_profitable_iters): Do not assume that peeling
25361         for the number of iterations will be needed for fully-masked loops.
25362         (vectorizable_reduction): Disable fully-masked loops.
25363         (vectorizable_live_operation): Likewise.
25364         (vect_halve_mask_nunits): New function.
25365         (vect_double_mask_nunits): Likewise.
25366         (vect_record_loop_mask): Likewise.
25367         (vect_get_loop_mask): Likewise.
25368         (vect_transform_loop): Handle the case in which the final loop
25369         iteration might handle a partial vector.  Call vect_set_loop_condition
25370         instead of slpeel_make_loop_iterate_ntimes.
25371         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
25372         (check_load_store_masking): New function.
25373         (prepare_load_store_mask): Likewise.
25374         (vectorizable_store): Handle fully-masked loops.
25375         (vectorizable_load): Likewise.
25376         (supportable_widening_operation): Use vect_halve_mask_nunits for
25377         booleans.
25378         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
25379         (vect_gen_while): New function.
25380         * config/aarch64/aarch64.md (umax<mode>3): New expander.
25381         (aarch64_uqdec<mode>): New insn.
25383 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25384             Alan Hayward  <alan.hayward@arm.com>
25385             David Sherwood  <david.sherwood@arm.com>
25387         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
25388         (reduc_xor_scal_optab): New optabs.
25389         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
25390         (reduc_xor_scal_@var{m}): Document.
25391         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
25392         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
25393         internal functions.
25394         * fold-const-call.c (fold_const_call): Handle them.
25395         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
25396         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
25397         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
25398         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
25399         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
25400         (UNSPEC_XORV): New unspecs.
25401         (optab): Add entries for them.
25402         (BITWISEV): New int iterator.
25403         (bit_reduc_op): New int attributes.
25405 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25406             Alan Hayward  <alan.hayward@arm.com>
25407             David Sherwood  <david.sherwood@arm.com>
25409         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
25410         * internal-fn.def (VEC_SHL_INSERT): New internal function.
25411         * optabs.def (vec_shl_insert_optab): New optab.
25412         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
25413         (duplicate_and_interleave): Likewise.
25414         * tree-vect-loop.c: Include internal-fn.h.
25415         (neutral_op_for_slp_reduction): New function, split out from
25416         get_initial_defs_for_reduction.
25417         (get_initial_def_for_reduction): Handle option 2 for variable-length
25418         vectors by loading the neutral value into a vector and then shifting
25419         the initial value into element 0.
25420         (get_initial_defs_for_reduction): Replace the code argument with
25421         the neutral value calculated by neutral_op_for_slp_reduction.
25422         Use gimple_build_vector for constant-length vectors.
25423         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
25424         but the first group_size elements have a neutral value.
25425         Use duplicate_and_interleave otherwise.
25426         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
25427         Update call to get_initial_defs_for_reduction.  Handle SLP
25428         reductions for variable-length vectors by creating one vector
25429         result for each scalar result, with the elements associated
25430         with other scalar results stubbed out with the neutral value.
25431         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
25432         Require IFN_VEC_SHL_INSERT for double reductions on
25433         variable-length vectors, or SLP reductions that have
25434         a neutral value.  Require can_duplicate_and_interleave_p
25435         support for variable-length unchained SLP reductions if there
25436         is no neutral value, such as for MIN/MAX reductions.  Also require
25437         the number of vector elements to be a multiple of the number of
25438         SLP statements when doing variable-length unchained SLP reductions.
25439         Update call to vect_create_epilog_for_reduction.
25440         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
25441         and remove initial values.
25442         (duplicate_and_interleave): Make public.
25443         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
25444         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
25446 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25447             Alan Hayward  <alan.hayward@arm.com>
25448             David Sherwood  <david.sherwood@arm.com>
25450         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
25451         (can_duplicate_and_interleave_p): New function.
25452         (vect_get_and_check_slp_defs): Take the vector of statements
25453         rather than just the current one.  Remove excess parentheses.
25454         Restriction rejectinon of vect_constant_def and vect_external_def
25455         for variable-length vectors to boolean types, or types for which
25456         can_duplicate_and_interleave_p is false.
25457         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
25458         (duplicate_and_interleave): New function.
25459         (vect_get_constant_vectors): Use gimple_build_vector for
25460         constant-length vectors and suitable variable-length constant
25461         vectors.  Use duplicate_and_interleave for other variable-length
25462         vectors.  Don't defer the update when inserting new statements.
25464 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25465             Alan Hayward  <alan.hayward@arm.com>
25466             David Sherwood  <david.sherwood@arm.com>
25468         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
25469         min_profitable_iters doesn't go negative.
25471 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25472             Alan Hayward  <alan.hayward@arm.com>
25473             David Sherwood  <david.sherwood@arm.com>
25475         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
25476         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
25477         * optabs.def (vec_mask_load_lanes_optab): New optab.
25478         (vec_mask_store_lanes_optab): Likewise.
25479         * internal-fn.def (MASK_LOAD_LANES): New internal function.
25480         (MASK_STORE_LANES): Likewise.
25481         * internal-fn.c (mask_load_lanes_direct): New macro.
25482         (mask_store_lanes_direct): Likewise.
25483         (expand_mask_load_optab_fn): Handle masked operations.
25484         (expand_mask_load_lanes_optab_fn): New macro.
25485         (expand_mask_store_optab_fn): Handle masked operations.
25486         (expand_mask_store_lanes_optab_fn): New macro.
25487         (direct_mask_load_lanes_optab_supported_p): Likewise.
25488         (direct_mask_store_lanes_optab_supported_p): Likewise.
25489         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
25490         parameter.
25491         (vect_load_lanes_supported): Likewise.
25492         * tree-vect-data-refs.c (strip_conversion): New function.
25493         (can_group_stmts_p): Likewise.
25494         (vect_analyze_data_ref_accesses): Use it instead of checking
25495         for a pair of assignments.
25496         (vect_store_lanes_supported): Take a masked_p parameter.
25497         (vect_load_lanes_supported): Likewise.
25498         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
25499         vect_store_lanes_supported and vect_load_lanes_supported.
25500         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
25501         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
25502         parameter.  Don't allow gaps for masked accesses.
25503         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
25504         and vect_load_lanes_supported.
25505         (get_load_store_type): Take a masked_p parameter and update
25506         call to get_group_load_store_type.
25507         (vectorizable_store): Update call to get_load_store_type.
25508         Handle IFN_MASK_STORE_LANES.
25509         (vectorizable_load): Update call to get_load_store_type.
25510         Handle IFN_MASK_LOAD_LANES.
25512 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25513             Alan Hayward  <alan.hayward@arm.com>
25514             David Sherwood  <david.sherwood@arm.com>
25516         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
25517         modes for SVE.
25518         * config/aarch64/aarch64-protos.h
25519         (aarch64_sve_struct_memory_operand_p): Declare.
25520         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
25521         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
25522         (VPRED, vpred): Handle SVE structure modes.
25523         * config/aarch64/constraints.md (Utx): New constraint.
25524         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
25525         (aarch64_sve_struct_nonimmediate_operand): New predicates.
25526         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
25527         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
25528         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
25529         structure modes.  Split into pieces after RA.
25530         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
25531         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
25532         New patterns.
25533         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
25534         SVE structure modes.
25535         (aarch64_classify_address): Likewise.
25536         (sizetochar): Move earlier in file.
25537         (aarch64_print_operand): Handle SVE register lists.
25538         (aarch64_array_mode): New function.
25539         (aarch64_sve_struct_memory_operand_p): Likewise.
25540         (TARGET_ARRAY_MODE): Redefine.
25542 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25543             Alan Hayward  <alan.hayward@arm.com>
25544             David Sherwood  <david.sherwood@arm.com>
25546         * target.def (array_mode): New target hook.
25547         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
25548         * doc/tm.texi: Regenerate.
25549         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
25550         * hooks.c (hook_optmode_mode_uhwi_none): New function.
25551         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
25552         targetm.array_mode.
25553         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
25554         type sizes.
25556 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25557             Alan Hayward  <alan.hayward@arm.com>
25558             David Sherwood  <david.sherwood@arm.com>
25560         * fold-const.c (fold_binary_loc): Check the argument types
25561         rather than the result type when testing for a vector operation.
25563 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25565         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
25566         * doc/tm.texi: Regenerate.
25568 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25569             Alan Hayward  <alan.hayward@arm.com>
25570             David Sherwood  <david.sherwood@arm.com>
25572         * doc/invoke.texi (-msve-vector-bits=): Document new option.
25573         (sve): Document new AArch64 extension.
25574         * doc/md.texi (w): Extend the description of the AArch64
25575         constraint to include SVE vectors.
25576         (Upl, Upa): Document new AArch64 predicate constraints.
25577         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
25578         enum.
25579         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
25580         (msve-vector-bits=): New option.
25581         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
25582         SVE when these are disabled.
25583         (sve): New extension.
25584         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
25585         modes.  Adjust their number of units based on aarch64_sve_vg.
25586         (MAX_BITSIZE_MODE_ANY_MODE): Define.
25587         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
25588         aarch64_addr_query_type.
25589         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
25590         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
25591         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
25592         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
25593         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
25594         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
25595         (aarch64_simd_imm_zero_p): Delete.
25596         (aarch64_check_zero_based_sve_index_immediate): Declare.
25597         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25598         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25599         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25600         (aarch64_sve_float_mul_immediate_p): Likewise.
25601         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25602         rather than an rtx.
25603         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
25604         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
25605         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
25606         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
25607         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
25608         (aarch64_regmode_natural_size): Likewise.
25609         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
25610         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
25611         left one place.
25612         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
25613         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
25614         for VG and the SVE predicate registers.
25615         (V_ALIASES): Add a "z"-prefixed alias.
25616         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
25617         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
25618         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
25619         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
25620         (REG_CLASS_NAMES): Add entries for them.
25621         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
25622         and the predicate registers.
25623         (aarch64_sve_vg): Declare.
25624         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
25625         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
25626         (REGMODE_NATURAL_SIZE): Define.
25627         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
25628         SVE macros.
25629         * config/aarch64/aarch64.c: Include cfgrtl.h.
25630         (simd_immediate_info): Add a constructor for series vectors,
25631         and an associated step field.
25632         (aarch64_sve_vg): New variable.
25633         (aarch64_dbx_register_number): Handle VG and the predicate registers.
25634         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
25635         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
25636         (VEC_ANY_DATA, VEC_STRUCT): New constants.
25637         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
25638         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
25639         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
25640         (aarch64_get_mask_mode): New functions.
25641         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
25642         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
25643         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
25644         predicate modes and predicate registers.  Explicitly restrict
25645         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
25646         to store a vector mode if it is recognized by
25647         aarch64_classify_vector_mode.
25648         (aarch64_regmode_natural_size): New function.
25649         (aarch64_hard_regno_caller_save_mode): Return the original mode
25650         for predicates.
25651         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
25652         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
25653         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
25654         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
25655         functions.
25656         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
25657         does not overlap dest if the function is frame-related.  Handle
25658         SVE constants.
25659         (aarch64_split_add_offset): New function.
25660         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
25661         them aarch64_add_offset.
25662         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
25663         and update call to aarch64_sub_sp.
25664         (aarch64_add_cfa_expression): New function.
25665         (aarch64_expand_prologue): Pass extra temporary registers to the
25666         functions above.  Handle the case in which we need to emit new
25667         DW_CFA_expressions for registers that were originally saved
25668         relative to the stack pointer, but now have to be expressed
25669         relative to the frame pointer.
25670         (aarch64_output_mi_thunk): Pass extra temporary registers to the
25671         functions above.
25672         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
25673         IP0 and IP1 values for SVE frames.
25674         (aarch64_expand_vec_series): New function.
25675         (aarch64_expand_sve_widened_duplicate): Likewise.
25676         (aarch64_expand_sve_const_vector): Likewise.
25677         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
25678         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
25679         into the register, rather than emitting a SET directly.
25680         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
25681         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
25682         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
25683         (offset_9bit_signed_scaled_p): New functions.
25684         (aarch64_replicate_bitmask_imm): New function.
25685         (aarch64_bitmask_imm): Use it.
25686         (aarch64_cannot_force_const_mem): Reject expressions involving
25687         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
25688         (aarch64_classify_index): Handle SVE indices, by requiring
25689         a plain register index with a scale that matches the element size.
25690         (aarch64_classify_address): Handle SVE addresses.  Assert that
25691         the mode of the address is VOIDmode or an integer mode.
25692         Update call to aarch64_classify_symbol.
25693         (aarch64_classify_symbolic_expression): Update call to
25694         aarch64_classify_symbol.
25695         (aarch64_const_vec_all_in_range_p): New function.
25696         (aarch64_print_vector_float_operand): Likewise.
25697         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
25698         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
25699         and the FP immediates 1.0 and 0.5.
25700         (aarch64_print_address_internal): Handle SVE addresses.
25701         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
25702         (aarch64_regno_regclass): Handle predicate registers.
25703         (aarch64_secondary_reload): Handle big-endian reloads of SVE
25704         data modes.
25705         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
25706         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
25707         (aarch64_convert_sve_vector_bits): New function.
25708         (aarch64_override_options): Use it to handle -msve-vector-bits=.
25709         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25710         rather than an rtx.
25711         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
25712         Handle SVE vector and predicate modes.  Accept VL-based constants
25713         that need only one temporary register, and VL offsets that require
25714         no temporary registers.
25715         (aarch64_conditional_register_usage): Mark the predicate registers
25716         as fixed if SVE isn't available.
25717         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
25718         Return true for SVE vector and predicate modes.
25719         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
25720         rather than an unsigned int.  Handle SVE modes.
25721         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
25722         SVE modes.
25723         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
25724         if SVE is enabled.
25725         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25726         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25727         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25728         (aarch64_sve_float_mul_immediate_p): New functions.
25729         (aarch64_sve_valid_immediate): New function.
25730         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
25731         Explicitly reject structure modes.  Check for INDEX constants.
25732         Handle PTRUE and PFALSE constants.
25733         (aarch64_check_zero_based_sve_index_immediate): New function.
25734         (aarch64_simd_imm_zero_p): Delete.
25735         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
25736         vector modes.  Accept constants in the range of CNT[BHWD].
25737         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
25738         ask for an Advanced SIMD mode.
25739         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
25740         (aarch64_simd_vector_alignment): Handle SVE predicates.
25741         (aarch64_vectorize_preferred_vector_alignment): New function.
25742         (aarch64_simd_vector_alignment_reachable): Use it instead of
25743         the vector size.
25744         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
25745         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
25746         functions.
25747         (MAX_VECT_LEN): Delete.
25748         (expand_vec_perm_d): Add a vec_flags field.
25749         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
25750         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
25751         (aarch64_evpc_ext): Don't apply a big-endian lane correction
25752         for SVE modes.
25753         (aarch64_evpc_rev): Rename to...
25754         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
25755         (aarch64_evpc_rev_global): New function.
25756         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
25757         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
25758         MAX_VECT_LEN.
25759         (aarch64_evpc_sve_tbl): New function.
