2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
[official-gcc.git] / gcc / ChangeLog
blob54f36a1278df73739cdd4ffc16d819ae2d8010b2
1 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
2         Richard Biener  <rguenther@suse.de>
4         * bitmap.h: Update data structure documentation, including a
5         description of bitmap views as either linked-lists or splay trees.
6         (struct bitmap_element_def): Update comments for splay tree bitmaps.
7         (struct bitmap_head_def): Likewise.
8         (bitmap_list_view, bitmap_tree_view): New prototypes.
9         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
10         tree_form fields.
11         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
12         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
13         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
14         released bitmap element here.
15         (bitmap_element_free): Remove.
16         (bitmap_elt_clear_from): Work on splay tree bitmaps.
17         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
18         this function similar ones such that linked-list bitmap implementation
19         functions are grouped.
20         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
21         and moved for grouping.
22         (bitmap_list_insert_element_after): Renamed from
23         bitmap_elt_insert_after, and moved for grouping.
24         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
25         (bitmap_tree_link_left, bitmap_tree_link_right,
26         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
27         bitmap_tree_link_element, bitmap_tree_unlink_element,
28         bitmap_tree_find_element): New functions for splay-tree bitmap
29         implementation.
30         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
31         Renamed and moved, see above entries.
32         (bitmap_tree_listify_from): New function to convert part of a splay
33         tree bitmap to a linked-list bitmap.
34         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
35         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
36         (bitmap_find_bit): Remove.
37         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
38         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
39         Handle splay tree bitmaps.
40         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
41         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
42         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
43         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
44         bitmap_intersect_compl_p, bitmap_ior_and_compl,
45         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
46         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
47         corresponding changes to use linked-list specific bitmap_element
48         manipulation functions as applicable for efficiency.
49         (bitmap_tree_to_vec): New function.
50         (debug_bitmap_elt_file): New function split out from ...
51         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
52         (bitmap_print): Likewise.
54         PR tree-optimization/63155
55         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
56         SSA edge worklists.
57         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
58         in tree-view.
60 2018-10-22  Martin Liska  <mliska@suse.cz>
62         PR tree-optimization/87686
63         Revert
64         2018-08-29  Martin Liska  <mliska@suse.cz>
66         * tree-switch-conversion.c (switch_conversion::expand):
67         Strenghten assumption about gswitch statements.
69 2018-10-22  Martin Liska  <mliska@suse.cz>
71         * ipa-icf.c (sem_item::compare_attributes): Remove.
72         (sem_item::compare_referenced_symbol_properties): Use
73         attribute_list_equal instead.
74         (sem_function::equals_wpa): Likewise.
75         * ipa-icf.h: Remove compare_attributes.
77 2018-10-22  Richard Biener  <rguenther@suse.de>
79         PR middle-end/87682
80         * mem-stats.h (mem_usage::operator==): Fix pasto.
82 2018-10-22  Richard Biener  <rguenther@suse.de>
84         PR tree-optimization/87640
85         * tree-vrp.c (set_value_range_with_overflow): Decompose
86         incomplete result.
87         (extract_range_from_binary_expr_1): Adjust.
89 2018-10-22  Martin Jambor  <mjambor@suse.cz>
91         * tree-eh.h (stmt_could_throw_p): Add function parameter.
92         (stmt_can_throw_external): Likewise.
93         (stmt_can_throw_internal): Likewise.
94         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
95         (lower_eh_constructs_2): Likewise.
96         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
97         (stmt_can_throw_external): Likewise.
98         (stmt_can_throw_internal): Likewise.
99         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
100         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
101         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
102         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
103         (pass_lower_eh_dispatch::execute): Pass cfun to
104         stmt_can_throw_external.
105         (cleanup_empty_eh): Likewise.
106         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
107         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
108         stmt_can_throw_external instead of pushing it to cfun.
109         (symbol_table::create_edge): Likewise.
110         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
111         stmt_can_throw_internal.
112         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
113         to stmt_could_throw_p.
114         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
115         stmt_can_throw_internal.
116         (pass_store_merging::execute): Likewise.
117         * gimple-ssa-strength-reduction.c
118         (find_candidates_dom_walker::before_dom_children): Pass cfun to
119         stmt_could_throw_p.
120         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
121         stmt_can_throw_internal.
122         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
123         to stmt_can_throw_external.
124         (check_stmt): Pass cfun to stmt_could_throw_p.
125         (check_stmt): Pass cfun to stmt_can_throw_external.
126         (pass_nothrow::execute): Likewise.
127         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
128         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
129         stmt_can_throw_internal.
130         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
131         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
132         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
133         * tree-complex.c (expand_complex_libcall): Pass cfun to
134         stmt_could_throw_p and to stmt_can_throw_internal.
135         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
136         * tree-inline.c (copy_edges_for_bb): Likewise.
137         (maybe_move_debug_stmts_to_successors): Likewise.
138         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
139         stmt_could_throw_p.
140         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
141         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
142         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
143         stmt_can_throw_internal.
144         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
145         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
146         stmt_could_throw_p.
147         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
148         stmt_can_throw_internal.
149         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
150         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
151         stmt_could_throw_p.
152         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
153         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
154         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
155         (convert_mult_to_fma_1): Likewise.
156         (convert_to_divmod): Likewise.
157         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
158         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
159         * tree-ssa-propagate.c
160         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
161         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
162         (maybe_optimize_range_tests): Likewise.
163         (linearize_expr_tree): Likewise.
164         (reassociate_bb): Likewise.
165         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
166         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
167         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
168         (handle_char_store): Likewise.
169         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
170         stmt_can_throw_internal.
171         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
172         stmt_could_throw_p.
173         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
174         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
175         (vectorizable_simd_clone_call): Likewise.
176         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
177         (gimple_stringop_fixed_value): Likewise.
179 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
181         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
182         literal pool references.
183         (s390_check_qrst_address): Adapt to the new behavior of
184         s390_loadrelative_operand_p ().
186 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
188         PR target/72782
189         * config/i386/sse.md (*andnot<mode>3_bcst): New.
191 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
193         PR target/72782
194         * config/i386/sse.md (*<code><mode>3_bcst): New.
196 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
198         PR target/72782
199         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
200         V4DI, V16SI and V8DI.
201         (*sub<mode>3<mask_name>_bcst): New.
202         (*add<mode>3<mask_name>_bcst): Likewise.
204 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
206         PR target/72782
207         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
208         __builtin_ia32_vfnmsubpd512_mask.
209         (_mm512_mask_fnmsub_round_pd): Likewise.
210         (_mm512_fnmsub_pd): Likewise.
211         (_mm512_mask_fnmsub_pd): Likewise.
212         (_mm512_maskz_fnmsub_round_pd): Use
213         __builtin_ia32_vfnmsubpd512_maskz.
214         (_mm512_maskz_fnmsub_pd): Likewise.
215         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
216         (_mm512_mask_fnmsub_round_ps): Likewise.
217         (_mm512_fnmsub_ps): Likewise.
218         (_mm512_mask_fnmsub_ps): Likewise.
219         (_mm512_maskz_fnmsub_round_ps): Use
220         __builtin_ia32_vfnmsubps512_maskz.
221         (_mm512_maskz_fnmsub_ps): Likewise.
222         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
223         __builtin_ia32_vfnmsubpd256_mask.
224         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
225         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
226         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
227         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
228         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
229         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
230         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
231         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
232         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
233         __builtin_ia32_vfnmsubpd.
234         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
235         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
236         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
237         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
238         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
239         * config/i386/i386-builtin.def: Add
240         __builtin_ia32_vfnmsubpd256_mask,
241         __builtin_ia32_vfnmsubpd256_maskz,
242         __builtin_ia32_vfnmsubpd128_mask,
243         __builtin_ia32_vfnmsubpd128_maskz,
244         __builtin_ia32_vfnmsubps256_mask,
245         __builtin_ia32_vfnmsubps256_maskz,
246         __builtin_ia32_vfnmsubps128_mask,
247         __builtin_ia32_vfnmsubps128_maskz,
248         __builtin_ia32_vfnmsubpd512_mask,
249         __builtin_ia32_vfnmsubpd512_maskz,
250         __builtin_ia32_vfnmsubps512_mask,
251         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
252         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
253         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
254         __builtin_ia32_vfnmsubpd256.
255         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
256         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
257         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
258         Likewise.
259         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
260         Likewise.
261         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
262         Likewise.
263         (fmai_vmfnmsub_<mode><round_name>): Likewise.
265 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
267         PR target/72782
268         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
269         __builtin_ia32_vfnmaddpd512_mask.
270         (_mm512_mask_fnmadd_round_pd): Likewise.
271         (_mm512_fnmadd_pd): Likewise.
272         (_mm512_mask_fnmadd_pd): Likewise.
273         (_mm512_maskz_fnmadd_round_pd): Use
274         __builtin_ia32_vfnmaddpd512_maskz.
275         (_mm512_maskz_fnmadd_pd): Likewise.
276         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
277         (_mm512_mask_fnmadd_round_ps): Likewise.
278         (_mm512_fnmadd_ps): Likewise.
279         (_mm512_mask_fnmadd_ps): Likewise.
280         (_mm512_maskz_fnmadd_round_ps): Use
281         __builtin_ia32_vfnmaddps512_maskz.
282         (_mm512_maskz_fnmadd_ps): Likewise.
283         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
284         __builtin_ia32_vfnmaddpd256_mask.
285         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
286         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
287         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
288         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
289         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
290         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
291         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
292         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
293         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
294         __builtin_ia32_vfnmaddpd.
295         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
296         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
297         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
298         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
299         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
300         * config/i386/i386-builtin.def: Add
301         __builtin_ia32_vfnmaddpd256_mask,
302         __builtin_ia32_vfnmaddpd256_maskz,
303         __builtin_ia32_vfnmaddpd128_mask,
304         __builtin_ia32_vfnmaddpd128_maskz,
305         __builtin_ia32_vfnmaddps256_mask,
306         __builtin_ia32_vfnmaddps256_maskz,
307         __builtin_ia32_vfnmaddps128_mask,
308         __builtin_ia32_vfnmaddps128_maskz,
309         __builtin_ia32_vfnmaddpd512_mask,
310         __builtin_ia32_vfnmaddpd512_maskz,
311         __builtin_ia32_vfnmaddps512_mask,
312         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
313         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
314         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
315         __builtin_ia32_vfnmaddpd256.
316         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
317         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
318         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
319         Likewise.
320         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
321         Likewise.
322         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
323         Likewise.
324         (fmai_vmfnmadd_<mode><round_name>): Likewise.
326 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
328         PR target/72782
329         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
330         __builtin_ia32_vfmsubpd512_mask.
331         (_mm512_mask_fmsub_round_pd): Likewise.
332         (_mm512_fmsub_pd): Likewise.
333         (_mm512_mask_fmsub_pd): Likewise.
334         (_mm512_maskz_fmsub_round_pd): Use
335         __builtin_ia32_vfmsubpd512_maskz.
336         (_mm512_maskz_fmsub_pd): Likewise.
337         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
338         (_mm512_mask_fmsub_round_ps): Likewise.
339         (_mm512_fmsub_ps): Likewise.
340         (_mm512_mask_fmsub_ps): Likewise.
341         (_mm512_maskz_fmsub_round_ps): Use
342         __builtin_ia32_vfmsubps512_maskz.
343         (_mm512_maskz_fmsub_ps): Likewise.
344         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
345         __builtin_ia32_vfmsubpd256_mask.
346         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
347         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
348         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
349         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
350         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
351         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
352         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
353         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
354         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
355         __builtin_ia32_vfmsubpd.
356         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
357         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
358         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
359         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
360         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
361         * config/i386/i386-builtin.def: Add
362         __builtin_ia32_vfmsubpd256_mask,
363         __builtin_ia32_vfmsubpd256_maskz,
364         __builtin_ia32_vfmsubpd128_mask,
365         __builtin_ia32_vfmsubpd128_maskz,
366         __builtin_ia32_vfmsubps256_mask,
367         __builtin_ia32_vfmsubps256_maskz,
368         __builtin_ia32_vfmsubps128_mask,
369         __builtin_ia32_vfmsubps128_maskz,
370         __builtin_ia32_vfmsubpd512_mask,
371         __builtin_ia32_vfmsubpd512_maskz,
372         __builtin_ia32_vfmsubps512_mask,
373         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
374         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
375         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
376         __builtin_ia32_vfmsubpd256.
377         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
378         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
379         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
380         Likewise.
381         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
382         Likewise.
383         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
384         Likewise.
385         (fmai_vmfmsub_<mode><round_name>): Likewise.
387 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
389         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
390         Remove plus.  Renamed to ...
391         (*sub<mode>3<mask_name>_bcst): This.
392         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
393         (*add<mode>3<mask_name>_bcst): This.
395 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
397         PR target/72782
398         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
400 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
402         PR target/87662
403         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
404         (_mm_or_epi32): Likewise.
405         (_mm256_xor_epi32): Likewise.
406         (_mm_xor_epi32): Likewise.
407         (_mm256_or_epi64): Likewise.
408         (_mm_or_epi64): Likewise.
409         (_mm256_xor_epi64): Likewise.
410         (_mm_xor_epi64): Likewise.
412 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
414         PR target/72782
415         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
417 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
419         PR middle-end/87647
420         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
422 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
424         * doc/ux.texi: Move @section directly after @node.
426 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
428         PR middle-end/85488
429         PR middle-end/87649
430         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
431         depend closely nested inside of loop with ordered clause with
432         a parameter.
434 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
436         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
437         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
438         * doc/ux.texi: New file.
440 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
442         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
443         be the first CR field allocated.
445 2018-10-19  Richard Biener  <rguenther@suse.de>
447         PR target/87657
448         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
449         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
451 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
453         PR target/72782
454         * config/i386/sse.md
455         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
456         (*add<mode>3<mask_name>_bcst_2): Likewise.
458 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
460         * config/i386/sse.md
461         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
462         Replace nonimmediate_operand with register_operand.
463         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
464         Likewise.
465         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
466         Likewise.
468 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
470         PR rtl-optimization/87596
471         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
472         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
473         for instructions in FROM..TO range.
475 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
477         * cfgexpand.c (expand_one_var): Use specific wording in error message
478         for non-local frame variables.
479         * stor-layout.c (layout_decl): Do not issue a warning for them.
481 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
483         * haifa-sched.c (priority): Add force_recompute parameter.
484         (apply_replacement): Call priority () with force_recompute = true.
485         (restore_pattern): Likewise.
487 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
489         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
490         HOST_BITS_PER_WIDE_INT.
491         (test_vector_ops_duplicate): Likewise.
493 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
495         PR target/72782
496         * config/i386/sse.md (VF_AVX512): New.
497         (avx512bcst): Likewise.
498         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
499         Likewise.
500         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
501         Likewise.
502         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
503         Likewise.
505 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
507         * doc/invoke.texi (-dumpversion): Improve grammar.
508         (-dumpfullversion): Make more consistent with -dumpversion.
510 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
512         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
513         Set JUMP_LABEL to the jump insn.
514         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
515         Predict emitted jump and add label to jump insn.
517 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
519         PR tree-optimization/87562
520         * input.c (get_substring_ranges_for_loc): Use
521         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
522         getting the linemap for the endpoint.  Verify that it's either
523         in the same linemap as the start point's spelling location, or
524         at least in the same file.
526 2018-10-18  Richard Biener  <rguenther@suse.de>
528         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
529         feed width-specific load/store costs through ix86_vec_cost.
530         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
531         (k8_cost): Likewise.
532         (bdver_cost): Likewise.
533         (znver1_cost): Likewise.
534         (btver1_cost): Likewise.
535         (btver2_cost): Likewise.
537 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
539         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
540         to simplify subreg of vec_merge.
542 2018-10-18  Richard Biener  <rguenther@suse.de>
544         * config/i386/i386.c: Fix costing of vector FMA.
546 2018-10-18  Richard Biener  <rguenther@suse.de>
548         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
549         and argument.
550         (ix86_builtin_vectorization_cost): For vec_construct properly
551         cost insertion into SSE regs.
552         (...): Adjust calls to ix86_vec_cost.
554 2018-10-18  Richard Biener  <rguenther@suse.de>
556         PR middle-end/87087
557         Revert
558         2018-02-07  Richard Biener  <rguenther@suse.de>
560         PR tree-optimization/84204
561         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
562         this place.
564 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
566         PR target/87537
567         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
568         of vec_duplicate.
569         (test_vector_ops_duplicate): Add test for a scalar subreg of a
570         VEC_MERGE of a VEC_DUPLICATE.
572 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
574         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
575         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
576         * doc/standards.texi (C Language): Document C2X.
577         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
578         (rl78_option_override): Handle "GNU C2X" language name.
580 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
582         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
583         Document C17 as published in 2018.
585 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
587         PR middle-end/87623
588         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
589         bail out if both sides do not have the same storage order.
591 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
593         * bitmap.c (bitmap_head::dump): New.
594         * bitmap.h (bitmap_head): Add dump().
595         * gimple-ssa-evrp-analyze.c
596         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
597         (evrp_range_analyzer::set_ssa_range_info): Same.
598         (evrp_range_analyzer::record_ranges_from_phis): Same.
599         (evrp_range_analyzer::record_ranges_from_stmt): Same.
600         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
601         * gimple-ssa-sprintf.c (get_int_range): Same.
602         (format_integer): Same.
603         (sprintf_dom_walker::handle_gimple_call): Same.
604         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
605         (ipcp_vr_lattice::top_p): Same.
606         (ipcp_vr_lattice::bottom_p): Same.
607         (ipcp_vr_lattice::set_to_bottom): Same.
608         (ipa_vr_operation_and_type_effects): Same.
609         (propagate_vr_across_jump_function): Same.
610         (ipcp_store_vr_results): Same.
611         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
612         (ipa_print_node_jump_functions_for_edge): Same.
613         (ipa_get_value_range): Same.
614         (ipa_compute_jump_functions_for_edge): Same.
615         (ipa_write_jump_function): Same.
616         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
617         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
618         Same.
619         * vr-values.c (set_value_range_to_nonnegative): Same.
620         (set_value_range_to_truthvalue): Same.
621         (vr_values::get_value_range): Same.
622         (vr_values::set_defs_to_varying): Same.
623         (vr_values::update_value_range): Same.
624         (symbolic_range_based_on_p): Same.
625         (vr_values::op_with_boolean_value_range_p): Same.
626         (vr_values::extract_range_for_var_from_comparison_expr): Same.
627         (vr_values::extract_range_from_ssa_name): Same.
628         (vr_values::extract_range_from_binary_expr): Same.
629         (vr_values::extract_range_from_unary_expr): Same.
630         (vr_values::extract_range_from_cond_expr): Same.
631         (vr_values::extract_range_from_comparison): Same.
632         (vr_values::check_for_binary_op_overflow): Same.
633         (vr_values::extract_range_basic): Same.
634         (vr_values::extract_range_from_assignment): Same.
635         (compare_ranges): Same.
636         (compare_range_with_value): Same.
637         (vr_values::adjust_range_with_scev): Same.
638         (vrp_valueize): Same.
639         (vrp_valueize_1): Same.
640         (vr_values::get_vr_for_comparison): Same.
641         (vr_values::compare_name_with_value): Same.
642         (vr_values::compare_names): Same.
643         (vr_values::vrp_evaluate_conditional): Same.
644         (find_case_label_ranges): Same.
645         (vr_values::vrp_visit_switch_stmt): Same.
646         (vr_values::extract_range_from_phi_node): Same.
647         (vr_values::simplify_div_or_mod_using_ranges): Same.
648         (vr_values::simplify_bit_ops_using_ranges): Same.
649         (test_for_singularity): Same.
650         (range_fits_type_p): Same.
651         (vr_values::simplify_cond_using_ranges_1): Same.
652         (vr_values::simplify_switch_using_ranges): Same.
653         (vr_values::simplify_float_conversion_using_ranges): Same.
654         (vr_values::two_valued_val_range_p): Same.
655         (vr_values::add_equivalence): Move to value_range::equiv_add.
656         * vr-values.h (vr_values::add_equivalence): Remove.
657         (VR_INITIALIZER): Remove.
658         * tree-vrp.c (value_range::set): New.
659         (value_range::equiv_add): New.
660         (value_range::value_range): New.
661         (value_range::deep_copy): New.
662         (value_range::check): New.
663         (value_range::equal_p): New.
664         (value_range::ignore_equivs_equal_p): New.
665         (value_range::operator==): New.
666         (value_range::operator!=): New.
667         (value_range::symbolic_p): New.
668         (value_range::numeric_p): New.
669         (value_range::set_undefined): New.
670         (value_range::set_varying): New.
671         (value_range::may_contain_p): New.
672         (value_range::equiv_clear): New.
673         (value_range::singleton_p): New.
674         (value_range::intersect): New.
675         (value_range::dump): New.
676         (value_range::set_and_canonicalize): New.
677         (set_value_range): Adjust for value_range API.
678         (set_value_range_to_undefined): Same.
679         (set_value_range_to_varying): Same.
680         (set_and_canonicalize_value_range): Same.
681         (set_value_range_to_nonnull): Same.
682         (set_value_range_to_null): Same.
683         (range_is_null): Same.
684         (range_is_nonnull): Same.
685         (range_int_cst_p): Same.
686         (range_int_cst_singleton_p): Same.
687         (symbolic_range_p): Same.
688         (range_includes_zero_p): Same.
689         (value_range_constant_singleton): Same.
690         (vrp_set_zero_nonzero_bits): Same.
691         (ranges_from_anti_range): Same.
692         (extract_range_into_wide_ints): Same.
693         (extract_range_from_multiplicative_op): Same.
694         (set_value_range_with_overflow): Same.
695         (extract_range_from_binary_expr_1): Same.
696         (extract_range_from_unary_expr): Same.
697         (dump_value_range): Same.
698         (debug_value_range): Same.
699         (vrp_prop::check_array_ref): Same.
700         (vrp_prop::check_mem_ref): Same.
701         (vrp_prop::vrp_initialize): Same.
702         (vrp_prop::visit_stmt): Same.
703         (intersect_ranges): Same.
704         (vrp_prop::visit_phi): Same.
705         (vrp_prop::vrp_finalize): Same.
706         (determine_value_range_1): Same.
707         (determine_value_range): Same.
708         (vrp_intersect_ranges_1): Rename to...
709         (vrp_intersect_1): this.
710         (vrp_intersect_ranges): Rename to...
711         (value_range::intersect_helper): ...this.
712         (vrp_meet_1): Rename to...
713         (value_range::union_helper): ...this.
714         (vrp_meet): Rename to...
715         (value_range::union_): ...this.
716         (copy_value_range): Remove.
717         * tree-vrp.h (struct value_range): Rewrite into a proper class.
718         (value_range::vrtype): New.
719         (value_range::type): New.
720         (value_range::equiv): New.
721         (value_range::min): New.
722         (value_range::max): New.
723         (value_range::varying_p): New.
724         (value_range::undefined_p): New.
725         (value_range::null_p): New.
726         (value_range::equiv_add): New.
727         (copy_value_range): Remove.
729 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
731         * Makefile.in (SELFTEST_TARGETS): New.
732         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
733         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
734         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
735         c/Make-lang.in.
736         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
737         (selftest-c++-gdb, selftest-c++-valgrind): Move to
738         cp/Make-lang.in.
739         * configure: Regenerate.
740         * configure.ac (selftest_languages): New.
742 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
744         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
745         overflow wraps argument.
746         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
747         wide_int_range_multiplicative_op.
748         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
749         overflow wraps argument.
750         (wide_int_range_multiplicative_op): Same.
751         (wide_int_range_lshift): Same.
752         (wide_int_range_div): Same.
753         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
754         (wide_int_range_lshift): Same.
755         (wide_int_range_div): Same.
757 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
759         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
760         use sign as argument.
761         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
762         wide_int_range_shift_undefined_p.
764 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
766         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
767         Rename to...
768         (@despeculate_copy<ALLI_TI:mode>): ... This.
769         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
770         switch statement.
772 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
774         * config.gcc: Obsolete *-*-solaris2.10*.
775         * doc/install.texi (Specific, *-*-solaris2*): Document it.
777 2018-10-12  Jeff Law  <law@redhat.com>
779         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
780         reg + sym +- const_int addressing modes.
782 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
784         * common.opt (fdiagnostics-minimum-margin-width=): New option.
785         * diagnostic-show-locus.c (layout::layout): Apply the minimum
786         margin width.
787         (layout::start_annotation_line): Only print up to 3 of the
788         margin character, to avoid touching the left-hand side.
789         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
790         minimum margin width, as set by test_diagnostic_context's ctor.
791         (selftest::test_fixit_insert_containing_newline): Likewise.
792         (selftest::test_fixit_insert_containing_newline_2): Likewise.
793         (selftest::test_line_numbers_multiline_range): Clear
794         dc.min_margin_width.
795         * diagnostic.c (diagnostic_initialize): Initialize
796         min_margin_width.
797         * diagnostic.h (struct diagnostic_context): Add field
798         "min_margin_width".
799         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
800         * opts.c (common_handle_option): Handle
801         OPT_fdiagnostics_minimum_margin_width_.
802         * selftest-diagnostic.c
803         (selftest::test_diagnostic_context::test_diagnostic_context):
804         Initialize min_margin_width to 6.
805         * toplev.c (general_init): Initialize global_dc->min_margin_width.
807 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
809         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
810         Fix usage of "error_at_rich_loc" in the comment.
812 2018-10-15  Renlin Li  <renlin.li@arm.com>
814         PR target/87563
815         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
816         if-conversioned loop when it contains ifn with types not
817         supported by backend.
818         * internal-fn.c (expand_direct_optab_fn): Add an assert.
819         (direct_internal_fn_supported_p): New helper function.
820         * internal-fn.h (direct_internal_fn_supported_p): Declare.
822 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
824         PR target/87572
825         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
826         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
827         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
829 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
831         PR tree-optimization/87022
832         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
833         bits in dist vector rather than the first one.
835 2018-10-15  Richard Biener  <rguenther@suse.de>
837         PR middle-end/87610
838         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
839         (visit_loadstore): When a used restrict tag escaped verify that
840         the points-to solution of "other" pointers do not include
841         escaped.
842         (compute_dependence_clique): If a used restrict tag escaped
843         communicated that down to visit_loadstore.
845 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
847         * config/s390/s390.c (s390_expand_vec_init): Force vector element
848         into reg if it isn't a general operand.
850 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
852         PR target/87599
853         * config/i386/sse.md (*vec_dupv2di): Add register source to
854         movddup.
856 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
858         PR target/87572
859         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
860         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
861         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
862         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
864 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
866         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
867         (notice_args_size): Set it in the current trace if no insn that can
868         throw internally has been seen yet.
869         (connect_traces): When connecting args_size between traces, allow the
870         incoming values not to match if there is an insn setting it before the
871         first insn that can throw internally; in that case, force the creation
872         of a CFI note on this latter insn.
874 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
876         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
877         the base class, because using the injected-class-name was not clearly
878         specified until DR 176.
880 2018-10-12  Paul Koning  <ni1d@arrl.net>
882         * config/pdp11/pdp11.md (doloop_end): New expander.
883         (doloop_end_insn): renamed from "doloop_end".
884         (addqi3): New pattern.
885         (subqi3): New pattern.
886         * config/pdp11/predicates.md (incdec_operand): New predicate.
888 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
890         PR middle-end/81376
891         * real.c (format_helper::can_represent_integral_type_p): New function
892         * real.h (format_helper::can_represent_integral_type_p): Ditto.
893         * match.pd: New pattern.
895 2018-10-12  Alexandre Oliva <oliva@adacore.com>
897         * configure.ac: Introduce --enable-large-address-aware
898         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
899         * doc/install.texi: Document it.
900         * configure, config.in: Rebuilt.
901         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
902         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
903         (LINK_SPEC): Insert it.
904         * config/i386/mingw-264.h: Likewise.
906         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
908 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
910         PR rtl-optimization/87600
911         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
913 2018-10-12  Paul Koning  <ni1d@arrl.net>
915         * doc/md.texi (doloop_end): Document that the pattern code may
916         need to check operand mode.
918 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
920         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
921         to zero-extend between int and floating-point registers.
922         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
923         ldp into floating-point registers.  Add type and arch attributes.
924         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
925         Use f_loads for type attribute.
927 2018-10-11  Martin Sebor  <msebor@redhat.com>
929         * doc/extend.texi (attribute packed): Correct typos.
931 2018-10-11  Martin Sebor  <msebor@redhat.com>
933         * doc/extend.texi (attribute flatten): Mention interaction with
934         noinline.
936 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
938         PR target/87156
939         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
940         Set new_decl virtual flag to zero.
942 2018-10-11  Martin Sebor  <msebor@redhat.com>
944         PR middle-end/87593
945         * doc/extend.texi (attribute format_arg): Discuss using multiple
946         attributes on a single function.
948 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
950         PR tree-optimization/86829
951         * match.pd (sin (atan (x))): New simplification rules.
952         (cos (atan (x))): Likewise.
953         * real.c (build_sinatan_real): New function.
954         * real.h (build_sinatan_real): Prototype.
956 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
958         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
959         function.
960         (fold_mergeeo_helper): New helper function.
961         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
962         intrinsics.  Correct some whitespace indentation issues.
964 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
966         PR target/87511
967         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
968         Use HOST_WIDE_INT_1U for shift.
970 2018-10-11  Doug Rupp  <rupp@adacore.com>
971             Olivier Hainque  <hainque@adacore.com>
973         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
974         Pass --relax to the linker for RTPs.
975         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
977 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
978             Jan Hubicka  <jh@suse.cz>
979             Martin Jambor  <mjambor@suse.cz>
981         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
982         the same elements are repeated rather than printing all of them.
983         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
984         "repeated" elements.
985         * read-rtl-function.c (test_loading_repeat): New function.
986         (read_rtl_function_c_tests): Call test_loading_repeat.
987         * rtl-tests.c (test_dumping_repeat): New function.
988         (rtl_tests_c_tests): Call test_dumping_repeat.
990 2018-10-11  Richard Biener  <rguenther@suse.de>
992         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
993         bdver?_cost): Unify to ...
994         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
995         * config/i386/i386.c (processor_cost_table): Adjust.
997 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
999         PR middle-end/87574
1000         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1001         the thunk when expanding to GIMPLE.
1003 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1005         * varasm.c (mergeable_string_section): Don't try to move zero-length
1006         strings to the merge section.