25760         (aarch64_expand_vec_perm_const_1): Update after rename of
25761         aarch64_evpc_rev.  Handle SVE permutes too, trying
25762         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
25763         than aarch64_evpc_tbl.
25764         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
25765         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
25766         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
25767         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
25768         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
25769         (aarch64_expand_sve_vcond): New functions.
25770         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
25771         of aarch64_vector_mode_p.
25772         (aarch64_dwarf_poly_indeterminate_value): New function.
25773         (aarch64_compute_pressure_classes): Likewise.
25774         (aarch64_can_change_mode_class): Likewise.
25775         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
25776         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
25777         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
25778         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
25779         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
25780         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
25781         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
25782         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
25783         constraints.
25784         (Dn, Dl, Dr): Accept const as well as const_vector.
25785         (Dz): Likewise.  Compare against CONST0_RTX.
25786         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
25787         of "vector" where appropriate.
25788         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
25789         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
25790         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
25791         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
25792         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
25793         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
25794         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
25795         (v_int_equiv): Extend to SVE modes.
25796         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
25797         mode attributes.
25798         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
25799         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
25800         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
25801         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
25802         (SVE_COND_FP_CMP): New int iterators.
25803         (perm_hilo): Handle the new unpack unspecs.
25804         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
25805         attributes.
25806         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
25807         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
25808         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
25809         (aarch64_equality_operator, aarch64_constant_vector_operand)
25810         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
25811         (aarch64_sve_nonimmediate_operand): Likewise.
25812         (aarch64_sve_general_operand): Likewise.
25813         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
25814         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
25815         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
25816         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
25817         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
25818         (aarch64_sve_float_arith_immediate): Likewise.
25819         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
25820         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
25821         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
25822         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
25823         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
25824         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
25825         (aarch64_sve_float_arith_operand): Likewise.
25826         (aarch64_sve_float_arith_with_sub_operand): Likewise.
25827         (aarch64_sve_float_mul_operand): Likewise.
25828         (aarch64_sve_vec_perm_operand): Likewise.
25829         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
25830         (aarch64_mov_operand): Accept const_poly_int and const_vector.
25831         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
25832         as well as const_vector.
25833         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
25834         in file.  Use CONST0_RTX and CONSTM1_RTX.
25835         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
25836         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
25837         Use aarch64_simd_imm_zero.
25838         * config/aarch64/aarch64-sve.md: New file.
25839         * config/aarch64/aarch64.md: Include it.
25840         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
25841         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
25842         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
25843         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
25844         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
25845         (sve): New attribute.
25846         (enabled): Disable instructions with the sve attribute unless
25847         TARGET_SVE.
25848         (movqi, movhi): Pass CONST_POLY_INT operaneds through
25849         aarch64_expand_mov_immediate.
25850         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
25851         CNT[BHSD] immediates.
25852         (movti): Split CONST_POLY_INT moves into two halves.
25853         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
25854         Split additions that need a temporary here if the destination
25855         is the stack pointer.
25856         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
25857         (*add<mode>3_poly_1): New instruction.
25858         (set_clobber_cc): New expander.
25860 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25862         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
25863         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
25864         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
25865         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
25866         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
25867         Change innermode from fixed_mode_size to machine_mode.
25868         (simplify_subreg): Update call accordingly.  Handle a constant-sized
25869         subreg of a variable-length CONST_VECTOR.
25871 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25872             Alan Hayward  <alan.hayward@arm.com>
25873             David Sherwood  <david.sherwood@arm.com>
25875         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
25876         (add_offset_to_base): New function, split out from...
25877         (create_mem_ref): ...here.  When handling a scale other than 1,
25878         check first whether the address is valid without the offset.
25879         Add it into the base if so, leaving the index and scale as-is.
25881 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
25883         PR c++/83778
25884         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
25885         fold_for_warn before checking if arg2 is INTEGER_CST.
25887 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
25889         * config/rs6000/predicates.md (load_multiple_operation): Delete.
25890         (store_multiple_operation): Delete.
25891         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
25892         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
25893         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
25894         guarded by TARGET_STRING.
25895         (rs6000_output_load_multiple): Delete.
25896         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
25897         OPTION_MASK_STRING / TARGET_STRING handling.
25898         (print_operand) <'N', 'O'>: Add comment that these are unused now.
25899         (const rs6000_opt_masks) <"string">: Change mask to 0.
25900         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
25901         (MASK_STRING): Delete.
25902         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
25903         parts.  Simplify.
25904         (load_multiple): Delete.
25905         (*ldmsi8): Delete.
25906         (*ldmsi7): Delete.
25907         (*ldmsi6): Delete.
25908         (*ldmsi5): Delete.
25909         (*ldmsi4): Delete.
25910         (*ldmsi3): Delete.
25911         (store_multiple): Delete.
25912         (*stmsi8): Delete.
25913         (*stmsi7): Delete.
25914         (*stmsi6): Delete.
25915         (*stmsi5): Delete.
25916         (*stmsi4): Delete.
25917         (*stmsi3): Delete.
25918         (movmemsi_8reg): Delete.
25919         (corresponding unnamed define_insn): Delete.
25920         (movmemsi_6reg): Delete.
25921         (corresponding unnamed define_insn): Delete.
25922         (movmemsi_4reg): Delete.
25923         (corresponding unnamed define_insn): Delete.
25924         (movmemsi_2reg): Delete.
25925         (corresponding unnamed define_insn): Delete.
25926         (movmemsi_1reg): Delete.
25927         (corresponding unnamed define_insn): Delete.
25928         * config/rs6000/rs6000.opt (mno-string): New.
25929         (mstring): Replace by deprecation warning stub.
25930         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25932 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
25934         * regrename.c (regrename_do_replace): If replacing the same
25935         reg multiple times, try to reuse last created gen_raw_REG.
25937         PR debug/81155
25938         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25939         main to workaround a bug in GDB.
25941 2018-01-12  Tom de Vries  <tom@codesourcery.com>
25943         PR target/83737
25944         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25946 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
25948         PR rtl-optimization/80481
25949         * ira-color.c (get_cap_member): New function.
25950         (allocnos_conflict_by_live_ranges_p): Use it.
25951         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25952         (setup_slot_coalesced_allocno_live_ranges): Ditto.
25954 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
25956         PR target/83628
25957         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25958         (*saddl_se_1): Ditto.
25959         (*ssubsi_1): Ditto.
25960         (*ssubl_se_1): Ditto.
25962 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25964         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25965         rather than wi::to_widest for DR_INITs.
25966         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25967         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25968         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25969         INTEGER_CSTs.
25970         (vect_analyze_group_access_1): Note that here.
25972 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25974         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25975         polynomial type sizes.
25977 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25979         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25980         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25981         (gimple_add_tmp_var): Likewise.
25983 2018-01-12  Martin Liska  <mliska@suse.cz>
25985         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25986         (gimple_alloc_sizes): Likewise.
25987         (dump_gimple_statistics): Use PRIu64 in printf format.
25988         * gimple.h: Change uint64_t to int.
25990 2018-01-12  Martin Liska  <mliska@suse.cz>
25992         * tree-core.h: Use uint64_t instead of int.
25993         * tree.c (tree_node_counts): Likewise.
25994         (tree_node_sizes): Likewise.
25995         (dump_tree_statistics): Use PRIu64 in printf format.
25997 2018-01-12  Martin Liska  <mliska@suse.cz>
25999         * Makefile.in: As qsort_chk is implemented in vec.c, add
26000         vec.o to linkage of gencfn-macros.
26001         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
26002         passing the info to record_node_allocation_statistics.
26003         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
26004         and pass the info.
26005         * ggc-common.c (struct ggc_usage): Add operator== and use
26006         it in operator< and compare function.
26007         * mem-stats.h (struct mem_usage): Likewise.
26008         * vec.c (struct vec_usage): Remove operator< and compare
26009         function. Can be simply inherited.
26011 2018-01-12  Martin Jambor  <mjambor@suse.cz>
26013         PR target/81616
26014         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
26015         * tree-ssa-math-opts.c: Include domwalk.h.
26016         (convert_mult_to_fma_1): New function.
26017         (fma_transformation_info): New type.
26018         (fma_deferring_state): Likewise.
26019         (cancel_fma_deferring): New function.
26020         (result_of_phi): Likewise.
26021         (last_fma_candidate_feeds_initial_phi): Likewise.
26022         (convert_mult_to_fma): Added deferring logic, split actual
26023         transformation to convert_mult_to_fma_1.
26024         (math_opts_dom_walker): New type.
26025         (math_opts_dom_walker::after_dom_children): New method, body moved
26026         here from pass_optimize_widening_mul::execute, added deferring logic
26027         bits.
26028         (pass_optimize_widening_mul::execute): Moved most of code to
26029         math_opts_dom_walker::after_dom_children.
26030         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
26031         * config/i386/i386.c (ix86_option_override_internal): Added
26032         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
26034 2018-01-12  Richard Biener  <rguenther@suse.de>
26036         PR debug/83157
26037         * dwarf2out.c (gen_variable_die): Do not reset old_die for
26038         inline instance vars.
26040 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
26042         PR target/81819
26043         * config/rx/rx.c (rx_is_restricted_memory_address):
26044         Handle SUBREG case.
26046 2018-01-12  Richard Biener  <rguenther@suse.de>
26048         PR tree-optimization/80846
26049         * target.def (split_reduction): New target hook.
26050         * targhooks.c (default_split_reduction): New function.
26051         * targhooks.h (default_split_reduction): Declare.
26052         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
26053         target requests first reduce vectors by combining low and high
26054         parts.
26055         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
26056         (get_vectype_for_scalar_type_and_size): Export.
26057         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
26058         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
26059         * doc/tm.texi: Regenerate.
26060         * config/i386/i386.c (ix86_split_reduction): Implement
26061         TARGET_VECTORIZE_SPLIT_REDUCTION.
26063 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26065         PR target/83368
26066         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
26067         in PIC mode except for TARGET_VXWORKS_RTP.
26068         * config/sparc/sparc.c: Include cfgrtl.h.
26069         (TARGET_INIT_PIC_REG): Define.
26070         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
26071         (sparc_pic_register_p): New predicate.
26072         (sparc_legitimate_address_p): Use it.
26073         (sparc_legitimize_pic_address): Likewise.
26074         (sparc_delegitimize_address): Likewise.
26075         (sparc_mode_dependent_address_p): Likewise.
26076         (gen_load_pcrel_sym): Remove 4th parameter.
26077         (load_got_register): Adjust call to above.  Remove obsolete stuff.
26078         (sparc_expand_prologue): Do not call load_got_register here.
26079         (sparc_flat_expand_prologue): Likewise.
26080         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
26081         (sparc_use_pseudo_pic_reg): New function.
26082         (sparc_init_pic_reg): Likewise.
26083         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26084         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
26086 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
26088         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
26089         Add item for branch_cost.
26091 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26093         PR rtl-optimization/83565
26094         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
26095         not extend the result to a larger mode for rotate operations.
26096         (num_sign_bit_copies1): Likewise.
26098 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26100         PR target/40411
26101         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
26102         -symbolic.
26103         Use values-Xc.o for -pedantic.
26104         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
26106 2018-01-12  Martin Liska  <mliska@suse.cz>
26108         PR ipa/83054
26109         * ipa-devirt.c (final_warning_record::grow_type_warnings):
26110         New function.
26111         (possible_polymorphic_call_targets): Use it.
26112         (ipa_devirt): Likewise.
26114 2018-01-12  Martin Liska  <mliska@suse.cz>
26116         * profile-count.h (enum profile_quality): Use 0 as invalid
26117         enum value of profile_quality.
26119 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26121         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26122         -mext-string options.
26124 2018-01-12  Richard Biener  <rguenther@suse.de>
26126         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26127         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26128         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26129         Likewise.
26130         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26132 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
26134         * configure.ac (--with-long-double-format): Add support for the
26135         configuration option to change the default long double format on
26136         PowerPC systems.
26137         * config.gcc (powerpc*-linux*-*): Likewise.
26138         * configure: Regenerate.
26139         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26140         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26141         used without modification.
26143 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26145         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26146         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26147         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26148         MISC_BUILTIN_SPEC_BARRIER.
26149         (rs6000_init_builtins): Likewise.
26150         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26151         enum value.
26152         (speculation_barrier): New define_insn.
26153         * doc/extend.texi: Document __builtin_speculation_barrier.
26155 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26157         PR target/83203
26158         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26159         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26160         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26161         iterators.
26162         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
26163         integral modes instead of "ss" and "sd".
26164         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
26165         vectors with 32-bit and 64-bit elements.
26166         (vecdupssescalarmodesuffix): New mode attribute.
26167         (vec_dup<mode>): Use it.
26169 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
26171         PR target/83330
26172         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
26173         frame if argument is passed on stack.
26175 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26177         PR target/82682
26178         * ree.c (combine_reaching_defs): Optimize also
26179         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26180         reg2=any_extend(exp); reg1=reg2;, formatting fix.
26182 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26184         PR middle-end/83189
26185         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26187 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26189         PR middle-end/83718
26190         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26191         after they are computed.
26193 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
26195         PR tree-optimization/83695
26196         * gimple-loop-linterchange.cc
26197         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26198         reset cached scev information after interchange.
26199         (pass_linterchange::execute): Remove call to scev_reset_htab.
26201 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26203         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26204         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26205         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26206         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26207         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26208         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26209         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26210         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26211         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26212         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26213         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26214         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26215         (V_lane_reg): Likewise.
26216         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26217         New define_expand.
26218         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26219         (vfmal_lane_low<mode>_intrinsic,
26220         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26221         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
26222         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
26223         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
26224         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
26225         vfmsl_lane_high<mode>_intrinsic): New define_insns.
26227 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26229         * config/arm/arm-cpus.in (fp16fml): New feature.
26230         (ALL_SIMD): Add fp16fml.
26231         (armv8.2-a): Add fp16fml as an option.
26232         (armv8.3-a): Likewise.
26233         (armv8.4-a): Add fp16fml as part of fp16.
26234         * config/arm/arm.h (TARGET_FP16FML): Define.
26235         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
26236         when appropriate.
26237         * config/arm/arm-modes.def (V2HF): Define.
26238         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
26239         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
26240         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
26241         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
26242         vfmsl_low, vfmsl_high): New set of builtins.
26243         * config/arm/iterators.md (PLUSMINUS): New code iterator.
26244         (vfml_op): New code attribute.
26245         (VFMLHALVES): New int iterator.
26246         (VFML, VFMLSEL): New mode attributes.
26247         (V_reg): Define mapping for V2HF.
26248         (V_hi, V_lo): New mode attributes.
26249         (VF_constraint): Likewise.
26250         (vfml_half, vfml_half_selector): New int attributes.
26251         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
26252         define_expand.
26253         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
26254         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
26255         New define_insn.
26256         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
26257         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
26258         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
26259         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
26260         documentation.
26261         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
26262         Document new effective target and option set.
26264 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26266         * config/arm/arm-cpus.in (armv8_4): New feature.
26267         (ARMv8_4a): New fgroup.
26268         (armv8.4-a): New arch.
26269         * config/arm/arm-tables.opt: Regenerate.
26270         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
26271         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
26272         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
26273         Add matching rules for -march=armv8.4-a and extensions.
26274         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
26276 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
26278         PR target/81821
26279         * config/rx/rx.md (BW): New mode attribute.
26280         (sync_lock_test_and_setsi): Add mode suffix to insn output.
26282 2018-01-11  Richard Biener  <rguenther@suse.de>
26284         PR tree-optimization/83435
26285         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
26286         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
26287         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
26289 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26290             Alan Hayward  <alan.hayward@arm.com>
26291             David Sherwood  <david.sherwood@arm.com>
26293         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
26294         field.
26295         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
26296         (aarch64_print_address_internal): Use it to check for a zero offset.
26298 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26299             Alan Hayward  <alan.hayward@arm.com>
26300             David Sherwood  <david.sherwood@arm.com>
26302         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
26303         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
26304         Return a poly_int64 rather than a HOST_WIDE_INT.
26305         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
26306         rather than a HOST_WIDE_INT.
26307         * config/aarch64/aarch64.h (aarch64_frame): Protect with
26308         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
26309         hard_fp_offset, frame_size, initial_adjust, callee_offset and
26310         final_offset from HOST_WIDE_INT to poly_int64.
26311         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
26312         to_constant when getting the number of units in an Advanced SIMD
26313         mode.
26314         (aarch64_builtin_vectorized_function): Check for a constant number
26315         of units.
26316         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
26317         GET_MODE_SIZE.
26318         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
26319         attribute instead of GET_MODE_NUNITS.
26320         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
26321         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
26322         GET_MODE_SIZE for fixed-size registers.
26323         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
26324         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
26325         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
26326         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
26327         (aarch64_print_operand, aarch64_print_address_internal)
26328         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
26329         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
26330         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
26331         Handle polynomial GET_MODE_SIZE.
26332         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
26333         wider than SImode without modification.
26334         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
26335         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
26336         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
26337         passing and returning SVE modes.
26338         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
26339         rather than GEN_INT.
26340         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
26341         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
26342         (aarch64_allocate_and_probe_stack_space): Likewise.
26343         (aarch64_layout_frame): Cope with polynomial offsets.
26344         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
26345         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
26346         polynomial offsets.
26347         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
26348         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
26349         poly_int64 rather than a HOST_WIDE_INT.
26350         (aarch64_get_separate_components, aarch64_process_components)
26351         (aarch64_expand_prologue, aarch64_expand_epilogue)
26352         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
26353         (aarch64_anchor_offset): New function, split out from...
26354         (aarch64_legitimize_address): ...here.
26355         (aarch64_builtin_vectorization_cost): Handle polynomial
26356         TYPE_VECTOR_SUBPARTS.
26357         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
26358         GET_MODE_NUNITS.
26359         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
26360         number of elements from the PARALLEL rather than the mode.
26361         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
26362         rather than GET_MODE_BITSIZE.
26363         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
26364         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
26365         (aarch64_expand_vec_perm_const_1): Handle polynomial
26366         d->perm.length () and d->perm elements.
26367         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
26368         Apply to_constant to d->perm elements.
26369         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
26370         polynomial CONST_VECTOR_NUNITS.
26371         (aarch64_move_pointer): Take amount as a poly_int64 rather
26372         than an int.
26373         (aarch64_progress_pointer): Avoid temporary variable.
26374         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
26375         the mode attribute instead of GET_MODE.
26377 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26378             Alan Hayward  <alan.hayward@arm.com>
26379             David Sherwood  <david.sherwood@arm.com>
26381         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
26382         x exists before using it.
26383         (aarch64_add_constant_internal): Rename to...
26384         (aarch64_add_offset_1): ...this.  Replace regnum with separate
26385         src and dest rtxes.  Handle the case in which they're different,
26386         including when the offset is zero.  Replace scratchreg with an rtx.
26387         Use 2 additions if there is no spare register into which we can
26388         move a 16-bit constant.
26389         (aarch64_add_constant): Delete.
26390         (aarch64_add_offset): Replace reg with separate src and dest
26391         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
26392         Use aarch64_add_offset_1.
26393         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
26394         an rtx rather than an int.  Take the delta as a poly_int64
26395         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
26396         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
26397         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
26398         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
26399         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
26400         and aarch64_add_sp.
26401         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
26402         aarch64_add_constant.
26404 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26406         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
26407         Use scalar_float_mode.
26409 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26411         * config/aarch64/aarch64-simd.md
26412         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
26413         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
26414         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
26415         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
26416         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
26417         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
26418         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
26419         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
26420         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
26421         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
26423 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26425         PR target/83514
26426         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
26427         targ_options->x_arm_arch_string is non NULL.
26429 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
26431         * config/aarch64/aarch64.h
26432         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
26434 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
26436         PR target/82096
26437         * expmed.c (emit_store_flag_force): Swap if const op0
26438         and change VOIDmode to mode of op0.
26440 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26442         PR rtl-optimization/83761
26443         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
26444         than bytes to mode_for_size.
26446 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26448         PR middle-end/83189
26449         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
26450         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
26451         profile.
26453 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26455         PR middle-end/83575
26456         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
26457         when in layout mode.
26458         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
26459         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
26460         partition fixup.
26462 2018-01-10  Michael Collison  <michael.collison@arm.com>
26464         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
26465         * config/aarch64/aarch64-option-extension.def: Add
26466         AARCH64_OPT_EXTENSION of 'fp16fml'.
26467         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26468         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
26469         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
26470         * config/aarch64/constraints.md (Ui7): New constraint.
26471         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
26472         (VFMLA_SEL_W): Ditto.
26473         (f16quad): Ditto.
26474         (f16mac1): Ditto.
26475         (VFMLA16_LOW): New int iterator.
26476         (VFMLA16_HIGH): Ditto.
26477         (UNSPEC_FMLAL): New unspec.
26478         (UNSPEC_FMLSL): Ditto.
26479         (UNSPEC_FMLAL2): Ditto.
26480         (UNSPEC_FMLSL2): Ditto.
26481         (f16mac): New code attribute.
26482         * config/aarch64/aarch64-simd-builtins.def
26483         (aarch64_fmlal_lowv2sf): Ditto.
26484         (aarch64_fmlsl_lowv2sf): Ditto.
26485         (aarch64_fmlalq_lowv4sf): Ditto.
26486         (aarch64_fmlslq_lowv4sf): Ditto.
26487         (aarch64_fmlal_highv2sf): Ditto.
26488         (aarch64_fmlsl_highv2sf): Ditto.
26489         (aarch64_fmlalq_highv4sf): Ditto.
26490         (aarch64_fmlslq_highv4sf): Ditto.
26491         (aarch64_fmlal_lane_lowv2sf): Ditto.
26492         (aarch64_fmlsl_lane_lowv2sf): Ditto.
26493         (aarch64_fmlal_laneq_lowv2sf): Ditto.
26494         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
26495         (aarch64_fmlalq_lane_lowv4sf): Ditto.
26496         (aarch64_fmlsl_lane_lowv4sf): Ditto.
26497         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
26498         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
26499         (aarch64_fmlal_lane_highv2sf): Ditto.
26500         (aarch64_fmlsl_lane_highv2sf): Ditto.
26501         (aarch64_fmlal_laneq_highv2sf): Ditto.
26502         (aarch64_fmlsl_laneq_highv2sf): Ditto.
26503         (aarch64_fmlalq_lane_highv4sf): Ditto.
26504         (aarch64_fmlsl_lane_highv4sf): Ditto.
26505         (aarch64_fmlalq_laneq_highv4sf): Ditto.
26506         (aarch64_fmlsl_laneq_highv4sf): Ditto.
26507         * config/aarch64/aarch64-simd.md:
26508         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
26509         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26510         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
26511         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26512         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
26513         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
26514         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
26515         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
26516         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
26517         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
26518         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
26519         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
26520         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
26521         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
26522         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
26523         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
26524         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
26525         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
26526         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
26527         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
26528         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
26529         (vfmlsl_low_u32): Ditto.
26530         (vfmlalq_low_u32): Ditto.
26531         (vfmlslq_low_u32): Ditto.
26532         (vfmlal_high_u32): Ditto.
26533         (vfmlsl_high_u32): Ditto.
26534         (vfmlalq_high_u32): Ditto.
26535         (vfmlslq_high_u32): Ditto.
26536         (vfmlal_lane_low_u32): Ditto.
26537         (vfmlsl_lane_low_u32): Ditto.
26538         (vfmlal_laneq_low_u32): Ditto.
26539         (vfmlsl_laneq_low_u32): Ditto.
26540         (vfmlalq_lane_low_u32): Ditto.
26541         (vfmlslq_lane_low_u32): Ditto.
26542         (vfmlalq_laneq_low_u32): Ditto.
26543         (vfmlslq_laneq_low_u32): Ditto.
26544         (vfmlal_lane_high_u32): Ditto.
26545         (vfmlsl_lane_high_u32): Ditto.
26546         (vfmlal_laneq_high_u32): Ditto.
26547         (vfmlsl_laneq_high_u32): Ditto.
26548         (vfmlalq_lane_high_u32): Ditto.
26549         (vfmlslq_lane_high_u32): Ditto.
26550         (vfmlalq_laneq_high_u32): Ditto.
26551         (vfmlslq_laneq_high_u32): Ditto.
26552         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
26553         (AARCH64_FL_FOR_ARCH8_4): New.
26554         (AARCH64_ISA_F16FML): New ISA flag.
26555         (TARGET_F16FML): New feature flag for fp16fml.
26556         (doc/invoke.texi): Document new fp16fml option.
26558 2018-01-10  Michael Collison  <michael.collison@arm.com>
26560         * config/aarch64/aarch64-builtins.c:
26561         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
26562         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26563         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
26564         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
26565         (AARCH64_ISA_SHA3): New ISA flag.
26566         (TARGET_SHA3): New feature flag for sha3.
26567         * config/aarch64/iterators.md (sha512_op): New int attribute.
26568         (CRYPTO_SHA512): New int iterator.
26569         (UNSPEC_SHA512H): New unspec.
26570         (UNSPEC_SHA512H2): Ditto.
26571         (UNSPEC_SHA512SU0): Ditto.
26572         (UNSPEC_SHA512SU1): Ditto.
26573         * config/aarch64/aarch64-simd-builtins.def
26574         (aarch64_crypto_sha512hqv2di): New builtin.
26575         (aarch64_crypto_sha512h2qv2di): Ditto.
26576         (aarch64_crypto_sha512su0qv2di): Ditto.
26577         (aarch64_crypto_sha512su1qv2di): Ditto.
26578         (aarch64_eor3qv8hi): Ditto.
26579         (aarch64_rax1qv2di): Ditto.
26580         (aarch64_xarqv2di): Ditto.
26581         (aarch64_bcaxqv8hi): Ditto.
26582         * config/aarch64/aarch64-simd.md:
26583         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
26584         (aarch64_crypto_sha512su0qv2di): Ditto.
26585         (aarch64_crypto_sha512su1qv2di): Ditto.
26586         (aarch64_eor3qv8hi): Ditto.
26587         (aarch64_rax1qv2di): Ditto.
26588         (aarch64_xarqv2di): Ditto.
26589         (aarch64_bcaxqv8hi): Ditto.
26590         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
26591         (vsha512h2q_u64): Ditto.
26592         (vsha512su0q_u64): Ditto.
26593         (vsha512su1q_u64): Ditto.
26594         (veor3q_u16): Ditto.
26595         (vrax1q_u64): Ditto.
26596         (vxarq_u64): Ditto.
26597         (vbcaxq_u16): Ditto.
26598         * config/arm/types.md (crypto_sha512): New type attribute.
26599         (crypto_sha3): Ditto.
26600         (doc/invoke.texi): Document new sha3 option.
26602 2018-01-10  Michael Collison  <michael.collison@arm.com>
26604         * config/aarch64/aarch64-builtins.c:
26605         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
26606         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26607         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
26608         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
26609         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
26610         (AARCH64_ISA_SM4): New ISA flag.
26611         (TARGET_SM4): New feature flag for sm4.
26612         * config/aarch64/aarch64-simd-builtins.def
26613         (aarch64_sm3ss1qv4si): Ditto.
26614         (aarch64_sm3tt1aq4si): Ditto.
26615         (aarch64_sm3tt1bq4si): Ditto.
26616         (aarch64_sm3tt2aq4si): Ditto.
26617         (aarch64_sm3tt2bq4si): Ditto.
26618         (aarch64_sm3partw1qv4si): Ditto.
26619         (aarch64_sm3partw2qv4si): Ditto.
26620         (aarch64_sm4eqv4si): Ditto.
26621         (aarch64_sm4ekeyqv4si): Ditto.
26622         * config/aarch64/aarch64-simd.md:
26623         (aarch64_sm3ss1qv4si): Ditto.
26624         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
26625         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
26626         (aarch64_sm4eqv4si): Ditto.
26627         (aarch64_sm4ekeyqv4si): Ditto.
26628         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
26629         (sm3part_op): Ditto.
26630         (CRYPTO_SM3TT): Ditto.
26631         (CRYPTO_SM3PART): Ditto.
26632         (UNSPEC_SM3SS1): New unspec.
26633         (UNSPEC_SM3TT1A): Ditto.
26634         (UNSPEC_SM3TT1B): Ditto.
26635         (UNSPEC_SM3TT2A): Ditto.
26636         (UNSPEC_SM3TT2B): Ditto.
26637         (UNSPEC_SM3PARTW1): Ditto.
26638         (UNSPEC_SM3PARTW2): Ditto.
26639         (UNSPEC_SM4E): Ditto.
26640         (UNSPEC_SM4EKEY): Ditto.
26641         * config/aarch64/constraints.md (Ui2): New constraint.
26642         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
26643         * config/arm/types.md (crypto_sm3): New type attribute.
26644         (crypto_sm4): Ditto.
26645         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
26646         (vsm3tt1aq_u32): Ditto.
26647         (vsm3tt1bq_u32): Ditto.
26648         (vsm3tt2aq_u32): Ditto.
26649         (vsm3tt2bq_u32): Ditto.
26650         (vsm3partw1q_u32): Ditto.
26651         (vsm3partw2q_u32): Ditto.
26652         (vsm4eq_u32): Ditto.
26653         (vsm4ekeyq_u32): Ditto.
26654         (doc/invoke.texi): Document new sm4 option.
26656 2018-01-10  Michael Collison  <michael.collison@arm.com>
26658         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
26659         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
26660         (AARCH64_FL_FOR_ARCH8_4): New.
26661         (AARCH64_FL_V8_4): New flag.
26662         (doc/invoke.texi): Document new armv8.4-a option.
26664 2018-01-10  Michael Collison  <michael.collison@arm.com>
26666         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26667         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
26668         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
26669         * config/aarch64/aarch64-option-extension.def: Add
26670         AARCH64_OPT_EXTENSION of 'sha2'.
26671         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
26672         (crypto): Disable sha2 and aes if crypto disabled.