1008 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
1010         PR target/87573
1011         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1013 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
1015         PR target/87550
1016         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1017         to special_args set.
1019 2018-10-10  Richard Biener  <rguenther@suse.de>
1021         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1022         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1023         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1024         and recursing and pattern terminating the recursion on SSE
1025         vector width using ix86_expand_reduc.
1026         (reduc_sminmax_scal_<mode>): Split into part reducing to half
1027         width and recursing and SSE2 vector variant doing the final
1028         reduction with ix86_expand_reduc.
1029         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1030         with terminating the recursion at AVX level, splitting that
1031         to SSE there.
1033 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
1035         * genmatch.c (error_cb): Rename to...
1036         (diagnostic_cb): ...this, converting int params to enums.
1037         (fatal_at): Update for renaming.
1038         (warning_at): Likewise.
1039         (main): Likewise.
1040         * input.c (selftest::ebcdic_execution_charset::apply):
1041         Update for renaming of...
1042         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1043         to...
1044         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1045         converting level and reason to enums.
1046         (class selftest::lexer_error_sink): Rename to...
1047         (class selftest::lexer_test_options): ...this, renaming field
1048         "m_errors" to "m_diagnostics".
1049         (selftest::lexer_test_options::apply): Update for renaming of...
1050         (selftest::lexer_test_options::on_error): ...this, renaming to...
1051         (selftest::lexer_test_options::on_diagnostic): ...this
1052         converting level and reason to enums.
1053         (selftest::test_lexer_string_locations_raw_string_unterminated):
1054         Update for renamings.
1055         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
1056         "reason".
1058 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
1060         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1061         * config/rs6000/pmmintrin.h: New file.
1063 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
1065         PR tree-optimization/86659
1066         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1068 2018-10-09  Richard Biener  <rguenther@suse.de>
1070         PR tree-optimization/63155
1071         * tree-ssa-structalias.c: Include tree-ssa.h.
1072         (get_constraint_for_ssa_var): For undefs return nothing_id.
1073         (find_func_aliases): Cleanup PHI handling.
1075 2018-10-09  Richard Biener  <rguenther@suse.de>
1077         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1078         replacements.
1080 2018-10-09  Martin Liska  <mliska@suse.cz>
1082         * asan.c (asan_emit_stack_protection): If a stack variable
1083         is located in a same file as current function, then emit
1084         line info into variable definition string.
1086 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1088         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1089         information.
1091 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1093         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1094         on the thunk.
1096 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1098         PR tree-optimization/86659
1099         * gimple-match.h (struct gimple_match_op): Add reverse field.
1100         (gimple_match_op::set_op): New overloaded method.
1101         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1102         the REF_REVERSE_STORAGE_ORDER flag on the value.
1103         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
1104         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
1105         
1106 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
1108         PR middle-end/63155
1109         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
1110         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
1112 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
1114         PR target/87517
1115         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
1116         Defined with __builtin_ia32_vfmaddsubpd512_mask.
1118 2018-10-08  Richard Biener  <rguenther@suse.de>
1120         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
1121         cost the same as AVX128 ones.
1123 2018-10-08  Paul Koning  <ni1d@arrl.net>
1125         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
1126         (expand_block_move): New function.
1127         * config/pdp11/pdp11.c (output_block_move): Remove.
1128         (expand_block_move): New function.
1129         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
1130         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
1131         (*movmemhi1): Remove.
1133 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
1135         * config/s390/2827.md: Increase latencies for some FP instructions.
1137 2018-10-08  Richard Biener  <rguenther@suse.de>
1139         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1140         Open a dump scope.
1141         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
1142         * tree-vectorizer.h (dump_stmt_cost): Adjust.
1143         (add_stmt_cost): Dump return value of the hook.
1145 2018-10-08  Richard Biener  <rguenther@suse.de>
1147         PR tree-optimization/63155
1148         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
1149         (ssa_propagation_engine::ssa_propagate): Remove redundant
1150         bitmap bit clearing.
1152 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
1154         PR rtl-optimization/86939
1155         PR rtl-optimization/87479
1156         * ira.h (non_conflicting_reg_copy_p): New prototype.
1157         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
1158         (make_hard_regno_dead): Don't add conflicts for register
1159         ignore_reg_for_conflicts.
1160         (make_object_dead): Likewise.
1161         (non_conflicting_reg_copy_p): New function.
1162         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
1163         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
1164         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
1165         (make_hard_regno_dead): Don't add conflicts for register
1166         ignore_reg_for_conflicts.  Remove special conflict handling of
1167         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
1168         check_pic_pseudo_p and update callers.
1169         (mark_pseudo_dead): Don't add conflicts for register
1170         ignore_reg_for_conflicts.
1171         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
1173 2018-10-05  Andrew Waterman  <andrew@sifive.com>
1174             Jim Wilson  <jimw@sifive.com>
1176         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
1177         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
1178         new pattern using HONOR_SNANS that emits one extra instruction.
1180 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1182         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
1183         patterns): Merge SI and DI patterns to a GPR pattern.
1184         (unnamed define_insn and define_split for record form of that): Merge
1185         to a single define_insn_and_split pattern.
1187 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
1189         PR c++/56856
1190         * input.c (expand_location_to_spelling_point): Add param "aspect"
1191         and use rather than hardcoding LOCATION_ASPECT_CARET.
1192         (get_substring_ranges_for_loc): Handle the case of a single token
1193         within a macro expansion.
1194         * input.h (expand_location_to_spelling_point): Add "aspect" param,
1195         defaulting to LOCATION_ASPECT_CARET.
1197 2018-10-05  Paul Koning  <ni1d@arrl.net>
1199         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
1200         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
1201         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1202         (pdp11_guard_type): New function.
1204 2018-10-05  Paul Koning  <ni1d@arrl.net>
1206         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
1207         * config/pdp11/pdp11.opt (mfloat32): Remove.
1208         (mfloat64): Remove.
1209         * doc/invoke.texi (pdp11 -mfloat32): Remove:
1210         (pdp11 -mfloat64): Remove.
1212 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1214         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
1215         (*cmp<mode>_cc_i387): Ditto.
1216         (*cmpu<mode>_cc_i387): Ditto.
1217         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1218         * config/i386/i386.c (ix86_expand_fp_compare): Remove
1219         "scratch" argument.
1220         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
1221         Emit x86_sahf_1 pattern.
1222         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
1223         (ix86_expand_carry_flag_compare): Ditto.
1225 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1227         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
1228         to reg_or_0_operand.  Add "C" constraint.
1229         (*cmpxf_cc_i387): Ditto.
1230         (*cmp<mode>_i387): Change operand 2 predicate
1231         to nonimm_or_0_operand.  Add "C" constraint.
1232         (*cmp<mode>_cc_i387): Ditto.
1233         (*cmp<mode>_0_i387): Remove insn pattern.
1234         (*cmp<mode>_0_cc_i387): Ditto.
1236 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1238         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
1239         * config/i386/predicates.md (nonimm_or_0_operand): Rename
1240         from vector_move_operand.  Update all uses.
1242 2018-10-05  Martin Sebor  <msebor@redhat.com>
1244         PR tree-optimization/87490
1245         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
1246         consistently.
1248 2018-10-05  Richard Biener  <rguenther@suse.de>
1250         PR tree-optimization/63155
1251         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
1252         vertical space in dumpfiles.
1253         * tree-ssa-propagate.h
1254         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
1255         * tree-ssa-propagate.c (cfg_blocks_back): New global.
1256         (ssa_edge_worklist_back): Likewise.
1257         (curr_order): Likewise.
1258         (cfg_blocks_get): Remove abstraction.
1259         (cfg_blocks_add): Likewise.
1260         (cfg_blocks_empty_p): Likewise.
1261         (add_ssa_edge): Add to current or next worklist based on
1262         RPO index.
1263         (add_control_edge): Likewise.
1264         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
1265         into ...
1266         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
1267         iteration from CFG and SSA edge worklist so we process
1268         everything in RPO order, prioritizing forward progress
1269         over iteration.
1270         (ssa_prop_init): Allocate new worklists, do not dump
1271         immediate uses.
1272         (ssa_prop_fini): Free new worklists.
1274 2018-10-05  Richard Biener  <rguenther@suse.de>
1276         * tree-core.h (tree_block::abstract_flag): Remove.
1277         (tree_block::block_num): Make full 32bits.
1278         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
1279         * tree.h (BLOCK_ABSTRACT): Remove.
1280         * dwarf2out.c (gen_lexical_block_die): Remove dead code
1281         resulting from BLOCK_ABSTRACT being always false.
1282         (gen_inlined_subroutine_die): Likewise.
1283         (gen_block_die): Likewise.
1284         * tree.c (block_ultimate_origin): Likewise.
1285         * tree-pretty-print.c (dump_block_node): Remove code dealing
1286         with BLOCK_ABSTRACT.
1287         * tree-ssa-live.c (dump_scope_block): Likewise.
1288         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
1289         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
1291 2018-10-05   Richard Biener  <rguenther@suse.de>
1293         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
1294         is asked for initialize mode to the component mode of the
1295         vector type.
1297 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
1299         PR target/87522
1300         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
1301         assembler for -mavx.
1302         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
1304 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1306         PR target/87509
1307         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
1308         RS6000_BTM_DFP.
1309         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
1310         to be DImode.  When using mffscrn, force the operand to a register.
1312 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
1314         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
1315         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
1316         X87MODEF mode iterator.
1317         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
1318         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
1319         X87MODEF mode iterator.
1321 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
1323         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
1324         -Wno-prio-ctor-dtor.
1326 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1328         * Makefile.in (OBJS): Add opt-problem.o.
1329         * dump-context.h: Include "selftest.h.
1330         (selftest::temp_dump_context): New forward decl.
1331         (class dump_context): Make friend of class
1332         selftest::temp_dump_context.
1333         (dump_context::dump_loc_immediate): New decl.
1334         (class dump_pretty_printer): Move here from dumpfile.c.
1335         (class temp_dump_context): Move to namespace selftest.
1336         (temp_dump_context::temp_dump_context): Add param
1337         "forcibly_enable_dumping".
1338         (selftest::verify_dumped_text):
1339         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
1340         (selftest::verify_item):
1341         (ASSERT_IS_TEXT): Move here from dumpfile.c.
1342         (ASSERT_IS_TREE): Likewise.
1343         (ASSERT_IS_GIMPLE): Likewise.
1344         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
1345         to...
1346         (dump_context::dump_loc_immediate): ...this new function.
1347         (class dump_pretty_printer): Move to dump-context.h.
1348         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
1349         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
1350         (temp_dump_context::temp_dump_context): Move to "selftest"
1351         namespace.  Add param "forcibly_enable_dumping", and use it to
1352         conditionalize the use of m_pp;
1353         (selftest::verify_dumped_text): Make non-static.
1354         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
1355         (selftest::verify_item): Make non-static.
1356         (ASSERT_IS_TEXT): Move to dump-context.h.
1357         (ASSERT_IS_TREE): Likewise.
1358         (ASSERT_IS_GIMPLE): Likewise.
1359         (selftest::test_capture_of_dump_calls): Pass "true" for new
1360         param of temp_dump_context.
1361         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
1362         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
1363         TDF_COMPARE_DEBUG.
1364         * opt-problem.cc: New file.
1365         * opt-problem.h: New file.
1366         * optinfo-emit-json.cc
1367         (selftest::test_building_json_from_dump_calls): Pass "true" for
1368         new param of temp_dump_context.
1369         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
1370         (optinfo::emit_for_opt_problem): New function.
1371         (optinfo::emit): Clarity which emit_item is used.
1372         * optinfo.h (optinfo::get_dump_location): New accessor.
1373         (optinfo::emit_for_opt_problem): New decl.
1374         (optinfo::emit): Make const.
1375         * selftest-run-tests.c (selftest::run_tests): Call
1376         selftest::opt_problem_cc_tests.
1377         * selftest.h (selftest::opt_problem_cc_tests): New decl.
1378         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
1379         bool to opt_result, converting fprintf messages to
1380         opt_result::failure_at calls.  Add "stmt" param for use by the
1381         failure_at calls.
1382         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
1383         (runtime_alias_check_p): Convert return type from bool to
1384         opt_result, converting dump_printf calls to
1385         opt_result::failure_at, using the statement DDR_A for their
1386         location.
1387         (find_data_references_in_stmt): Convert return type from bool to
1388         opt_result, converting "return false" to opt_result::failure_at
1389         with a new message.
1390         * tree-data-ref.h: Include "opt-problem.h".
1391         (dr_analyze_innermost): Convert return type from bool to opt_result,
1392         and add a const gimple * param.
1393         (find_data_references_in_stmt): Convert return type from bool to
1394         opt_result.
1395         (runtime_alias_check_p): Likewise.
1396         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
1397         dr_analyze_innermost.
1398         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
1399         Convert return type from bool to opt_result, adding a message for
1400         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
1401         (vect_analyze_data_ref_dependence): Convert return type from bool
1402         to opt_result.  Change sense of return type from "false"
1403         effectively meaning "no problems" to "false" meaning a problem,
1404         so that "return false" becomes "return opt_result::success".
1405         Convert "return true" calls to opt_result::failure_at, using
1406         the location of statement A rather than vect_location.
1407         (vect_analyze_data_ref_dependences): Convert return type from bool
1408         to opt_result.
1409         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
1410         calls to opt_result::failure_at, using the stmt location rather
1411         than vect_location.
1412         (vect_verify_datarefs_alignment): Convert return type from bool
1413         to opt_result.
1414         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
1415         into multiple more-tightly-scoped copies.
1416         (vect_analyze_data_refs_alignment): Convert return type from bool
1417         to opt_result.
1418         (vect_analyze_data_ref_accesses): Likewise, converting a
1419         "return false" to a "return opt_result::failure_at", adding a
1420         new message.
1421         (vect_prune_runtime_alias_test_list): Convert return type from
1422         bool to opt_result, converting dump_printf_loc to
1423         opt_result::failure_at.  Add a %G to show the pertinent statement,
1424         and use the stmt's location rather than vect_location.
1425         (vect_find_stmt_data_reference): Convert return type from
1426         bool to opt_result, converting dump_printf_loc to
1427         opt_result::failure_at, using stmt's location.
1428         (vect_analyze_data_refs):  Convert return type from bool to
1429         opt_result.  Convert "return false" to "return
1430         opt_result::failure_at", adding messages as needed.
1431         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
1432         type from bool to opt_result.
1433         (vect_determine_vf_for_stmt): Likewise.
1434         (vect_determine_vectorization_factor): Likewise, converting
1435         dump_printf_loc to opt_result::failure_at, using location of phi
1436         rather than vect_location.
1437         (vect_analyze_loop_form_1): Convert return type from bool to
1438         opt_result, converting dump_printf_loc calls, retaining the use of
1439         vect_location.
1440         (vect_analyze_loop_form): Convert return type from loop_vec_info
1441         to opt_loop_vec_info.
1442         (vect_analyze_loop_operations): Convert return type from bool to
1443         opt_result, converting dump_printf_loc calls, using the location
1444         of phi/stmt rather than vect_location where available.  Convert
1445         various "return false" to "return opt_result::failure_at" with
1446         "unsupported phi" messages.
1447         (vect_get_datarefs_in_loop): Convert return type from bool to
1448         opt_result.  Add a message for the
1449         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
1450         (vect_analyze_loop_2): Convert return type from bool to
1451         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
1452         each "goto again;", adding new messages where needed.
1453         Add "unsupported grouped {store|load}" messages.
1454         (vect_analyze_loop): Convert return type from loop_vec_info to
1455         opt_loop_vec_info.
1456         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
1457         bool to opt_result.
1458         * tree-vect-stmts.c (process_use): Likewise, converting
1459         dump_printf_loc call and using stmt location, rather than
1460         vect_location.
1461         (vect_mark_stmts_to_be_vectorized): Likeise.
1462         (vect_analyze_stmt): Likewise, adding a %G.
1463         (vect_get_vector_types_for_stmt): Convert return type from bool to
1464         opt_result, converting dump_printf_loc calls and using stmt
1465         location, rather than vect_location.
1466         (vect_get_mask_type_for_stmt): Convert return type from tree to
1467         opt_tree, converting dump_printf_loc calls and using stmt location.
1468         * tree-vectorizer.c: Include "opt-problem.h.
1469         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
1470         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
1471         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
1472         enabled, use it to report at the top level "couldn't vectorize
1473         loop" followed by the problem.
1474         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
1475         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
1476         to opt_result.
1477         (vect_analyze_stmt): Likewise.
1478         (vect_get_vector_types_for_stmt): Likewise.
1479         (tree vect_get_mask_type_for_stmt): Likewise.
1480         (vect_analyze_data_ref_dependences): Likewise.
1481         (vect_enhance_data_refs_alignment): Likewise.
1482         (vect_analyze_data_refs_alignment): Likewise.
1483         (vect_verify_datarefs_alignment): Likewise.
1484         (vect_analyze_data_ref_accesses): Likewise.
1485         (vect_prune_runtime_alias_test_list): Likewise.
1486         (vect_find_stmt_data_reference): Likewise.
1487         (vect_analyze_data_refs): Likewise.
1488         (vect_analyze_loop): Convert return type from loop_vec_info to
1489         opt_loop_vec_info.
1490         (vect_analyze_loop_form): Likewise.
1491         (vect_analyze_slp): Convert return type from bool to opt_result.
1493 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1495         * doc/invoke.texi (-fopt-info): Document new "internals"
1496         sub-option.
1497         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
1498         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
1499         MSG_ALL_KINDS.
1500         (optinfo_verbosity_options): Add "internals".
1501         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
1502         (dump_context::apply_dump_filter_p): New member function.
1503         (dump_context::dump_loc): Use apply_dump_filter_p rather than
1504         explicitly masking the dump_kind.
1505         (dump_context::begin_scope): Increment the scope depth first.  Use
1506         apply_dump_filter_p rather than explicitly masking the dump_kind.
1507         (dump_context::emit_item): Use apply_dump_filter_p rather than
1508         explicitly masking the dump_kind.
1509         (dump_dec): Likewise.
1510         (dump_hex): Likewise.
1511         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
1512         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
1513         (opt_info_switch_p): Update handling of default
1514         MSG_OPTIMIZED_LOCATIONS to cope with default of
1515         MSG_PRIORITY_USER_FACING.
1516         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
1517         masking the dump_kind.
1518         (selftest::test_capture_of_dump_calls): Update test_dump_context
1519         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
1520         than MSG_ALL.  Generalize scope test to be run at all four
1521         combinations of with/without MSG_PRIORITY_USER_FACING and
1522         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
1523         for each of the two values.
1524         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
1525         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
1526         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
1527         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
1528         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
1529         with MSG_PRIORITY_*.
1530         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
1531         dump messages as MSG_PRIORITY_USER_FACING.
1532         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
1533         about the interaction with MSG_PRIORITY_*.
1535 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1537         * varasm.c (output_constant): Add new parameter merge_strings.
1538         Make strings properly zero terminated in merge string sections.
1539         (mergeable_string_section): Don't fail if the last char is non-zero.
1540         (assemble_variable_contents): Handle merge string sections.
1541         (assemble_variable): Likewise.
1542         (assemble_constant_contents): Likewise.
1543         (output_constant_def_contents): Likewise.
1544         (output_constructor_array_range,
1545         output_constructor_regular_field): Adjust call to output_constant.
1546         (output_object_block): Adjust call to assemble_constant_contents
1547         and assemble_variable_contents.
1549 2018-10-04  Martin Liska  <mliska@suse.cz>
1551         PR c/87483
1552         * cgraphunit.c (process_function_and_variable_attributes):
1553         Warn about a function with alias attribute and a body.
1555 2018-10-04  Martin Liska  <mliska@suse.cz>
1557         PR ipa/82625
1558         * multiple_target.c (redirect_to_specific_clone): New function.
1559         (ipa_target_clone): Use it.
1560         * tree-inline.c: Fix comment.
1562 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
1564         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
1565         fields.
1566         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
1567         (gcc::dump_manager::register_pass): New member function, adapted
1568         from loop body in gcc::pass_manager::register_pass, adding a
1569         call to update_dfi_for_opt_info.
1570         (gcc::dump_manager::opt_info_enable_passes): Store the
1571         -fopt-info options into the new fields.  Move the loop
1572         bodies into...
1573         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
1574         function.
1575         * dumpfile.h (struct opt_pass): New forward decl.
1576         (gcc::dump_manager::register_pass): New decl.
1577         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
1578         (class gcc::dump_manager): Add fields "m_optgroup_flags",
1579         "m_optinfo_flags", and "m_optinfo_filename".
1580         * passes.c (gcc::pass_manager::register_pass): Move all of the
1581         dump-handling code to gcc::dump_manager::register_pass.
1583 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
1585         PR rtl-optimization/87466
1586         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
1587         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
1588         * doc/tm.texi: Regenerate.
1589         * ira-lives.c (process_bb_node_lives): Use the new target hook.
1590         * lra-lives.c (process_bb_lives): Likewise.
1591         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
1592         Define.
1594 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
1596         * params.c (add_params): Fix initialization.
1598 2018-10-04  Martin Liska  <mliska@suse.cz>
1600         PR gcov-profile/84107
1601         * tree-profile.c (init_ic_make_global_vars):
1602         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
1603         Come up with new ic_tuple* variables.  Emit
1604         __gcov_indirect_call{,_topn} variables.
1605         (gimple_gen_ic_profiler): Access the variable
1606         and emit gimple.
1607         (gimple_gen_ic_func_profiler): Access
1608         __gcov_indirect_call.callee field.
1609         (gimple_init_gcov_profiler): Use ptr_type_node.
1610         * value-prof.c (gimple_ic): Use ptr_type_node.
1612 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1614         PR tree-optimization/85787
1615         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
1616         into this function and add support for detecting multiple phis.
1617         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
1619 2018-10-04  Martin Liska  <mliska@suse.cz>
1621         PR ipa/87491
1622         * ipa-inline.c (inline_to_all_callers_1):
1623         Call ultimate_alias_target for node being inlined.
1625 2018-10-03  Jeff Law  <law@redhat.com>
1627         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
1628         target's wchar_t.
1629         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
1630         * tree.h (get_typenode_from_name): Prototype.
1632 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
1634         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
1635         Change operand 2 predicate to nonimmediate_operand.
1636         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1638 2018-10-03  Martin Sebor  <msebor@redhat.com>
1639             Jeff Law  <law@redhat.com>
1641         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
1642         initialize it.
1643         (get_string_length): Detect unterminated arrays.
1644         (format_string): Same.
1645         (format_directive): Warn about unterminated arrays.
1646         (handle_gimple_call): Mark statements with no_warning as needed.
1648 2018-10-03  Jim Wilson  <jimw@sifive.com>
1650         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
1651         also define __riscv_abi_rve.  Delete trailing white space.
1653 2018-10-03  Paul Koning  <ni1d@arrl.net>
1655         Enable LRA register allocator for PDP11.
1656         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
1657         (R): Likewise.
1658         (D): Likewise.
1659         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
1660         * config/pdp11/pdp11.opt (-mlra): New option.
1661         * doc/invoke.texi (PDP-11 Options): Document -mlra.
1663 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
1665         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
1666         (*<absneg:code>extend<mode>xf2): Ditto.
1668 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
1670         PR tree-optimization/87415
1671         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
1672         precision fields.
1674 2018-10-02  Jeff Law  <law@redhat.com>
1676         * gimple-fold.c (get_range_strlen): Only set *nonstr when
1677         an unterminated string is discovered.  Bubble up range
1678         even for unterminated strings.
1679         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
1680         indicates the string was not terminated via NONSTR.
1682 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
1684         * tree-vrp.c (extract_range_from_unary_expr): Special case all
1685         pointer conversions.
1686         Do not do anything special for anti-ranges.
1688 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
1690         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
1691         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
1693 2018-10-03  Martin Liska  <mliska@suse.cz>
1695         PR gcov-profile/86109
1696         * coverage.c (coverage_begin_function): Do not
1697         mark lambdas as artificial.
1698         * tree-core.h (struct GTY): Remove tm_clone_flag
1699         and introduce new lambda_function.
1700         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
1702 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
1704         PR target/87474
1705         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
1706         P8_VECTOR and VSX are enabled.
1708 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
1710         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
1711         0x3907 as CPU model number.
1713 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
1715         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
1716         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
1717         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
1718         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
1719         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
1720         * config/s390/s390.md: Likewise. Rename also the cpu attribute
1721         value from arch12 to z14.
1723 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
1725         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
1726         (isinfxf2): Ditto.
1727         (isinf<mode>2): Ditto.
1729 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
1731         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
1732         before emitting fxam.  Perform calculations in XFmode.
1734 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
1736         * match.pd (((X /[ex] A) +- B) * A): New transformation.
1738 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
1740         PR middle-end/87319
1741         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
1742         * tree.c (signed_or_unsigned_type_for): Handle complex.
1744 2018-10-02  Jeff Law  <law@redhat.com>
1746         * gimple-fold.c (get_range_strlen): Remove dead code.
1748 2018-10-02  Martin Sebor  <msebor@redhat.com>
1749             Jeff Law  <law@redhat.com>
1751         * builtins.c (unterminated_array): Add new arguments.
1752         If argument is not terminated, bubble up size and exact
1753         state to callers.
1754         (expand_builtin_strnlen): Detect, avoid expanding
1755         and diagnose unterminated arrays.
1756         (c_strlen): Fill in offset of start of unterminated strings.
1757         * builtins.h (unterminated_array): Update prototype.
1759 2018-10-02  Richard Biener  <rguenther@suse.de>
1761         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
1762         of haddv4df, first reduce to SSE width and exploit the fact
1763         that we only need element zero with the reduction result.
1764         (reduc_plus_scal_v2df): Likewise.
1766 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
1768         * dojump.h (do_jump): Delete.
1769         (do_jump_1): Likewise.
1770         (split_comparison): Move around.
1771         * dojump.c (do_jump): Make static.
1772         (do_jump_1): Likewise.
1773         (jumpifnot): Move around.
1774         (jumpifnot_1): Likewise.
1775         (jumpif): Likewise.
1776         (jumpif_1): Likewise.
1777         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
1779 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
1781         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
1782         insns in the delay slot and add_insn_after for the jump insn.
1784 2018-10-02  Richard Biener  <rguenther@suse.de>
1786         * tree-inline.c (expand_call_inline): Use the location of
1787         the callee declaration for the inline-entry marker.
1788         * final.c (notice_source_line): Remove special-casing of
1789         NOTE_INSN_INLINE_ENTRY.
1791 2018-10-01  Carl Love  <cel@us.ibm.com>
1793         PR 69431
1794         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
1795         (__builtin_mtfsb0): New.
1796         (__builtin_mtfsb1): New.
1797         ( __builtin_set_fpscr_rn): New.
1798         (__builtin_set_fpscr_drn): New.
1799         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
1800         (rs6000_expand_set_fpscr_rn_builtin): Add.
1801         (rs6000_expand_set_fpscr_drn_builtin): Add.
1802         (rs6000_expand_builtin): Add case statement entries for
1803         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
1804         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
1805         RS6000_BUILTIN_MFFSL.
1806         (rs6000_init_builtins): Add ftype initialization and def_builtin
1807         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1808         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1809         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1810         rs6000_mffscdrn): Add define_insn.
1811         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1812         * doc/extend.texi: Add documentation for the builtins.
1814 2018-10-01  Richard Biener  <rguenther@suse.de>
1816         PR tree-optimization/87465
1817         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1818         causing branch miscounts.
1820 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1822         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1823         aarch64_option_default_param):  New.
1824         (params.h): Include.
1825         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1826         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1827         stack-clash protection validation code.
1829 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1831         * params.c (validate_param): New.
1832         (add_params): Use it.
1833         (set_param_value): Refactor param validation into validate_param.
1834         (diagnostic.h): Include.
1835         * diagnostic.h (diagnostic_ready_p): New.
1837 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1839         * params.c (set_param_value):
1840         Add index of parameter being validated.
1841         * common/common-target.def (option_validate_param): New.
1842         * common/common-targhooks.h (default_option_validate_param): New.
1843         * common/common-targhooks.c (default_option_validate_param): New.
1844         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1845         * doc/tm.texi: Regenerate.
1847 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1849         PR target/86486
1850         * config/aarch64/aarch64.c (aarch64_override_options_internal):
1851         Add validation for stack-clash parameters and set defaults.
1853 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1855         PR target/86486
1856         * configure.ac: Add stack-clash-protection-guard-size.
1857         * doc/install.texi: Document it.
1858         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1859         * params.def: Update comment for guard-size.
1860         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1861         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1862         * configure: Regenerate.
1864 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1866         PR target/86486
1867         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1868         STACK_DYNAMIC_OFFSET): New.
1869         * config/aarch64/aarch64.c (aarch64_layout_frame):
1870         Update outgoing args size.
1871         (aarch64_stack_clash_protection_alloca_probe_range,
1872         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1874 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1876         PR target/86486
1877         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1878         probe ranges.
1879         * target.def (stack_clash_protection_alloca_probe_range): New.
1880         (stack_clash_protection_final_dynamic_probe): Remove.
1881         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1882         (default_stack_clash_protection_final_dynamic_probe): Remove.
1883         * targhooks.c: Likewise.
1884         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1885         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1886         * doc/tm.texi: Regenerate.
1888 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1890         PR target/86486
1891         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1892         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
1893         aarch64_clamp_to_uimm12_shift): New.
1894         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
1895         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
1897 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
1899         PR target/86486
1900         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
1902 2018-10-01  Jeff Law  <law@redhat.com>
1903             Richard Sandiford <richard.sandiford@linaro.org>
1904             Tamar Christina  <tamar.christina@arm.com>
1906         PR target/86486
1907         * config/aarch64/aarch64.md
1908         (probe_stack_range): Add k (SP) constraint.
1909         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
1910         STACK_CLASH_MAX_UNROLL_PAGES): New.
1911         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
1912         stack probes for stack clash.
1913         (aarch64_allocate_and_probe_stack_space): New.
1914         (aarch64_expand_prologue): Use it.
1915         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
1916         (aarch64_sub_sp): Add emit_move_imm optional param.
1918 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
1920         PR tree-optimization/87261
1921         * match.pd: Remove trailing whitespace.
1922         Add (x & y) | ~(x | y) -> ~(x ^ y),
1923         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
1925 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1927         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1928         constraints.