26673         (crypto): Enable aes and sha2 if enabled.
26674         (simd): Disable sha2 and aes if simd disabled.
26675         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
26676         New flags.
26677         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
26678         (TARGET_SHA2): New feature flag for sha2.
26679         (TARGET_AES): New feature flag for aes.
26680         * config/aarch64/aarch64-simd.md:
26681         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
26682         conditional on TARGET_AES.
26683         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
26684         (aarch64_crypto_sha1hsi): Make pattern conditional
26685         on TARGET_SHA2.
26686         (aarch64_crypto_sha1hv4si): Ditto.
26687         (aarch64_be_crypto_sha1hv4si): Ditto.
26688         (aarch64_crypto_sha1su1v4si): Ditto.
26689         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
26690         (aarch64_crypto_sha1su0v4si): Ditto.
26691         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
26692         (aarch64_crypto_sha256su0v4si): Ditto.
26693         (aarch64_crypto_sha256su1v4si): Ditto.
26694         (doc/invoke.texi): Document new aes and sha2 options.
26696 2018-01-10  Martin Sebor  <msebor@redhat.com>
26698         PR tree-optimization/83781
26699         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
26700         as string arrays.
26702 2018-01-11  Martin Sebor  <msebor@gmail.com>
26703             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26705         PR tree-optimization/83501
26706         PR tree-optimization/81703
26708         * tree-ssa-strlen.c (get_string_cst): Rename...
26709         (get_string_len): ...to this.  Handle global constants.
26710         (handle_char_store): Adjust.
26712 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
26713             Jim Wilson  <jimw@sifive.com>
26715         * config/riscv/riscv-protos.h (riscv_output_return): New.
26716         * config/riscv/riscv.c (struct machine_function): New naked_p field.
26717         (riscv_attribute_table, riscv_output_return),
26718         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
26719         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
26720         (riscv_compute_frame_info): Only compute frame->mask if not a naked
26721         function.
26722         (riscv_expand_prologue): Add early return for naked function.
26723         (riscv_expand_epilogue): Likewise.
26724         (riscv_function_ok_for_sibcall): Return false for naked function.
26725         (riscv_set_current_function): New.
26726         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
26727         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
26728         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
26729         * doc/extend.texi (RISC-V Function Attributes): New.
26731 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
26733         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
26734         check for 128-bit long double before checking TCmode.
26735         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
26736         128-bit long doubles before checking TFmode or TCmode.
26737         (FLOAT128_IBM_P): Likewise.
26739 2018-01-10  Martin Sebor  <msebor@redhat.com>
26741         PR tree-optimization/83671
26742         * builtins.c (c_strlen): Unconditionally return zero for the empty
26743         string.
26744         Use -Warray-bounds for warnings.
26745         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
26746         for non-constant array indices with COMPONENT_REF, arrays of
26747         arrays, and pointers to arrays.
26748         (gimple_fold_builtin_strlen): Determine and set length range for
26749         non-constant character arrays.
26751 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
26753         PR middle-end/81897
26754         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
26755         empty blocks.
26757 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
26759         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
26761 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
26763         PR target/83399
26764         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
26765         VECTOR_MEM_ALTIVEC_OR_VSX_P.
26766         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
26767         indexed_or_indirect_operand predicate.
26768         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
26769         (*vsx_le_perm_load_v8hi): Likewise.
26770         (*vsx_le_perm_load_v16qi): Likewise.
26771         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
26772         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
26773         (*vsx_le_perm_store_v8hi): Likewise.
26774         (*vsx_le_perm_store_v16qi): Likewise.
26775         (eight unnamed splitters): Likewise.
26777 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
26779         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
26780         * config/rs6000/emmintrin.h: Likewise.
26781         * config/rs6000/mmintrin.h: Likewise.
26782         * config/rs6000/xmmintrin.h: Likewise.
26784 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
26786         PR c++/43486
26787         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
26788         "public_flag".
26789         * tree.c (tree_nop_conversion): Return true for location wrapper
26790         nodes.
26791         (maybe_wrap_with_location): New function.
26792         (selftest::check_strip_nops): New function.
26793         (selftest::test_location_wrappers): New function.
26794         (selftest::tree_c_tests): Call it.
26795         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
26796         (maybe_wrap_with_location): New decl.
26797         (EXPR_LOCATION_WRAPPER_P): New macro.
26798         (location_wrapper_p): New inline function.
26799         (tree_strip_any_location_wrapper): New inline function.
26801 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
26803         PR target/83735
26804         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
26805         stack_realign_offset for the largest alignment of stack slot
26806         actually used.
26807         (ix86_find_max_used_stack_alignment): New function.
26808         (ix86_finalize_stack_frame_flags): Use it.  Set
26809         max_used_stack_alignment if we don't realign stack.
26810         * config/i386/i386.h (machine_function): Add
26811         max_used_stack_alignment.
26813 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
26815         * config/arm/arm.opt (-mbranch-cost): New option.
26816         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
26817         account.
26819 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
26821         PR target/83629
26822         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
26823         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
26825 2018-01-10  Richard Biener  <rguenther@suse.de>
26827         PR debug/83765
26828         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
26829         early out so it also covers the case where we have a non-NULL
26830         origin.
26832 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
26834         PR tree-optimization/83753
26835         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
26836         for non-strided grouped accesses if the number of elements is 1.
26838 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26840         PR target/81616
26841         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
26842         * i386.h (TARGET_USE_GATHER): Define.
26843         * x86-tune.def (X86_TUNE_USE_GATHER): New.
26845 2018-01-10  Martin Liska  <mliska@suse.cz>
26847         PR bootstrap/82831
26848         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
26849         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
26850         partitioning.
26851         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
26852         CLEANUP_NO_PARTITIONING is not set.
26854 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
26856         * doc/rtl.texi: Remove documentation of (const ...) wrappers
26857         for vectors, as a partial revert of r254296.
26858         * rtl.h (const_vec_p): Delete.
26859         (const_vec_duplicate_p): Don't test for vector CONSTs.
26860         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
26861         * expmed.c (make_tree): Likewise.
26863         Revert:
26864         * common.md (E, F): Use CONSTANT_P instead of checking for
26865         CONST_VECTOR.
26866         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
26867         checking for CONST_VECTOR.
26869 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26871         PR middle-end/83575
26872         * predict.c (force_edge_cold): Handle in more sane way edges
26873         with no prediction.
26875 2018-01-09  Carl Love  <cel@us.ibm.com>
26877         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
26878         V4SI, V4SF types.
26879         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
26880         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
26881         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
26882         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
26883         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
26884         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
26885         * config/rs6000/rs6000-protos.h: Add extern defition for
26886         rs6000_generate_float2_double_code.
26887         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
26888         function.
26889         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
26890         (float2_v2df): Add define_expand.
26892 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
26894         PR target/83628
26895         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
26896         op_mode in the force_to_mode call.
26898 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26900         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
26901         instead of checking each element individually.
26902         (aarch64_evpc_uzp): Likewise.
26903         (aarch64_evpc_zip): Likewise.
26904         (aarch64_evpc_ext): Likewise.
26905         (aarch64_evpc_rev): Likewise.
26906         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
26907         instead of checking each element individually.  Return true without
26908         generating rtl if
26909         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
26910         whether all selected elements come from the same input, instead of
26911         checking each element individually.  Remove calls to gen_rtx_REG,
26912         start_sequence and end_sequence and instead assert that no rtl is
26913         generated.
26915 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26917         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
26918         order of HIGH and CONST checks.
26920 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26922         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
26923         if the destination isn't an SSA_NAME.
26925 2018-01-09  Richard Biener  <rguenther@suse.de>
26927         PR tree-optimization/83668
26928         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26929         move prologue...
26930         (canonicalize_loop_form): ... here, renamed from ...
26931         (canonicalize_loop_closed_ssa_form): ... this and amended to
26932         swap successor edges for loop exit blocks to make us use
26933         the RPO order we need for initial schedule generation.
26935 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
26937         PR tree-optimization/64811
26938         * match.pd: When optimizing comparisons with Inf, avoid
26939         introducing or losing exceptions from comparisons with NaN.
26941 2018-01-09  Martin Liska  <mliska@suse.cz>
26943         PR sanitizer/82517
26944         * asan.c (shadow_mem_size): Add gcc_assert.
26946 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
26948         Don't save registers in main().
26950         PR target/83738
26951         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26952         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26953         * config/avr/avr.c (avr_set_current_function): Don't error if
26954         naked, OS_task or OS_main are specified at the same time.
26955         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26956         OS_main.
26957         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26958         attribute.
26959         * common/config/avr/avr-common.c (avr_option_optimization_table):
26960         Switch on -mmain-is-OS_task for optimizing compilations.
26962 2018-01-09  Richard Biener  <rguenther@suse.de>
26964         PR tree-optimization/83572
26965         * graphite.c: Include cfganal.h.
26966         (graphite_transform_loops): Connect infinite loops to exit
26967         and remove fake edges at the end.
26969 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26971         * ipa-inline.c (edge_badness): Revert accidental checkin.
26973 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26975         PR ipa/80763
26976         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26977         symbols; not inline clones.
26979 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
26981         PR target/83507
26982         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26983         hard registers.  Formatting fixes.
26985         PR preprocessor/83722
26986         * gcc.c (try_generate_repro): Pass
26987         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26988         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26989         do_report_bug.
26991 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
26992             Kito Cheng  <kito.cheng@gmail.com>
26994         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26995         (riscv_leaf_function_p): Delete.
26996         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26998 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27000         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
27001         function.
27002         (do_ifelse): New function.
27003         (do_isel): New function.
27004         (do_sub3): New function.
27005         (do_add3): New function.
27006         (do_load_mask_compare): New function.
27007         (do_overlap_load_compare): New function.
27008         (expand_compare_loop): New function.
27009         (expand_block_compare): Call expand_compare_loop() when appropriate.
27010         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
27011         option description.
27012         (-mblock-compare-inline-loop-limit): New option.
27014 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27016         PR target/83677
27017         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
27018         Reverse order of second and third operands in first alternative.
27019         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
27020         of first and second elements in UNSPEC_VPERMR vector.
27021         (altivec_expand_vec_perm_le): Likewise.
27023 2018-01-08  Jeff Law  <law@redhat.com>
27025         PR rtl-optimizatin/81308
27026         * tree-switch-conversion.c (cfg_altered): New file scoped static.
27027         (process_switch): If group_case_labels makes a change, then set
27028         cfg_altered.
27029         (pass_convert_switch::execute): If a switch is converted, then
27030         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
27032         PR rtl-optimization/81308
27033         * recog.c (split_all_insns): Conditionally cleanup the CFG after
27034         splitting insns.
27036 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
27038         PR target/83663 - Revert r255946
27039         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
27040         generation for cases where splatting a value is not useful.
27041         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
27042         across a vec_duplicate and a paradoxical subreg forming a vector
27043         mode to a vec_concat.
27045 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27047         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
27048         -march=armv8.3-a variants.
27049         * config/arm/t-multilib: Likewise.
27050         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
27052 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27054         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
27055         to generate rtl.
27056         (cceq_ior_compare_complement): Give it a name so I can use it, and
27057         change boolean_or_operator predicate to boolean_operator so it can
27058         be used to generate a crand.
27059         (eqne): New code iterator.
27060         (bd/bd_neg): New code_attrs.
27061         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
27062         a single define_insn.
27063         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
27064         decrement (bdnzt/bdnzf/bdzt/bdzf).
27065         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
27066         with the new names of the branch decrement patterns, and added the
27067         names of the branch decrement conditional patterns.
27069 2018-01-08  Richard Biener  <rguenther@suse.de>
27071         PR tree-optimization/83563
27072         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
27073         cache.
27075 2018-01-08  Richard Biener  <rguenther@suse.de>
27077         PR middle-end/83713
27078         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
27080 2018-01-08  Richard Biener  <rguenther@suse.de>
27082         PR tree-optimization/83685
27083         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
27084         references to abnormals.
27086 2018-01-08  Richard Biener  <rguenther@suse.de>
27088         PR lto/83719
27089         * dwarf2out.c (output_indirect_strings): Handle empty
27090         skeleton_debug_str_hash.
27091         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
27093 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27095         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
27096         (emit_store_direct): Likewise.
27097         (arc_trampoline_adjust_address): Likewise.
27098         (arc_asm_trampoline_template): New function.
27099         (arc_initialize_trampoline): Use asm_trampoline_template.
27100         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
27101         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
27102         * config/arc/arc.md (flush_icache): Delete pattern.
27104 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27106         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
27107         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
27108         munaligned-access.
27110 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27112         PR target/83681
27113         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
27114         by not USED_FOR_TARGET.
27115         (make_pass_resolve_sw_modes): Likewise.
27117 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27119         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27120         USED_FOR_TARGET.
27122 2018-01-08  Richard Biener  <rguenther@suse.de>
27124         PR middle-end/83580
27125         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27127 2018-01-08  Richard Biener  <rguenther@suse.de>
27129         PR middle-end/83517
27130         * match.pd ((t * 2) / 2) -> t): Add missing :c.
27132 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
27134         PR middle-end/81897
27135         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27136         basic blocks with a small number of successors.
27137         (convert_control_dep_chain_into_preds): Improve handling of
27138         forwarder blocks.
27139         (dump_predicates): Split apart into...
27140         (dump_pred_chain): ...here...
27141         (dump_pred_info): ...and here.
27142         (can_one_predicate_be_invalidated_p): Add debugging printfs.
27143         (can_chain_union_be_invalidated_p): Improve check for invalidation
27144         of paths.
27145         (uninit_uses_cannot_happen): Avoid unnecessary if
27146         convert_control_dep_chain_into_preds yielded nothing.
27148 2018-01-06  Martin Sebor  <msebor@redhat.com>
27150         PR tree-optimization/83640
27151         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27152         subtracting negative offset from size.
27153         (builtin_access::overlap): Adjust offset bounds of the access to fall
27154         within the size of the object if possible.
27156 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
27158         PR rtl-optimization/83699
27159         * expmed.c (extract_bit_field_1): Restrict the vector usage of
27160         extract_bit_field_as_subreg to cases in which the extracted
27161         value is also a vector.
27163         * lra-constraints.c (process_alt_operands): Test for the equivalence
27164         substitutions when detecting a possible reload cycle.
27166 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
27168         PR debug/83480
27169         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
27170         by default if flag_selective_schedling{,2}.  Formatting fixes.
27172         PR rtl-optimization/83682
27173         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
27174         if it has non-VECTOR_MODE element mode.
27175         (vec_duplicate_p): Likewise.
27177         PR middle-end/83694
27178         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27179         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27181 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27183         PR target/83604
27184         * config/i386/i386-builtin.def
27185         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27186         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27187         Require also OPTION_MASK_ISA_AVX512F in addition to
27188         OPTION_MASK_ISA_GFNI.
27189         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27190         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27191         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27192         to OPTION_MASK_ISA_GFNI.
27193         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27194         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27195         OPTION_MASK_ISA_AVX512BW.
27196         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27197         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27198         addition to OPTION_MASK_ISA_GFNI.
27199         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27200         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27201         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27202         to OPTION_MASK_ISA_GFNI.
27203         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27204         a requirement for all ISAs rather than any of them with a few
27205         exceptions.
27206         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27207         processing.
27208         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27209         bitmasks to be enabled with 3 exceptions, instead of requiring any
27210         enabled ISA with lots of exceptions.
27211         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27212         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27213         Change avx512bw in isa attribute to avx512f.
27214         * config/i386/sgxintrin.h: Add license boilerplate.
27215         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
27216         to __AVX512F__ and __AVX512VL to __AVX512VL__.
27217         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27218         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27219         defined.
27220         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27221         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
27222         temporarily sse2 rather than sse if not enabled already.
27224         PR target/83604
27225         * config/i386/sse.md (VI248_VLBW): Rename to ...
27226         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
27227         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
27228         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
27229         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
27230         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
27231         mode iterator instead of VI248_VLBW.
27233 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
27235         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
27236         (record_modified): Skip clobbers; add debug output.
27237         (param_change_prob): Use sreal frequencies.
27239 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27241         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
27242         punt for user-aligned variables.
27244 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27246         * tree-chrec.c (chrec_contains_symbols): Return true for
27247         POLY_INT_CST.
27249 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
27251         PR target/82439
27252         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
27253         of (x|y) == x for BICS pattern.
27255 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27257         PR tree-optimization/83605
27258         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
27259         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
27260         can throw.
27262 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27264         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
27265         * config/epiphany/rtems.h: New file.
27267 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27268             Uros Bizjak  <ubizjak@gmail.com>
27270         PR target/83554
27271         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
27272         QIreg_operand instead of register_operand predicate.
27273         * config/i386/i386.c (ix86_rop_should_change_byte_p,
27274         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
27275         comments instead of -fmitigate[-_]rop.
27277 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27279         PR bootstrap/81926
27280         * cgraphunit.c (symbol_table::compile): Switch to text_section
27281         before calling assembly_start debug hook.
27282         * run-rtl-passes.c (run_rtl_passes): Likewise.
27283         Include output.h.
27285 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27287         * tree-vrp.c (extract_range_from_binary_expr_1): Check
27288         range_int_cst_p rather than !symbolic_range_p before calling
27289         extract_range_from_multiplicative_op_1.
27291 2018-01-04  Jeff Law  <law@redhat.com>
27293         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
27294         redundant test in assertion.
27296 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27298         * doc/rtl.texi: Document machine_mode wrapper classes.
27300 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27302         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
27303         using tree_to_uhwi.
27305 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27307         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
27308         the VEC_PERM_EXPR fold to fail.
27310 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27312         PR debug/83585
27313         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
27314         to switched_sections.
27316 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27318         PR target/83680
27319         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
27320         test for d.testing.
27322 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
27324         PR target/83387
27325         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
27326         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
27328 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27330         PR debug/83666
27331         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
27332         is BLKmode and bitpos not zero or mode change is needed.
27334 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27336         PR target/83675
27337         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
27338         TARGET_VIS2.
27340 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
27342         PR target/83628
27343         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
27344         instead of MULT rtx.  Update all corresponding splitters.
27345         (*saddl_se): Ditto.
27346         (*ssub<modesuffix>): Ditto.
27347         (*ssubl_se): Ditto.
27348         (*cmp_sadd_di): Update split patterns.
27349         (*cmp_sadd_si): Ditto.
27350         (*cmp_sadd_sidi): Ditto.
27351         (*cmp_ssub_di): Ditto.
27352         (*cmp_ssub_si): Ditto.
27353         (*cmp_ssub_sidi): Ditto.
27354         * config/alpha/predicates.md (const23_operand): New predicate.
27355         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
27356         Look for ASHIFT, not MULT inner operand.
27357         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
27359 2018-01-04  Martin Liska  <mliska@suse.cz>
27361         PR gcov-profile/83669
27362         * gcov.c (output_intermediate_file): Add version to intermediate
27363         gcov file.
27364         * doc/gcov.texi: Document new field 'version' in intermediate
27365         file format. Fix location of '-k' option of gcov command.
27367 2018-01-04  Martin Liska  <mliska@suse.cz>
27369         PR ipa/82352
27370         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
27372 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27374         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
27376 2018-01-03  Martin Sebor  <msebor@redhat.com>
27378         PR tree-optimization/83655
27379         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
27380         checking calls with invalid arguments.
27382 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27384         * tree-vect-stmts.c (vect_get_store_rhs): New function.
27385         (vectorizable_mask_load_store): Delete.
27386         (vectorizable_call): Return false for masked loads and stores.
27387         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
27388         instead of gimple_assign_rhs1.
27389         (vectorizable_load): Handle IFN_MASK_LOAD.
27390         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
27392 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27394         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
27395         split out from..,
27396         (vectorizable_mask_load_store): ...here.
27397         (vectorizable_load): ...and here.
27399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27401         * tree-vect-stmts.c (vect_build_all_ones_mask)
27402         (vect_build_zero_merge_argument): New functions, split out from...
27403         (vectorizable_load): ...here.
27405 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27407         * tree-vect-stmts.c (vect_check_store_rhs): New function,
27408         split out from...
27409         (vectorizable_mask_load_store): ...here.
27410         (vectorizable_store): ...and here.
27412 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27414         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
27415         split out from...
27416         (vectorizable_mask_load_store): ...here.
27418 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27420         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
27421         (vect_model_store_cost): Take a vec_load_store_type instead of a
27422         vect_def_type.
27423         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
27424         (vect_model_store_cost): Take a vec_load_store_type instead of a
27425         vect_def_type.
27426         (vectorizable_mask_load_store): Update accordingly.
27427         (vectorizable_store): Likewise.
27428         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
27430 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27432         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
27433         IFN_MASK_LOAD calls here rather than...
27434         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
27436 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27437             Alan Hayward  <alan.hayward@arm.com>
27438             David Sherwood  <david.sherwood@arm.com>
27440         * expmed.c (extract_bit_field_1): For vector extracts,
27441         fall back to extract_bit_field_as_subreg if vec_extract
27442         isn't available.
27444 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27445             Alan Hayward  <alan.hayward@arm.com>
27446             David Sherwood  <david.sherwood@arm.com>
27448         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
27449         they are variable or constant sized.
27450         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
27451         slots for constant-sized data.
27453 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27454             Alan Hayward  <alan.hayward@arm.com>
27455             David Sherwood  <david.sherwood@arm.com>
27457         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
27458         handling COND_EXPRs with boolean comparisons, try to find a better
27459         basis for the mask type than the boolean itself.
27461 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27463         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
27464         is calculated and how it can be overridden.
27465         * genmodes.c (max_bitsize_mode_any_mode): New variable.
27466         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
27467         if defined.
27468         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
27469         if nonzero.
27471 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27472             Alan Hayward  <alan.hayward@arm.com>
27473             David Sherwood  <david.sherwood@arm.com>
27475         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
27476         Remove the mode argument.
27477         (aarch64_simd_valid_immediate): Remove the mode and inverse
27478         arguments.
27479         * config/aarch64/iterators.md (bitsize): New iterator.
27480         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
27481         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
27482         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
27483         aarch64_simd_valid_immediate.
27484         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
27485         (aarch64_reg_or_bic_imm): Likewise.
27486         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
27487         with an insn_type enum and msl with a modifier_type enum.
27488         Replace element_width with a scalar_mode.  Change the shift
27489         to unsigned int.  Add constructors for scalar_float_mode and
27490         scalar_int_mode elements.
27491         (aarch64_vect_float_const_representable_p): Delete.
27492         (aarch64_can_const_movi_rtx_p)
27493         (aarch64_simd_scalar_immediate_valid_for_move)
27494         (aarch64_simd_make_constant): Update call to
27495         aarch64_simd_valid_immediate.
27496         (aarch64_advsimd_valid_immediate_hs): New function.
27497         (aarch64_advsimd_valid_immediate): Likewise.
27498         (aarch64_simd_valid_immediate): Remove mode and inverse
27499         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
27500         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
27501         and aarch64_float_const_representable_p on the result.
27502         (aarch64_output_simd_mov_immediate): Remove mode argument.
27503         Update call to aarch64_simd_valid_immediate and use of
27504         simd_immediate_info.
27505         (aarch64_output_scalar_simd_mov_immediate): Update call
27506         accordingly.
27508 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27509             Alan Hayward  <alan.hayward@arm.com>
27510             David Sherwood  <david.sherwood@arm.com>
27512         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
27513         (mode_nunits): Likewise CONST_MODE_NUNITS.
27514         * machmode.def (ADJUST_NUNITS): Document.
27515         * genmodes.c (mode_data::need_nunits_adj): New field.
27516         (blank_mode): Update accordingly.
27517         (adj_nunits): New variable.
27518         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
27519         parameter.
27520         (emit_mode_size_inline): Set need_bytesize_adj for all modes
27521         listed in adj_nunits.
27522         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
27523         listed in adj_nunits.  Don't emit case statements for such modes.
27524         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
27525         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
27526         nothing if adj_nunits is nonnull.
27527         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
27528         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
27529         (emit_mode_fbit): Update use of print_maybe_const_decl.
27530         (emit_move_size): Likewise.  Treat the array as non-const
27531         if adj_nunits.
27532         (emit_mode_adjustments): Handle adj_nunits.
27534 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27536         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
27537         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
27538         (VECTOR_MODES): Use it.
27539         (make_vector_modes): Take the prefix as an argument.
27541 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27542             Alan Hayward  <alan.hayward@arm.com>
27543             David Sherwood  <david.sherwood@arm.com>
27545         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
27546         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
27547         for MODE_VECTOR_BOOL.
27548         * machmode.def (VECTOR_BOOL_MODE): Document.
27549         * genmodes.c (VECTOR_BOOL_MODE): New macro.
27550         (make_vector_bool_mode): New function.
27551         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
27552         MODE_VECTOR_BOOL.
27553         * lto-streamer-in.c (lto_input_mode_table): Likewise.
27554         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
27555         Likewise.
27556         * stor-layout.c (int_mode_for_mode): Likewise.
27557         * tree.c (build_vector_type_for_mode): Likewise.
27558         * varasm.c (output_constant_pool_2): Likewise.
27559         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
27560         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
27561         for MODE_VECTOR_BOOL.
27562         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
27563         of mode class checks.
27564         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
27565         instead of a list of mode class checks.
27566         (expand_vector_scalar_condition): Likewise.
27567         (type_for_widest_vector_mode): Handle BImode as an inner mode.
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         * machmode.h (mode_size): Change from unsigned short to
27574         poly_uint16_pod.
27575         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
27576         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27577         or if measurement_type is not polynomial.
27578         (fixed_size_mode::includes_p): Check for constant-sized modes.
27579         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
27580         return a poly_uint16 rather than an unsigned short.
27581         (emit_mode_size): Change the type of mode_size from unsigned short
27582         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
27583         (emit_mode_adjustments): Cope with polynomial vector sizes.
27584         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27585         for GET_MODE_SIZE.
27586         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27587         for GET_MODE_SIZE.
27588         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
27589         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
27590         * caller-save.c (setup_save_areas): Likewise.
27591         (replace_reg_with_saved_mem): Likewise.
27592         * calls.c (emit_library_call_value_1): Likewise.
27593         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
27594         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
27595         (gen_lowpart_for_combine): Likewise.
27596         * convert.c (convert_to_integer_1): Likewise.
27597         * cse.c (equiv_constant, cse_insn): Likewise.
27598         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
27599         (cselib_subst_to_values): Likewise.
27600         * dce.c (word_dce_process_block): Likewise.
27601         * df-problems.c (df_word_lr_mark_ref): Likewise.
27602         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
27603         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
27604         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
27605         (rtl_for_decl_location): Likewise.
27606         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
27607         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
27608         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
27609         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
27610         (expand_expr_real_1): Likewise.
27611         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
27612         (pad_below): Likewise.
27613         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27614         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
27615         * ira.c (get_subreg_tracking_sizes): Likewise.
27616         * ira-build.c (ira_create_allocno_objects): Likewise.
27617         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
27618         (ira_sort_regnos_for_alter_reg): Likewise.
27619         * ira-costs.c (record_operand_costs): Likewise.
27620         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
27621         (resolve_simple_move): Likewise.
27622         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
27623         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
27624         (lra_constraints): Likewise.
27625         (CONST_POOL_OK_P): Reject variable-sized modes.
27626         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
27627         (add_pseudo_to_slot, lra_spill): Likewise.
27628         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27629         * optabs-query.c (get_best_extraction_insn): Likewise.
27630         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27631         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
27632         (expand_mult_highpart, valid_multiword_target_p): Likewise.
27633         * recog.c (offsettable_address_addr_space_p): Likewise.
27634         * regcprop.c (maybe_mode_change): Likewise.
27635         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
27636         * regrename.c (build_def_use): Likewise.
27637         * regstat.c (dump_reg_info): Likewise.
27638         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
27639         (find_reloads, find_reloads_subreg_address): Likewise.
27640         * reload1.c (eliminate_regs_1): Likewise.
27641         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
27642         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
27643         (simplify_binary_operation_1, simplify_subreg): Likewise.
27644         * targhooks.c (default_function_arg_padding): Likewise.
27645         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
27646         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
27647         (verify_gimple_assign_ternary): Likewise.
27648         * tree-inline.c (estimate_move_cost): Likewise.
27649         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27650         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
27651         (get_address_cost_ainc): Likewise.
27652         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
27653         (vect_supportable_dr_alignment): Likewise.
27654         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27655         (vectorizable_reduction): Likewise.
27656         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
27657         (vectorizable_operation, vectorizable_load): Likewise.
27658         * tree.c (build_same_sized_truth_vector_type): Likewise.
27659         * valtrack.c (cleanup_auto_inc_dec): Likewise.
27660         * var-tracking.c (emit_note_insn_var_location): Likewise.
27661         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
27662         (ADDR_VEC_ALIGN): Likewise.
27664 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27665             Alan Hayward  <alan.hayward@arm.com>
27666             David Sherwood  <david.sherwood@arm.com>
27668         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
27669         unsigned short.
27670         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27671         or if measurement_type is polynomial.
27672         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
27673         * combine.c (make_extraction): Likewise.
27674         * dse.c (find_shift_sequence): Likewise.
27675         * dwarf2out.c (mem_loc_descriptor): Likewise.
27676         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
27677         (extract_bit_field, extract_low_bits): Likewise.
27678         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
27679         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
27680         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
27681         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
27682         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27683         * reload.c (find_reloads): Likewise.
27684         * reload1.c (alter_reg): Likewise.
27685         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
27686         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
27687         * tree-if-conv.c (predicate_mem_writes): Likewise.
27688         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
27689         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
27690         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
27691         * valtrack.c (dead_debug_insert_temp): Likewise.
27692         * varasm.c (mergeable_constant_section): Likewise.