1929         (ashlsi3_insn): Update instruction constraints.
1930         (ashrsi3_insn): Likewise.
1931         (rotrsi3): Likewise.
1932         (add_shift): Likewise.
1933         * config/arc/constraints.md (Csz): New 32 bit constraint. It
1934         avoids placing in the limm field small constants which, otherwise,
1935         could end into a small instruction.
1937 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
1939         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1940         destination register is not odd-even.
1941         (umaddsidi4_split): Likewise.
1943 2018-10-01  Richard Biener  <rguenther@suse.de>
1945         * tree-inline.c (expand_call_inline): Store origin of fn
1946         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1947         * tree.c (block_ultimate_origin): Simplify and do some
1948         checking.
1950 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
1952         * config/i386/mmx.md (EMMS): New int iterator.
1953         (emms): New int attribute.
1954         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1955         EMMS int iterator.  Explicitly declare clobbers.
1956         (mmx_emms): Remove expander.
1957         (mmx_femms): Ditto.
1958         * config/i386/predicates.md (emms_operation): Remove predicate.
1959         (vzeroall_pattern): New predicate.
1960         (vzeroupper_pattern): Rename from vzeroupper_operation.
1961         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1962         vzeroupper_pattern and vzeroall_pattern predicates.
1964 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
1966         PR rtl-optimization/86939
1967         * ira-lives.c (make_hard_regno_born): Rename from this...
1968         (make_hard_regno_live): ... to this.  Remove update to conflict
1969         information.  Update function comment.
1970         (make_hard_regno_dead): Add conflict information update.  Update
1971         function comment.
1972         (make_object_born): Rename from this...
1973         (make_object_live): ... to this.  Remove update to conflict information.
1974         Update function comment.
1975         (make_object_dead):  Add conflict information update.  Update function
1976         comment.
1977         (mark_pseudo_regno_live): Call make_object_live.
1978         (mark_pseudo_regno_subword_live): Likewise.
1979         (mark_hard_reg_dead): Update function comment.
1980         (mark_hard_reg_live): Call make_hard_regno_live.
1981         (process_bb_node_lives): Likewise.
1982         * lra-lives.c (make_hard_regno_born): Rename from this...
1983         (make_hard_regno_live): ... to this.  Remove update to conflict
1984         information.  Remove now uneeded check_pic_pseudo_p argument.
1985         Update function comment.
1986         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1987         to conflict information.  Update function comment.
1988         (mark_pseudo_live): Remove update to conflict information.  Update
1989         function comment.
1990         (mark_pseudo_dead): Add conflict information update.
1991         (mark_regno_live): Call make_hard_regno_live.
1992         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1993         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1995 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
1997         PR target/87370
1998         * config/i386/i386.c (construct_container): Use TImode for
1999         BLKmode values in 2 integer registers.
2001 2018-09-29  Jeff Law  <law@redhat.com>
2003         * builtins.c (unterminated_array): Pass in c_strlen_data * to
2004         c_strlen rather than just a tree *.
2005         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2006         Update recursive calls appropriately.  If caller did not provide a
2007         suitable data pointer, create a local one.  When a non-terminated
2008         string is discovered, bubble up information about the string via the
2009         c_strlen_data object.
2010         * builtins.h (c_strlen): Update prototype.
2011         (c_strlen_data): New structure.
2012         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2013         For a type 2 call, if c_strlen indicates a non-terminated string
2014         use the length of the non-terminated string.
2015         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2017 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
2019         PR target/87467
2020         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2021         __m512d type for __A argument rather than __m512.
2023 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
2025         * match.pd (simple_comparison): Don't optimize if either operand is
2026         a function pointer when target needs function pointer canonicalization.
2028 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2030         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2031         power5 .. power9 to remove indirection.
2032         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2033         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2034         ASM_CPU_476_SPEC): Delete.
2035         (ASM_CPU_SPEC): Adjust.
2036         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2037         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2039 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2041         * config.in: Delete HAVE_AS_DCI.
2042         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2043         * config/rs6000/rs6000.h: Ditto.
2044         * configure.ac: Delete HAVE_AS_DCI.
2045         * configure: Regenerate.
2047 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2049         * config.in (HAVE_AS_LWSYNC): Delete.
2050         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2051         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
2052         do it as a .long .
2053         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2054         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
2055         as a .long .
2056         * configure.ac: Delete HAVE_AS_LWSYNC.
2057         * configure: Regenerate.
2059 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2060             Pierre-Marie de Rodat  <derodat@adacore.com>
2062         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2063         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2064         (cgraph_node::create_thunk): Add indirect_offset parameter.
2065         (thunk_adjust): Likewise.
2066         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2067         and initialize the corresponding field with it.
2068         (cgraph_node::dump): Dump indirect_offset field.
2069         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2070         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2071         (thunk_adjust): Add indirect_offset parameter and deal with it.
2072         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2073         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
2074         or if the thunk is external or local.  Fix formatting.  Do not chain
2075         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
2076         if any, in the GIMPLE representation.
2077         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2078         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2079         (input_node): Read indirect_offset field.
2080         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2081         call to thunk_adjust.
2082         * tree-nested.c (struct nesting_info): Add thunk_p field.
2083         (create_nesting_tree): Set it.
2084         (convert_all_function_calls): Copy static chain from targets to thunks.
2085         (finalize_nesting_tree_1): Return early for thunks.
2086         (unnest_nesting_tree_1): Do not finalize thunks.
2087         (gimplify_all_functions): Do not gimplify thunks.
2089 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2091         * opt-suggestions.c (option_proposer::build_option_suggestions):
2092         Release "option_values".
2094 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2096         * coverage.c (get_coverage_counts): Convert problem-reporting dump
2097         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2098         * dumpfile.c (kind_as_string): New function.
2099         (dump_loc): Rather than a hardcoded prefix of "note: ", use
2100         kind_as_string to vary the prefix based on dump_kind.
2101         (selftest::test_capture_of_dump_calls): Update for above.
2103 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2105         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2106         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
2108 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2110         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
2111         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
2112         INVALID_REGNUM instead of FPSR_REG.
2113         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
2114         * config/i386/i386.md: Update comment of FP compares.
2115         (fldenv): Do not clobber FPSR_REG.
2117 2018-09-28  Richard Biener  <rguenther@suse.de>
2119         * tree.h (BLOCK_ORIGIN): New.
2120         * omp-expand.c (grid_expand_target_grid_body): Assign
2121         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
2122         * tree-inline.c (remap_block): Likewise.
2123         * auto-profile.c (get_function_decl_from_block): Simplify
2124         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
2125         * langhooks.c (lhd_print_error_function): Likewise.
2126         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
2127         Likewise.
2128         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2129         * tree.c (block_nonartificial_location): Likewise.
2130         (block_ultimate_origin): Likewise.
2131         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
2132         no longer needed LTO case.
2134 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
2135             Jan Hubicka  <jh@suse.cz>
2136             Martin Jambor  <mjambor@suse.cz>
2138         * simplify-rtx.c (simplify_merge_mask): New function.
2139         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
2140         same masks are used in op1 or op2.
2141         (test_vec_merge): New function.
2142         (test_vector_ops): Call test_vec_merge.
2144 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2146         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
2147         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
2148         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
2149         (cypress_costs): Set it.
2150         (supersparc_costs): Likewise.
2151         (hypersparc_costs): Likewise.
2152         (leon_cost): Likewise.
2153         (leon3_costs): Likewise.
2154         (sparclet_costs): Likewise.
2155         (ultrasparc_costs): Likewise.
2156         (ultrasparc_costs): Likewise.
2157         (niagara_costs): Likewise.
2158         (niagara2_costs): Likewise.
2159         (niagara3_costs): Likewise.
2160         (niagara4_costs): Likewise.
2161         (niagara7_costs): Likewise.
2162         (m8_costs): Likewise.
2163         (TARGET_CAN_FOLLOW_JUMP): Define.
2164         (pass_work_around_errata::gate): Minor tweak.
2165         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
2166         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
2167         Automaitcally clear MASK_FSMULD mask for V7 processors.
2168         (sparc_can_follow_jump): New static function.
2169         (output_ubranch): Deal with CROSSING_JUMP_P.
2170         (sparc_use_sched_lookahead): Rewrite using switch statement.
2171         (sparc_issue_rate): Reorder.
2172         (sparc_branch_cost): New function.
2174 2018-09-27  Martin Sebor  <msebor@redhat.com>
2176         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
2177         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
2178         (int_fits_type_p): Same.
2180 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
2182         * config/i386/i386.md (FPCR_REG): Remove.
2183         (UNSPEC_FLDCW): Remove.
2184         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
2185         (x86_fldcw_1): Remove insn pattern.
2186         (fnstenv): Do not clobber FPCR_REG.
2187         (fldenv): Ditto.
2188         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
2189         (CALL_USED_REGISTERS): Ditto.
2190         (REG_ALLOC_ORDER): Ditto.
2191         (REG_CLASS_CONTENTS): Ditto.
2192         (HI_REGISTER_NAMES): Ditto.
2193         (ADDITIONAL_REGISTER_NAMES): Use defines instead
2194         of numerical constants.
2195         * config/i386/i386.c (regclass_map): Remove fpsr register.
2196         (dbx_register_map): Ditto.
2197         (dbx64_register_map): Ditto.
2198         (svr4_dbx_register_map): Ditto.
2199         (print_reg): Do not handle FPCR_REG.
2201 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
2203         PR target/87149
2204         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
2205         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
2206         Delete, always treat as true.
2207         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2208         Ditto.  Simplify remaining code.
2209         * config/powerpcspe/powerpcspe.h: Ditto.
2210         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
2211         Simplify remaining code.
2212         (rs6000_expand_builtin): Ditto.
2213         * config/rs6000/rs6000.h: Ditto.
2214         * configure.ac: Ditto.
2215         * configure: Regenerate.
2217 2018-09-27  Martin Liska  <mliska@suse.cz>
2219         * coverage.c (get_coverage_counts): Revert the formatting
2220         of missing profile opt info.
2222 2018-09-27  Richard Biener  <rguenther@suse.de>
2224         PR debug/37801
2225         PR debug/87440
2226         * dwarf2out.c (set_block_origin_self): Do not mark outermost
2227         block as we do not output that.
2228         (gen_inlined_subroutine_die): Elide the originally outermost
2229         block, matching what we do for concrete instances.
2230         (decls_for_scope): Add parameter specifying whether to recurse
2231         to subblocks.
2233 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
2234             Tom de Vries  <tom@codesourcery.com>
2236         PR 82089
2238         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
2239         STORE_FLAG_VALUE == 1.
2241 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
2243         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
2244         constant definitions.
2245         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
2246         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
2247         ("speculation_barrier"): New expander definition.
2249 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
2251         PR gcov-profile/86957
2252         * common.opt: New warning option -Wmissing-profile.
2253         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
2254         * doc/invoke.texi: Document -Wmissing-profile.
2256 2018-09-26  Jim Wilson  <jimw@sifive.com>
2258         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
2259         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
2261 2018-09-26  Martin Sebor  <msebor@redhat.com>
2263         * tree.c (zerop): Change return type to bool.
2264         (integer_zerop, integer_onep, integer_each_onep): Same.
2265         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2266         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2267         (real_onep, real_minus_onep, chain_index): Same.
2268         (print_type_hash_statistics, type_list_equal): Same.
2269         * tree.h (zerop): Same.
2270         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
2271         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2272         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2273         (real_onep, real_minus_onep, chain_index): Same.
2274         (print_type_hash_statistics, type_list_equal): Same.
2276 2018-09-26  Jim Wilson  <jimw@sifive.com>
2278         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
2280 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
2282         PR target/87414
2283         * config/i386/i386.c: Include debug.h and dwarf2out.h.
2284         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2285         call.
2287 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
2289         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
2291 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2293         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
2294         and FP_SECOND_SSE_REGS.
2295         (REG_CLASS_NAMES): Ditto.
2296         (REG_CLASS_CONTENTS): Ditto.
2297         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
2298         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
2299         (ix86_preferred_output_reload_class): Ditto.
2300         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
2301         clobber constraint to "=&f".
2302         (fix_truncdi_i387): Ditto.
2303         (lrintxfdi2): Ditto.
2304         (fistdi2_<rounding>): Ditto.
2305         (fpremxf4_i387): Change "=u" constraint to "=f".
2306         (fprem1xf4_i387): Ditto.
2307         (sincosxf3): Ditto.
2308         (fptanxf4_i387): Ditto.
2309         (fxtractxf3_i387): Ditto.
2310         (fscalexf4_i387): Ditto.
2311         (atan2xf3): Change "u" constraint to "f".
2312         (fyl2xxf3_i387): Ditto.
2313         (fyl2xp1xf3_i387): Ditto.
2315 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2317         PR target/87439
2318         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
2319         for removed I387_MASK_PM entity.
2322 2018-09-26  Jeff Law  <law@redhat.com>
2323         Revert
2324         2018-09-26  Alexey Neyman  <stilor@att.net>
2326         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2327         headers are no longer pulled in by <isl/val.h>.
2329 2018-09-26  Richard Biener  <rguenther@suse.de>
2331         PR debug/87443
2332         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
2333         or concrete instance DIE to the tree.  Create abstract origin
2334         attributes also for concrete instances.
2336 2018-09-26  Alexey Neyman  <stilor@att.net>
2338         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2339         headers are no longer pulled in by <isl/val.h>.
2341 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
2343         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
2344         Use new helper functions.
2345         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
2346         Use new helper functions.
2347         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
2348         aarch_mm_needs_release): New declarations.
2349         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
2350         aarch_mm_needs_release): New.
2352 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
2354         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
2355         (arm32_output_mi_thunk): Deal with long calls.
2357 2018-09-26  Richard Biener  <rguenther@suse.de>
2359         PR debug/87428
2360         PR debug/87362
2361         * tree-inline.c (expand_call_inline): When the location
2362         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2363         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2364         the inserted BLOCK to make inlined_function_outer_scope_p
2365         recognize it.
2366         * dwarf2out.c (add_call_src_coords_attributes): Do not add
2367         coords for reserved locations.
2369 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2371         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
2372         (*call_indirect_nonlocal_sysv<mode>): Ditto.
2373         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
2374         (*sibcall_nonlocal_sysv<mode>): Ditto.
2375         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
2376         (<bd>_<mode>): Ditto.
2377         (<bd>tf_<mode>): Ditto.
2379 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2381         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
2382         control string as a list of templates instead of as C code.
2383         (*altivec_movti): Ditto.
2384         * config/rs6000/darwin.md (movdf_low_di): Ditto.
2386 2018-09-25  Jim Wilson  <jimw@sifive.com>
2388         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
2389         when target symbol is weak.
2391 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2393         PR c/87387
2394         * builtins.c (unterminated_array): Simplify.
2395         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
2396         where pointer arithmetic is safe.
2398 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
2400         PR target/86987
2401         * config/rs6000/altivec.md (altivec_vspltb): Use
2402         const_0_to_15_operand instead of u5bit_cint_operand.
2403         (*altivec_vspltb_internal): Ditto.
2404         (altivec_vspltb_direct): Ditto.
2405         (altivec_vsplth): Use const_0_to_7_operand instead of
2406         u5bit_cint_operand.
2407         (*altivec_vsplth_internal): Ditto.
2408         (altivec_vsplth_direct): Ditto.
2409         (altivec_vspltw): Use const_0_to_3_operand instead of
2410         u5bit_cint_operand.
2411         (*altivec_vspltw_internal): Ditto.
2412         (altivec_vspltw_direct): Ditto.
2413         (altivec_vspltsf): Ditto.
2414         (*altivec_vspltsf_internal): Ditto.
2415         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
2416         various splats with the proper size immediate.  Reorder the various
2417         cases by ascending size of immediate, and put all such together.
2419 2018-09-25  Richard Biener  <rguenther@suse.de>
2421         PR debug/83941
2422         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
2423         GC-ification.
2424         (maybe_create_die_with_external_ref): Do not create
2425         DW_TAG_imported_unit here.
2426         (add_abstract_origin_attribute): Handle external BLOCK refs.
2427         (dwarf2out_abstract_function): Simplify LTO case.
2428         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
2429         rather than using maybe_create_die_with_external_ref.
2431 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
2433         PR target/71278
2434         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
2435         (frndintxf2_mask_pm_i387): Ditto.
2436         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
2437         Enable for !flag_trapping_math.
2438         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
2439         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
2440         Change operand 1 predicate to nonimmediate_operand.
2441         (attr "i387_cw"): Remove mask_pm.
2442         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
2443         (enum ix86_entity): Remove I387_MASK_PM.
2444         * config/i386/i386.c (ix86_i387_mode_needed): Do not
2445         handle I387_MASK_PM.
2446         (ix86_mode_needed): Ditto.
2447         (ix86_mode_after): Ditto.
2448         (ix86_mode_entry): Ditto.
2449         (ix86_mode_exit): Ditto.
2450         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
2452 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
2454         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
2455         to_update_switch_stmts to vNULL instead of calling create on them
2456         immediately.
2458 2018-09-25  Richard Biener  <rguenther@suse.de>
2460         PR tree-optimization/87402
2461         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
2462         (visit_phi): Re-instantiate handling of supposed to be VARYING
2463         but non-VARYING backedge value.
2465 2018-09-25  Richard Biener  <rguenther@suse.de>
2467         PR debug/83941
2468         * dwarf2out.c (struct sym_off_pair): New.
2469         (external_die_map): New global.
2470         (lookup_decl_die): When in LTO create DIEs lazily from the
2471         external_die_map.
2472         (lookup_block_die): New function, create DIEs lazily in LTO.
2473         (equate_block_to_die): New function.
2474         (dwarf2out_die_ref_for_decl): During WPA get the association
2475         from the external DIE map.
2476         (dwarf2out_register_external_die): Record mapping into the
2477         external DIE map.
2478         (maybe_create_die_with_external_ref): New function split out from
2479         DIE generation part of old dwarf2out_register_external_die.
2480         (add_abstract_origin_attribute): Do not return the DIE.  When
2481         in LTO reference externals directly.
2482         (dwarf2out_abstract_function): When in LTO ignore calls for
2483         decls with external DIEs (already present abstract instances).
2484         (gen_call_site_die): Adjust.
2485         (add_high_low_attributes): Likewise.
2486         (gen_lexical_block_die): Likewise.
2487         (gen_inlined_subroutine_die): Likewie.
2488         (gen_block_die): Likewise.
2489         (dwarf2out_inline_entry): Likewise.
2490         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
2491         DIEs.
2493 2018-09-25  Martin Liska  <mliska@suse.cz>
2495         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
2496         integers and not by a float value.
2498 2018-09-25  Martin Liska  <mliska@suse.cz>
2500         PR fortran/87394
2501         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
2502         instead of NULL.
2503         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
2504         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
2505         * godump.c (go_format_type): Remove extra parenthesis.
2507 2018-09-25  Martin Liska  <mliska@suse.cz>
2509         * alias.c (set_dest_equal_p): Remove unused function.
2510         * config/i386/i386.c (def_builtin_pure2): Likewise.
2511         * diagnostic-show-locus.c (class layout): Remove
2512         unused field.
2513         (layout::layout): Likewise here.
2514         * dump-context.h (class temp_dump_context): Likewise.
2515         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
2516         (add_AT_loclistsptr): Likewise.
2517         (add_AT_offset): Likewise.
2518         (get_AT_hi_pc): Likewise.
2519         (is_comdat_die): Likewise.
2520         (type_is_enum): Likewise.
2521         (ceiling): Likewise.
2522         (add_AT_vms_delta): Likewise.
2523         (is_class_die): Likewise.
2524         * edit-context.c (class line_event): Remove unused field.
2525         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
2526         unused function.
2527         * ipa-cp.c (ipa_get_vr_lat): Likewise.
2528         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
2529         (ok_for_base_p_nonstrict): Likewise.
2530         * tree-chrec.c (is_not_constant_evolution): Likewise.
2531         (chrec_fold_poly_cst): Likewise.
2532         * tree-if-conv.c (has_pred_critical_p): Likewise.
2533         * tree-ssa-coalesce.c (print_exprs): Likewise.
2534         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
2535         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
2536         * tree-vrp.c (value_ranges_intersect_p): Likewise.
2537         (value_range_nonnegative_p): Likewise.
2539 2018-09-25  Martin Liska  <mliska@suse.cz>
2541         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2542         Do not handle "GNU Pascal".
2543         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2544         Likewise.
2545         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
2546         from documentation. Likewise.
2547         * dbxout.c (dbxout_range_type): Likewise.
2548         * doc/cpp.texi: Likewise.
2549         * doc/extend.texi: Likewise.
2550         * doc/frontends.texi: Likewise.
2551         * doc/invoke.texi: Remove Pascal entry.
2552         * tree.def (CLEANUP_POINT_EXPR): Likewise.
2553         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
2555 2018-09-25  Martin Liska  <mliska@suse.cz>
2557         PR middle-end/86078
2558         * doc/invoke.texi: Document all parameters and remove default
2559         of the parameters.
2561 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2563         PR bootstrap/87417
2564         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
2565         contains HOST_WIDE_INTs when computing its size.
2567 2018-09-24  Jim Wilson  <jimw@sifive.com>
2569         PR target/87391
2570         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
2571         not TARGET_RVE.
2572         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
2574 2018-09-24  Andrew Pinski  <apinski@marvell.com>
2576         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2577         access prev before checking it for NULLness in the
2578         AARCH64_FUSE_CMP_BRANCH case.
2580 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
2582         PR target/82699
2583         * config/i386/i386.c (rest_of_insert_endbranch): Set
2584         endbr_queued_at_entrance to true and don't insert ENDBR if
2585         x86_function_profiler will be called.
2586         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2587         is true.
2588         * config/i386/i386.h (machine_function): Add
2589         endbr_queued_at_entrance.
2591 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2593         * genattrtab.c (mk_attr_alt): Use alternative_mask.
2594         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
2595         types.
2596         (check_attr_test): Use alternative_mask.
2597         (get_attr_value): Likewise.
2598         (compute_alternative_mask): Use alternative_mask and XWINT.
2599         (make_alternative_compare): Use alternative_mask.
2600         (attr_alt_subset_p): Use XWINT.
2601         (attr_alt_subset_of_compl_p): Likewise.
2602         (attr_alt_intersection): Use alternative_mask and XWINT.
2603         (attr_alt_union): Likewise.
2604         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
2605         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
2606         (simplify_test_exp): Use alternative_mask and XWINT.
2607         (write_test_expr): Use alternative_mask and XWINT, adjust bit
2608         number calculation to support 64 bits.  Generate code that
2609         checks 64-bit masks.
2610         (main): Use alternative_mask.
2611         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
2613 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2615         PR target/80080
2616         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
2617         RETURN+USE when returning via %r14.
2619 2018-09-24  Martin Liska  <mliska@suse.cz>
2621         * gcov.c (output_lines): Print colorization legend
2622         for both flag_use_colors and flag_use_hotness_colors.
2623         Reword the help.
2625 2018-09-24  Martin Liska  <mliska@suse.cz>
2627         * coverage.c (get_coverage_counts): Use warning_at
2628         with current_function_decl location. Use %qD in warning
2629         message.
2631 2018-09-24  Martin Liska  <mliska@suse.cz>
2633         * memory-block.h (memory_block_pool::release): Annotate with
2634         valgrind that the memory is not accessible.
2636 2018-09-24  Martin Liska  <mliska@suse.cz>
2638         PR sanitizer/85774
2639         * asan.c: Make asan_handled_variables extern.
2640         * asan.h: Likewise.
2641         * cfgexpand.c (expand_stack_vars): Make sure
2642         a representative is unpoison if another
2643         variable in the partition is handled by
2644         use-after-scope sanitization.
2646 2018-09-24  Richard Biener  <rguenther@suse.de>
2648         PR tree-optimization/63155
2649         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2650         the worklist when the edge of the respective argument isn't
2651         executable.
2653 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2655         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
2656         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
2657         (MASK_CLASS_P): Update for rename.
2658         (MAYBE_MASK_CLASS_P): Ditto.
2659         (REG_CLASS_NAMES): Update.
2660         (REG_CLASS_CONTENT): Update.
2661         * config/i386/i386.c (regclass_map): Update for MASK_REG
2662         and ALL_MASK_REGS rename.
2663         * config/i386/constraints.md (Yk): Update for rename.
2664         (k): Ditto.
2666 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2668         * config/i386/i386.h (enum reg_class): Remove
2669         EVEX_SSE_REGS and MOD4_SSE_REGS.
2670         (REG_CLASS_NAMES): Update.
2671         (REG_CLASS_CONTENT): Update.
2672         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
2673         registers as ALL_SSE_REGS.
2674         (ix86_additional_allocno_class_p): Remove.
2675         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
2676         (ix86_register_priority): Lower priority of EVEX SSE registers.
2677         Use IN_RANGE macro where appropriate.
2678         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
2679         AVX-5124VNNIW checks.
2680         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
2681         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
2682         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
2683         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
2684         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
2685         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
2686         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
2687         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
2688         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
2689         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
2690         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
2691         * config/i386/constraints.md (Yh): Remove.
2693 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
2695         * config/i386/i386.c (regclass_map): Declare integer REX registers
2696         as GENERAL_REGS.
2698 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
2700         * doc/service.texi (Service): Switch the fsf.org link to https.
2702 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
2704         PR target/86798
2705         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2706         Define to speculation_safe_value_not_needed.
2708 2018-09-21  Florian Weimer  <fweimer@redhat.com>
2710         PR middle-end/81035
2711         * doc/extend.texi (Common Function Attributes): Mention that
2712         noreturn suppresses tail call optimization.
2714 2018-09-21  Jeff Law  <law@redhat.com>
2716         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
2717         vr_values::cleanup_edges_and_switches.
2718         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
2719         vr_values class.
2720         (identify_jump_threads): Remove EDGE_IGNORE handling.
2721         (execute_vrp): Move handling of to_remove_edges and
2722         to_update_switch_stmts into vr_values class member functions.
2723         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
2724         (to_update_switch_stmts): Likewise.
2725         * vr-values.c: Include cfghooks.h.
2726         (vr_values::vr_values): Initialize to_remove_edges and
2727         to_update_switch_stmts.
2728         (vr_values::~vr_values): Verify to_remove_edges and
2729         to_update_switch_stmts are empty.
2730         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
2731         (vr_values::cleanup_edges_and_switches): New member function.
2732         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
2733         function.  Add new data members.
2735 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
2737         PR tree-optimization/87309
2738         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
2739         calls with pflags and alt_flags.
2740         (selftest::test_capture_of_dump_calls): Add test of interaction of
2741         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
2743 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2745         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
2747 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2749         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
2751 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2753         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
2754         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
2755         (VXWORKS_LIBS_RTP): Minor reordering.
2757 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2759         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
2760         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
2762 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2764         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
2765         (PTRDIFF_TYPE): Likewise.
2767 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2769         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
2770         triplet, similar to support for VxWorks7.
2771         * config/vxworks-dummy.h: Provide a default definition
2772         of TARGET_VXWORKS64 to 0.
2774 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2776         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
2777         * config/vxworks-dummy.h: here.
2779 2018-09-21  Olivier Hainque  <hainque@adacore.com>
2781         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
2783 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
2784             Bo Zhou  <zbo.zhou@hisilicon.com>
2786         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
2787         * config/aarch64/aarch64-tune.md: Regenerated.
2788         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
2789         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
2790         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
2792 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
2793             Julian Brown  <julian@codesourcery.com>
2795         * builtins.c (get_builtin_sync_mem): Handle address spaces.
2797 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2799         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2800         if the call takes a static chain.
2802 2018-09-21  Martin Liska  <mliska@suse.cz>
2804         * auto-profile.c (autofdo_source_profile::read): Do not
2805         set sum_all.
2806         (read_profile): Do not add working sets.
2807         (read_autofdo_file): Remove sum_all.
2808         (afdo_callsite_hot_enough_for_early_inline): Remove const
2809         qualifier.
2810         * coverage.c (struct counts_entry): Remove gcov_summary.
2811         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2812         do not support GCOV_TAG_PROGRAM_SUMMARY.
2813         (get_coverage_counts): Remove summary and expected
2814         arguments.
2815         * coverage.h (get_coverage_counts): Likewise.
2816         * doc/gcov-dump.texi: Remove -w option.
2817         * gcov-dump.c (dump_working_sets): Remove.
2818         (main): Do not support '-w' option.
2819         (print_usage): Likewise.
2820         (tag_summary): Likewise.
2821         * gcov-io.c (gcov_write_summary): Do not dump
2822         histogram.
2823         (gcov_read_summary): Likewise.
2824         (gcov_histo_index): Remove.
2825         (gcov_histogram_merge): Likewise.
2826         (compute_working_sets): Likewise.
2827         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2828         it not obsolete.
2829         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2830         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2831         (GCOV_HISTOGRAM_SIZE): Remove.
2832         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2833         (struct gcov_summary): Simplify rapidly just
2834         to runs and sum_max fields.
2835         (gcov_histo_index): Remove.
2836         (NUM_GCOV_WORKING_SETS): Likewise.
2837         (compute_working_sets): Likewise.
2838         * gcov-tool.c (print_overlap_usage_message): Remove
2839         trailing empty line.
2840         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2841         (output_lines): Remove program related line.
2842         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2843         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2844         histogram.
2845         (input_profile_summary): Do not read it.
2846         (merge_profile_summaries): And do not merge it.
2847         (input_symtab): Do not call removed function.
2848         * modulo-sched.c (sms_schedule): Do not print sum_max.
2849         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2850         removed when histogram method was invented.
2851         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2852         mode.
2853         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2854         GCOV coding style.
2855         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2856         and dump selected value.
2857         * profile.c (add_working_set): Remove.
2858         (get_working_sets): Likewise.
2859         (find_working_set): Likewise.
2860         (get_exec_counts): Do not work with working sets.
2861         (read_profile_edge_counts): Do not inform as sum_max is removed.
2862         (compute_branch_probabilities): Likewise.
2863         (compute_value_histograms): Remove argument for call of
2864         get_coverage_counts.
2865         * profile.h: Do not make gcov_summary const.
2867 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
2869         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2871 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
2873         PR tree-optimization/86990
2874         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2875         Check that the entire merged store group is made of constants only for
2876         overlapping stores.
2878 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
2880         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2881         (VTABLE_VERIFICATION_SPEC): Likewise.
2882         (SANITIZER_EARLY_SPEC): Likewise.