27693         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
27695 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27696             Alan Hayward  <alan.hayward@arm.com>
27697             David Sherwood  <david.sherwood@arm.com>
27699         * expr.c (expand_assignment): Cope with polynomial mode sizes
27700         when assigning to a CONCAT.
27702 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27703             Alan Hayward  <alan.hayward@arm.com>
27704             David Sherwood  <david.sherwood@arm.com>
27706         * machmode.h (mode_precision): Change from unsigned short to
27707         poly_uint16_pod.
27708         (mode_to_precision): Return a poly_uint16 rather than an unsigned
27709         short.
27710         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
27711         or if measurement_type is not polynomial.
27712         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
27713         in which the mode is already known to be a scalar_int_mode.
27714         * genmodes.c (emit_mode_precision): Change the type of mode_precision
27715         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
27716         initializer.
27717         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27718         for GET_MODE_PRECISION.
27719         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27720         for GET_MODE_PRECISION.
27721         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
27722         as polynomial.
27723         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
27724         (expand_field_assignment, make_extraction): Likewise.
27725         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
27726         (get_last_value): Likewise.
27727         * convert.c (convert_to_integer_1): Likewise.
27728         * cse.c (cse_insn): Likewise.
27729         * expr.c (expand_expr_real_1): Likewise.
27730         * lra-constraints.c (simplify_operand_subreg): Likewise.
27731         * optabs-query.c (can_atomic_load_p): Likewise.
27732         * optabs.c (expand_atomic_load): Likewise.
27733         (expand_atomic_store): Likewise.
27734         * ree.c (combine_reaching_defs): Likewise.
27735         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
27736         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
27737         * tree.h (type_has_mode_precision_p): Likewise.
27738         * ubsan.c (instrument_si_overflow): Likewise.
27740 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27741             Alan Hayward  <alan.hayward@arm.com>
27742             David Sherwood  <david.sherwood@arm.com>
27744         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
27745         polynomial numbers of units.
27746         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
27747         (valid_vector_subparts_p): New function.
27748         (build_vector_type): Remove temporary shim and take the number
27749         of units as a poly_uint64 rather than an int.
27750         (build_opaque_vector_type): Take the number of units as a
27751         poly_uint64 rather than an int.
27752         * tree.c (build_vector_from_ctor): Handle polynomial
27753         TYPE_VECTOR_SUBPARTS.
27754         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
27755         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
27756         (build_vector_from_val): If the number of units is variable,
27757         use build_vec_duplicate_cst for constant operands and
27758         VEC_DUPLICATE_EXPR otherwise.
27759         (make_vector_type): Remove temporary is_constant ().
27760         (build_vector_type, build_opaque_vector_type): Take the number of
27761         units as a poly_uint64 rather than an int.
27762         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
27763         VECTOR_CST_NELTS.
27764         * cfgexpand.c (expand_debug_expr): Likewise.
27765         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
27766         (store_constructor, expand_expr_real_1): Likewise.
27767         (const_scalar_mask_from_tree): Likewise.
27768         * fold-const-call.c (fold_const_reduction): Likewise.
27769         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
27770         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
27771         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
27772         (fold_relational_const): Likewise.
27773         (native_interpret_vector): Likewise.  Change the size from an
27774         int to an unsigned int.
27775         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
27776         TYPE_VECTOR_SUBPARTS.
27777         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
27778         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
27779         duplicating a non-constant operand into a variable-length vector.
27780         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
27781         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
27782         * ipa-icf.c (sem_variable::equals): Likewise.
27783         * match.pd: Likewise.
27784         * omp-simd-clone.c (simd_clone_subparts): Likewise.
27785         * print-tree.c (print_node): Likewise.
27786         * stor-layout.c (layout_type): Likewise.
27787         * targhooks.c (default_builtin_vectorization_cost): Likewise.
27788         * tree-cfg.c (verify_gimple_comparison): Likewise.
27789         (verify_gimple_assign_binary): Likewise.
27790         (verify_gimple_assign_ternary): Likewise.
27791         (verify_gimple_assign_single): Likewise.
27792         * tree-pretty-print.c (dump_generic_node): Likewise.
27793         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27794         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
27795         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
27796         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
27797         (vect_shift_permute_load_chain): Likewise.
27798         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
27799         (expand_vector_condition, optimize_vector_constructor): Likewise.
27800         (lower_vec_perm, get_compute_type): Likewise.
27801         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27802         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
27803         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
27804         (vect_recog_mask_conversion_pattern): Likewise.
27805         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
27806         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
27807         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27808         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
27809         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
27810         (vectorizable_shift, vectorizable_operation, vectorizable_store)
27811         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
27812         (supportable_widening_operation): Likewise.
27813         (supportable_narrowing_operation): Likewise.
27814         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
27815         Likewise.
27816         * varasm.c (output_constant): Likewise.
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         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
27823         so that both the length == 3 and length != 3 cases set up their
27824         own permute vectors.  Add comments explaining why we know the
27825         number of elements is constant.
27826         (vect_permute_load_chain): Likewise.
27828 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27829             Alan Hayward  <alan.hayward@arm.com>
27830             David Sherwood  <david.sherwood@arm.com>
27832         * machmode.h (mode_nunits): Change from unsigned char to
27833         poly_uint16_pod.
27834         (ONLY_FIXED_SIZE_MODES): New macro.
27835         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
27836         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
27837         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
27838         New typedefs.
27839         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
27840         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
27841         or if measurement_type is not polynomial.
27842         * genmodes.c (ZERO_COEFFS): New macro.
27843         (emit_mode_nunits_inline): Make mode_nunits_inline return a
27844         poly_uint16.
27845         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
27846         Use ZERO_COEFFS when emitting initializers.
27847         * data-streamer.h (bp_pack_poly_value): New function.
27848         (bp_unpack_poly_value): Likewise.
27849         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27850         for GET_MODE_NUNITS.
27851         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27852         for GET_MODE_NUNITS.
27853         * tree.c (make_vector_type): Remove temporary shim and make
27854         the real function take the number of units as a poly_uint64
27855         rather than an int.
27856         (build_vector_type_for_mode): Handle polynomial nunits.
27857         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
27858         * emit-rtl.c (const_vec_series_p_1): Likewise.
27859         (gen_rtx_CONST_VECTOR): Likewise.
27860         * fold-const.c (test_vec_duplicate_folding): Likewise.
27861         * genrecog.c (validate_pattern): Likewise.
27862         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
27863         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27864         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
27865         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
27866         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
27867         * rtlanal.c (subreg_get_info): Likewise.
27868         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27869         (vect_grouped_load_supported): Likewise.
27870         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
27871         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
27872         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
27873         (simplify_const_unary_operation, simplify_binary_operation_1)
27874         (simplify_const_binary_operation, simplify_ternary_operation)
27875         (test_vector_ops_duplicate, test_vector_ops): Likewise.
27876         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
27877         instead of CONST_VECTOR_NUNITS.
27878         * varasm.c (output_constant_pool_2): Likewise.
27879         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
27880         explicit-encoded elements in the XVEC for variable-length vectors.
27882 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27884         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
27886 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27887             Alan Hayward  <alan.hayward@arm.com>
27888             David Sherwood  <david.sherwood@arm.com>
27890         * coretypes.h (fixed_size_mode): Declare.
27891         (fixed_size_mode_pod): New typedef.
27892         * builtins.h (target_builtins::x_apply_args_mode)
27893         (target_builtins::x_apply_result_mode): Change type to
27894         fixed_size_mode_pod.
27895         * builtins.c (apply_args_size, apply_result_size, result_vector)
27896         (expand_builtin_apply_args_1, expand_builtin_apply)
27897         (expand_builtin_return): Update accordingly.
27899 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27901         * cse.c (hash_rtx_cb): Hash only the encoded elements.
27902         * cselib.c (cselib_hash_rtx): Likewise.
27903         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
27904         CONST_VECTOR encoding.
27906 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27907             Jeff Law  <law@redhat.com>
27909         PR target/83641
27910         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
27911         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
27912         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
27913         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
27915         PR target/83641
27916         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
27917         explicitly probe *sp in a noreturn function if there were any callee
27918         register saves or frame pointer is needed.
27920 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27922         PR debug/83621
27923         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27924         BLKmode for ternary, binary or unary expressions.
27926         PR debug/83645
27927         * var-tracking.c (delete_vta_debug_insn): New inline function.
27928         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27929         insns from get_insns () to NULL instead of each bb separately.
27930         Use delete_vta_debug_insn.  No longer static.
27931         (vt_debug_insns_local, variable_tracking_main_1): Adjust
27932         delete_vta_debug_insns callers.
27933         * rtl.h (delete_vta_debug_insns): Declare.
27934         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27935         instead of variable_tracking_main.
27937 2018-01-03  Martin Sebor  <msebor@redhat.com>
27939         PR tree-optimization/83603
27940         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27941         arguments past the endof the argument list in functions declared
27942         without a prototype.
27943         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27944         Avoid checking when arguments are null.
27946 2018-01-03  Martin Sebor  <msebor@redhat.com>
27948         PR c/83559
27949         * doc/extend.texi (attribute const): Fix a typo.
27950         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27951         issuing -Wsuggest-attribute for void functions.
27953 2018-01-03  Martin Sebor  <msebor@redhat.com>
27955         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27956         offset_int::from instead of wide_int::to_shwi.
27957         (maybe_diag_overlap): Remove assertion.
27958         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27959         * gimple-ssa-sprintf.c (format_directive): Same.
27960         (parse_directive): Same.
27961         (sprintf_dom_walker::compute_format_length): Same.
27962         (try_substitute_return_value): Same.
27964 2018-01-03  Jeff Law  <law@redhat.com>
27966         PR middle-end/83654
27967         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27968         non-constant residual for zero at runtime and avoid probing in
27969         that case.  Reorganize code for trailing problem to mirror handling
27970         of the residual.
27972 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27974         PR tree-optimization/83501
27975         * tree-ssa-strlen.c (get_string_cst): New.
27976         (handle_char_store): Call get_string_cst.
27978 2018-01-03  Martin Liska  <mliska@suse.cz>
27980         PR tree-optimization/83593
27981         * tree-ssa-strlen.c: Include tree-cfg.h.
27982         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27983         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27984         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27985         to false.
27986         (strlen_dom_walker::before_dom_children): Call
27987         gimple_purge_dead_eh_edges. Dump tranformation with details
27988         dump flags.
27989         (strlen_dom_walker::before_dom_children): Update call by adding
27990         new argument cleanup_eh.
27991         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27993 2018-01-03  Martin Liska  <mliska@suse.cz>
27995         PR ipa/83549
27996         * cif-code.def (VARIADIC_THUNK): New enum value.
27997         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27998         thunks.
28000 2018-01-03  Jan Beulich  <jbeulich@suse.com>
28002         * sse.md (mov<mode>_internal): Tighten condition for when to use
28003         vmovdqu<ssescalarsize> for TI and OI modes.
28005 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28007         Update copyright years.
28009 2018-01-03  Martin Liska  <mliska@suse.cz>
28011         PR ipa/83594
28012         * ipa-visibility.c (function_and_variable_visibility): Skip
28013         functions with noipa attribure.
28015 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28017         * gcc.c (process_command): Update copyright notice dates.
28018         * gcov-dump.c (print_version): Ditto.
28019         * gcov.c (print_version): Ditto.
28020         * gcov-tool.c (print_version): Ditto.
28021         * gengtype.c (create_file): Ditto.
28022         * doc/cpp.texi: Bump @copying's copyright year.
28023         * doc/cppinternals.texi: Ditto.
28024         * doc/gcc.texi: Ditto.
28025         * doc/gccint.texi: Ditto.
28026         * doc/gcov.texi: Ditto.
28027         * doc/install.texi: Ditto.
28028         * doc/invoke.texi: Ditto.
28030 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28032         * vector-builder.h (vector_builder::m_full_nelts): Change from
28033         unsigned int to poly_uint64.
28034         (vector_builder::full_nelts): Update prototype accordingly.
28035         (vector_builder::new_vector): Likewise.
28036         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
28037         (vector_builder::operator ==): Likewise.
28038         (vector_builder::finalize): Likewise.
28039         * int-vector-builder.h (int_vector_builder::int_vector_builder):
28040         Take the number of elements as a poly_uint64 rather than an
28041         unsigned int.
28042         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
28043         from unsigned int to poly_uint64.
28044         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
28045         (vec_perm_indices::new_vector): Likewise.
28046         (vec_perm_indices::length): Likewise.
28047         (vec_perm_indices::nelts_per_input): Likewise.
28048         (vec_perm_indices::input_nelts): Likewise.
28049         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
28050         number of elements per input as a poly_uint64 rather than an
28051         unsigned int.  Use the original encoding for variable-length
28052         vectors, rather than clamping each individual element.
28053         For the second and subsequent elements in each pattern,
28054         clamp the step and base before clamping their sum.
28055         (vec_perm_indices::series_p): Handle polynomial element counts.
28056         (vec_perm_indices::all_in_range_p): Likewise.
28057         (vec_perm_indices_to_tree): Likewise.
28058         (vec_perm_indices_to_rtx): Likewise.
28059         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
28060         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
28061         (tree_vector_builder::new_binary_operation): Handle polynomial
28062         element counts.  Return false if we need to know the number
28063         of elements at compile time.
28064         * fold-const.c (fold_vec_perm): Punt if the number of elements
28065         isn't known at compile time.
28067 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28069         * vec-perm-indices.h (vec_perm_builder): Change element type
28070         from HOST_WIDE_INT to poly_int64.
28071         (vec_perm_indices::element_type): Update accordingly.
28072         (vec_perm_indices::clamp): Handle polynomial element_types.
28073         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28074         (vec_perm_indices::all_in_range_p): Likewise.
28075         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
28076         than shwi trees.
28077         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
28078         polynomial vec_perm_indices element types.
28079         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
28080         * fold-const.c (fold_vec_perm): Likewise.
28081         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
28082         * tree-vect-generic.c (lower_vec_perm): Likewise.
28083         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28084         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
28085         element type to HOST_WIDE_INT.
28087 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28088             Alan Hayward  <alan.hayward@arm.com>
28089             David Sherwood  <david.sherwood@arm.com>
28091         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
28092         rather than an int.  Use plus_constant.
28093         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
28094         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
28096 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28097             Alan Hayward  <alan.hayward@arm.com>
28098             David Sherwood  <david.sherwood@arm.com>
28100         * calls.c (emit_call_1, expand_call): Change struct_value_size from
28101         a HOST_WIDE_INT to a poly_int64.
28103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28104             Alan Hayward  <alan.hayward@arm.com>
28105             David Sherwood  <david.sherwood@arm.com>
28107         * calls.c (load_register_parameters): Cope with polynomial
28108         mode sizes.  Require a constant size for BLKmode parameters
28109         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
28110         forces a parameter to be padded at the lsb end in order to
28111         fill a complete number of words, require the parameter size
28112         to be ordered wrt UNITS_PER_WORD.