2883         (SANITIZER_SPEC): Likewise.
2884         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2885         * doc/invoke.texi (Link Options): Document -r.
2887 2018-09-20  Richard Biener <rguenther@suse.de>
2889         PR middle-end/87054
2890         * gimplify.c (gimplify_expr): Retain alignment of
2891         addressable lvalue in dereference.
2893 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
2895         PR bootstrap/87013
2896         * configure.ac: Check for .loc is_stmt support.
2897         * configure, config.in: Rebuilt.
2898         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2899         if not supported.
2901 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2903         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
2904         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
2905         -misel=no.
2907 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
2909         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2910         VECTOR_OTHER.
2911         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2912         case VECTOR_OTHER.
2914 2018-09-20  Marek Polacek  <polacek@redhat.com>
2916         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
2918 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
2920         PR tree-optimization/87288
2921         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2922         into account when determining PEELING_FOR_NITERS.
2924 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
2926         PR tree-optimization/86877
2927         * tree-vect-loop.c (vect_analyze_loop_2): Call
2928         vect_verify_datarefs_alignment.
2930 2018-09-19  Marek Polacek  <polacek@redhat.com>
2932         * doc/invoke.texi: Document -Wclass-conversion.
2934 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
2936         * config/pa/pa.c (pa_adjust_priority): Delete.
2937         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2939         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2940         (atomic_storehi): Likewise.
2941         (atomic_storesi): Likewise.
2942         (atomic_loaddi): Restore compare and swap exchange loop code.
2944 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
2946         PR rtl-optimization/86902
2947         * combine.c (try_combine): When changing the CC mode used, don't change
2948         an unrelated mode in other_insn to that new CC mode.
2950 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
2952         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2953         with %T in place of calls to dump_generic_expr.
2954         (prune_runtime_alias_test_list): Likewise.
2955         (create_runtime_alias_checks): Likewise.
2956         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2957         (vect_analyze_data_ref_dependence): Likewise.
2958         (vect_slp_analyze_data_ref_dependence): Likewise.
2959         (vect_record_base_alignment): Likewise.  Use %G in place of call
2960         to dump_gimple_stmt.
2961         (vect_compute_data_ref_alignment): Likewise.
2962         (verify_data_ref_alignment): Likewise.
2963         (vect_find_same_alignment_drs): Likewise.
2964         (vect_analyze_group_access_1): Likewise.
2965         (vect_analyze_data_ref_accesses): Likewise.
2966         (dependence_distance_ge_vf): Likewise.
2967         (dump_lower_bound): Likewise.
2968         (vect_prune_runtime_alias_test_list): Likewise.
2969         (vect_find_stmt_data_reference): Likewise.
2970         (vect_analyze_data_refs): Likewise.
2971         (vect_create_addr_base_for_vector_ref): Likewise.
2972         (vect_create_data_ref_ptr): Likewise.
2973         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2974         (vect_can_advance_ivs_p): Likewise.
2975         (vect_update_ivs_after_vectorizer): Likewise.
2976         (vect_gen_prolog_loop_niters): Likewise.
2977         (vect_prepare_for_masked_peels): Likewise.
2978         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2979         (vect_determine_vectorization_factor): Likewise.
2980         (vect_is_simple_iv_evolution): Likewise.
2981         (vect_analyze_scalar_cycles_1): Likewise.
2982         (vect_analyze_loop_operations): Likewise.
2983         (report_vect_op): Likewise.
2984         (vect_is_slp_reduction): Likewise.
2985         (check_reduction_path): Likewise.
2986         (vect_is_simple_reduction): Likewise.
2987         (vect_create_epilog_for_reduction): Likewise.
2988         (vect_finalize_reduction:): Likewise.
2989         (vectorizable_induction): Likewise.
2990         (vect_transform_loop_stmt): Likewise.
2991         (vect_transform_loop): Likewise.
2992         (optimize_mask_stores): Likewise.
2993         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2994         (vect_split_statement): Likewise.
2995         (vect_recog_over_widening_pattern): Likewise.
2996         (vect_recog_average_pattern): Likewise.
2997         (vect_determine_min_output_precision_1): Likewise.
2998         (vect_determine_precisions_from_range): Likewise.
2999         (vect_determine_precisions_from_users): Likewise.
3000         (vect_mark_pattern_stmts): Likewise.
3001         (vect_pattern_recog_1): Likewise.
3002         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3003         (vect_record_max_nunits): Likewise.
3004         (vect_build_slp_tree_1): Likewise.
3005         (vect_build_slp_tree_2): Likewise.
3006         (vect_print_slp_tree): Likewise.
3007         (vect_analyze_slp_instance): Likewise.
3008         (vect_detect_hybrid_slp_stmts): Likewise.
3009         (vect_detect_hybrid_slp_1): Likewise.
3010         (vect_slp_analyze_operations): Likewise.
3011         (vect_slp_analyze_bb_1): Likewise.
3012         (vect_transform_slp_perm_load): Likewise.
3013         (vect_schedule_slp_instance): Likewise.
3014         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3015         (vect_mark_stmts_to_be_vectorized): Likewise.
3016         (vect_init_vector_1): Likewise.
3017         (vect_get_vec_def_for_operand): Likewise.
3018         (vect_finish_stmt_generation_1): Likewise.
3019         (vect_check_load_store_mask): Likewise.
3020         (vectorizable_call): Likewise.
3021         (vectorizable_conversion): Likewise.
3022         (vectorizable_operation): Likewise.
3023         (vectorizable_load): Likewise.
3024         (vect_analyze_stmt): Likewise.
3025         (vect_is_simple_use): Likewise.
3026         (vect_get_vector_types_for_stmt): Likewise.
3027         (vect_get_mask_type_for_stmt): Likewise.
3028         * tree-vectorizer.c (increase_alignment): Likewise.
3030 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
3032         * doc/rtl.texi: Adjust vec_select description.
3033         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3034         non-constant selectors.
3036 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
3038         * config/aarch64/aarch64-protos.h
3039         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3040         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3041         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3042         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3043         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3044         (AARCH64_FL_PROFILE): Move index so flags are ordered.
3045         (AARCH64_ISA_RCPC8_4): New flag.
3046         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
3047         to aarch64_offset_9bit_signed_unscaled_p.
3048         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
3049         and use stlur.
3050         * config/aarch64/constraints.md (Ust): New constraint.
3051         * config/aarch64/predicates.md.
3052         (aarch64_9bit_offset_memory_operand): New predicate.
3053         (aarch64_rcpc_memory_operand): New predicate.
3055 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
3057         PR rtl-optimization/87361
3058         * rtlanal.c (nonzero_bits1): Revert accidental change.
3060 2018-09-19  Richard Biener  <rguenther@suse.de>
3062         PR tree-optimization/87349
3063         PR tree-optimization/87342
3064         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3066 2018-09-18  Marek Polacek  <polacek@redhat.com>
3068         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3069         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3071 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3073         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3075 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3077         PR rtl-optimization/86882
3078         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3080 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
3082         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3083         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3085 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
3087         PR other/87353
3088         * doc/invoke.texi (Link Options): Fix formatting and grammar.
3090 2018-09-18  Richard Biener  <rguenther@suse.de>
3092         PR middle-end/63155
3093         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3094         (compute_samebase_partition_bases): Likewise.
3095         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3096         (gimple_can_coalesce_p): Simplify.
3098 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
3100         Handle a library implementation of ffs calling __builtin_ffs.
3101         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3102         (mmix_init_libfuncs): New function: make __builtin_ffs expand
3103         to __ffsdi2.
3105 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3107         * diagnostic-show-locus.c (class layout_range): Add field
3108         "m_original_idx".
3109         (layout_range::layout_range): Add "original_idx" param and use it
3110         to initialize new field.
3111         (make_range): Use 0 for original_idx.
3112         (layout::layout): Pass in index to calls to
3113         maybe_add_location_range.
3114         (layout::maybe_add_location_range): Add param "original_idx" and
3115         pass it on to layout_range.
3116         (layout::print_any_labels): Pass on range->m_original_idx to
3117         get_text call.
3118         (gcc_rich_location::add_location_if_nearby): Use 0 for
3119         original_idx.
3120         * gcc-rich-location.h (text_range_label::get_text): Update for new
3121         param.
3122         (range_label_for_type_mismatch::get_text): Likewise.
3124 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3126         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
3128 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3130         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
3131         format_string_diagnostic_t.
3132         (fmtwarn_n): Likewise.
3133         * substring-locations.c
3134         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
3135         (format_warning_n_va): Convert to...
3136         (format_string_diagnostic_t::emit_warning_n_va): ...this.
3137         (format_warning_va): Convert to...
3138         (format_string_diagnostic_t::emit_warning_va): ...this.
3139         (format_warning_at_substring): Convert to...
3140         (format_string_diagnostic_t::emit_warning): ...this.
3141         (format_warning_at_substring_n): Convert to...
3142         (format_string_diagnostic_t::emit_warning_n): ...this.
3143         * substring-locations.h (class format_string_diagnostic_t): New
3144         class.
3145         (format_warning_va): Convert to
3146         format_string_diagnostic_t::emit_warning_va.
3147         (format_warning_n_va): Convert to
3148         format_string_diagnostic_t::emit_warning_n_va.
3149         (format_warning_at_substring): Convert to
3150         format_string_diagnostic_t::emit_warning.
3151         (format_warning_at_substring_n): Convert to
3152         format_string_diagnostic_t::emit_warning_n.
3154 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
3155             Bernd Schmidt <bernds_cb1@t-online.de>
3157         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
3158         SImode args.
3160 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3162         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
3163         operand 0 predicate to nonimmediate operand.
3164         (rint<mode>2_frndint): Remove insn pattern.
3165         (rint<mode>2): Change operand 1 predicate to general_operand.
3166         Extend operand 1 to XFmode and generate rintxf2 insn.
3167         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
3168         Do not use X87MODEF mode macro.
3169         (frndintxf2_<rounding>_i387): Rename from
3170         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
3171         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
3172         to XFmode and generate significandxf3 insn.
3174 2018-09-17  Richard Biener  <rguenther@suse.de>
3176         PR tree-optimization/87328
3177         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
3178         visiting unexecutable backedges when not iterating.
3179         (do_rpo_vn): Mark all edges not executable even when not
3180         iterating.
3182 2018-09-17  Martin Jambor  <mjambor@suse.cz>
3184         PR c/63886
3185         * doc/invoke.texi (Warning Options): Likewise.
3187 2018-09-17  Richard Biener  <rguenther@suse.de>
3189         PR tree-optimization/87301
3190         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
3191         clean EH info from leftover copy assignments.
3193 2018-09-17  Martin Liska  <mliska@suse.cz>
3195         PR gcov-profile/85871
3196         * gcov.c (output_intermediate_file): Fix out of bounds
3197         access.
3199 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
3201         * config/arc/arc.c: Object attributes for core4 not reflected
3202         correctly.
3203         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
3204         core3).
3206 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
3208         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
3210 2018-09-17  Martin Liska  <mliska@suse.cz>
3212         * doc/gcov.texi: Document new option --use-hotness-colors.
3213         * gcov.c (struct source_info): Declare new field.
3214         (source_info::source_info): Set default for maximum_count.
3215         (print_usage): Add new -q option.
3216         (process_args): Process it.
3217         (accumulate_line_info): Save src->maximum_count.
3218         (output_line_beginning): Make color line number if
3219         flag_use_hotness_colors is set.
3220         (output_line_details): Pass default argument value.
3221         (output_lines): Pass src->maximum_count.
3223 2018-09-17  Martin Liska  <mliska@suse.cz>
3225         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
3226         Use processor_names table.
3227         * config/i386/i386.c (ix86_default_align): Use
3228         processor_cost_table for alignment values.
3229         (ix86_option_override_internal): Use processor_names.
3230         (ix86_function_specific_print): Likewise.
3231         * config/i386/i386.h (struct processor_costs):
3232         Add alignment values.
3233         (struct ptt): Remove and replace with const char *.
3234         * config/i386/x86-tune-costs.h (struct processor_costs):
3235         Declare default alignments for all costs.
3237 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
3239         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
3240         symbolics or VR_VARYING ranges for ABS_EXPR.
3241         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
3242         when range will wrap.
3244 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
3246         PR middle-end/86864
3247         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
3248         before and after a JUMP_TABLE_DATA.
3250 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
3252         PR middle-end/87188
3253         * dojump.c (do_compare_and_jump): Canonicalize function pointers
3254         when one operand is a function pointer.  Use POINTER_TYPE_P and
3255         FUNC_OR_METHOD_TYPE_P.
3256         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3257         * fold-const.c (build_range_check): Likewise.
3258         * match.pd (simple_comparison): Likewise.
3260 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
3262         PR c/82967
3263         * spellcheck.c (get_edit_distance_cutoff): New function.
3264         (selftest::test_edit_distance_unit_test_oneway): Rename to...
3265         (selftest::test_get_edit_distance_one_way): ...this.
3266         (selftest::test_get_edit_distance_unit): Rename to...
3267         (selftest::test_get_edit_distance_both_ways): ...this.
3268         (selftest::test_edit_distances): Move tests to this new function,
3269         and test some more pairs of strings.  Update for above renaming.
3270         (selftest::get_old_cutoff): New function.
3271         (selftest::test_get_edit_distance_cutoff): New function.
3272         (selftest::assert_suggested_for): New function.
3273         (ASSERT_SUGGESTED_FOR): New macro.
3274         (selftest::assert_not_suggested_for): New function.
3275         (ASSERT_NOT_SUGGESTED_FOR): New macro.
3276         (selftest::test_suggestions): New function.
3277         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
3278         tests to selftest::test_edit_distances and call it.  Add calls to
3279         selftest::test_get_edit_distance_cutoff and
3280         selftest::test_suggestions.
3281         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
3282         (best_match::consider): Replace hard-coded cutoff calculation with
3283         a call to...
3284         (best_match::get_cutoff): New declaration.
3285         (best_match::get_best_meaningful_candidate): Likewise.
3287 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3289         * builtins.c (fold_builtin_strlen): Remove TODO comment.
3291 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3293         revert:
3294         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3296         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3297         terminated string literal.
3299 2018-09-14  Martin Sebor  <msebor@redhat.com>
3301         * builtins.c (unterminated_array): Handle ARRAY_REF.
3302         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
3303         * builtins.h (unterminated_array): Declare extern.
3304         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
3305         arrays.
3306         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
3307         calls.
3309 2018-09-14  Martin Sebor  <msebor@redhat.com>
3310             Jeff Law  <law@redhat.com>
3312         * builtins.c (unterminated_array): New.
3313         (expand_builtin_strcpy): Adjust.
3314         (expand_builtin_strcpy_args): Detect unterminated arrays.
3315         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
3316         unterminated arrays.
3317         * gimple-fold.h (get_maxval_strlen): Add argument.
3318         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
3320         * gimple-fold.c (get_range_strlen): Add argument.
3321         (get_maxval_strlen): Adjust.
3322         * gimple-fold.h (get_range_strlen): Add argument.
3324 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
3326         * config/i386/movdirintrin.h: Fix copyright year.
3328 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
3330         * reg-stack.c: Include regs.h.
3331         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
3332         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
3333         FIRST_STACK_REG, not DFmode.
3334         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
3335         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
3336         (change stack): Default register mode to the reg_raw_mode of
3337         FIRST_STACK_REG, not DFmode.
3338         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
3339         (*swapxf): Rename from swapxf.
3341 2018-09-14  Carl Love  <cel@us.ibm.com>
3343         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
3344         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
3346 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
3348         PR target/87224
3349         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
3350         alternatives.
3352 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
3354         PR target/85628
3355         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
3357 2018-09-14  Jason Merrill  <jason@redhat.com>
3359         Fix --enable-gather-detailed-mem-stats.
3360         * hash-table.c (hash_table_usage): Change from variable to function.
3361         * hash-table.h: Adjust.
3362         * Makefile.in: Add missing dependencies on hash-table.h.
3364 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3366         PR tree-optimization/87259
3367         PR lto/87283
3368         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
3369         execute_cse_reciprocals_1 has tried transforming.
3371 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
3373         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
3374         VR_VARYING for PLUS/MINUS_EXPR.
3376 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
3378         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
3379         formatting.
3381 2018-09-14  Richard Biener  <rguenther@suse.de>
3383         PR middle-end/63155
3384         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
3385         bits for the merged partition.
3387 2018-09-13  Martin Sebor  <msebor@redhat.com>
3388             Bernd Edlinger  <bernd.edlinger@hotmail.de>
3390         * builtins.h (c_srlen): Add argument.
3391         * builtins.c (warn_string_no_nul): New function.
3392         (c_strlen): Add argument and use it.  Update recursive calls.
3393         Pass DECL argument to string_constant to get info on non
3394         terminated strings.  Update *NONSTR as needed.
3395         (fold_builtin_strlen): Add argument to calls to c_strlen.
3396         Warn for unterminated arrays.
3397         (warn_string_no_null): Add prototype.
3398         * expr.c (string_constant): Update arguments.  Update recursive
3399         calls appropriately.  Detect missing NUL terminator and outermost
3400         declaration its missing in.
3401         Improve checks for arrays with nonzero lower bound or elements
3402         that are not a single byte.  Simplify offset computation.
3403         Simplify checks for non-NUL terminated strings.
3404         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
3405         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
3407 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3409         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
3410         correctly.
3411         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
3412         argument.  Fix range checks.
3413         * fold-const.h (c_getstr): Adjust protoype.
3414         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
3415         string is constant but contains no NUL byte.
3417         * expr.c (string_constant): Adjust function comment.
3418         Remove bogus check for zero termination.
3420         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
3422         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
3423         (get_constant_size): Don't make STRING_CSTs larger than they are.
3424         (check_string_literal): New check function for STRING_CSTs.
3425         (output_constant): Use it.
3427 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
3429         PR target/86812
3430         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
3432 2018-09-13  Richard Biener  <rguenther@suse.de>
3434         PR tree-optimization/87263
3435         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
3436         (struct unwind_state): Add max_rpo field.
3437         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
3438         Compute max_rpo, the max RPO number a block can be backwards reached
3439         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
3440         separating it from the iterating mode.
3442 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
3444         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
3445         (rfs_decision): New scheduling decision.
3447 2018-09-13  Richard Biener  <rguenther@suse.de>
3449         PR bootstrap/87134
3450         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
3451         (vn_nary_op_insert_pieces_predicated): Do not write useless
3452         valid_dominated_by_p entry outside of the allocated storage.
3454 2018-09-13  Omar Sandoval  <osandov@osandov.com>
3455             Tom de Vries  <tdevries@suse.de>
3457         PR debug/86985
3458         * dwarf2out.c (is_c): New function.
3459         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
3461 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
3463         PR target/85628
3464         * config/aarch64/aarch64.md (*aarch64_bfxil):
3465         Define.
3466         * config/aarch64/constraints.md (Ulc): Define.
3467         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
3468         Define.
3469         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
3470         New function.
3472 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
3474         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
3475         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
3476         aarch64_layout_frame call.
3477         (aarch64_expand_epilogue): Likewise.
3478         (aarch64_initial_elimination_offset): Likewise.
3479         (aarch64_get_separate_components): Likewise.
3480         (aarch64_use_return_insn_p): Likewise.
3481         (aarch64_layout_frame): Remove unneeded check.
3483 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3485         * configure.ac: Only append
3486         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
3487         gcc_config_arguments if it was never reconfigured or last reconfigure
3488         was with different arguments.
3489         * configure: Regenerated.
3491 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3492             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3494         PR middle-end/87290
3495         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
3496         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
3498 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
3500         PR tree-optimization/87287
3501         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
3502         X % C == 0 to X % (unsigned) C == 0 optimization to ...
3503         * match.pd (X % C == 0): ... here.  New optimization.
3505 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3507         PR middle-end/82853
3508         * expr.h (maybe_optimize_mod_cmp): Declare.
3509         * expr.c (mod_inv): New function.
3510         (maybe_optimize_mod_cmp): New function.
3511         (do_store_flag): Use it.
3512         * cfgexpand.c (expand_gimple_cond): Likewise.
3514 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
3515             Julian Brown  <julian@codesourcery.com>
3517         PR middle-end/86336
3518         * gimplify.c (gimplify_scan_omp_clauses): Set
3519         target_firstprivatize_array_bases in OpenACC parallel and kernels
3520         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
3521         OpenACC data regions.
3523 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
3525         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
3526         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
3528 2018-09-12  Richard Biener  <rguenther@suse.de>
3530         PR tree-optimization/87280
3531         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
3532         edge but unreachable target.
3533         (do_rpo_vn): For conservatively handling a PHI only mark
3534         the backedge executable but not the block reachable.
3536 2018-09-12  Richard Biener  <rguenther@suse.de>
3538         PR tree-optimization/87266
3539         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
3540         visited blocks.
3542 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
3544         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
3545         constants.
3546         ("trunc<BFP:mode><DFP_ALL:mode>2")
3547         ("trunc<DFP_ALL:mode><BFP:mode>2")
3548         ("extend<BFP:mode><DFP_ALL:mode>2")
3549         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
3550         according to the target operand type.
3552 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3553             Andreas Krebbel  <krebbel@linux.ibm.com>
3555         PR tree-optimization/86844
3556         * gimple-ssa-store-merging.c
3557         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
3558         there are any overlapping stores in between them, make sure they are
3559         also coalesced or we give up completely.
3561 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
3563         PR middle-end/87248
3564         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
3565         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
3567 2018-09-12  Tom de Vries  <tdevries@suse.de>
3569         * common.opt (gdescribe-dies): Add option.
3570         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
3571         attribute for artifical and nameless decls.
3572         (dwarf2out_register_external_die): Add description attribute to
3573         external reference die.
3574         (add_desc_attribute): New functions.
3575         (gen_subprogram_die): Add description attribute to
3576         DW_TAG_call_site_parameter.
3577         * tree-pretty-print.c (print_generic_expr_to_str): New function.
3578         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
3579         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
3580         -gno-describe-dies.
3581         (@item -gdescribe-dies): Add.
3583 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
3585         * tree-vrp.c (vrp_shift_undefined_p): Remove.
3586         (extract_range_from_binary_expr_1: Call
3587         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
3588         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
3589         depend on sign.
3591 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
3593         * gimple-ssa-warn-alloca.c
3594         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
3595         field for ALLOCA_BOUND_*_LARGE.
3597 2018-09-11  Nathan Sidwell  <nathan@acm.org>
3599         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
3601 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3603         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
3604         for clobbers.  Remove obsolete comment.
3606 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3608         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
3609         mpxchk, mpxld and mpxst types.
3610         (define_attr length_immediate): Remove all processing of mpx types.
3611         (define_attr prefix_0f): Ditto.
3612         (define_attr memory): Ditto.
3614 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
3616         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
3617         (log<mode>2): Change operand 1 predicate to general_operand.
3618         Extend operand 1 to XFmode and generate logxf3 insn.
3619         (log10<mode>2): Change operand 1 predicate to general_operand.
3620         Extend operand 1 to XFmode and generate log10xf3 insn.
3621         (log2<mode>2): Change operand 1 predicate to general_operand.
3622         Extend operand 1 to XFmode and generate log2xf3 insn.
3623         (fyl2xp1_extend<mode>xf3_i387): Remove.
3624         (log1p<mode>2): Change operand 1 predicate to general_operand.
3625         Extend operand 1 to XFmode and generate log1pxf3 insn.
3626         (fxtract_extend<mode>xf3_i387): Remove.
3627         (logb<mode>2): Change operand 1 predicate to general_operand.
3628         Extend operand 1 to XFmode and generate logbxf3 insn.
3629         (ilogb<mode>2): Change operand 1 predicate to general_operand.
3630         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
3631         (significand<mode>2): Change operand 1 predicate to general_operand.
3632         Extend operand 1 to XFmode and generate significandxf3 insn.
3634 2018-09-11  Nathan Sidwell  <nathan@acm.org>
3636         * gcc.c (perror_with_name, pfatal_with_name): Delete.
3637         (load_specs): Use fatal_error.
3638         (DELETE_IF_ORDINARY, process_command): Use error.
3639         (execute, run_attempt): Use fatal_error.
3641 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
3643         * diagnostic-core.h (sorry_at): New prototype.
3644         * diagnostic.c (sorry_at): New function.
3646 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
3648         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
3649         by zero as VR_UNDEFINED.
3651 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
3653         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
3654         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
3655         (<sincos>mode2): New expander.
3656         (sincos_extend<mode>xf3_i387): Remove insn pattern.
3657         (sincos -> sin, cos splitters): Remove splitter patterns.
3658         (sincos<mode>3): Change operand 2 predicate to general_operand.
3659         Extend operand 2 to XFmode and generate sincosxf3 insn.
3660         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
3661         Change operand 3 predicate to const1_operand.
3662         (fptan_extend<mode>xf4_i387): Remove insn pattern.
3663         (tanxf2): Update operands in the call to fptanxf4_i387.
3664         (tan<mode>2): Change operand 1 predicate to general_operand.
3665         Extend operand 1 to XFmode and generate tanxf3 insn.
3666         (atan2xf3): Rename from *fpatanxf3_i387.
3667         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
3668         (atan2xf3): Remove expander.
3669         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
3670         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
3671         (atan<mode>2): Change operand 1 predicate to general_operand.
3672         Extend operand 1 to XFmode and generate atanxf3 insn.
3674 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
3676         * config/i386/i386.md (x87/SSE constant load splitter): Use
3677         memory_operand instead of nonimmediate_operand for input operand
3678         predicate.
3680 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
3682         * config/i386/i386.md (float partial SSE register stall splitter): Move
3683         splitter near its instruction pattern.
3684         (float_extend partial SSE register stall splitter): Ditto.
3685         (float_truncate partial SSE register stall splitter): Ditto.
3687 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
3689         PR target/86794
3690         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3691         to speculation_safe_value_not_needed.
3693         PR target/85666
3694         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
3695         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
3696         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
3697         leaf_function_p, instead use has_hard_reg_initial_val.
3699 2018-09-09  Nathan Sidwell  <nathan@acm.org>
3701         * gcc.h (pfatal_with_name): Don't declare here.
3702         * gcc.c (pfatal_with_name): Make static.
3704 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
3706         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
3707         earlyclobber.
3709 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
3711         PR rtl-optimization/85458
3712         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
3713         priority hook to reduce the priority of EXPR.
3715 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
3717         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
3718         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
3719         with X87_ENABLE_FLOAT.  Remove preparation code.
3720         (*float<SWI48:mode><MODEF:mode>2): Rename from
3721         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
3722         math using "enabled" attribute.
3723         (*floatdi<MODEF:mode>2_i387): Rename from
3724         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
3725         enable for 32bit targets only.
3726         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
3727         splitter.
3728         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
3729         as operand 1 predicate.  Rewrite as define_insn_and_split.
3730         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
3732 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
3734         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
3735         to fallthru to FLOAT case.
3737 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3739         PR target/86731
3740         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
3741         around folding of vec_sl to handle out of range shift values.
3743 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
3745         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
3746         Update callers to gen_fix_trunc<mode>_i387_fisttp
3747         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
3748         nonimmediate_operand.
3749         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
3750         and corresponding splitters.
3751         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
3752         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
3753         (fix_truncdi_i387_with_temp): Remove insn pattern
3754         and corresponding splitters.
3755         (fix_trunc<mode>_i387): Change operand 0 predicate to
3756         nonimmediate_operand.
3757         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
3758         and corresponding splitters.
3759         (*fistdi2_1): Remove.
3760         (fistdi2): Ditto.
3761         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
3762         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
3763         (*fist<mode>2_1): Remove.
3764         (fist<mode>2): Ditto.
3765         (fist<mode>2_with_temp): Remove insn pattern and corresponding
3766         splitters.
3767         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
3768         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
3769         (fistdi2_<rounding>): Change operand 0 predicate to
3770         nonimmediate_operand.
3771         (fistdi2_<rounding>_with_temp): Remove insn pattern
3772         and corresponding splitters.
3773         (fist<mode>2_<rounding>): Change operand 0 predicate to
3774         nonimmediate_operand.
3775         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
3776         and corresponding splitters.
3778         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
3780 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3782         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
3783         the init value.
3785 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3787         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
3788         early gimple folding of vec_splat().
3789         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
3790         * gimple-fold.h: Add an extern define for tree_vec_extract().
3792 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
3794         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
3795         wrappers around TREE_TYPE comparisons.
3797 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3799         PR target/80080
3800         * config/s390/predicates.md: Add nonsym_memory_operand.
3801         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
3802         contains a SYMBOL_REF, load it into an intermediate pseudo.
3803         (s390_emit_compare_and_swap): Legitimize operand.
3804         * config/s390/s390.md: Use the new nonsym_memory_operand
3805         with UNSPECV_CAS patterns.
3807 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3809         PR target/80080
3810         * config/s390/s390-passes.def: New file.
3811         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3812         declaration.
3813         (make_pass_s390_early_mach): Add declaration.
3814         * config/s390/s390.c (make_pass_s390_early_mach):
3815         (s390_option_override): Remove dynamic registration.
3816         * config/s390/t-s390: Add s390-passes.def.
3818 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3820         * config/s390/s390.c (s390_decompose_constant_pool_ref):
3821         Remove UNSPEC_LTREL_BASE check.
3822         (annotate_constant_pool_refs): Likewise.
3823         (find_constant_pool_ref): Likewise.
3824         (find_ltrel_base): Removed.
3825         (replace_ltrel_base): Removed.
3826         (s390_mainpool_finish): Remove replace_ltrel_base call.
3827         (s390_chunkify_start): Remove pending LTREL_BASE logic.
3828         (s390_chunkify_finish): Remove replace_ltrel_base call.
3829         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3831 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
3833         PR target/86779
3834         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3835         to speculation_safe_value_not_needed.
3837 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
3838             Bernd Schmidt  <bernds_cb1@t-online.de>
3840         * config/nvptx/nvptx-opts.h: New file.
3841         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3842         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3843         (ASM_SPEC): Define.
3844         (TARGET_SM35): New macro.
3845         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3846         correct predicate.
3847         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3848         values.
3849         (misa=): New option.
3850         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3852 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
3854         * config/i386/i386.md (truncdfsf2): Remove expander.
3855         (truncdfsf2_with_temp): Ditto.
3856         (truncxf<mode>2): Ditto.
3857         (*truncdfsf_fast_mixed): Remove insn pattern.
3858         (*truncdfsf_fast_i387): Ditto.
3859         (*truncdfsf_mixed): Ditto.
3860         (*truncdfsf_i387): Ditto.
3861         (*truncdfsf2_i387_1): Ditto.
3862         (*truncxfsf2_mixed): Ditto.
3863         (*truncxfdf2_mixed): Ditto.
3864         (*truncxf<mode>2_i387_noop): Ditto. Update callers
3865         to call gen_truncxf<mode>2 instead.
3866         (*truncxf<mode>2_i387): Remove.
3867         (reg->reg splitters): Remove splitter pattern.
3868         (reg->mem splitters): Ditto.
3870         (truncdfsf2): New insn pattern.
3871         (truncxf<mode>2): Ditto.
3873 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3875         * tree-ssa-math-opts.c (is_mult_by): New function.
3876         (is_square_of): Use the above.
3877         (optimize_recip_sqrt): New function.
3878         (pass_cse_reciprocals::execute): Use the above.
3880 2018-09-05  Richard Biener  <rguenther@suse.de>
3882         PR bootstrap/87134
3883         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3884         to zero-init the emplaced vec.
3886 2018-09-05  Martin Liska  <mliska@suse.cz>
3888         PR tree-optimization/87205
3889         * tree-switch-conversion.c (pass_lower_switch::execute):
3890         Group cases for switch statements.
3892 2018-09-05  Richard Biener  <rguenther@suse.de>
3894         PR tree-optimization/87217
3895         * tree-ssa-sccvn.c (vuse_valueize): New.
3896         (vn_reference_lookup_pieces): Use it.
3897         (vn_reference_lookup): Likewise.
3899 2018-09-05  Nathan Sidwell  <nathan@acm.org>
3901         PR c++/87137
3902         * stor-layout.c (place_field): Scan forwards to check last
3903         bitfield when ms_bitfield_placement is in effect.
3905 2018-09-05  Richard Biener  <rguenther@suse.de>
3907         PR bootstrap/87225
3908         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3909         return.
3911 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
3912             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
3914         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
3915         * config.gcc (extra_objs): Build it.
3916         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
3917         Likewise.
3918         * config/aarch64/aarch64-passes.def
3919         (pass_tag_collision_avoidance): New pass.
3920         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
3921         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
3922         (aarch64_classify_address): Remove static qualifier.
3923         (aarch64_address_info, aarch64_address_type): Move to...
3924         * config/aarch64/aarch64-protos.h: ... here.
3925         (make_pass_tag_collision_avoidance): New function.
3926         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3927         New tuning flag.
3929 2018-09-05  Martin Liska  <mliska@suse.cz>
3931         * doc/gcov.texi: Update documentation of humar
3932         readable mode.
3933         * gcov.c (format_count): Print one decimal place, it provides
3934         more fine number of situations like '1G' vs. '1.4G'.
3936 2018-09-05  Martin Liska  <mliska@suse.cz>
3938         PR target/87164
3939         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3940         * optc-gen.awk: Allow 'Var' for Deprecated options in order
3941         to generate a MASK value.
3943 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
3945         PR debug/86593
3946         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3947         if frame pointer isn't used.
3948         (compute_frame_pointer_to_fb_displacement): Likewise.
3950 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
3952         PR target/87198
3953         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3954         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3955         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3956         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3957         and OPTION_MASK_ISA_XSAVEC_UNSET.
3959 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
3961         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3962         XOR operations in NAND case.
3964 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3966         * wide-int-range.cc (wide_int_range_convert): New.
3967         * wide-int-range.h (wide_int_range_convert): New.
3968         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3969         code into wide_int_range_convert.
3970         (extract_range_into_wide_ints): Do not munge anti range constants
3971         into the entire domain.  Just return the range back.
3973 2018-09-04  Martin Liska  <mliska@suse.cz>
3975         * genmatch.c (output_line_directive): Add new argument
3976         fnargs.
3977         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3979 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
3981         * doc/invoke.texi (Option Summary): Add whitespace.
3983         * doc/invoke.texi (Option Summary): Add -Waligned-new.
3985 2018-09-04  Richard Biener  <rguenther@suse.de>
3987         PR tree-optimization/87211
3988         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3989         backedge value we're supposed to treat as VARYING also number
3990         the PHI to VARYING in case it got a different value-number already.
3992 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
3994         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3995         (extract_range_from_binary_expr_1): Do not call
3996         vrp_can_optimize_bit_op.
3997         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3998         static.
3999         (wide_int_range_get_mask_and_bounds): New.
4000         (wide_int_range_optimize_bit_op): New.
4001         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4002         (wide_int_range_bit_and): Same.
4003         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4004         (wide_int_range_optimize_bit_op): New.
4005         (wide_int_range_get_mask_and_bounds): New.
4007 2018-09-04  Richard Biener  <rguenther@suse.de>
4009         PR tree-optimization/87176
4010         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4011         variable.  When value-numbering a virtual PHI node make sure
4012         to not value-number to the backedge value.
4014 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4016         * doc/extend.texi (Long Long, Hex Floats): Document support for
4017         long long and hex floats in more recent versions of ISO C++.
4019 2018-09-03  Richard Biener  <rguenther@suse.de>
4021         PR tree-optimization/87177
4022         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4023         cleanup.
4025 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4027         * bb-reorder.c (edge_order): Convert to C-qsort-style
4028         tri-state comparator.
4029         (reorder_basic_blocks_simple): Change std::stable_sort to
4030         gcc_stablesort.
4032 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4034         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4035         tri-state comparator.
4036         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4038 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4040         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
4041         (mergesort): ... here as maximum count for using netsort.
4042         (gcc_qsort): Set nlim to 3 if stable sort is requested.
4043         (gcc_stablesort): New.
4044         * system.h (gcc_stablesort): Declare.
4046 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4048         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
4049         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
4050         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
4052 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
4054         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
4055         lxsdx and stxsdx alternatives.
4056         (*mov<mode>_hardfloat64): Ditto.
4057         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4059 2018-09-03  Richard Biener  <rguenther@suse.de>
4061         PR tree-optimization/87200
4062         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4063         simplify result.
4065 2018-09-03  Martin Liska  <mliska@suse.cz>
4067         PR tree-optimization/87201
4068         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4069         Fix parenthesis in an expression.
4071 2018-09-03  Richard Biener  <rguenther@suse.de>
4073         PR tree-optimization/87197
4074         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4075         visited.  CSE the VN_INFO hashtable lookup.
4077         PR tree-optimization/87169
4078         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4079         iterating make sure there's no extra backedges from irreducible
4080         regions feeding the header.  Mark the destination block
4081         executable.
4083 2018-09-03  Martin Liska  <mliska@suse.cz>
4085         PR driver/83193
4086         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4087         * common/common-targhooks.c (default_get_valid_option_values):
4088         New function.
4089         * common/common-targhooks.h (default_get_valid_option_values):
4090         Likewise.
4091         * common/config/i386/i386-common.c: Move processor_target_table
4092         from i386.c.
4093         (ix86_get_valid_option_values): New function.
4094         (TARGET_GET_VALID_OPTION_VALUES): New macro.
4095         * config/i386/i386.c (struct ptt): Move to i386-common.c.
4096         (PTA_*): Move all defined masks into i386-common.c.
4097         (ix86_function_specific_restore): Use new processor_cost_table.
4098         * config/i386/i386.h (struct ptt): Moved from i386.c.
4099         (struct pta): Likewise.
4100         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4101         * doc/tm.texi.in: Likewise.
4102         * opt-suggestions.c (option_proposer::suggest_option):
4103         Pass prefix to build_option_suggestions.
4104         (option_proposer::get_completions): Likewise.
4105         (option_proposer::build_option_suggestions): Use the new target
4106         hook.
4107         * opts.c (struct option_help_tuple): New struct.
4108         (print_filtered_help): Use the new target hook.
4110 2018-09-03  Martin Liska  <mliska@suse.cz>
4112         PR middle-end/59521
4113         * predict.c (set_even_probabilities): Add likely_edges
4114         argument and handle cases where we have precisely one
4115         likely edge.
4116         (combine_predictions_for_bb): Catch also likely_edges.
4117         (tree_predict_by_opcode): Handle gswitch statements.
4118         * tree-cfg.h (find_case_label_for_value): New declaration.
4119         (find_taken_edge_switch_expr): Likewise.
4120         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4121         Find pivot in decision tree based on probabily, not by number of
4122         nodes.
4124 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
4126         * doc/standards.texi (Standards): Update Objective-C reference.
4128 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4130         * doc/install.texi (Prerequisites): Update link for MPC.
4132 2018-09-01  Michael Matz  <matz@suse.de>
4134         PR tree-optimization/87074
4135         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
4136         PHIs for outer-loop uses.
4138 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4140         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
4141         * doc/invoke.texi (C Dialect Options): Ditto.
4143 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4145         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
4147 2018-08-31  Richard Biener  <rguenther@suse.de>
4149         PR tree-optimization/87168
4150         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
4151         (rpo_elim::eliminate_avail): When OP was not visited it must
4152         be available.
4154 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
4156         * tree-vrp.c (copy_value_range): Convert param "from" from
4157         "value_range *" to "const value_range *".
4158         (range_is_null): Likewise for param "vr".
4159         (range_int_cst_p): Likewise.
4160         (range_int_cst_singleton_p): Likewise.
4161         (symbolic_range_p): Likewise.
4162         (value_ranges_intersect_p): Likewise for both params.
4163         (value_range_nonnegative_p): Likewise for param "vr".
4164         (value_range_constant_singleton): Likewise.
4165         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
4166         (extract_range_into_wide_ints): Likewise for param "vr".
4167         (extract_range_from_multiplicative_op): Likewise for params "vr0"
4168         and "vr1".
4169         (vrp_can_optimize_bit_op): Likewise.
4170         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
4171         "vr1_".
4172         (extract_range_from_unary_expr): Likewise.
4173         (debug_value_range): Likewise for param "vr".
4174         (value_range::dump): Add "const" qualifier.
4175         (vrp_prop::check_array_ref): Convert local "vr" from
4176         "value_range *" to "const value_range *".
4177         (vrp_prop::check_mem_ref): Likewise.
4178         (vrp_prop::visit_stmt): Likewise for local "old_vr".
4179         (vrp_intersect_ranges_1): Likewise for param "vr_1".
4180         (vrp_intersect_ranges): Likewise.
4181         (simplify_stmt_for_jump_threading): Likewise for local "vr".
4182         (vrp_prop::vrp_finalize): Likewise.
4183         * tree-vrp.h (value_range::dump): Add "const" qualifier.
4184         (vrp_intersect_ranges): Add "const" qualifier to params as above.
4185         (extract_range_from_unary_expr): Likewise.
4186         (value_range_constant_singleton): Likewise.
4187         (symbolic_range_p): Likewise.
4188         (copy_value_range): Likewise.
4189         (extract_range_from_binary_expr_1): Likewise.
4190         (range_int_cst_p): Likewise.
4191         (vrp_set_zero_nonzero_bits): Likewise.
4192         (range_int_cst_singleton_p): Likewise.
4194 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
4196         * config/aarch64/arm_neon.h (vabsd_s64): New.
4197         (vnegd_s64): Likewise.
4199 2018-08-31  Martin Jambor  <mjambor@suse.cz>
4201         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
4203 2018-08-31  Martin Liska  <mliska@suse.cz>
4205         * ipa-icf.c (sem_item::add_type): Use
4206         sem_item::m_type_hash_cache.
4207         * ipa-icf.h: Move the cache from sem_item_optimizer
4208         to sem_item.
4210 2018-08-31  Nathan Sidwell  <nathan@acm.org>
4212         * doc/extend.texi (Backwards Compatibility): Remove implicit
4213         extern C leeway of () being (...).
4215 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4217         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
4219 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
4221         PR target/86684
4222         PR target/87149
4223         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
4225 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
4227         PR middle-end/87138
4228         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
4229         gen_int_mode.  Formatting fixes.
4231 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
4233         * target.def (custom_function_descriptors): Improve documentation.
4234         * doc/tm.texi.in (Trampolines): Expand discussion of function
4235         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
4236         beginning of the section.
4237         * doc/tm.texi: Regenerated.
4239 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
4241         * cfg.h (class auto_edge_flag): Spell out the template-id of the
4242         base class in the initializer list.  This is a workaround for
4243         building with older GCC.
4244         (class auto_bb_flag): Likewise.
4246 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
4248         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
4249         (altivec_vcmpequ<VI_char>_p): Remove star.
4250         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
4251         vector load modes.
4252         (expand_strncmp_vec_sequence): New function.
4253         (emit_final_str_compare_vec): New function.
4254         (expand_strn_compare): Add support for vector strncmp.
4255         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
4256         length specification to bytes.
4257         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
4258         (vcmpnezb_p): New pattern.
4259         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
4260         for option -mstring-compare-inline-limit.
4262 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
4264         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
4265         (PTA_SKYLAKE): Add PTA_AES.
4266         (PTA_GOLDMONT): Likewise.
4268 2018-08-29  Jan Hubicka  <jh@suse.cz>
4270         PR lto/86517
4271         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
4272         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
4274 2018-08-29  Jan Hubicka  <jh@suse.cz>
4276         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
4277         TYPE_STUB_DECL.
4278         (hash_tree): Do not visit TYPE_STUB_DECL.
4279         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
4280         stream TYPE_STUB_DECL.
4281         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
4282         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
4283         after free_lang_data.
4284         (type_in_anonymous_namespace_p): Likewise.
4286 2018-08-29  Jan Hubicka  <jh@suse.cz>
4288         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4289         comment that it has to be even number.
4290         (class sreal): Change m_sig type to int32_t.
4291         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4292         int64_t for temporary calculations.
4293         (sreal_verify_basics): Drop one bit from minimum and maximum.
4295 2018-08-30  Richard Biener  <rguenther@suse.de>
4297         PR tree-optimization/87147
4298         * tree-ssa-sccvn.c (SSA_VISITED): New function.
4299         (visit_phi): When the degenerate result is from the backedge and
4300         we didn't visit its definition yet drop to VARYING.
4301         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
4303 2018-08-29  Jan Hubicka  <jh@suse.cz>
4305         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
4306         DECL_VINDEX.
4307         (hash_tree): Likewise.
4309 2018-08-29  Jan Hubicka  <jh@suse.cz>
4311         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
4312         and TYPE_NEXT_REF_TO.
4314 2018-08-29  Jan Hubicka  <jh@suse.cz>
4316         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4317         comment that it has to be even number.
4318         (class sreal): Change m_sig type to int32_t.
4319         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4320         int64_t for temporary calculations.
4321         (sreal_verify_basics): Drop one bit from minimum and maximum.
4323 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
4325         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
4327 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
4329         PR middle-end/86995
4330         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
4331         if to_add is negative.
4333 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4335         PR middle-end/87053
4336         * builtins.c (c_strlen): Improve range checks.
4338 2018-08-29  Martin Sebor  <msebor@redhat.com>
4339             Jeff Law  <law@redhat.com>
4341         PR tree-optimization/86714
4342         PR tree-optimization/86711
4343         * builtins.c (c_strlen): Add arguments to call to string_constant.
4344         * expr.c (string_constant): Add argument.  Detect missing nul
4345         terminator and outermost declaration it's missing in.
4346         * expr.h (string_constant): Add argument.
4347         * fold-const.c (read_from_constant_string): Add arguments to call to
4348         string_constant.
4349         (c_getstr): Likewise.
4350         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4351         to string_constant.
4352         * tree-ssa-strlen.c (get_stridx): Likewise.
4354 2018-08-29  Jan Hubicka  <jh@suse.cz>
4356         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
4357         Do not stream DECL_VINDEX.
4358         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
4359         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
4360         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
4362 2018-08-29  Richard Biener  <rguenther@suse.de>
4364         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
4365         virtual operands that are not default defs to honor region
4366         boundaries.
4367         (rpo_vn_valueize): Remove ineffective code here.
4369 2018-08-29  Richard Biener  <rguenther@suse.de>
4371         PR tree-optimization/87132
4372         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
4373         when skipping defs reachable over backedges.
4375 2018-08-29  Richard Biener  <rguenther@suse.de>
4377         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
4378         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
4379         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
4380         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
4381         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
4382         (vn_reference_lookup_3): Remove use of const_parms.
4383         (free_rpo_vn): Do not free const_parms.
4384         (do_rpo_vn): Do not call init_const_parms.
4385         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
4386         SSA_NAME_POINTS_TO_READONLY_MEMORY.
4387         (call_may_clobber_ref_p_1): Likewise.
4389 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
4391         PR other/86726
4392         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
4393         (-O): Ditto.
4394         (-ftree-scev-cprop): Document.
4396 2018-08-29  Jan Hubicka  <jh@suse.cz>
4398         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
4399         parameters.
4400         (sreal constructor): Update.
4401         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
4402         sreal:operator/): Update.
4404 2018-08-29  Martin Liska  <mliska@suse.cz>
4406         * tree-switch-conversion.c (switch_conversion::expand):
4407         Strenghten assumption about gswitch statements.
4409 2018-08-29  Richard Biener  <rguenther@suse.de>
4411         PR tree-optimization/87117
4412         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
4413         re-value-number released SSA VDEFs.
4415 2018-08-29  Richard Biener  <rguenther@suse.de>
4417         PR tree-optimization/87126
4418         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
4420 2018-08-28  Jim Wilson  <jimw@sifive.com>
4422         * config/riscv/pic.md: Rewrite.
4423         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
4424         invalid address.
4425         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
4426         (SOFTF, default_load, softload, softstore): New.
4428 2018-08-28  Jeff Law  <law@redhat.com>
4430         * fold-const.c (fold_binary_loc): Remove recently added assert.
4432 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
4434         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
4435         to OP parmeter of generated function.
4437 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
4439         PR tree-optimization/87009
4440         * match.pd: Add boolean optimizations.
4442 2018-08-28  Martin Sebor  <msebor@redhat.com>
4444         PR middle-end/86631
4445         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
4446         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
4447         (pass_walloca::gate): Use it.
4448         (alloca_call_type): Same.
4449         (pass_walloca::execute): Same.
4450         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
4452 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
4454         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
4455         GCC_VERSION for usage of "__gcc_dump_printf__" format from
4456         >= 3005 to >= 9000.
4458 2018-08-28  Richard Biener  <rguenther@suse.de>
4460         PR tree-optimization/87124
4461         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
4462         constants before looking up avail.
4464 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
4466         PR middle-end/87099
4467         * calls.c (maybe_warn_nonstring_arg): Punt early if
4468         warn_stringop_overflow is zero.  Don't call get_range_strlen
4469         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
4470         Swap comparison operands to have constants on rhs.  Only use
4471         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
4472         increment lenrng[0].
4474 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
4476         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
4477         use of tree_to_shwi.  Remove duplicated test for the size being
4478         a whole number of bytes.
4480 2018-08-28  Richard Biener  <rguenther@suse.de>
4482         PR tree-optimization/87117
4483         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
4484         Handle removed stmt without LHS (GIMPLE_NOP).
4486 2018-08-28  Richard Biener  <rguenther@suse.de>
4488         PR tree-optimization/87117
4489         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
4490         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
4492 2018-08-28  Richard Biener  <rguenther@suse.de>
4494         PR tree-optimization/87117
4495         * tree-ssa-pre.c (compute_avail): Do not make expressions
4496         with predicated values available.
4497         (get_expr_value_id): Assert we do not run into predicated value
4498         expressions.
4500 2018-08-28  Richard Biener  <rguenther@suse.de>
4502         PR tree-optimization/87117
4503         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
4504         get virtual operands.
4505         (get_expr_operands): Handle STRING_CST like other decls.
4507 2018-08-28  Martin Liska  <mliska@suse.cz>
4509         * tree.h: Update documentation of fndecl_built_in_p
4510         functions.
4513 2018-08-27  Jeff Law  <law@redhat.com>
4514         PR tree-optimization/87110
4515         * tree-ssa-dse.c (compute_trims): Handle non-constant
4516         TYPE_SIZE_UNIT.
4518 2018-08-27  Martin Sebor  <msebor@redhat.com>
4520         PR tree-optimization/86914
4521         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
4523 2018-08-27  Martin Sebor  <msebor@redhat.com>
4525         PR tree-optimization/87112
4526         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
4527         the type of the bound argument.
4529 2018-08-27  Jeff Law  <law@redhat.com>
4531         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
4532         type does not have a TYPE_SIZE_UNIT.
4534 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
4536         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
4537         with include of backend.h.
4539 2018-08-27  Richard Biener  <rguenther@suse.de>
4541         PR tree-optimization/86927
4542         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
4543         use const cond reduction code.
4545 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
4547         PR tree-optimization/85758
4548         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
4550 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
4552         PR c++/87091
4553         * diagnostic-show-locus.c (class layout_range): Update for
4554         conversion of show_caret_p to a tri-state.
4555         (layout_range::layout_range): Likewise.
4556         (make_range): Likewise.
4557         (layout::maybe_add_location_range): Likewise.
4558         (layout::should_print_annotation_line_p): Don't show annotation
4559         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
4560         (layout::get_state_at_point): Update for conversion of
4561         show_caret_p to a tri-state.  Bail out early for
4562         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
4563         underlining or source colorization.
4564         (gcc_rich_location::add_location_if_nearby): Update for conversion
4565         of show_caret_p to a tri-state.
4566         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
4567         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
4568         Likewise.
4569         (selftest::test_one_liner_labels): Likewise.
4570         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
4571         conversion of show_caret_p to a tri-state.
4572         * pretty-print.c (text_info::set_location): Likewise.
4573         * pretty-print.h (text_info::set_location): Likewise.
4574         * substring-locations.c (format_warning_n_va): Likewise.
4575         * tree-diagnostic.c (default_tree_printer): Likewise.
4576         * tree-pretty-print.c (newline_and_indent): Likewise.
4578 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
4580         PR c++/87091
4581         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
4582         line above for line-insertion fix-it hints.
4583         (selftest::test_fixit_insert_containing_newline): Update the
4584         expected results, and add a test with line-numbering enabled.
4586 2018-08-27  Martin Liska  <mliska@suse.cz>
4588         PR sanitizer/86962
4589         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
4590         params with DECL_HAS_VALUE_EXPR_P.
4592 2018-08-27  Martin Liska  <mliska@suse.cz>
4594         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
4595         selected expansion strategy.
4597 2018-08-27  Martin Liska  <mliska@suse.cz>
4599         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
4600         * builtins.c (is_builtin_fn): Likewise.
4601         * attribs.c (diag_attr_exclusions): Use new function
4602         fndecl_built_in_p and remove check for FUNCTION_DECL if
4603         possible.
4604         (builtin_mathfn_code): Likewise.
4605         (fold_builtin_expect): Likewise.
4606         (fold_call_expr): Likewise.
4607         (fold_builtin_call_array): Likewise.
4608         (fold_call_stmt): Likewise.
4609         (set_builtin_user_assembler_name): Likewise.
4610         (is_simple_builtin): Likewise.
4611         * calls.c (gimple_alloca_call_p): Likewise.
4612         (maybe_warn_nonstring_arg): Likewise.
4613         * cfgexpand.c (expand_call_stmt): Likewise.
4614         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
4615         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
4616         (cgraph_node::verify_node): Likewise.
4617         * cgraphclones.c (build_function_decl_skip_args): Likewise.
4618         (cgraph_node::create_clone): Likewise.
4619         * config/arm/arm.c (arm_insert_attributes): Likewise.
4620         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
4621         * dse.c (scan_insn): Likewise.
4622         * expr.c (expand_expr_real_1): Likewise.
4623         * fold-const.c (operand_equal_p): Likewise.
4624         (fold_binary_loc): Likewise.
4625         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
4626         * gimple-low.c (lower_stmt): Likewise.
4627         * gimple-pretty-print.c (dump_gimple_call): Likewise.
4628         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
4629         Likewise.
4630         * gimple.c (gimple_build_call_from_tree): Likewise.
4631         (gimple_call_builtin_p): Likewise.
4632         (gimple_call_combined_fn): Likewise.
4633         * gimplify.c (gimplify_call_expr): Likewise.
4634         (gimple_boolify): Likewise.
4635         (gimplify_modify_expr): Likewise.
4636         (gimplify_addr_expr): Likewise.
4637         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
4638         * ipa-cp.c (determine_versionability): Likewise.
4639         * ipa-fnsummary.c (compute_fn_summary): Likewise.
4640         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
4641         * ipa-split.c (visit_bb): Likewise.
4642         (split_function): Likewise.
4643         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4644         * lto-cgraph.c (input_node): Likewise.
4645         * lto-streamer-out.c (write_symbol): Likewise.
4646         * omp-low.c (setjmp_or_longjmp_p): Likewise.
4647         (lower_omp_1): Likewise.
4648         * predict.c (strip_predict_hints): Likewise.
4649         * print-tree.c (print_node): Likewise.
4650         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
4651         * trans-mem.c (is_tm_irrevocable): Likewise.
4652         (is_tm_load): Likewise.
4653         (is_tm_simple_load): Likewise.
4654         (is_tm_store): Likewise.
4655         (is_tm_simple_store): Likewise.
4656         (is_tm_abort): Likewise.
4657         (tm_region_init_1): Likewise.
4658         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
4659         * tree-cfg.c (verify_gimple_call): Likewise.
4660         (move_stmt_r): Likewise.
4661         (stmt_can_terminate_bb_p): Likewise.
4662         * tree-eh.c (lower_eh_constructs_2): Likewise.
4663         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
4664         * tree-inline.c (remap_gimple_stmt): Likewise.
4665         (copy_bb): Likewise.
4666         (estimate_num_insns): Likewise.
4667         (fold_marked_statements): Likewise.
4668         * tree-sra.c (scan_function): Likewise.
4669         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
4670         (optimize_stack_restore): Likewise.
4671         (pass_fold_builtins::execute): Likewise.
4672         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
4673         (mark_all_reaching_defs_necessary_1): Likewise.
4674         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
4675         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4676         (pass_forwprop::execute): Likewise.
4677         * tree-ssa-loop-im.c (stmt_cost): Likewise.
4678         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4679         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
4680         * tree-ssa-strlen.c (get_string_length): Likewise.
4681         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
4682         (find_func_aliases_for_call): Likewise.
4683         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
4684         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
4685         * tree-tailcall.c (find_tail_calls): Likewise.
4686         * tree.c (need_assembler_name_p): Likewise.
4687         (free_lang_data_in_decl): Likewise.
4688         (get_call_combined_fn): Likewise.
4689         * ubsan.c (is_ubsan_builtin_p): Likewise.
4690         * varasm.c (incorporeal_function_p): Likewise.
4691         * tree.h (DECL_BUILT_IN): Remove and replace with
4692         fndecl_built_in_p.
4693         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
4694         (fndecl_built_in_p): New.
4696 2018-08-27  Martin Liska  <mliska@suse.cz>
4698         PR tree-optimization/86847
4699         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
4700         Dump also subtree probability.
4701         (switch_decision_tree::do_jump_if_equal): New function.
4702         (switch_decision_tree::emit_case_nodes): Handle special
4703         situations in balanced tree that can be emitted much simpler.
4704         Fix calculation of probabilities that happen in tree expansion.
4705         * tree-switch-conversion.h (struct cluster): Add
4706         is_single_value_p.
4707         (struct simple_cluster): Likewise.
4708         (struct case_tree_node): Add new function has_child.
4709         (do_jump_if_equal): New.
4711 2018-08-27  Martin Liska  <mliska@suse.cz>
4713         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
4714         Add new argument to bit_test_cluster constructor.
4715         (bit_test_cluster::emit): Set bits really number of values
4716         handlel by a test.
4717         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
4718         probability argument.
4719         * tree-switch-conversion.h (struct bit_test_cluster):
4720         Add m_handles_entire_switch.
4722 2018-08-27  Martin Liska  <mliska@suse.cz>
4724         PR tree-optimization/86702
4725         * tree-switch-conversion.c (jump_table_cluster::emit):
4726         Make probabilities even for values in jump table
4727         according to number of cases handled.
4728         (switch_decision_tree::compute_cases_per_edge): Pass
4729         argument to reset_out_edges_aux function.
4730         (switch_decision_tree::analyze_switch_statement): Likewise.
4731         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
4732         Make it static.
4734 2018-08-27  Martin Liska  <mliska@suse.cz>
4736         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
4737         cfun argument explicitly.
4738         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
4739         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
4740         function gimple_switch_default_bb.
4741         (convert_switch_statements):
4742         (expand_builtins):
4743         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
4744         * stmt.c (label_to_block_fn): Use label_to_block and pass
4745         cfun argument explicitly and use gimple_switch_label_bb.
4746         (expand_case): Likewise.
4747         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
4748         cfun argument explicitly. Likewise.
4749         (make_edges_bb): Likewise.
4750         (make_cond_expr_edges): Likewise.
4751         (get_cases_for_edge): Likewise.
4752         (make_gimple_switch_edges): Likewise.
4753         (label_to_block_fn): Likewise.
4754         (label_to_block): Likewise.
4755         (make_goto_expr_edges): Likewise.
4756         (make_gimple_asm_edges): Likewise.
4757         (main_block_label): Likewise.
4758         (group_case_labels_stmt): Likewise.
4759         (find_taken_edge_computed_goto): Likewise.
4760         (find_taken_edge_switch_expr): Likewise.
4761         (gimple_verify_flow_info): Likewise.
4762         (gimple_redirect_edge_and_branch): Likewise.
4763         (gimple_switch_label_bb): New function.
4764         (gimple_switch_default_bb): Likewise.
4765         (gimple_switch_edge): Likewise.
4766         (gimple_switch_default_edge): Likewise.
4767         * tree-cfg.h (label_to_block_fn): Remove and replace ...
4768         (label_to_block): ... with this.
4769         (gimple_switch_label_bb): New.
4770         (gimple_switch_default_bb): Likewise.
4771         (gimple_switch_edge): Likewise.
4772         (gimple_switch_default_edge): Likewise.
4773         * tree-cfgcleanup.c (convert_single_case_switch): Use
4774         new gimple functions and pass new argument to label_to_block.
4775         (cleanup_control_flow_bb):
4776         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
4777         cfun argument explicitly.
4778         (make_eh_edges): Likewise.
4779         (redirect_eh_dispatch_edge): Likewise.
4780         (lower_resx): Likewise.
4781         (lower_eh_dispatch): Likewise.
4782         (maybe_remove_unreachable_handlers): Likewise.