28114 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28115             Alan Hayward  <alan.hayward@arm.com>
28116             David Sherwood  <david.sherwood@arm.com>
28118         * reload1.c (spill_stack_slot_width): Change element type
28119         from unsigned int to poly_uint64_pod.
28120         (alter_reg): Treat mode sizes as polynomial.
28122 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28123             Alan Hayward  <alan.hayward@arm.com>
28124             David Sherwood  <david.sherwood@arm.com>
28126         * reload.c (complex_word_subreg_p): New function.
28127         (reload_inner_reg_of_subreg, push_reload): Use it.
28129 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28130             Alan Hayward  <alan.hayward@arm.com>
28131             David Sherwood  <david.sherwood@arm.com>
28133         * lra-constraints.c (process_alt_operands): Reject matched
28134         operands whose sizes aren't ordered.
28135         (match_reload): Refer to this check here.
28137 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28138             Alan Hayward  <alan.hayward@arm.com>
28139             David Sherwood  <david.sherwood@arm.com>
28141         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28142         that the mode size is in the set {1, 2, 4, 8, 16}.
28144 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28145             Alan Hayward  <alan.hayward@arm.com>
28146             David Sherwood  <david.sherwood@arm.com>
28148         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28149         Use plus_constant instead of gen_rtx_PLUS.
28151 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28152             Alan Hayward  <alan.hayward@arm.com>
28153             David Sherwood  <david.sherwood@arm.com>
28155         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28156         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28157         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28158         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28159         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28160         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28161         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28162         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28163         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28164         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
28165         a poly_int64.
28166         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
28167         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
28168         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
28169         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
28170         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
28171         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
28172         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
28173         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
28174         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
28175         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28176         function.
28177         * expr.c (emit_move_resolve_push): Treat the input and result
28178         of PUSH_ROUNDING as a poly_int64.
28179         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28180         (emit_push_insn): Likewise.
28181         * lra-eliminations.c (mark_not_eliminable): Likewise.
28182         * recog.c (push_operand): Likewise.
28183         * reload1.c (elimination_effects): Likewise.
28184         * rtlanal.c (nonzero_bits1): Likewise.
28185         * calls.c (store_one_arg): Likewise.  Require the padding to be
28186         known at compile time.
28188 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28189             Alan Hayward  <alan.hayward@arm.com>
28190             David Sherwood  <david.sherwood@arm.com>
28192         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28193         Use plus_constant instead of gen_rtx_PLUS.
28195 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28196             Alan Hayward  <alan.hayward@arm.com>
28197             David Sherwood  <david.sherwood@arm.com>
28199         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28200         rather than an int.
28202 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28203             Alan Hayward  <alan.hayward@arm.com>
28204             David Sherwood  <david.sherwood@arm.com>
28206         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28207         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28208         via stack temporaries.  Treat the mode size as polynomial too.
28210 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28211             Alan Hayward  <alan.hayward@arm.com>
28212             David Sherwood  <david.sherwood@arm.com>
28214         * expr.c (expand_expr_real_2): When handling conversions involving
28215         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28216         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
28217         as a poly_uint64 too.
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         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
28225 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28226             Alan Hayward  <alan.hayward@arm.com>
28227             David Sherwood  <david.sherwood@arm.com>
28229         * combine.c (can_change_dest_mode): Handle polynomial
28230         REGMODE_NATURAL_SIZE.
28231         * expmed.c (store_bit_field_1): Likewise.
28232         * expr.c (store_constructor): Likewise.
28233         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
28234         and polynomial REGMODE_NATURAL_SIZE.
28235         (gen_lowpart_common): Likewise.
28236         * reginfo.c (record_subregs_of_mode): Likewise.
28237         * rtlanal.c (read_modify_subreg_p): Likewise.
28239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28240             Alan Hayward  <alan.hayward@arm.com>
28241             David Sherwood  <david.sherwood@arm.com>
28243         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
28244         numbers of elements.
28246 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28247             Alan Hayward  <alan.hayward@arm.com>
28248             David Sherwood  <david.sherwood@arm.com>
28250         * match.pd: Cope with polynomial numbers of vector elements.
28252 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28253             Alan Hayward  <alan.hayward@arm.com>
28254             David Sherwood  <david.sherwood@arm.com>
28256         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
28257         in a POINTER_PLUS_EXPR.
28259 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28260             Alan Hayward  <alan.hayward@arm.com>
28261             David Sherwood  <david.sherwood@arm.com>
28263         * omp-simd-clone.c (simd_clone_subparts): New function.
28264         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
28265         (ipa_simd_modify_function_body): Likewise.
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         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
28272         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
28273         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
28274         (expand_vector_condition, vector_element): Likewise.
28275         (subparts_gt): New function.
28276         (get_compute_type): Use subparts_gt.
28277         (count_type_subparts): Delete.
28278         (expand_vector_operations_1): Use subparts_gt instead of
28279         count_type_subparts.
28281 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28282             Alan Hayward  <alan.hayward@arm.com>
28283             David Sherwood  <david.sherwood@arm.com>
28285         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
28286         (vect_compile_time_alias): ...this new function.  Do the calculation
28287         on poly_ints rather than trees.
28288         (vect_prune_runtime_alias_test_list): Update call accordingly.
28290 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28291             Alan Hayward  <alan.hayward@arm.com>
28292             David Sherwood  <david.sherwood@arm.com>
28294         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
28295         numbers of units.
28296         (vect_schedule_slp_instance): Likewise.
28298 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28299             Alan Hayward  <alan.hayward@arm.com>
28300             David Sherwood  <david.sherwood@arm.com>
28302         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
28303         constant and extern definitions for variable-length vectors.
28304         (vect_get_constant_vectors): Note that the number of units
28305         is known to be constant.
28307 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28308             Alan Hayward  <alan.hayward@arm.com>
28309             David Sherwood  <david.sherwood@arm.com>
28311         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
28312         of units as polynomial.  Choose between WIDE and NARROW based
28313         on multiple_p.
28315 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28316             Alan Hayward  <alan.hayward@arm.com>
28317             David Sherwood  <david.sherwood@arm.com>
28319         * tree-vect-stmts.c (simd_clone_subparts): New function.
28320         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
28322 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28323             Alan Hayward  <alan.hayward@arm.com>
28324             David Sherwood  <david.sherwood@arm.com>
28326         * tree-vect-stmts.c (vectorizable_call): Treat the number of
28327         vectors as polynomial.  Use build_index_vector for
28328         IFN_GOMP_SIMD_LANE.
28330 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28331             Alan Hayward  <alan.hayward@arm.com>
28332             David Sherwood  <david.sherwood@arm.com>
28334         * tree-vect-stmts.c (get_load_store_type): Treat the number of
28335         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
28336         for variable-length vectors.
28337         (vectorizable_mask_load_store): Treat the number of units as
28338         polynomial, asserting that it is constant if the condition has
28339         already been enforced.
28340         (vectorizable_store, vectorizable_load): Likewise.
28342 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28343             Alan Hayward  <alan.hayward@arm.com>
28344             David Sherwood  <david.sherwood@arm.com>
28346         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
28347         of units as polynomial.  Punt if we can't tell at compile time
28348         which vector contains the final result.
28350 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28351             Alan Hayward  <alan.hayward@arm.com>
28352             David Sherwood  <david.sherwood@arm.com>
28354         * tree-vect-loop.c (vectorizable_induction): Treat the number
28355         of units as polynomial.  Punt on SLP inductions.  Use an integer
28356         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
28357         cast of such a series for variable-length floating-point
28358         reductions.
28360 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28361             Alan Hayward  <alan.hayward@arm.com>
28362             David Sherwood  <david.sherwood@arm.com>
28364         * tree.h (build_index_vector): Declare.
28365         * tree.c (build_index_vector): New function.
28366         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
28367         of units as polynomial, forcibly converting it to a constant if
28368         vectorizable_reduction has already enforced the condition.
28369         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
28370         to create a {1,2,3,...} vector.
28371         (vectorizable_reduction): Treat the number of units as polynomial.
28372         Choose vectype_in based on the largest scalar element size rather
28373         than the smallest number of units.  Enforce the restrictions
28374         relied on above.
28376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28377             Alan Hayward  <alan.hayward@arm.com>
28378             David Sherwood  <david.sherwood@arm.com>
28380         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
28381         number of units as polynomial.
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         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
28388         * target.def (autovectorize_vector_sizes): Return the vector sizes
28389         by pointer, using vector_sizes rather than a bitmask.
28390         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
28391         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
28392         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
28393         Likewise.
28394         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
28395         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
28396         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
28397         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
28398         * omp-general.c (omp_max_vf): Likewise.
28399         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28400         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
28401         * tree-vect-loop.c (vect_analyze_loop): Likewise.
28402         * tree-vect-slp.c (vect_slp_bb): Likewise.
28403         * doc/tm.texi: Regenerate.
28404         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
28405         to a poly_uint64.
28406         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
28407         the vector size as a poly_uint64 rather than an unsigned int.
28408         (current_vector_size): Change from an unsigned int to a poly_uint64.
28409         (get_vectype_for_scalar_type): Update accordingly.
28410         * tree.h (build_truth_vector_type): Take the size and number of
28411         units as a poly_uint64 rather than an unsigned int.
28412         (build_vector_type): Add a temporary overload that takes
28413         the number of units as a poly_uint64 rather than an unsigned int.
28414         * tree.c (make_vector_type): Likewise.
28415         (build_truth_vector_type): Take the number of units as a poly_uint64
28416         rather than an unsigned int.
28418 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28419             Alan Hayward  <alan.hayward@arm.com>
28420             David Sherwood  <david.sherwood@arm.com>
28422         * target.def (get_mask_mode): Take the number of units and length
28423         as poly_uint64s rather than unsigned ints.
28424         * targhooks.h (default_get_mask_mode): Update accordingly.
28425         * targhooks.c (default_get_mask_mode): Likewise.
28426         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
28427         * doc/tm.texi: Regenerate.
28429 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28430             Alan Hayward  <alan.hayward@arm.com>
28431             David Sherwood  <david.sherwood@arm.com>
28433         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
28434         * omp-general.c (omp_max_vf): Likewise.
28435         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
28436         (expand_omp_simd): Handle polynomial safelen.
28437         * omp-low.c (omplow_simd_context): Add a default constructor.
28438         (omplow_simd_context::max_vf): Change from int to poly_uint64.
28439         (lower_rec_simd_input_clauses): Update accordingly.
28440         (lower_rec_input_clauses): Likewise.
28442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28443             Alan Hayward  <alan.hayward@arm.com>
28444             David Sherwood  <david.sherwood@arm.com>
28446         * tree-vectorizer.h (vect_nunits_for_cost): New function.
28447         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
28448         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
28449         (vect_analyze_slp_cost): Likewise.
28450         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
28451         (vect_model_load_cost): Likewise.
28453 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28454             Alan Hayward  <alan.hayward@arm.com>
28455             David Sherwood  <david.sherwood@arm.com>
28457         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
28458         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
28459         from an unsigned int * to a poly_uint64_pod *.
28460         (calculate_unrolling_factor): New function.
28461         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
28463 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28464             Alan Hayward  <alan.hayward@arm.com>
28465             David Sherwood  <david.sherwood@arm.com>
28467         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
28468         from an unsigned int to a poly_uint64.
28469         (_loop_vec_info::slp_unrolling_factor): Likewise.
28470         (_loop_vec_info::vectorization_factor): Change from an int
28471         to a poly_uint64.
28472         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
28473         (vect_get_num_vectors): New function.
28474         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
28475         (vect_get_num_copies): Use vect_get_num_vectors.
28476         (vect_analyze_data_ref_dependences): Change max_vf from an int *
28477         to an unsigned int *.
28478         (vect_analyze_data_refs): Change min_vf from an int * to a
28479         poly_uint64 *.
28480         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28481         than an unsigned HOST_WIDE_INT.
28482         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
28483         (vect_analyze_data_ref_dependence): Change max_vf from an int *
28484         to an unsigned int *.
28485         (vect_analyze_data_ref_dependences): Likewise.
28486         (vect_compute_data_ref_alignment): Handle polynomial vf.
28487         (vect_enhance_data_refs_alignment): Likewise.
28488         (vect_prune_runtime_alias_test_list): Likewise.
28489         (vect_shift_permute_load_chain): Likewise.
28490         (vect_supportable_dr_alignment): Likewise.
28491         (dependence_distance_ge_vf): Take the vectorization factor as a
28492         poly_uint64 rather than an unsigned HOST_WIDE_INT.
28493         (vect_analyze_data_refs): Change min_vf from an int * to a
28494         poly_uint64 *.
28495         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
28496         vfm1 as a poly_uint64 rather than an int.  Make the same change
28497         for the returned bound_scalar.
28498         (vect_gen_vector_loop_niters): Handle polynomial vf.
28499         (vect_do_peeling): Likewise.  Update call to
28500         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
28501         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
28502         be constant.
28503         * tree-vect-loop.c (vect_determine_vectorization_factor)
28504         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
28505         (vect_get_known_peeling_cost): Likewise.
28506         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
28507         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
28508         (vect_transform_loop): Likewise.  Use the lowest possible VF when
28509         updating the upper bounds of the loop.
28510         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
28511         rather than an int.
28512         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
28513         polynomial unroll factors.
28514         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
28515         (vect_make_slp_decision): Likewise.
28516         (vect_supported_load_permutation_p): Likewise, and polynomial
28517         vf too.
28518         (vect_analyze_slp_cost): Handle polynomial vf.
28519         (vect_slp_analyze_node_operations): Likewise.
28520         (vect_slp_analyze_bb_1): Likewise.
28521         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28522         than an unsigned HOST_WIDE_INT.
28523         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
28524         (vectorizable_load): Handle polynomial vf.
28525         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
28526         a poly_uint64.
28527         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
28529 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28530             Alan Hayward  <alan.hayward@arm.com>
28531             David Sherwood  <david.sherwood@arm.com>
28533         * match.pd: Handle bit operations involving three constants
28534         and try to fold one pair.
28536 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28538         * tree-vect-loop-manip.c: Include gimple-fold.h.
28539         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
28540         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
28541         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
28542         Add a path that uses a step of VF instead of 1, but disable it
28543         for now.
28544         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
28545         and niters_no_overflow parameters.  Update calls to
28546         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
28547         Create a new SSA name if the latter choses to use a ste other
28548         than zero, and return it via niters_vector_mult_vf_var.
28549         * tree-vect-loop.c (vect_transform_loop): Update calls to
28550         vect_do_peeling, vect_gen_vector_loop_niters and
28551         slpeel_make_loop_iterate_ntimes.
28552         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
28553         (vect_gen_vector_loop_niters): Update declarations after above changes.
28555 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
28557         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
28558         128-bit round to integer instructions.
28559         (ceil<mode>2): Likewise.
28560         (btrunc<mode>2): Likewise.
28561         (round<mode>2): Likewise.
28563 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28565         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
28566         unaligned VSX load/store on P8/P9.
28567         (expand_block_clear): Allow the use of unaligned VSX
28568         load/store on P8/P9.