4783         (unsplit_eh): Likewise.
4784         (cleanup_empty_eh): Likewise.
4785         (verify_eh_edges): Likewise.
4786         (verify_eh_dispatch_edge): Likewise.
4787         * tree-ssa-dom.c (record_edge_info): Likewise.
4788         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
4789         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
4790         (thread_through_normal_block): Likewise.
4791         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
4792         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
4793         * tree-switch-conversion.c (switch_conversion::collect): Use new
4794         gimple functions.
4795         (switch_conversion::check_final_bb): Likewise.
4796         (switch_conversion::gather_default_values): Pass new argument
4797         to label_to_block.
4798         (switch_conversion::build_constructors): Likewise.
4799         (switch_decision_tree::compute_cases_per_edge): Use new
4800         gimple_switch_edge function.
4801         (switch_decision_tree::analyze_switch_statement): Pass new argument
4802         to label_to_block.
4803         (switch_decision_tree::try_switch_expansion): Use
4804         gimple_switch_default_edge.
4805         * tree-vrp.c (find_switch_asserts): Pass new argument
4806         to label_to_block.
4807         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4808         (vr_values::simplify_switch_using_ranges): Likewise.
4810 2018-08-27  Richard Biener  <rguenther@suse.de>
4812         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4813         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4815         * tree-ssa-sccvn.h (struct vn_pval): New structure.
4816         (struct vn_nary_op_s): Add unwind_to member.  Add
4817         predicated_values flag and put result into a union together
4818         with a linked list of vn_pval.
4819         (struct vn_ssa_aux): Add name member to make maintaining
4820         a map of SSA name to vn_ssa_aux possible.  Remove no longer
4821         needed info, dfsnum, low, visited, on_sccstack, use_processed
4822         and range_info_anti_range_p members.
4823         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4824         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4825         New functions.
4826         (vn_valueize): New global.
4827         (vn_context_bb): Likewise.
4828         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4829         VN_INFO_PTR_INFO): Remove.
4830         * tree-ssa-sccvn.c: ... (rewrite)
4831         (pass_fre::execute): For -O2+ initialize loops and run
4832         RPO VN in optimistic mode (iterating).  For -O1 and -Og
4833         run RPO VN in non-optimistic mode.
4834         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4835         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4836         * doc/invoke.texi (sccvn-max-scc-size): Remove.
4837         (rpo-vn-max-loop-depth): Document.
4838         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4839         when valuezing the VUSE signals we walked out of the region.
4840         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4841         (phi_translate): Set VN context block to use for availability
4842         lookup.
4843         (compute_avail): Likewise.
4844         (pre_valueize): New function.
4845         (pass_pre::execute): Adjust to the RPO VN API.
4847         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4848         (propagate_constants_for_unrolling): Remove.
4849         (tree_unroll_loops_completely): Perform value-numbering
4850         on the unrolled bodies loop parent.
4852 2018-08-27  Richard Biener  <rguenther@suse.de>
4854         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4855         for partial antic compute.
4857 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
4859         PR rtl-optimization/87065
4860         * combine.c (simplify_if_then_else): Formatting fix.
4861         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4862         check.
4863         (known_cond): Don't return const_true_rtx for vector modes.  Use
4864         CONST0_RTX instead of const0_rtx.  Formatting fixes.
4866 2018-08-27  Martin Liska  <mliska@suse.cz>
4868         PR gcov-profile/87069
4869         * gcov.c (process_file): Record files already processed
4870         and warn about a file being processed multiple times.
4872 2018-08-27  Martin Liska  <mliska@suse.cz>
4874         PR driver/83193
4875         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4876         Set default values for x_aarch64_*_string strings.
4877         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4878         prefix.  For -mabi do not print '=ABI' in help and use
4879         <option_value> format for -msve-vector-bits and -moverride
4880         options.
4882 2018-08-26  Jeff Law  <law@redhat.com>
4884         * config/mips/frame-header-opt.c: Include "backend.h" rather than
4885         "cfg.h"
4887 2018-08-26  Marek Polacek  <polacek@redhat.com>
4889         PR c++/87029, Implement -Wredundant-move.
4890         * doc/invoke.texi: Document -Wredundant-move.
4892 2018-08-25  Martin Sebor  <msebor@redhat.com>
4894         PR tree-optimization/87059
4895         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
4896         to the same type as the other.
4897         * fold-const.c (fold_binary_loc): Assert expectation.
4899 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
4901         * config/darwin.c (machopic_legitimize_pic_address): Clean up
4902         extraneous parentheses, dead code section and formatting.
4904 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
4906         PR c++/87091
4907         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
4908         wide enough for jumps in the line-numbering to be visible.
4909         (layout::print_gap_in_line_numbering): New member function.
4910         (layout::calculate_line_spans): When using line numbering, merge
4911         line spans that are only 1 line apart.
4912         (diagnostic_show_locus): When printing line numbers, show gaps in
4913         line numbering directly, rather than printing headers.
4914         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
4915         line-numbering with multiple line spans.
4916         (selftest::test_fixit_insert_containing_newline_2): Add test of
4917         line-numbering, in which the spans are close enough to be merged.
4919 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
4921         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
4922         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
4923         * tree-vrp.c (range_is_nonnull): Remove.
4924         (range_includes_zero_p): Accept value_range instead of min/max.
4925         (extract_range_from_binary_expr_1): Do not early bail on
4926         POINTER_PLUS_EXPR.
4927         Use range_includes_zero_p instead of range_is_nonnull.
4928         (extract_range_from_unary_expr): Use range_includes_zero_p instead
4929         of range_is_nonnull.
4930         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
4931         special case VR_ANTI_RANGE.
4932         (vrp_finalize): Same.
4933         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4934         instead of min/max.
4935         (range_is_nonnull): Remove.
4936         * vr-values.c (vrp_stmt_computes_nonzero): Use
4937         range_includes_zero_p instead of range_is_nonnull.
4938         (extract_range_basic): Pass value_range to range_includes_zero_p
4939         instead of range_is_nonnull.
4941 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
4943         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4944         * emit-rtl.h (rtl_data): Remove return_bnd.
4945         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4946         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4947         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4948         (POINTER_BOUNDS_MODE): Remove definition.
4949         (make_pointer_bounds_mode): Remove.
4950         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4951         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4952         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4953         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4954         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4955         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4956         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4958         * config/i386/i386-modes.def (BND32, BND64): Remove.
4959         * config/i386/i386.c (dbx_register_map): Remove bound registers.
4960         (dbx64_register_map): Ditto.
4961         (svr4_dbx_register_map): Ditto.
4962         (indirect_thunk_bnd_needed): Remove.
4963         (indirect_thunks_bnd_used): Ditto.
4964         (indirect_return_bnd_needed): Ditto.
4965         (indirect_return_via_cx_bnd): Ditto.
4966         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4967         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4968         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
4969         (output_indirect_thunk_function): Remove handling of
4970         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4971         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4972         (ix86_save_reg): Remove handling of crtl->return_bnd.
4973         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4974         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4975         and UNSPEC_BNDLX_ADDR.
4976         (ix86_output_indirect_branch_via_reg): Remove handling of
4977         indirect_thunk_prefix_bnd.
4978         (ix86_output_indirect_branch_via_push): Ditto.
4979         (ix86_output_function_return): Ditto.
4980         (ix86_output_indirect_function_return): Ditto.
4981         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4982         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4983         (CALL_USED_REGISTERS): Ditto.
4984         (REG_ALLOC_ORDER): Update for removal of bound registers.
4985         (HI_REGISTER_NAMES): Ditto.
4986         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4987         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4988         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4989         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4990         (FIRST_PSEUDO_REG): Update.
4991         (BND): Remove mode iterator.
4992         * config/i386/predicates.md (bnd_mem_operator): Remove.
4994 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
4996         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4997         vectors.
4999 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5001         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5002         the case in which the permute needs only a single element and
5003         repeats for every vector of the result.  Extend that case to
5004         handle variable-length vectors.
5005         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5007 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
5009         PR debug/79342
5010         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5011         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5013 2018-08-24  Richard Biener  <rguenther@suse.de>
5015         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5016         bb_flags_allocated members.
5017         (auto_flag): New RAII class for allocating flags.
5018         (auto_edge_flag): New RAII class for allocating edge flags.
5019         (auto_bb_flag): New RAII class for allocating bb flags.
5020         * cfgloop.c (verify_loop_structure): Allocate temporary edge
5021         flag dynamically.
5022         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5023         in favor of temporarily allocated BB flag.
5024         * hsa-brig.c: Re-order includes.
5025         * hsa-dump.c: Likewise.
5026         * hsa-regalloc.c: Likewise.
5027         * print-rtl.c: Likewise.
5028         * profile-count.c: Likewise.
5030 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
5032         PR target/86989
5033         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5034         the TOC register.
5036 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5038         PR 87073/bootstrap
5039         * wide-int-range.cc (wide_int_range_div): Do not ignore result
5040         from wide_int_range_multiplicative_op.
5042 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5044         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5045         "permutaion".
5047 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5049         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
5050         to 'expanded'.
5052 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
5054         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
5055         full GENERIC expression used for replacement.
5057 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
5059         * tree-vrp.c (abs_extent_range): Remove.
5060         (extract_range_into_wide_ints): Pass wide ints by reference.
5061         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5062         Pass wide ints by reference in all calls to
5063         extract_range_into_wide_ints.
5064         * wide-int-range.cc (wide_int_range_div): New.
5065         * wide-int-range.h (wide_int_range_div): New.
5066         (wide_int_range_includes_zero_p): New.
5067         (wide_int_range_zero_p): New.
5069 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
5071         * config/aarch64/aarch64.md (arches): New enum.
5072         (arch): New enum attr.
5073         (arch_enabled): New attr.
5074         (enabled): Now uses arch_enabled only.
5075         (simd, sve, fp16): Removed attribute.
5076         (fp): Attr now defined in terms of 'arch'.
5077         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5078         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5079         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5080         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5081         attributes into 'arch'.
5082         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5083         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5084         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5085         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5086         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5087         'simd' attribute into 'arch'.
5088         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5089         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5090         Convert use of 'fp' attribute to 'arch'.
5091         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5092         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5093         into 'arch'.
5094         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5095         (different modes) Merge 'fp' and 'simd' into 'arch'.
5096         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5097         'simd' into 'arch'.
5099 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
5101         PR rtl-optimization/87026
5102         * expmed.c (canonicalize_comparison): If we can no longer create
5103         pseudoregisters, don't.
5105 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
5107         PR target/86951
5108         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
5109         prototype.
5110         * config/arm/arm.c (speculation_barrier_libfunc): New static
5111         variable.
5112         (arm_init_libfuncs): Initialize it.
5113         (arm_emit_speculation_barrier): New function.
5114         * config/arm/arm.md (speculation_barrier): Call
5115         arm_emit_speculation_barrier for architectures that do not have
5116         DSB or ISB.
5117         (speculation_barrier_insn): Only match on Armv7 or later.
5119 2018-08-23  Richard Biener  <rguenther@suse.de>
5121         PR middle-end/87024
5122         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
5123         calls.
5125 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5127         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
5128         of single-vector TBLs.
5129         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
5130         one input is given.
5132 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5134         PR target/85910
5135         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
5136         aarch64_evpc_tbl guard.
5138 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5140         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
5141         behaviour.
5143 2018-08-22  Martin Sebor  <msebor@redhat.com>
5145         PR middle-end/87052
5146         * tree-pretty-print.c (pretty_print_string): Add argument.
5147         (dump_generic_node): Call to pretty_print_string with string size.
5149 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
5151         PR rtl-optimization/86771
5152         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
5153         of two SETs into those two SETs, one to be placed at i2, if that SETs
5154         destination is modified between i2 and i3.
5156 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5158         PR tree-optimization/86725
5159         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
5160         function.
5161         (vect_analyze_scalar_cycles_1): Check it.
5163 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5165         PR tree-optimization/86725
5166         * tree-vect-loop.c (vect_is_simple_reduction): When treating
5167         an outer loop phi as a double reduction, make sure that the
5168         single user of the phi result is an inner loop phi.
5170 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5172         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
5173         grouped stores with gaps to a strided group.
5175 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5177         * tree-vect-stmts.c (get_group_load_store_type)
5178         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
5179         first statement in a group.
5181 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5183         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
5184         the sequence used in gcc/gcc.c.
5186 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5188         PR other/704
5189         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
5190         building it.
5192 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5194         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
5195         Darwin10-specific unwinder-shim.
5196         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
5197         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
5198         New to cater for Darwin10 Rosetta.
5200 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5202         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
5203         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
5205 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5207         PR bootstrap/81033
5208         PR target/81733
5209         PR target/52795
5210         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
5211         (dwarf2out_switch_text_section): Generate a local label for the
5212         second function sub-section and apply it as the second FDE start
5213         label.
5214         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
5215         second sub-section start.
5217 2018-08-22  Richard Biener  <rguenther@suse.de>
5219         PR tree-optimization/86988
5220         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
5222 2018-08-22  Richard Biener  <rguenther@suse.de>
5224         PR tree-optimization/86945
5225         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
5227 2018-08-22  Alexandre Oliva <oliva@adacore.com>
5229         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
5230         a comment about how uses of r2 for .sdata2 come about.
5232 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
5234         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
5236 2018-08-21  Marek Polacek  <polacek@redhat.com>
5238         PR c++/86981, Implement -Wpessimizing-move.
5239         * doc/invoke.texi: Document -Wpessimizing-move.
5241 2018-08-21  Jan Hubicka  <jh@suse.cz>
5243         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
5244         * tree.h (is_redundant_typedef): Remove.
5245         * dwarf2out.c (is_redundant_typedef): Turn into static function.
5247 2018-08-21  Jan Hubicka  <jh@suse.cz>
5249         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
5250         when possible.
5252 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
5254         * expmed.c (extract_low_bits): Reject invalid subregs early.
5256 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5258         PR middle-end/86121
5259         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
5260         behaviour.
5262 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5264         * config/vxworks.h: Guard vxworks_asm_out_constructor and
5265         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
5266         * config/vxworks.c: Likewise.
5268 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5270         * config/vxworks.c: Set targetm.have_ctors_dtors
5271         if HAVE_INITFINI_ARRAY_SUPPORT.
5272         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
5273         if HAVE_INITFINI_ARRAY_SUPPORT.
5275 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5277         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
5278         default search path for VxWorks < 7.
5280 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5282         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
5283         (get_format_string): Refer to c_getstr.
5285 2018-08-21  Tom de Vries  <tdevries@suse.de>
5287         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5288         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5289         (debuginfo_early_stop): Declare.
5290         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5291         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5292         (debuginfo_early_stop): New function.
5293         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
5294         and debuginfo_early_stop.
5295         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
5296         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
5297         (general_init): Call debuginfo_early_init.
5298         (finalize): Call debuginfo_fini.
5299         (do_compile): Call debuginfo_init.
5300         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
5301         -fdump-early-debug.
5302         (@item -fdump-debug, @item -fdump-earlydebug): Add.
5304 2018-08-21  Tom de Vries  <tdevries@suse.de>
5306         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
5307         flag_dump_noaddr and flag_dump_unnumbered.
5309 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
5311         * wide-int-range.cc (wide_int_range_abs): New.
5312         (wide_int_range_order_set): Rename from wide_int_range_min_max.
5313         * wide-int-range.h (wide_int_range_abs): New.
5314         (wide_int_range_min_max): New.
5315         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
5316         case to call wide_int_range_abs.
5317         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
5318         (extract_range_from_abs_expr): Delete.
5320 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
5322         PR target/87033
5323         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
5324         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
5325         for indexed loads.
5327 2018-08-20  Nathan Sidwell  <nathan@acm.org>
5328             Jeff Law <law@redhat.com>
5330         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
5331         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
5333 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
5335         PR other/84889
5336         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
5337         (decl_attributes): Likewise.
5338         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
5339         instance.
5340         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5341         * diagnostic-core.h (class auto_diagnostic_group): New class.
5342         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
5343         (diagnostic_report_diagnostic): Handle the first diagnostics within
5344         a group.
5345         (emit_diagnostic): Add auto_diagnostic_group instance.
5346         (inform): Likewise.
5347         (inform_n): Likewise.
5348         (warning): Likewise.
5349         (warning_at): Likewise.
5350         (warning_n): Likewise.
5351         (pedwarn): Likewise.
5352         (permerror): Likewise.
5353         (error): Likewise.
5354         (error_n): Likewise.
5355         (error_at): Likewise.
5356         (sorry): Likewise.
5357         (fatal_error): Likewise.
5358         (internal_error): Likewise.
5359         (internal_error_no_backtrace): Likewise.
5360         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
5361         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
5362         * diagnostic.h (struct diagnostic_context): Add fields
5363         "diagnostic_group_nesting_depth",
5364         "diagnostic_group_emission_count", "begin_group_cb",
5365         "end_group_cb".
5366         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
5367         Add auto_diagnostic_group instance(s).
5368         (find_explicit_erroneous_behavior): Likewise.
5369         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
5370         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
5371         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
5372         (gimplify_va_arg_expr): Likewise.
5373         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
5374         (HSA_SORRY_AT): Likewise.
5375         * ipa-devirt.c (compare_virtual_tables): Likewise.
5376         (warn_odr): Likewise.
5377         * multiple_target.c (expand_target_clones): Likewise.
5378         * opts-common.c (cmdline_handle_error): Likewise.
5379         * reginfo.c (globalize_reg): Likewise.
5380         * substring-locations.c (format_warning_n_va): Likewise.
5381         * tree-inline.c (expand_call_inline): Likewise.
5382         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
5383         * tree-ssa-loop-niter.c
5384         (do_warn_aggressive_loop_optimizations): Likewise.
5385         * tree-ssa-uninit.c (warn_uninit): Likewise.
5386         * tree.c (warn_deprecated_use): Likewise.
5388 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
5390         PR target/87014
5391         * config/i386/i386.md (eh_return): Always update EH return
5392         address in word_mode.
5394 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
5396         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
5397         TARGET_SPLIT_COMPLEX_ARG is defined.
5399 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5401         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
5403 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5405         PR target/86984
5406         * expr.c (expand_assignment): Assert that bitpos is positive.
5407         (store_field): Likewise
5408         (expand_expr_real_1): Make sure that bitpos is positive.
5409         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
5410         integer overflow.
5412 2018-08-20  Nathan Sidwell  <nathan@acm.org>
5414         * Makefile.in (CPP_ID_DATA_H): Delete.
5415         (CPP_INTERNAL_H): Don't add it.
5416         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
5417         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
5419 2018-08-20  Richard Biener  <rguenther@suse.de>
5421         PR tree-optimization/78655
5422         * tree-vrp.c (extract_range_from_binary_expr_1): Make
5423         pointer + offset nonnull if either operand is nonnull work.
5425 2018-08-20  Tom de Vries  <tdevries@suse.de>
5427         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
5428         unless the referenced die describes the added property using
5429         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
5430         Otherwise, add a DW_AT_location to the referenced die.
5432 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
5434         PR target/86994
5435         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
5436         register_operand when calling ix86_set_reg_reg_cost.
5437         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
5438         Set *total to 0 for operands that satisfy x86_64_immediate_operand
5439         predicate and to 1 otherwise.
5441 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
5443         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
5444         emit a diagnostic that it is not supported and reset the option.
5445         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
5446         supported and consume the option.  (ASM_FINAL_SPEC): New.
5448 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
5450         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
5451         a sentence.
5453 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
5455         C-SKY port: Documentation
5457         * doc/extend.texi (C-SKY Function Attributes): New section.
5458         * doc/invoke.texi (Option Summary): Add C-SKY options.
5459         (C-SKY Options): New section.
5460         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
5462 2018-08-17  Jojo  <jijie_rong@c-sky.com>
5463             Huibin Wang  <huibin_wang@c-sky.com>
5464             Sandra Loosemore  <sandra@codesourcery.com>
5465             Chung-Lin Tang  <cltang@codesourcery.com>
5467         C-SKY port: Backend implementation
5469         * config/csky/*: New.
5470         * common/config/csky/*: New.
5472 2018-08-17  Jojo  <jijie_rong@c-sky.com>
5473             Huibin Wang  <huibin_wang@c-sky.com>
5474             Sandra Loosemore  <sandra@codesourcery.com>
5475             Chung-Lin Tang  <cltang@codesourcery.com>
5476             Andrew Jenner  <andrew@codesourcery.com>
5478         C-SKY port: Configury
5480         * config.gcc (csky-*-*): New.
5481         * configure.ac: Add csky to targets for dwarf2 debug_line support.
5482         * configure: Regenerated.
5484 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
5486         * dump-context.h: Include "dumpfile.h".
5487         (dump_context::dump_printf_va): Convert final param from va_list
5488         to va_list *.  Convert from ATTRIBUTE_PRINTF to
5489         ATTRIBUTE_GCC_DUMP_PRINTF.
5490         (dump_context::dump_printf_loc_va): Likewise.
5491         * dumpfile.c: Include "stringpool.h".
5492         (make_item_for_dump_printf_va): Delete.
5493         (make_item_for_dump_printf): Delete.
5494         (class dump_pretty_printer): New class.
5495         (dump_pretty_printer::dump_pretty_printer): New ctor.
5496         (dump_pretty_printer::emit_items): New member function.
5497         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
5498         function.
5499         (dump_pretty_printer::emit_item): New member function.
5500         (dump_pretty_printer::stash_item): New member function.
5501         (dump_pretty_printer::format_decoder_cb): New member function.
5502         (dump_pretty_printer::decode_format): New member function.
5503         (dump_context::dump_printf_va): Reimplement in terms of
5504         dump_pretty_printer.
5505         (dump_context::dump_printf_loc_va): Convert final param from va_list
5506         to va_list *.
5507         (dump_context::begin_scope): Reimplement call to
5508         make_item_for_dump_printf.
5509         (dump_printf): Update for change to dump_printf_va.
5510         (dump_printf_loc): Likewise.
5511         (selftest::test_capture_of_dump_calls): Convert "stmt" from
5512         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
5513         with %T, %E, and %G.
5514         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
5515         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
5516         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
5517         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
5518         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
5519         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
5520         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
5521         within a dump_printf_loc call to "%wu".
5522         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
5523         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
5524         missing space after "=".
5525         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
5526         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
5527         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
5528         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
5529         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
5530         duplicate "vectorized" from message.
5532 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
5534         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
5535         polyNxK_t element's TYPE_STRING_FLAG.
5537 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
5539         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
5540         (they were unnamed before).  Fix comments.
5542 2018-08-17  Nathan Sidwell  <nathan@acm.org>
5544         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
5546 2018-08-17  Richard Biener  <rguenther@suse.de>
5548         PR tree-optimization/86841
5549         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
5551 2018-08-17  Martin Liska  <mliska@suse.cz>
5553         * common.opt: Remove Warn, Init and Report for options with
5554         Ignore/Deprecated flag. Warning is done automatically for
5555         Deprecated flags.
5556         * config/i386/i386.opt: Likewise.
5557         * config/ia64/ia64.opt: Likewise.
5558         * config/rs6000/rs6000.opt: Likewise.
5559         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5560         Remove usage of flag_check_pointer_bounds.
5561         * lto-wrapper.c (merge_and_complain): Do not handle
5562         OPT_fcheck_pointer_bounds.
5563         (append_compiler_options): Likewise.
5564         * opt-functions.awk: Do not handle Deprecated.
5565         * optc-gen.awk: Check that Var, Report and Init are not
5566         used for an option with Ignore/Deprecated flag.
5567         * opts-common.c (decode_cmdline_option): Do not report
5568         CL_ERR_DEPRECATED.
5569         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
5570         options.
5571         * opts.h (struct cl_option): Remove cl_deprecated flag.
5572         (CL_ERR_DEPRECATED): Remove error enum value.
5574 2018-08-17  Richard Biener  <rguenther@suse.de>
5576         PR middle-end/86505
5577         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
5578         across a va-arg-pack using call adjust its return value accordingly.
5580 2018-08-16  Martin Sebor  <msebor@redhat.com>
5582         PR tree-optimization/86853
5583         * gimple-ssa-sprintf.c (struct format_result): Rename member.
5584         (struct fmtresult): Add member and initialize it in ctors.
5585         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
5586         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
5587         (format_directive): Set POSUNDER4K when MAYFAIL is set.
5588         (parse_directive): Handle %C same as %c.
5589         (sprintf_dom_walker::compute_format_length): Adjust.
5590         (is_call_safe): Adjust.
5592 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5594         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
5595         for determining how to count the elements.
5596         * builtins.h (c_strlen): Adjust prototype.
5597         * expr.c (string_constant): Add new parameter mem_size.
5598         Set *mem_size appropriately.
5599         * expr.h (string_constant): Adjust protoype.
5600         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
5601         * gimple-fold.h (get_range_strlen): Adjust prototype.
5602         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
5603         (format_string): Call get_string_length with eltsize.
5605 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
5607         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
5608         to emit the span, rather than setting it as the prefix.
5610 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
5612         * diagnostic-show-locus.c (layout::start_annotation_line): Add
5613         "margin_char" parameter, defaulting to space.  Use it in place
5614         of pp_space for the initial part of the margin.
5615         (layout::print_leading_fixits): Use '+' when filling the margin
5616         of line-insertion fix-it hints.
5618 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5620         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
5621         Delete.
5623 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5625         * config/rs6000/altivec.md: Don't set length attribute to the default
5626         value.
5627         * config/rs6000/darwin.md: Ditto.
5628         * config/rs6000/dfp.md: Ditto.
5629         * config/rs6000/htm.md: Ditto.
5630         * config/rs6000/rs6000.md: Ditto.
5631         * config/rs6000/sync.md: Ditto.
5632         * config/rs6000/vsx.md: Ditto.
5634 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5636         * config/rs6000/altivec.md: Don't set length attribute to the default
5637         value, for branch instructions.
5638         * config/rs6000/darwin.md: Ditto.
5639         * config/rs6000/rs6000.md: Ditto.
5641 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
5643         * config/rs6000/rs6000.md (length): Always define as const_int 4.
5644         (unnamed conditional branch define_insn): Set length to 4 or 8
5645         depending on offset.
5646         (<bd>_<mode>): Similar, for alternative 0.
5647         (<bd>tf_<mode>): Ditto.
5649 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5651         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
5653 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
5655         * doc/rtl.texi: Replace old RTX class names with new names.
5658 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
5660         * expmed.h (canonicalize_comparison): New declaration.
5661         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
5662         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
5663         * optabs.c (prepare_cmp_insn): Likewise.
5664         * rtl.h (unsigned_condition_p): New function which checks if a
5665         comparison operator is unsigned.
5667 2018-08-16  Nathan Sidwell  <nathan@acm.org>
5669         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
5670         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
5672 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5674         PR target/84711
5675         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
5676         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
5677         (mov<mov>): ..this and enable unconditionally.
5679 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
5681         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
5683 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
5685         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
5686         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
5687         "Common" with "Target".
5689 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
5691         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
5692         * doc/invoke.texi (mmitigate-rop): Remove.
5693         * config/i386/i386.c: Do not include "regrename.h".
5694         (ix86_rop_should_change_byte_p, reg_encoded_number)
5695         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
5696         Remove.
5697         (ix86_reorg): Remove call to ix86_mitigate_rop.
5698         * config/i386/i386.md (attr "modrm_class"): Remove.
5699         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
5700         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
5701         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
5703 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
5705         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
5706         allow folding of mergeh() and mergel() for the float and double types.
5707         (fold_mergehl_helper): Rework to handle building a permute tree
5708         for float vectors.
5710 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
5712         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
5713         for TARGET_SSE.
5715 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
5717         * common.opt (fdiagnostics-show-labels): New option.
5718         * diagnostic-show-locus.c (class layout_range): Add field
5719         "m_label".
5720         (class layout): Add field "m_show_labels_p".
5721         (layout_range::layout_range): Add param "label" and use it to
5722         initialize m_label.
5723         (make_range): Pass in NULL for new "label" param of layout_range's
5724         ctor.
5725         (layout::layout): Initialize m_show_labels_p.
5726         (layout::maybe_add_location_range): Pass in loc_range->m_label
5727         when constructing layout_range instances.
5728         (struct line_label): New struct.
5729         (layout::print_any_labels): New member function.
5730         (layout::print_line): Call it if label-printing is enabled.
5731         (selftest::test_one_liner_labels): New test.
5732         (selftest::test_diagnostic_show_locus_one_liner): Call it.
5733         * diagnostic.c (diagnostic_initialize): Initialize
5734         context->show_labels_p.
5735         * diagnostic.h (struct diagnostic_context): Add field
5736         "show_labels_p".
5737         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5738         -fno-diagnostics-show-labels.
5739         * dwarf2out.c (gen_producer_string): Add
5740         OPT_fdiagnostics_show_labels to the ignored options.
5741         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
5742         param.
5743         (gcc_rich_location::maybe_add_expr): Likewise.
5744         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
5745         label" param, defaulting to NULL.
5746         (gcc_rich_location::add_expr): Add "label" param.
5747         (gcc_rich_location::maybe_add_expr): Likewise.
5748         (class text_range_label): New class.
5749         (class range_label_for_type_mismatch): New class.
5750         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
5751         of format_warning_va.
5752         (fmtwarn_n): Likewise for new params of format_warning_n_va.
5753         * lto-wrapper.c (merge_and_complain): Add
5754         OPT_fdiagnostics_show_labels to the "pick one setting" options.
5755         (append_compiler_options): Likewise to the dropped options.
5756         (append_diag_options): Likewise to the passed-on options.
5757         * opts.c (common_handle_option): Handle the new option.
5758         * selftest-diagnostic.c
5759         (test_diagnostic_context::test_diagnostic_context): Enable
5760         show_labels_p.
5761         * substring-locations.c: Include "gcc-rich-location.h".
5762         (format_warning_n_va): Add "fmt_label" and "param_label" params
5763         and use them as appropriate.
5764         (format_warning_va): Add "fmt_label" and "param_label" params,
5765         passing them on to format_warning_n_va.
5766         (format_warning_at_substring): Likewise.
5767         (format_warning_at_substring_n): Likewise.
5768         * substring-locations.h (format_warning_va): Add "fmt_label" and
5769         "param_label" params.
5770         (format_warning_n_va): Likewise.
5771         (format_warning_at_substring): Likewise.
5772         (format_warning_at_substring_n): Likewise.
5773         * toplev.c (general_init): Initialize global_dc->show_labels_p.