28570 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28572         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
28573         New function.
28574         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
28575         swap associated with both a load and a store.
28577 2018-01-02  Andrew Waterman  <andrew@sifive.com>
28579         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
28580         * config/riscv/riscv.md (clear_cache): Use it.
28582 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
28584         * web.c: Remove out-of-date comment.
28586 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28588         * expr.c (fixup_args_size_notes): Check that any existing
28589         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
28590         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
28591         (emit_single_push_insn): ...here.
28593 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28595         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
28596         (const_vector_encoded_nelts): New function.
28597         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
28598         (const_vector_int_elt, const_vector_elt): Declare.
28599         * emit-rtl.c (const_vector_int_elt_1): New function.
28600         (const_vector_elt): Likewise.
28601         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
28602         of CONST_VECTOR_ELT.
28604 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28606         * expr.c: Include rtx-vector-builder.h.
28607         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
28608         directly on the tree encoding.
28609         (const_vector_from_tree): Likewise.
28610         * optabs.c: Include rtx-vector-builder.h.
28611         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
28612         sequence of "u" values.
28613         * vec-perm-indices.c: Include rtx-vector-builder.h.
28614         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
28615         directly on the vec_perm_indices encoding.
28617 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28619         * doc/rtl.texi (const_vector): Describe new encoding scheme.
28620         * Makefile.in (OBJS): Add rtx-vector-builder.o.
28621         * rtx-vector-builder.h: New file.
28622         * rtx-vector-builder.c: Likewise.
28623         * rtl.h (rtx_def::u2): Add a const_vector field.
28624         (CONST_VECTOR_NPATTERNS): New macro.
28625         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
28626         (CONST_VECTOR_DUPLICATE_P): Likewise.
28627         (CONST_VECTOR_STEPPED_P): Likewise.
28628         (CONST_VECTOR_ENCODED_ELT): Likewise.
28629         (const_vec_duplicate_p): Check for a duplicated vector encoding.
28630         (unwrap_const_vec_duplicate): Likewise.
28631         (const_vec_series_p): Check for a non-duplicated vector encoding.
28632         Say that the function only returns true for integer vectors.
28633         * emit-rtl.c: Include rtx-vector-builder.h.
28634         (gen_const_vec_duplicate_1): Delete.
28635         (gen_const_vector): Call gen_const_vec_duplicate instead of
28636         gen_const_vec_duplicate_1.
28637         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
28638         (gen_const_vec_duplicate): Use rtx_vector_builder.
28639         (gen_const_vec_series): Likewise.
28640         (gen_rtx_CONST_VECTOR): Likewise.
28641         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
28642         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28643         Build a new vector rather than modifying a CONST_VECTOR in-place.
28644         (handle_special_swappables): Update call accordingly.
28645         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
28646         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28647         Build a new vector rather than modifying a CONST_VECTOR in-place.
28648         (handle_special_swappables): Update call accordingly.
28650 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28652         * simplify-rtx.c (simplify_const_binary_operation): Use
28653         CONST_VECTOR_ELT instead of XVECEXP.
28655 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28657         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
28658         the selector elements to be different from the data elements
28659         if the selector is a VECTOR_CST.
28660         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
28661         ssizetype for the selector.
28663 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28665         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
28666         before testing each element individually.
28667         * tree-vect-generic.c (lower_vec_perm): Likewise.
28669 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28671         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
28672         * selftest-run-tests.c (selftest::run_tests): Call it.
28673         * vector-builder.h (vector_builder::operator ==): New function.
28674         (vector_builder::operator !=): Likewise.
28675         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
28676         (vec_perm_indices::all_from_input_p): New function.
28677         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28678         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
28679         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
28680         instead of reading the VECTOR_CST directly.  Detect whether both
28681         vector inputs are the same before constructing the vec_perm_indices,
28682         and update the number of inputs argument accordingly.  Use the
28683         utility functions added above.  Only construct sel2 if we need to.
28685 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28687         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
28688         the broadcast of the low byte.
28689         (expand_mult_highpart): Use an explicit encoding for the permutes.
28690         * optabs-query.c (can_mult_highpart_p): Likewise.
28691         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
28692         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28693         (vectorizable_bswap): Likewise.
28694         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
28695         explicit encoding for the power-of-2 permutes.
28696         (vect_permute_store_chain): Likewise.
28697         (vect_grouped_load_supported): Likewise.
28698         (vect_permute_load_chain): Likewise.
28700 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28702         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
28703         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
28704         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
28705         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28706         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
28707         (vect_gen_perm_mask_any): Likewise.
28709 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28711         * int-vector-builder.h: New file.
28712         * vec-perm-indices.h: Include int-vector-builder.h.
28713         (vec_perm_indices): Redefine as an int_vector_builder.
28714         (auto_vec_perm_indices): Delete.
28715         (vec_perm_builder): Redefine as a stand-alone class.
28716         (vec_perm_indices::vec_perm_indices): New function.
28717         (vec_perm_indices::clamp): Likewise.
28718         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
28719         (vec_perm_indices::new_vector): New function.
28720         (vec_perm_indices::new_expanded_vector): Update for new
28721         vec_perm_indices class.
28722         (vec_perm_indices::rotate_inputs): New function.
28723         (vec_perm_indices::all_in_range_p): Operate directly on the
28724         encoded form, without computing elided elements.
28725         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
28726         encoding.  Update for new vec_perm_indices class.
28727         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
28728         the given vec_perm_builder.
28729         (expand_vec_perm_var): Update vec_perm_builder constructor.
28730         (expand_mult_highpart): Use vec_perm_builder instead of
28731         auto_vec_perm_indices.
28732         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
28733         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
28734         or double series encoding as appropriate.
28735         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
28736         vec_perm_indices instead of auto_vec_perm_indices.
28737         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28738         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28739         (vect_permute_store_chain): Likewise.
28740         (vect_grouped_load_supported): Likewise.
28741         (vect_permute_load_chain): Likewise.
28742         (vect_shift_permute_load_chain): Likewise.
28743         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28744         (vect_transform_slp_perm_load): Likewise.
28745         (vect_schedule_slp_instance): Likewise.
28746         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28747         (vectorizable_mask_load_store): Likewise.
28748         (vectorizable_bswap): Likewise.
28749         (vectorizable_store): Likewise.
28750         (vectorizable_load): Likewise.
28751         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
28752         vec_perm_indices instead of auto_vec_perm_indices.  Use
28753         tree_to_vec_perm_builder to read the vector from a tree.
28754         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
28755         vec_perm_builder instead of a vec_perm_indices.
28756         (have_whole_vector_shift): Use vec_perm_builder and
28757         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
28758         truncation to calc_vec_perm_mask_for_shift.
28759         (vect_create_epilog_for_reduction): Likewise.
28760         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
28761         from auto_vec_perm_indices to vec_perm_indices.
28762         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28763         instead of changing individual elements.
28764         (aarch64_vectorize_vec_perm_const): Use new_vector to install
28765         the vector in d.perm.
28766         * config/arm/arm.c (expand_vec_perm_d::perm): Change
28767         from auto_vec_perm_indices to vec_perm_indices.
28768         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28769         instead of changing individual elements.
28770         (arm_vectorize_vec_perm_const): Use new_vector to install
28771         the vector in d.perm.
28772         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
28773         Update vec_perm_builder constructor.
28774         (rs6000_expand_interleave): Likewise.
28775         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
28776         (rs6000_expand_interleave): Likewise.
28778 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28780         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
28781         to qimode could truncate the indices.
28782         * optabs.c (expand_vec_perm_var): Likewise.
28784 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28786         * Makefile.in (OBJS): Add vec-perm-indices.o.
28787         * vec-perm-indices.h: New file.
28788         * vec-perm-indices.c: Likewise.
28789         * target.h (vec_perm_indices): Replace with a forward class
28790         declaration.
28791         (auto_vec_perm_indices): Move to vec-perm-indices.h.
28792         * optabs.h: Include vec-perm-indices.h.
28793         (expand_vec_perm): Delete.
28794         (selector_fits_mode_p, expand_vec_perm_var): Declare.
28795         (expand_vec_perm_const): Declare.
28796         * target.def (vec_perm_const_ok): Replace with...
28797         (vec_perm_const): ...this new hook.
28798         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
28799         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
28800         * doc/tm.texi: Regenerate.
28801         * optabs.def (vec_perm_const): Delete.
28802         * doc/md.texi (vec_perm_const): Likewise.
28803         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
28804         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
28805         expand_vec_perm for constant permutation vectors.  Assert that
28806         the mode of variable permutation vectors is the integer equivalent
28807         of the mode that is being permuted.
28808         * optabs-query.h (selector_fits_mode_p): Declare.
28809         * optabs-query.c: Include vec-perm-indices.h.
28810         (selector_fits_mode_p): New function.
28811         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
28812         is defined, instead of checking whether the vec_perm_const_optab
28813         exists.  Use targetm.vectorize.vec_perm_const instead of
28814         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
28815         fit in the vector mode before using a variable permute.
28816         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
28817         vec_perm_indices instead of an rtx.
28818         (expand_vec_perm): Replace with...
28819         (expand_vec_perm_const): ...this new function.  Take the selector
28820         as a vec_perm_indices rather than an rtx.  Also take the mode of
28821         the selector.  Update call to shift_amt_for_vec_perm_mask.
28822         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
28823         Use vec_perm_indices::new_expanded_vector to expand the original
28824         selector into bytes.  Check whether the indices fit in the vector
28825         mode before using a variable permute.
28826         (expand_vec_perm_var): Make global.
28827         (expand_mult_highpart): Use expand_vec_perm_const.
28828         * fold-const.c: Includes vec-perm-indices.h.
28829         * tree-ssa-forwprop.c: Likewise.
28830         * tree-vect-data-refs.c: Likewise.
28831         * tree-vect-generic.c: Likewise.
28832         * tree-vect-loop.c: Likewise.
28833         * tree-vect-slp.c: Likewise.
28834         * tree-vect-stmts.c: Likewise.
28835         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
28836         Delete.
28837         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
28838         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
28839         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
28840         (aarch64_vectorize_vec_perm_const): ...this new function.
28841         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28842         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28843         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
28844         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
28845         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28846         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28847         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
28848         into...
28849         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
28850         check for NEON modes.
28851         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
28852         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
28853         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
28854         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
28855         into...
28856         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
28857         the old VEC_PERM_CONST conditions.
28858         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
28859         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
28860         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
28861         (ia64_vectorize_vec_perm_const_ok): Merge into...
28862         (ia64_vectorize_vec_perm_const): ...this new function.
28863         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
28864         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
28865         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
28866         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
28867         * config/mips/mips.c (mips_expand_vec_perm_const)
28868         (mips_vectorize_vec_perm_const_ok): Merge into...
28869         (mips_vectorize_vec_perm_const): ...this new function.
28870         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
28871         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
28872         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
28873         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
28874         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
28875         (rs6000_expand_vec_perm_const): Delete.
28876         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
28877         Delete.
28878         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28879         (altivec_expand_vec_perm_const_le): Take each operand individually.
28880         Operate on constant selectors rather than rtxes.
28881         (altivec_expand_vec_perm_const): Likewise.  Update call to
28882         altivec_expand_vec_perm_const_le.
28883         (rs6000_expand_vec_perm_const): Delete.
28884         (rs6000_vectorize_vec_perm_const_ok): Delete.
28885         (rs6000_vectorize_vec_perm_const): New function.
28886         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28887         an element count and rtx array.
28888         (rs6000_expand_extract_even): Update call accordingly.
28889         (rs6000_expand_interleave): Likewise.
28890         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
28891         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
28892         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
28893         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
28894         (rs6000_expand_vec_perm_const): Delete.
28895         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28896         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28897         (altivec_expand_vec_perm_const_le): Take each operand individually.
28898         Operate on constant selectors rather than rtxes.
28899         (altivec_expand_vec_perm_const): Likewise.  Update call to
28900         altivec_expand_vec_perm_const_le.
28901         (rs6000_expand_vec_perm_const): Delete.
28902         (rs6000_vectorize_vec_perm_const_ok): Delete.
28903         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
28904         reference to the SPE evmerge intructions.
28905         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28906         an element count and rtx array.
28907         (rs6000_expand_extract_even): Update call accordingly.
28908         (rs6000_expand_interleave): Likewise.
28909         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
28910         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
28911         new function.
28912         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28914 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28916         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
28917         vector mode and that that mode matches the mode of the data
28918         being permuted.
28919         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
28920         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
28921         directly using expand_vec_perm_1 when forcing selectors into
28922         registers.
28923         (expand_vec_perm_var): New function, split out from expand_vec_perm.
28925 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28927         * optabs-query.h (can_vec_perm_p): Delete.
28928         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28929         * optabs-query.c (can_vec_perm_p): Split into...
28930         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28931         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28932         particular selector is valid.
28933         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28934         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28935         (vect_grouped_load_supported): Likewise.
28936         (vect_shift_permute_load_chain): Likewise.
28937         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28938         (vect_transform_slp_perm_load): Likewise.
28939         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28940         (vectorizable_bswap): Likewise.
28941         (vect_gen_perm_mask_checked): Likewise.
28942         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
28943         implementations of variable permutation vectors into account
28944         when deciding which selector to use.
28945         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28946         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28947         with a false third argument.
28948         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28949         to test whether the constant selector is valid and can_vec_perm_var_p
28950         to test whether a variable selector is valid.
28952 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28954         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28955         * optabs-query.c (can_vec_perm_p): Likewise.
28956         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28957         instead of vec_perm_indices.
28958         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28959         (vect_gen_perm_mask_checked): Likewise,
28960         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28961         (vect_gen_perm_mask_checked): Likewise,
28963 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28965         * optabs-query.h (qimode_for_vec_perm): Declare.
28966         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28967         (qimode_for_vec_perm): ...this new function.
28968         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28970 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28972         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28973         does not have a conditional at the top.
28975 2018-01-02  Richard Biener  <rguenther@suse.de>
28977         * ipa-inline.c (big_speedup_p): Fix expression.
28979 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28981         PR target/81616
28982         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28983         for generic 4->6.
28985 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28987         PR target/81616
28988         Generic tuning.
28989         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28990         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28991         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28992         cond_taken_branch_cost 3->4.
28994 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
28996         PR tree-optimization/83581
28997         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28998         TODO_cleanup_cfg if any changes have been made.
29000         PR middle-end/83608
29001         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
29002         convert_modes if target mode has the right side, but different mode
29003         class.
29005         PR middle-end/83609
29006         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
29007         last argument when extracting from CONCAT.  If either from_real or
29008         from_imag is NULL, use expansion through memory.  If result is not
29009         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
29010         the parts directly to inner mode, if even that fails, use expansion
29011         through memory.
29013         PR middle-end/83623
29014         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
29015         check for bswap in mode rather than HImode and use that in expand_unop
29016         too.
29018 Copyright (C) 2018 Free Software Foundation, Inc.
29020 Copying and distribution of this file, with or without modification,
29021 are permitted in any medium without royalty provided the copyright
29022 notice and this notice are preserved.