5775 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
5777         PR testsuite/86519
5778         * builtins.c (expand_builtin_memcmp): Do not expand the call
5779         when overflow is detected.
5781 2018-08-15  Martin Sebor  <msebor@redhat.com>
5783         PR tree-optimization/71625
5784         * config/aarch64/aarch64-builtins.c
5785         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
5787 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
5789         * config/s390/s390.c (s390_reorg): Remove loop.
5791 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
5793         * config/darwin.c
5794          (darwin_function_switched_text_sections): Delete.
5795         * gcc/config/darwin.h
5796          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
5798 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
5800         PR target/81685
5801         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
5802         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
5803         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
5805 2018-08-15  Martin Liska  <mliska@suse.cz>
5807         PR tree-optimization/86925
5808         * predict.c (expr_expected_value_1): When taking
5809         later predictor, assign also probability.
5810         Use fold_build2_initializer_loc in order to fold
5811         the expression in -frounding-math.
5813 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
5815         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5816         patterns.
5817         (expand_vec_perm_1): Try the new method.
5819 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
5821         PR target/86547
5822         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5823         Check whether lra_live_max_point is 0 before dividing.
5825 2018-08-14  Martin Sebor  <msebor@redhat.com>
5827         PR tree-optimization/86650
5828         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5829         (vrp_prop::check_mem_ref): Same.
5831 2018-08-13  Liu Hao <lh_mouse@126.com>
5833         * pretty-print.c (eat_esc_sequence): Swap the foreground and
5834         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5835         and clear it thereafter, as it only works for DBCS.
5837 2018-08-13  Liu Hao <lh_mouse@126.com>
5839         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5840         handle returned by _get_osf_handle().
5842 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5844         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5845         for folding vec_perm.
5847 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
5849         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5850         Add support for gimple-folding of vec_pack() and vec_unpack()
5851         intrinsics.
5853 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
5855         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5856         vec_xst variants to the list.
5857         (rs6000_gimple_fold_builtin): Add support for folding unaligned
5858         vector loads and stores.
5860 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
5862         * config.gcc (rs6000-ibm-aix4.x): Delete.
5863         (rs6000-ibm-aix5.1): Delete.
5864         (rs6000-ibm-aix5.2): Delete.
5865         (rs6000-ibm-aix5.3): Delete.
5866         * config/rs6000/aix43.h: Delete.
5867         * config/rs6000/aix51.h: Delete.
5868         * config/rs6000/aix52.h: Delete.
5869         * config/rs6000/t-aix43: Delete.
5871 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
5873         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5874         New function.
5875         (s390_decompose_address): Factor out constant pool ref
5876         decomposition.
5878 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5880         * config/nds32/nds32-predicates.c
5881         (nds32_can_use_bclr_p): Change return type as bool.
5882         (nds32_can_use_bset_p): Ditto.
5883         (nds32_can_use_btgl_p): Ditto.
5884         (nds32_can_use_bitci_p): Ditto.
5885         * config/nds32/nds32-protos.h
5886         (nds32_can_use_bclr_p): Change declaration.
5887         (nds32_can_use_bset_p): Ditto.
5888         (nds32_can_use_btgl_p): Ditto.
5889         (nds32_can_use_bitci_p): Ditto.
5891 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5893         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
5894         Support -msched-prolog-epilog option.
5895         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
5897 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5899         * common/config/nds32/nds32-common.c
5900         (nds32_option_optimization_table): Enalbe -malways-align.
5902 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
5904         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
5905         extra_headers.
5906         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
5907         OPT_misr_secure_ case.
5908         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
5909         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
5910         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
5911         secure attribute.
5912         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
5913         (nds32_isr_info): New field security_level.
5914         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
5915         * config/nds32/nds32.md (return_internal): Consider critical attribute.
5916         * config/nds32/nds32.opt (misr-secure): New option.
5917         * config/nds32/nds32_init.inc: New file.
5918         * config/nds32/nds32_isr.h: New file.
5920 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
5922         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
5923         Update comment for atomic instructions.
5924         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
5925         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5926         Remove.
5927         (atomic_loaddi): Revise fence expansion to only emit fence prior to
5928         load for __ATOMIC_SEQ_CST model.
5929         (atomic_loaddi_1): Remove float register target.
5930         (atomic_storedi): Handle CONST_INT values.
5931         (atomic_storedi_1): Remove float register source.  Add special case
5932         for zero value.
5933         (memory_barrier): New expander and insn.
5935 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
5937         PR tree-optimization/86835
5938         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5939         new_stmt after def_gsi, make sure to insert new_square_stmt after
5940         that stmt, not 2 stmts before it.
5942 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
5944         PR target/82418
5945         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5946         instead of SWI48.
5948 2018-08-10  Martin Liska  <mliska@suse.cz>
5950         PR target/83610
5951         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5952         function type.
5953         * builtins.c (expand_builtin_expect_with_probability):
5954         New function.
5955         (expand_builtin_expect_with_probability): New function.
5956         (build_builtin_expect_predicate): Add new argumnet probability
5957         for BUILT_IN_EXPECT_WITH_PROBABILITY.
5958         (fold_builtin_expect):
5959         (fold_builtin_2):
5960         (fold_builtin_3):
5961         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5962         * builtins.h (fold_builtin_expect): Set new argument.
5963         * doc/extend.texi: Document __builtin_expect_with_probability.
5964         * doc/invoke.texi: Likewise.
5965         * gimple-fold.c (gimple_fold_call): Pass new argument.
5966         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5967         also BUILT_IN_EXPECT_WITH_PROBABILITY.
5968         * predict.c (get_predictor_value): New function.
5969         (expr_expected_value): Add new argument probability. Assume
5970         that predictor and probability are always non-null.
5971         (expr_expected_value_1): Likewise.  For __builtin_expect and
5972         __builtin_expect_with_probability set probability.  Handle
5973         combination in binary expressions.
5974         (tree_predict_by_opcode): Simplify code by simply calling
5975         get_predictor_value.
5976         (pass_strip_predict_hints::execute): Add handling of
5977         BUILT_IN_EXPECT_WITH_PROBABILITY.
5978         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5979         new predictor.
5980         * tree.h (DECL_BUILT_IN_P): New function.
5982 2018-08-10  Martin Liska  <mliska@suse.cz>
5984         PR tree-optimization/85799
5985         * passes.def: Add argument for pass_strip_predict_hints.
5986         * predict.c (class pass_strip_predict_hints): Add new argument
5987         early_p.
5988         (strip_predictor_early): New function.
5989         (pass_strip_predict_hints::execute): Call the function to
5990         strip predictors.
5991         (strip_predict_hints): New function.
5992         * predict.def: Fix comment.
5994 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
5996         * Makefile.in: Clarify which tm.texi to copy over to assert the
5997         right to grant a GFDL license for all.
5999 2018-08-09  Jeff Law  <law@redhat.com>
6001         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6002         unused variable.
6004 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
6006         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6007         prototype.
6009 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6011         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6012         reductions for variable-length vectors.
6014 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
6016         PR other/84889
6017         * common.opt (fdiagnostics-show-line-numbers): New option.
6018         * diagnostic-show-locus.c (class layout): Add fields
6019         "m_show_line_numbers_p" and "m_linenum_width";
6020         (num_digits): New function.
6021         (test_num_digits): New function.
6022         (layout::layout): Initialize new fields.  Update m_x_offset
6023         logic to handle any left margin.
6024         (layout::print_source_line): Print line number when requested.
6025         (layout::start_annotation_line): New member function.
6026         (layout::print_annotation_line): Call it.
6027         (layout::print_leading_fixits): Likewise.
6028         (layout::print_trailing_fixits): Likewise.  Update calls to
6029         move_to_column for new parameter.
6030         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6031         it to potentially call start_annotation_line.
6032         (layout::show_ruler): Call start_annotation_line.
6033         (selftest::test_line_numbers_multiline_range): New selftest.
6034         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6035         and selftest::test_line_numbers_multiline_range.
6036         * diagnostic.c (diagnostic_initialize): Initialize
6037         show_line_numbers_p.
6038         * diagnostic.h (struct diagnostic_context): Add field
6039         "show_line_numbers_p".
6040         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6041         -fno-diagnostics-show-line-numbers.
6042         * dwarf2out.c (gen_producer_string): Add
6043         OPT_fdiagnostics_show_line_numbers to the ignored options.
6044         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6045         one setting" options.
6046         (append_compiler_options): Likewise to the dropped options.
6047         (append_diag_options): Likewise to the passed-on options.
6048         * opts.c (common_handle_option): Handle the new option.
6049         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
6051 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6053         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
6054         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
6055         third argument of type "const signed char" to descriptions of
6056         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
6057         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
6058         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6059         __builtin_bcdsub_ov functions.
6061 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6063         PR tree-optimization/86858
6064         * tree-vect-loop.c (vect_is_simple_reduction): Restore
6065         flow_bb_inside_loop_p calls.
6067 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6069         PR tree-optimization/86871
6070         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6071         instead of gimple_assign_lhs.
6073 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
6075         PR target/86887
6076         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6077         register constraint to operand 0.
6078         (add<mode>3_carryinC): Likewise.
6079         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6081 2018-08-09  Martin Liska  <mliska@suse.cz>
6083         PR c/86895
6084         * common.opt: Remove extra line.
6086 2018-08-09  Martin Liska  <mliska@suse.cz>
6088         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6089         at the end of a line, make first letter capital and end up
6090         a sentence with a dot.
6091         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6092         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6093         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6094         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6095         (PARAM_MAX_ISL_OPERATIONS): Likewise.
6096         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6097         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6098         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6099         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6100         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6101         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6102         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6103         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
6104         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
6105         (PARAM_TREE_REASSOC_WIDTH): Likewise.
6106         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
6107         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
6108         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
6110 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
6112         PR target/84332
6113         * config/s390/s390.c (s390_option_override_internal): Reduce the
6114         stack-clash-protection-probe-interval param if it would be too big
6115         for z900.
6117 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
6119         PR target/46179
6120         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
6121         * config/m68k/m68k.c (handle_move_double): Don't call
6122         m68k_final_prescan_insn.
6123         (m68k_adjust_decorated_operand): Renamed from
6124         m68k_final_prescan_insn, remove first and third operand and
6125         simplify.
6126         (print_operand): Call it.
6127         (print_operand_address): Call it.
6129 2018-08-08  Nathan Sidwell  <nathan@acm.org>
6131         * diagnostic.c (diagnostic_report_current_module): Use
6132         linemap_included_from & linemap_included_from_linemap.
6134 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
6136         * config/aarch64/aarch64-cores.def: Add phecda core.
6137         * config/aarch64/aarch64-tune.md: Regenerate.
6138         * doc/invoke.texi: Add phecda core.
6140 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6142         PR target/85295
6143         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
6144         definitions.
6145         * config/s390/s390.md ("movti"): Add more alternatives for
6146         constant to GPR copies.
6148 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6150         * config/s390/s390.c: Fix whitespace damage throughout the file.
6151         * config/s390/s390.h: Likewise.
6152         * config/s390/tpf.h: Likewise.
6154 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6156         * config/s390/s390.c (s390_loadrelative_operand_p):
6157         Remove TARGET_CPU_ZARCH usages.
6158         (s390_rtx_costs): Likewise.
6159         (s390_legitimate_constant_p): Likewise.
6160         (s390_cannot_force_const_mem): Likewise.
6161         (legitimate_reload_constant_p): Likewise.
6162         (s390_preferred_reload_class): Likewise.
6163         (legitimize_pic_address): Likewise.
6164         (legitimize_tls_address): Likewise.
6165         (s390_split_branches): Removed.
6166         (s390_add_execute): Removed.
6167         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
6168         (s390_mainpool_start): Likewise.
6169         (s390_mainpool_finish): Likewise.
6170         (s390_mainpool_cancel): Removed.
6171         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
6172         (s390_chunkify_cancel): Likewise.
6173         (s390_return_addr_rtx): Likewise.
6174         (s390_register_info): Remove split_branches_pending_p uages.
6175         (s390_optimize_register_info): Likewise.
6176         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
6177         split_branches_pending_p usages.
6178         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
6179         (s390_load_got): Likewise.
6180         (s390_expand_split_stack_prologue): Likewise.
6181         (output_asm_nops): Likewise.
6182         (s390_function_profiler): Likewise.
6183         (s390_emit_call): Likewise.
6184         (s390_conditional_register_usage): Likewise.
6185         (s390_optimize_prologue): Likewise.
6186         (s390_reorg): Remove TARGET_CPU_ZARCH and
6187         split_branches_pending_p usages.
6188         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
6189         usages.
6190         (s390_output_indirect_thunk_function): Likewise.
6191         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
6192         (TARGET_CPU_ZARCH_P): Removed.
6193         (struct machine_function): Remove split_branches_pending_p.
6194         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
6196 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6198         * common/config/s390/s390-common.c (processor_flags_table):
6199         Remove flags.
6200         * config.gcc: Remove with_arch/with_tune support.
6201         * config/s390/2064.md: Remove cpu attribute comparisons.
6202         * config/s390/driver-native.c (s390_host_detect_local_cpu):
6203         Remove MTN.
6204         * config/s390/linux.h (ASM_SPEC):
6205         Remove -march support.
6206         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
6207         Use a table to get an arch level.
6208         * config/s390/s390-opts.h (enum processor_type):
6209         Remove enum values.
6210         * config/s390/s390.c
6211         (processor_table): Remove entries, add arch_level values.
6212         (s390_issue_rate): Remove cases.
6213         (s390_option_override): Adjust
6214         s390_option_override_internal() call.
6215         (s390_option_override_internal): Remove deprecation warning.
6216         (s390_valid_target_attribute_tree): Adjust
6217         s390_option_override_internal() call.
6218         * config/s390/s390.h (struct s390_processor):
6219         Share with s390-c.c, add arch_level field.
6220         * config/s390/s390.md:
6221         Remove occurrences in cpu attribute.
6222         * config/s390/s390.opt: Remove -march/-mtune support.
6223         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
6224         * doc/invoke.texi: Remove deprecation warning.
6226 2018-08-08  Luis Machado  <luis.machado@linaro.org>
6228         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
6229         global.
6230         (qdf24xx_tunings): Set vector cost structure to
6231         qdf24xx_vector_cost.
6233         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
6234         <register_sextend>: Set to 3.
6236 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
6238         PR target/86838
6239         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
6240         * config/aarch64/aarch64-simd.md
6241         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
6242         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
6243         (aarch64_frecpx<mode>): ...this new pattern.
6244         * config/aarch64/aarch64-simd-builtins.def: Remove comment
6245         about aarch64_frecp<FRECP:frecp_suffix><mode>.
6247 2018-08-07  Martin Liska  <mliska@suse.cz>
6249         PR middle-end/83023
6250         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
6251         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
6252         * predict.def (PRED_MALLOC_NONNULL): New predictor.
6253         * doc/extend.texi: Document that malloc attribute adds
6254         hit to compiler.
6256 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
6258         PR target/86785
6259         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6260         Define to speculation_safe_value_not_needed.
6262 2018-08-06  Jeff Law  <law@redhat.com>
6264         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
6265         the vr_values instance to cprop_into_stmt.
6266         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
6267         (cprop_operand): Also query EVRP to determine if OP is a constant.
6269 2018-08-06  Nathan Sidwell  <nathan@acm.org>
6271         * diagnostic.c (diagnostic_report_current_module): Reroll
6272         included-at loop.  Translate text.
6274 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
6276         * function-tests.c (selftest::test_expansion_to_rtl): Call
6277         free_after_compilation.
6279 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6281         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
6283 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
6285         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
6286         loops with memory block operations from getting unrolled.
6288 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
6290         PR target/86807
6291         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6292         Define to speculation_safe_value_not_needed.
6294 2018-08-06  Jeff Law  <law@redhat.com>
6296         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
6297         assert.
6299 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6301         PR target/86662
6302         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
6303         with all enabled __intN types.
6305         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
6307 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6309         * alias.c (record_set): Check for clobber high.
6310         * cfgexpand.c (expand_gimple_stmt): Likewise.
6311         * combine-stack-adj.c (single_set_for_csa): Likewise.
6312         * combine.c (find_single_use_1): Likewise.
6313         (set_nonzero_bits_and_sign_copies): Likewise.
6314         (get_combine_src_dest): Likewise.
6315         (is_parallel_of_n_reg_sets): Likewise.
6316         (try_combine): Likewise.
6317         (record_dead_and_set_regs_1): Likewise.
6318         (reg_dead_at_p_1): Likewise.
6319         (reg_dead_at_p): Likewise.
6320         * dce.c (deletable_insn_p): Likewise.
6321         (mark_nonreg_stores_1): Likewise.
6322         (mark_nonreg_stores_2): Likewise.
6323         * df-scan.c (df_find_hard_reg_defs): Likewise.
6324         (df_uses_record): Likewise.
6325         (df_get_call_refs): Likewise.
6326         * dwarf2out.c (mem_loc_descriptor): Likewise.
6327         * haifa-sched.c (haifa_classify_rtx): Likewise.
6328         * ira-build.c (create_insn_allocnos): Likewise.
6329         * ira-costs.c (scan_one_insn): Likewise.
6330         * ira.c (equiv_init_movable_p): Likewise.
6331         (rtx_moveable_p): Likewise.
6332         (interesting_dest_for_shprep): Likewise.
6333         * jump.c (mark_jump_label_1): Likewise.
6334         * postreload-gcse.c (record_opr_changes): Likewise.
6335         * postreload.c (reload_cse_simplify): Likewise.
6336         (struct reg_use): Add source expr.
6337         (reload_combine): Check for clobber high.
6338         (reload_combine_note_use): Likewise.
6339         (reload_cse_move2add): Likewise.
6340         (move2add_note_store): Likewise.
6341         * print-rtl.c (print_pattern): Likewise.
6342         * recog.c (decode_asm_operands): Likewise.
6343         (store_data_bypass_p): Likewise.
6344         (if_test_bypass_p): Likewise.
6345         * regcprop.c (kill_clobbered_value): Likewise.
6346         (kill_set_value): Likewise.
6347         * reginfo.c (reg_scan_mark_refs): Likewise.
6348         * reload1.c (maybe_fix_stack_asms): Likewise.
6349         (eliminate_regs_1): Likewise.
6350         (elimination_effects): Likewise.
6351         (mark_not_eliminable): Likewise.
6352         (scan_paradoxical_subregs): Likewise.
6353         (forget_old_reloads_1): Likewise.
6354         * reorg.c (find_end_label): Likewise.
6355         (try_merge_delay_insns): Likewise.
6356         (redundant_insn): Likewise.
6357         (own_thread_p): Likewise.
6358         (fill_simple_delay_slots): Likewise.
6359         (fill_slots_from_thread): Likewise.
6360         (dbr_schedule): Likewise.
6361         * resource.c (update_live_status): Likewise.
6362         (mark_referenced_resources): Likewise.
6363         (mark_set_resources): Likewise.
6364         * rtl.c (copy_rtx): Likewise.
6365         * rtlanal.c (reg_referenced_p): Likewise.
6366         (single_set_2): Likewise.
6367         (noop_move_p): Likewise.
6368         (note_stores): Likewise.
6369         * sched-deps.c (sched_analyze_reg): Likewise.
6370         (sched_analyze_insn): Likewise.
6372 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6374         * cse.c (invalidate_reg): New function extracted from...
6375         (invalidate): ...here.
6376         (canonicalize_insn): Check for clobber high.
6377         (invalidate_from_clobbers): invalidate clobber highs.
6378         (invalidate_from_sets_and_clobbers): Likewise.
6379         (count_reg_usage): Check for clobber high.
6380         (insn_live_p): Likewise.
6381         * cselib.c (cselib_expand_value_rtx_1):Likewise.
6382         (cselib_invalidate_regno): Check for clobber in setter.
6383         (cselib_invalidate_rtx): Pass through setter.
6384         (cselib_invalidate_rtx_note_stores):
6385         (cselib_process_insn): Check for clobber high.
6386         * cselib.h (cselib_invalidate_rtx): Add operand.
6388 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6390         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
6391         (mark_not_eliminable): Likewise.
6392         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
6393         * lra-lives.c (process_bb_lives): Check for clobber high.
6394         * lra.c (new_insn_reg): Remember clobber highs.
6395         (collect_non_operand_hard_regs): Check for clobber high.
6396         (lra_set_insn_recog_data): Likewise.
6397         (add_regs_to_insn_regno_info): Likewise.
6398         (lra_update_insn_regno_info): Likewise.
6400 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6402         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
6403         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
6405 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6407         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
6408         (copy_insn_1): Likewise.
6409         (gen_hard_reg_clobber_high): New gen function.
6410         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
6411         * genemit.c (gen_exp): Likewise.
6412         (gen_emit_seq): Pass through info.
6413         (gen_insn): Check for CLOBBER_HIGH.
6414         (gen_expand): Pass through info.
6415         (gen_split): Likewise.
6416         (output_add_clobbers): Likewise.
6417         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
6418         (remove_clobbers): Likewise.
6419         * rtl.h (gen_hard_reg_clobber_high): New declaration.
6421 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6423         * doc/rtl.texi (clobber_high): Add.
6424         (parallel): Add in clobber high
6425         * rtl.c (rtl_check_failed_code3): Add function.
6426         * rtl.def (CLOBBER_HIGH): Add expression.
6427         * rtl.h (RTL_CHECKC3): Add macro.
6428         (rtl_check_failed_code3): Add declaration.
6429         (XC3EXP): Add macro.
6431 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
6433         PR target/86386
6434         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
6435         cfun->machine->max_used_stack_alignment if needed.
6437 2018-08-04  Martin Sebor  <msebor@redhat.com>
6439         PR tree-optimization/86571
6440         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
6441         NaN output to 4.
6443 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
6445         PR target/86799
6446         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6447         Define.
6449 2018-08-03  Jeff Law  <law@redhat.com>
6451         PR target/86795
6452         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6453         Define to speculation_safe_value_not_needed.
6455 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
6457         * doc/gcov.texi (-x): Remove duplicate "to".
6458         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
6459         (-Wif-not-aligned): Remove duplicate "is".
6460         (-flto): Remove duplicate "the".
6461         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
6462         duplicate "v5.00.b".
6463         (MSP430 Options): Remove duplicate "and" from the description
6464         of "-mgprel-sec=regexp".
6465         (x86 Options): Remove duplicate copies of "vmldLog102" and
6466         vmlsLog104 from description of "-mveclibabi=type".
6468 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
6470         * internal-fn.h (first_commutative_argument): Declare.
6471         * internal-fn.c (first_commutative_argument): New function.
6472         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
6473         restrictions for pattern statements.  Use first_commutative_argument
6474         to look for commutative operands in calls to internal functions.
6476 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
6478         * Makefile.in (wide-int-range.o): New.
6479         * tree-vrp.c: Move all the wide_int_* functions to...
6480         * wide-int-range.cc: ...here.
6481         * tree-vrp.h: Move all the wide_int_* prototypes to...
6482         * wide-int-range.h: ...here.
6484 2018-08-03  Tom de Vries  <tdevries@suse.de>
6486         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
6487         UI_NONE.
6488         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
6489         * except.c (output_function_exception_table): Do early exit if
6490         targetm_common.except_unwind_info (&global_options) == UI_NONE.
6492 2018-08-03  Martin Liska  <mliska@suse.cz>
6494         * predict.c (dump_prediction): Change to 2 digits
6495         in fraction part.
6497 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
6499         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
6500         neon_dup_q to...
6501         (falkor_am_1_gtov_gtov): ... a new insn reservation.
6503 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
6505         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
6506         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
6507         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
6508         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
6509         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
6511 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
6513         * diagnostic-show-locus.c (diagnostic_show_locus): Use
6514         pp_take_prefix when saving the existing prefix.
6515         * diagnostic.c (diagnostic_append_note): Likewise.
6516         * langhooks.c (lhd_print_error_function): Likewise.
6517         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
6518         param's type.  Free the existing prefix.
6519         (pp_take_prefix): New function.
6520         (pretty_printer::pretty_printer): Drop the prefix parameter.
6521         Rename the length parameter to match the comment.
6522         (pretty_printer::~pretty_printer): Free the prefix.
6523         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
6524         parameter.
6525         (struct pretty_printer): Drop the "const" from "prefix" field's
6526         type and clarify memory management.
6527         (pp_set_prefix): Drop the "const" from the 2nd param.
6528         (pp_take_prefix): New decl.
6530 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
6532         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
6533         for word_mode_ok here instead of passing as argument.
6534         (expand_block_compare): Change select_block_compare_mode() call.
6535         (expand_strncmp_gpr_sequence): New function.
6536         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
6538 2018-08-02  Jeff Law  <law@redhat.com>
6540         PR target/86790
6541         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6542         Define to speculation_safe_value_not_needed.
6544         PR target/86784
6545         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6546         Define to speculation_safe_value_not_needed.
6548 2018-08-02  Tom de Vries  <tdevries@suse.de>
6550         PR target/86660
6551         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
6552         function.  Return UI_TARGET unconditionally.
6553         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
6554         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
6556 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6558         * genemit.c (print_overload_test): Fix typo.
6560 2018-08-02  Richard Biener  <rguenther@suse.de>
6562         PR tree-optimization/86816
6563         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
6564         which checks for value availability before querying it.
6565         (gvn_uses_equal): Use it.
6566         (same_succ_hash): Likewise.
6567         (gimple_equal_p): Likewise.
6569 2018-08-02  Nick Clifton  <nickc@redhat.com>
6571         PR target/86813
6572         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6573         Define to speculation_safe_value_not_needed.
6575         PR target/86810
6576         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6577         Define to speculation_safe_value_not_needed.
6579         PR target/86810
6580         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6581         Define to speculation_safe_value_not_needed.
6583         PR target/86803
6584         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6585         Define to speculation_safe_value_not_needed.
6587         PR target/86797
6588         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6589         Define to speculation_safe_value_not_needed.
6591         PR target/86791
6592         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6593         Define to speculation_safe_value_not_needed.
6595         PR target/86789
6596         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6597         Define to speculation_safe_value_not_needed.
6599         PR target/86787
6600         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6601         Define to speculation_safe_value_not_needed.
6603         PR target/86782
6604         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6605         speculation_safe_value_not_needed.
6607         PR target/86781
6608         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
6609         to speculation_safe_value_not_needed.
6611 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6613         * doc/md.texi: Expand the documentation of instruction names
6614         to mention port-local uses.  Document '@' in pattern names.
6615         * read-md.h (overloaded_instance, overloaded_name): New structs.
6616         (mapping): Declare.
6617         (md_reader::handle_overloaded_name): New member function.
6618         (md_reader::get_overloads): Likewise.
6619         (md_reader::m_first_overload): New member variable.
6620         (md_reader::m_next_overload_ptr): Likewise.
6621         (md_reader::m_overloads_htab): Likewise.
6622         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
6623         m_next_overload_ptr and m_overloads_htab.
6624         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
6625         (get_mode_token, get_code_token, get_int_token): New functions.
6626         (map_attr_string): Add an optional argument that passes back
6627         the associated iterator.
6628         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
6629         (md_reader::handle_overloaded_name, add_overload_instance): New
6630         functions.
6631         (apply_iterators): Handle '@' names.  Report an error if '@'
6632         is used without iterators.
6633         (initialize_iterators): Initialize the new iterator_group fields.
6634         * genopinit.c (handle_overloaded_code_for)
6635         (handle_overloaded_gen): New functions.
6636         (main): Use them to print declarations of maybe_code_for_* and
6637         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
6638         * genemit.c (print_overload_arguments, print_overload_test)
6639         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
6640         (main): Use it to print definitions of maybe_code_for_* and
6641         maybe_gen_* functions.
6642         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
6643         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
6644         instead of explicit mode checks.
6645         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
6646         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
6647         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
6648         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
6649         (aarch64_expand_compare_and_swap): Likewise
6650         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
6651         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
6652         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
6653         (aarch64_constant_pool_reload_icode): Delete.
6654         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
6655         instead of aarch64_constant_pool_reload_icode.  Use
6656         code_for_aarch64_reload_mov instead of explicit mode checks.
6657         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
6658         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
6659         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
6660         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
6661         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
6662         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
6663         (aarch64_atomic_load_op_code): Delete.
6664         (aarch64_emit_atomic_load_op): Likewise.
6665         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
6666         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
6667         instead of aarch64_emit_atomic_load_op.
6668         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
6669         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
6670         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
6671         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
6672         character before the pattern name.
6673         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
6674         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
6675         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
6676         (aarch64_frecps<mode>): Likewise.
6677         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
6678         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
6679         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
6680         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
6681         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
6683 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
6685         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
6686         Allow HFmode constants if TARGET_FP_F16INST.
6688 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
6690         PR target/86014
6691         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
6692         No longer check last store for clobber of address register.
6694 2018-08-02  Martin Liska  <mliska@suse.cz>
6696         PR gcov-profile/86817
6697         * gcov.c (process_all_functions): New function.
6698         (main): Call it.
6699         (process_file): Move functions processing to
6700         process_all_functions.
6702 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
6704         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
6705         "const" to the "gimple *" and "rtx_insn *" parameters.
6706         * dumpfile.h (dump_user_location_t::dump_user_location_t):
6707         Likewise.
6708         (dump_location_t::dump_location_t): Likewise.
6710 2018-08-01  Martin Sebor  <msebor@redhat.com>
6712         PR tree-optimization/86650
6713         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
6714         rather than a "gcall *".  Directly pass the data of interest
6715         to percent_K_format, rather than building a temporary CALL_EXPR
6716         to hold it.
6717         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
6718         (gimple_fold_builtin_strncat): Adjust.
6719         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
6720         gcall* argument with gimple*.
6721         * gimple-ssa-warn-restrict.c (check_call): Same.
6722         (wrestrict_dom_walker::before_dom_children): Same.
6723         (builtin_access::builtin_access): Same.
6724         (check_bounds_or_overlap): Same
6725         (maybe_diag_overlap): Same.
6726         (maybe_diag_offset_bounds): Same.
6727         * tree-diagnostic.c (default_tree_printer): Move usage of
6728         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
6729         to this callsite.
6730         * tree-pretty-print.c (percent_K_format): Add argument.
6731         * tree-pretty-print.h: Add argument.
6732         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
6733         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
6734         (maybe_diag_stxncpy_trunc): Same.
6735         (handle_builtin_stxncpy): Same.
6736         (handle_builtin_strcat): Same.
6738 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6740         * match.pd: Optimise pointer range checks.
6742 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6744         PR tree-optimization/86758
6745         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
6746         to remove pattern statements.
6748 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6750         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
6751         result of dfs_enumerate_from when constructing stmt_vec_infos,
6752         instead of additionally calling get_loop_body.
6754 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6756         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
6757         parameter.
6758         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
6759         When creating an iv, assert that the step is not known to be zero.
6760         (vect_setup_realignment): Update call accordingly.
6761         * tree-vect-stmts.c (vectorizable_store): Likewise.
6762         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
6764 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6766         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
6767         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
6768         (vectorizable_reduction): Likewise.
6769         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6770         (vect_detect_hybrid_slp_stmts): Likewise.
6771         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6773 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
6775         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
6776         (wide_int_set_zero_nonzero_bits): ...this.
6777         (zero_nonzero_bits_from_vr): Rename to...
6778         (vrp_set_zero_nonzero_bits): ...this.
6779         (extract_range_from_multiplicative_op_1): Abstract wide int
6780         code...
6781         (wide_int_range_multiplicative_op): ...here.
6782         (extract_range_from_binary_expr_1): Extract wide int binary
6783         operations into their own functions.
6784         (wide_int_range_lshift): New.
6785         (wide_int_range_can_optimize_bit_op): New.
6786         (wide_int_range_shift_undefined_p): New.
6787         (wide_int_range_bit_xor): New.
6788         (wide_int_range_bit_ior): New.
6789         (wide_int_range_bit_and): New.
6790         (wide_int_range_trunc_mod): New.
6791         (extract_range_into_wide_ints): New.
6792         (vrp_shift_undefined_p): New.
6793         (extract_range_from_multiplicative_op): New.
6794         (vrp_can_optimize_bit_op): New.
6795         * tree-vrp.h (value_range::dump): New.
6796         (wide_int_range_multiplicative_op): New.
6797         (wide_int_range_lshift):New.
6798         (wide_int_range_shift_undefined_p): New.
6799         (wide_int_range_bit_xor): New.
6800         (wide_int_range_bit_ior): New.
6801         (wide_int_range_bit_and): New.
6802         (wide_int_range_trunc_mod): New.
6803         (zero_nonzero_bits_from_bounds): Rename to...
6804         (wide_int_set_zero_nonzero_bits): ...this.
6805         (zero_nonzero_bits_from_vr): Rename to...
6806         (vrp_set_zero_nonzero_bits): ...this.
6807         (range_easy_mask_min_max): Rename to...
6808         (wide_int_range_can_optimize_bit_op): this.
6809         * vr-values.c (simplify_bit_ops_using_ranges): Rename
6810         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6812 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6814         * tree-vectorizer.h (vect_orig_stmt): New function.
6815         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6816         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6817         (vect_create_epilog_for_reduction): Likewise.
6818         (vectorizable_live_operation): Likewise.
6819         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6820         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6821         * tree-vect-stmts.c (vectorizable_call): Likewise.
6822         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6824 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6826         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6827         argument.
6828         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6829         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6830         (vect_transform_loop): Likewise.
6831         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6833 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6835         * tree-vectorizer.h (vect_schedule_slp): Return void.
6836         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6837         (vect_schedule_slp): Likewise.
6839 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6841         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6842         argument.
6843         (vect_transform_loop): Update calls accordingly.  Schedule SLP
6844         instances before the main loop, if any exist.
6846 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
6848         PR tree-optimization/86749
6849         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6850         If the lhs is used in a COND_EXPR, check that it is being used
6851         as the "then" or "else" value.
6853 2018-08-01  Tom de Vries  <tdevries@suse.de>
6855         PR target/86800
6856         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6857         speculation_safe_value_not_needed.
6859 2018-08-01  Richard Biener  <rguenther@suse.de>
6861         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6862         as base and offset.
6864 2018-08-01  Martin Liska  <mliska@suse.cz>
6866         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6867         format how successful transformation is dumped.
6868         (gimple_mod_pow2_value_transform): Likewise.
6869         (gimple_mod_subtract_transform): Likewise.
6870         (gimple_stringops_transform): Likewise.
6872 2018-08-01  Martin Liska  <mliska@suse.cz>
6874         PR value-prof/35543
6875         * value-prof.c (interesting_stringop_to_profile_p):
6876         Simplify the code and add BUILT_IN_MEMMOVE.
6877         (gimple_stringops_transform): Likewise.
6879 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
6881         * config/aarch64/aarch64-simd.md
6882         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6883         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6884         use GPI iterator instead of SI mode.
6886 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
6888         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6889         rs6000_speculation_barrier.
6890         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6891         new barrier pattern name.
6893 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
6895         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
6896         (speculation_barrier): New insn.
6898 2018-08-01  Richard Biener  <rguenther@suse.de>
6900         PR bootstrap/86724
6901         * graphite.h: Include isl/id.h and isl/space.h to allow build
6902         with ISL 0.20.
6904 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
6906         PR target/86651
6907         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6908         mode for COFF targets.
6909         * defaults.h (TARGET_COFF): Define.
6910         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
6911         TARGET_COFF): Define.
6912         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
6913         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
6914         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
6916 2018-07-31  Alexandre Oliva <oliva@adacore.com>
6918         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
6919         * gimple-streamer-out.c (output_bb): Save it.
6920         * lto-streamer-in.c (input_struct_function_base): Restore
6921         instance discriminator if available.  Create map on demand.
6922         * lto-streamer-out.c (output_struct_function_base): Save it if
6923         available.
6924         * final.c (decl_to_instance_map): Document LTO strategy.
6926 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
6927             Olivier Hainque  <hainque@adacore.com>
6929         * debug.h (decl_to_instance_map_t): New type.
6930         (decl_to_instance_map): Declare.
6931         (maybe_create_decl_to_instance_map): New inline function.
6932         * final.c (bb_discriminator, last_bb_discriminator): New statics,
6933         to track basic block discriminators.
6934         (final_start_function_1): Initialize them.
6935         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6936         bb_discriminator.
6937         (decl_to_instance_map): New variable.
6938         (map_decl_to_instance, maybe_set_discriminator): New functions.
6939         (notice_source_line): Set discriminator.
6941 2018-07-31  Ian Lance Taylor  <iant@golang.org>
6943         * targhooks.c (default_have_speculation_safe_value): Add
6944         ATTRIBUTE_UNUSED.
6946 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
6948         * dump-context.h: Include "pretty-print.h".
6949         (dump_context::refresh_dumps_are_enabled): New decl.
6950         (dump_context::emit_item): New decl.
6951         (class dump_context): Add fields "m_test_pp" and
6952         "m_test_pp_flags".
6953         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6954         (temp_dump_context::get_dumped_text): New decl.
6955         (class temp_dump_context): Add field "m_pp".
6956         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6957         (dump_context::refresh_dumps_are_enabled): ...and add a test for
6958         m_test_pp.
6959         (set_dump_file): Update for above change.
6960         (set_alt_dump_file): Likewise.
6961         (dump_loc): New overload, taking a pretty_printer *.
6962         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
6963         to any test pretty-printer.
6964         (make_item_for_dump_gimple_stmt): New function, adapted from
6965         optinfo::add_gimple_stmt.
6966         (dump_context::dump_gimple_stmt): Call it, and use the result,
6967         eliminating the direct usage of dump_file and alt_dump_file in
6968         favor of indirectly using them via emit_item.
6969         (make_item_for_dump_gimple_expr): New function, adapted from
6970         optinfo::add_gimple_expr.
6971         (dump_context::dump_gimple_expr): Call it, and use the result,
6972         eliminating the direct usage of dump_file and alt_dump_file in
6973         favor of indirectly using them via emit_item.
6974         (make_item_for_dump_generic_expr): New function, adapted from
6975         optinfo::add_tree.
6976         (dump_context::dump_generic_expr): Call it, and use the result,
6977         eliminating the direct usage of dump_file and alt_dump_file in
6978         favor of indirectly using them via emit_item.
6979         (make_item_for_dump_printf_va): New function, adapted from
6980         optinfo::add_printf_va.
6981         (make_item_for_dump_printf): New function.
6982         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6983         and use the result, eliminating the direct usage of dump_file and
6984         alt_dump_file in favor of indirectly using them via emit_item.
6985         (make_item_for_dump_dec): New function.
6986         (dump_context::dump_dec): Call it, and use the result,
6987         eliminating the direct usage of dump_file and alt_dump_file in
6988         favor of indirectly using them via emit_item.
6989         (make_item_for_dump_symtab_node): New function, adapted from
6990         optinfo::add_symtab_node.
6991         (dump_context::dump_symtab_node): Call it, and use the result,
6992         eliminating the direct usage of dump_file and alt_dump_file in
6993         favor of indirectly using them via emit_item.
6994         (dump_context::begin_scope): Reimplement, avoiding direct usage
6995         of dump_file and alt_dump_file in favor of indirectly using them
6996         via emit_item.
6997         (dump_context::emit_item): New member function.
6998         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6999         Set up test pretty-printer on the underlying context.  Call
7000         refresh_dumps_are_enabled.
7001         (temp_dump_context::~temp_dump_context): Call
7002         refresh_dumps_are_enabled.
7003         (temp_dump_context::get_dumped_text): New member function.
7004         (selftest::verify_dumped_text): New function.
7005         (ASSERT_DUMPED_TEXT_EQ): New macro.
7006         (selftest::test_capture_of_dump_calls): Run all tests twice, with
7007         and then without optinfo enabled.  Add uses of
7008         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
7009         * dumpfile.h: Update comment for the dump_* API.
7010         * optinfo-emit-json.cc
7011         (selftest::test_building_json_from_dump_calls): Update for new
7012         param for temp_dump_context ctor.
7013         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7014         and "m_owned" field.
7015         (optinfo_item::~optinfo_item): Likewise.
7016         (optinfo::add_item): New member function.
7017         (optinfo::emit): Update comment.
7018         (optinfo::add_string): Delete.
7019         (optinfo::add_printf): Delete.
7020         (optinfo::add_printf_va): Delete.
7021         (optinfo::add_gimple_stmt): Delete.
7022         (optinfo::add_gimple_expr): Delete.
7023         (optinfo::add_tree): Delete.
7024         (optinfo::add_symtab_node): Delete.
7025         (optinfo::add_dec): Delete.
7026         * optinfo.h (class dump_context): New forward decl.
7027         (optinfo::add_item): New decl.
7028         (optinfo::add_string): Delete.
7029         (optinfo::add_printf): Delete.
7030         (optinfo::add_printf_va): Delete.
7031         (optinfo::add_gimple_stmt): Delete.
7032         (optinfo::add_gimple_expr): Delete.
7033         (optinfo::add_tree): Delete.
7034         (optinfo::add_symtab_node): Delete.
7035         (optinfo::add_dec): Delete.
7036         (optinfo::add_poly_int): Delete.
7037         (optinfo_item::optinfo_item): Remove "owned" param.
7038         (class optinfo_item): Remove field "m_owned".
7040 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7042         PR middle-end/86705
7043         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7044         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7045         requested variable alignment.
7046         (expand_one_ssa_partition): Likewise.
7047         (expand_one_var): Likewise.
7049 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7051         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7052         to speculation_safe_value_not_needed.
7054 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7056         * targhooks.h (speculation_safe_value_not_needed): New prototype.
7057         * targhooks.c (speculation_safe_value_not_needed): New function.
7058         * target.def (have_speculation_safe_value): Update documentation.
7059         * doc/tm.texi: Regenerated.
7061 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7063         * config/aarch64/iterators.md (ALLI_TI): New iterator.
7064         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7065         expand.
7066         (despeculate_copy<ALLI:mode>_insn): New insn.
7067         (despeculate_copyti_insn): New insn.
7068         (despeculate_simple<ALLI:mode>): New insn
7069         (despeculate_simpleti): New insn.
7070         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7071         function.
7072         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7073         aarch64_speculation_safe_value.
7074         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7076 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7078         * config/aarch64/aarch64-speculation.cc: New file.
7079         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7080         before pass_reorder_blocks.
7081         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7082         prototype.
7083         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7084         X14 and X15 when tracking speculation.
7085         * config/aarch64/aarch64.md (register name constants): Add
7086         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7087         (unspec): Add UNSPEC_SPECULATION_TRACKER.
7088         (speculation_barrier): New insn attribute.
7089         (cmp<mode>): Allow SP in comparisons.
7090         (speculation_tracker): New insn.
7091         (speculation_barrier): Add speculation_barrier attribute.
7092         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7093         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7094         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7096 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7098         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7099         aarch64_track_speculation is true.
7100         (tb<optab><mode>1): Likewise.
7101         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7102         generate CB[N]Z when tracking speculation.
7103         (aarch64_split_compare_and_swap): Likewise.
7104         (aarch64_split_atomic_op): Likewise.
7106 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7108         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
7110 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7112         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
7113         (speculation_barrier): New insn.
7115 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7117         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
7118         * config/arm/arm.md (speculation_barrier): New expand.
7119         (speculation_barrier_insn): New pattern.
7121 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7123         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
7124         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
7125         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
7126         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
7127         list.
7128         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
7129         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
7130         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
7131         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
7132         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
7133         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
7134         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
7135         * builtins.c (expand_speculation_safe_value): New function.
7136         (expand_builtin): Call it.
7137         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
7138         * doc/extend.texi: Document __builtin_speculation_safe_value.
7139         * doc/md.texi: Document "speculation_barrier" pattern.
7140         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
7141         TARGET_HAVE_SPECULATION_SAFE_VALUE.
7142         * doc/tm.texi: Regenerated.
7143         * target.def (have_speculation_safe_value, speculation_safe_value): New
7144         hooks.
7145         * targhooks.c (default_have_speculation_safe_value): New function.
7146         (default_speculation_safe_value): New function.
7147         * targhooks.h (default_have_speculation_safe_value): Add prototype.
7148         (default_speculation_safe_value): Add prototype.
7150 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7152         * dump-context.h (dump_context::dump_loc): New decl.
7153         * dumpfile.c (dump_context::dump_loc): New member function.
7154         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
7155         and dump_gimple_stmt.
7156         (dump_context::dump_gimple_expr_loc): Likewise, using
7157         dump_gimple_expr.
7158         (dump_context::dump_generic_expr_loc): Likewise, using
7159         dump_generic_expr.
7160         (dump_context::dump_printf_loc_va): Likewise, using
7161         dump_printf_va.
7162         (dump_context::begin_scope): Explicitly using the global function
7163         "dump_loc", rather than the member function.
7165 2018-07-31  Martin Sebor  <msebor@redhat.com>
7167         PR tree-optimization/86741
7168         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
7170 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
7172         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
7173         depend on whether prefetch instructions will be emitted or not.
7174         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
7175         will be emitted or not.
7176         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
7177         (TARGET_SETMEM_PFD): New macros.
7179 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7181         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
7182         (NULL_STMT_VEC_INFO): Delete.
7183         (stmt_vec_info::operator*): Likewise.
7184         (stmt_vec_info::operator gimple *): Likewise.
7185         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
7186         of NULL_STMT_VEC_INFO.
7187         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7188         (vect_reassociating_reduction_p): Likewise.
7189         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7190         (vectorizable_store): Likewise.
7191         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
7192         (vec_info::free_stmt_vec_infos): Likewise.
7194 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7196         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
7197         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
7199 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7201         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
7202         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7203         (vec_info::free_stmt_vec_info): New private member functions.
7204         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
7205         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
7206         * tree-parloops.c (gather_scalar_reductions): Remove calls to
7207         set_stmt_vec_info_vec and free_stmt_vec_infos.
7208         * tree-vect-loop.c (_loop_vec_info): Remove call to
7209         set_stmt_vec_info_vec.
7210         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
7211         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
7212         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
7213         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7214         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
7215         assignments in {vec_info::,}new_stmt_vec_info that are redundant
7216         with the clearing in the xcalloc.
7217         (stmt_vec_info_vec): Delete.
7218         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
7219         (vectorize_loops): Likewise.
7220         (vec_info::~vec_info): Remove argument from call to
7221         free_stmt_vec_infos.
7222         (vec_info::add_stmt): Remove vinfo argument from call to
7223         new_stmt_vec_info.
7225 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7227         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
7228         rather than a gimple stmt.
7229         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
7230         information for pattern statements when passed the original
7231         statement; instead wait to be passed the pattern statement itself.
7232         Don't call set_vinfo_for_stmt here.
7233         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
7234         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
7235         stmt_vec_infos here.
7236         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
7237         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
7238         stmt_vec_infos entry.
7240 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7242         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
7243         * tree-vectorizer.c (vec_info::replace_stmt): New function.
7244         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
7245         * tree-vect-stmts.c (vectorizable_call): Likewise.
7246         (vectorizable_simd_clone_call): Likewise.
7248 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7250         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
7251         * tree-vectorizer.c (vec_info::remove_stmt): New function.
7252         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
7253         * tree-vect-loop.c (vect_transform_loop): Likewise.
7254         * tree-vect-slp.c (vect_schedule_slp): Likewise.
7255         * tree-vect-stmts.c (vect_remove_stores): Likewise.
7257 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7259         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
7260         (vect_dr_stmt): Delete.
7261         * tree-vectorizer.c (vec_info::lookup_dr): New function.
7262         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
7263         of DR_VECT_AUX.
7264         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7265         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
7266         (vect_verify_datarefs_alignment, vect_peeling_supportable)
7267         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7268         (vect_analyze_data_refs): Likewise.
7269         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
7270         argument.
7271         (vect_find_same_alignment_drs): Likewise.
7272         (vect_slp_analyze_node_dependences): Update calls accordingly.
7273         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7274         instead of DR_VECT_AUX.
7275         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
7276         of a vector data references.  Use vec_info::lookup_dr instead of
7277         DR_VECT_AUX.
7278         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
7279         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7280         instead of DR_VECT_AUX.
7282 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7284         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
7285         dr_vec_info.
7286         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7287         accordingly.
7288         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7289         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7290         (vect_gen_prolog_loop_niters): Likewise.
7292 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7294         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
7295         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
7296         (vect_known_alignment_in_bytes, vect_dr_behavior)
7297         (vect_get_scalar_dr_size): Take references as dr_vec_infos
7298         instead of data_references.  Update calls to other routines for
7299         which the same change has been made.
7300         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
7301         dr_vec_infos instead of stmt_vec_infos.
7302         (vect_analyze_data_ref_dependence): Update call accordingly.
7303         (vect_slp_analyze_data_ref_dependence)
7304         (vect_record_base_alignments): Use DR_VECT_AUX.
7305         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
7306         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
7307         (vector_alignment_reachable_p, vect_get_data_access_cost)
7308         (vect_peeling_supportable, vect_analyze_group_access_1)
7309         (vect_analyze_group_access, vect_analyze_data_ref_access)
7310         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
7311         (vect_compile_time_alias, vect_small_gap_p)
7312         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
7313         (vect_supportable_dr_alignment): Take references as dr_vec_infos
7314         instead of data_references.  Update calls to other routines for
7315         which the same change has been made.
7316         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
7317         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7318         (vect_slp_analyze_and_verify_node_alignment)
7319         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7320         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7321         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
7322         above changes.
7323         (_vect_peel_info::dr): Replace with...
7324         (_vect_peel_info::dr_info): ...this new field.
7325         (vect_peeling_hash_get_most_frequent)
7326         (vect_peeling_hash_choose_best_peeling): Update accordingly.
7327         (vect_peeling_hash_get_lowest_cost):
7328         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
7329         routines for which the same change has been made.
7330         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
7331         data_reference.
7332         * tree-vect-loop-manip.c (get_misalign_in_elems)
7333         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
7334         above changes.
7335         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7336         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
7337         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
7338         (get_group_load_store_type, get_negative_load_store_type)
7339         (vect_get_data_ptr_increment, vectorizable_store)
7340         (vectorizable_load): Likewise.
7341         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
7342         Update calls to other routines for which the same change has been made.
7344 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7346         * tree-vectorizer.h (vec_info::move_dr): New member function.
7347         (dataref_aux): Rename to...
7348         (dr_vec_info): ...this and add "dr" and "stmt" fields.
7349         (_stmt_vec_info::dr_aux): Update accordingly.
7350         (_stmt_vec_info::data_ref_info): Delete.
7351         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
7352         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
7353         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
7354         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
7355         of data_ref.
7356         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
7357         (STMT_VINFO_DR_INFO): New macro.
7358         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
7359         (set_dr_misalignment): Update after rename of dataref_aux.
7360         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
7361         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
7362         initialization of STMT_VINFO_DATA_REF.
7363         * tree-vectorizer.c (vec_info::move_dr): New function.
7364         * tree-vect-patterns.c (vect_recog_bool_pattern)
7365         (vect_recog_mask_conversion_pattern)
7366         (vect_recog_gather_scatter_pattern): Use it.
7367         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
7368         the "dr" and "stmt" fields of dr_vec_info instead of
7369         STMT_VINFO_DATA_REF.
7371 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7373         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
7374         (is_pattern_stmt_p): Use it.
7375         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
7376         on pattern statements.
7378 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7380         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
7381         original stmt as a stmt_vec_info rather than a gimple stmt.
7382         (vect_pattern_recog_1): Take the statement directly as a
7383         stmt_vec_info, rather than via a gimple_stmt_iterator.
7384         Update call to vect_mark_pattern_stmts.
7385         (vect_pattern_recog): Update calls accordingly.
7387 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7389         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
7390         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
7391         a vect_def_type for the first argument.
7392         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
7393         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
7394         operand if it isn't defined by a vectorized statement.
7395         (vect_build_gather_load_calls): Remove the mask_dt argument and
7396         update calls to vect_get_vec_def_for_stmt_copy.
7397         (vectorizable_bswap): Likewise the dt argument.
7398         (vectorizable_call): Update calls to vectorizable_bswap and
7399         vect_get_vec_def_for_stmt_copy.
7400         (vectorizable_simd_clone_call, vectorizable_assignment)
7401         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
7402         (vectorizable_comparison): Update calls to
7403         vect_get_vec_def_for_stmt_copy.
7404         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
7405         vect_is_simple_use.
7406         (vect_get_loop_based_defs): Remove dt argument and update call
7407         to vect_get_vec_def_for_stmt_copy.
7408         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
7409         and vect_get_vec_def_for_stmt_copy.
7410         (vectorizable_load): Update calls to vect_build_gather_load_calls
7411         and vect_get_vec_def_for_stmt_copy.
7412         * tree-vect-loop.c (vect_create_epilog_for_reduction)
7413         (vectorizable_reduction, vectorizable_live_operation): Update calls
7414         to vect_get_vec_def_for_stmt_copy.
7416 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7418         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
7419         and gimple stmt arguments with a stmt_vec_info.
7420         (vect_record_base_alignments): Update calls accordingly.
7421         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
7422         and gimple stmt arguments with a stmt_vec_info.
7423         (vect_build_slp_tree_1): Remove vinfo argument and update call
7424         to vect_record_max_nunits.
7425         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
7426         and vect_record_max_nunits.
7428 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7430         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
7431         file and take a stmt_vec_info instead of a gimple stmt.
7432         (supportable_widening_operation, vect_finish_replace_stmt)
7433         (vect_finish_stmt_generation, vect_get_store_rhs)
7434         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
7435         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
7436         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
7437         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
7438         (vect_create_data_ref_ptr, bump_vector_ptr)
7439         (vect_permute_store_chain, vect_setup_realignment)
7440         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
7441         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
7442         (vectorizable_reduction, vectorizable_induction)
7443         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
7444         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
7445         than gimple stmts as arguments.
7446         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
7447         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
7448         (can_group_stmts_p, vect_check_gather_scatter)
7449         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7450         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
7451         (vect_permute_load_chain, vect_shift_permute_load_chain)
7452         (vect_transform_grouped_load)
7453         (vect_record_grouped_load_vectors): Likewise.
7454         * tree-vect-loop.c (vect_fixup_reduc_chain)
7455         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
7456         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
7457         (vectorizable_reduction, vectorizable_induction)
7458         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
7459         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
7460         (vect_get_load_store_mask): Likewise.
7461         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
7462         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
7463         * tree-vect-stmts.c (vect_mark_relevant)
7464         (is_simple_and_all_uses_invariant)
7465         (exist_non_indexing_operands_for_use_p, process_use)
7466         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
7467         (vect_get_vec_def_for_operand, vect_get_vec_defs)
7468         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
7469         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
7470         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
7471         (get_negative_load_store_type, get_load_store_type)
7472         (vect_check_load_store_mask, vect_check_store_rhs)
7473         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
7474         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7475         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7476         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7477         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
7478         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
7479         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
7480         (supportable_widening_operation): Likewise.
7482 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7484         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
7485         a stmt_vec_info instead of a gcall.
7486         (vect_check_gather_scatter): Update call accordingly.
7487         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
7488         of a gphi.
7489         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
7490         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
7491         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
7492         instead of a gimple stmt.
7493         (vect_transform_loop): Update calls accordingly.
7494         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
7495         stmt_vec_infos instead of gimple stmts.
7496         (vect_analyze_slp_instance): Update use accordingly.
7497         * tree-vect-stmts.c (read_vector_array, write_vector_array)
7498         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
7499         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
7500         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
7501         (vect_gen_widened_results_half, vect_get_loop_based_defs)
7502         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
7503         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
7504         down to subroutines.
7506 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7508         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
7509         of the worklist from a vector of gimple stmts to a vector of
7510         stmt_vec_infos.
7511         * tree-vect-stmts.c (vect_mark_relevant, process_use)
7512         (vect_mark_stmts_to_be_vectorized): Likewise
7514 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7516         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
7517         statement before passing it to vect_analyze_stmt.
7518         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
7519         the chain of phi vector definitions.  Track the exit phi via its
7520         stmt_vec_info.
7521         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
7522         STMT_VINFO_REDUC_DEF.
7523         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
7524         stmt_vec_infos to handle the statement chains.
7525         (vect_get_slp_defs): Record the first statement in the node
7526         using a stmt_vec_info.
7527         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
7528         statements here and pass their stmt_vec_info down to subroutines.
7529         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
7530         down to vect_finish_stmt_generation.
7531         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
7532         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
7533         stmt_vec_infos to subroutines.
7534         (vect_remove_stores): Use stmt_vec_infos to handle the statement
7535         chains.
7537 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7539         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
7540         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
7541         (vect_permute_store_chain, vect_setup_realignment)
7542         (vect_permute_load_chain, vect_shift_permute_load_chain)
7543         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
7544         stmts internally, and when passing values to other vectorizer routines.
7545         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7546         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
7547         (vect_analyze_loop_operations, get_initial_def_for_reduction)
7548         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
7549         (vectorizable_reduction, vectorizable_induction)
7550         (vectorizable_live_operation, vect_transform_loop_stmt)
7551         (vect_transform_loop): Likewise.
7552         * tree-vect-patterns.c (vect_reassociating_reduction_p)
7553         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
7554         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
7555         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7556         (vect_slp_analyze_node_operations_1): Likewise.
7557         * tree-vect-stmts.c (vect_mark_relevant, process_use)
7558         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
7559         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
7560         (vect_finish_stmt_generation_1, get_group_load_store_type)
7561         (get_load_store_type, vect_build_gather_load_calls)
7562         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7563         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7564         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7565         (vectorizable_store, vectorizable_load, vectorizable_condition)
7566         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
7567         (supportable_widening_operation): Likewise.
7568         (vect_get_vector_types_for_stmt): Likewise.
7569         * tree-vectorizer.h (vect_dr_behavior): Likewise.
7571 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7573         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7574         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
7575         (vect_permute_store_chain, vect_permute_load_chain)
7576         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
7577         repeated stmt_vec_info lookups.
7578         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7579         (vect_update_ivs_after_vectorizer): Likewise.
7580         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
7581         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7582         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
7583         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7584         (vect_bb_slp_scalar_cost): Likewise.
7585         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
7587 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7589         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
7590         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
7591         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
7592         to get gassigns and gcalls, rather than operating on generc gimple
7593         stmts.
7594         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
7595         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
7596         (vectorizable_load, vect_analyze_stmt): Likewise.
7597         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
7599 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7601         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
7602         return stmt_vec_infos rather than gimple stmts.  Do not accept
7603         null arguments.
7604         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
7605         of a gimple stmt.
7606         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7607         Update use of get_later_stmt.
7608         (vect_get_constant_vectors): Update call accordingly.
7609         (vect_schedule_slp_instance): Likewise
7610         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
7611         (vect_slp_analyze_instance_dependence): Likewise.
7612         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
7614 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7616         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
7617         (stmt_info_for_cost::stmt_info): ...this new field.
7618         (add_stmt_costs): Update accordingly.
7619         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7620         (vect_get_known_peeling_cost): Likewise.
7621         (vect_estimate_min_profitable_iters): Likewise.
7622         * tree-vect-stmts.c (record_stmt_cost): Likewise.
7624 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7626         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
7627         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7628         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7629         accordingly.
7630         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
7632 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7634         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
7635         a gimple stmt to a stmt_vec_info.
7636         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7638 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7640         * tree-vectorizer.h (vec_info::grouped_stores): Change from
7641         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7642         (_loop_vec_info::reduction_chains): Likewise.
7643         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
7644         accordingly.
7645         * tree-vect-slp.c (vect_analyze_slp): Likewise.
7647 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7649         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
7650         a gimple stmt to a stmt_vec_info.
7651         (_stmt_vec_info::next_element): Likewise.
7652         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
7653         (vect_slp_analyze_and_verify_node_alignment)
7654         (vect_analyze_group_access_1, vect_analyze_group_access)
7655         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
7656         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
7657         (vect_supportable_dr_alignment): Update accordingly.
7658         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
7659         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
7660         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
7661         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7662         * tree-vect-slp.c (vect_build_slp_tree_1)
7663         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
7664         (vect_split_slp_store_group, vect_analyze_slp_instance)
7665         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
7666         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
7667         (get_group_load_store_type, get_load_store_type)
7668         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
7669         (vect_transform_stmt, vect_remove_stores): Likewise.
7671 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7673         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
7674         than a gimple stmt.
7675         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7676         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
7677         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
7678         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
7679         (vector_alignment_reachable_p, vect_get_data_access_cost)
7680         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
7681         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
7682         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7683         (vect_analyze_group_access_1, vect_analyze_group_access)
7684         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
7685         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
7686         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
7687         result of vect_dr_stmt and use the stmt_vec_info instead of
7688         the associated gimple stmt.
7689         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7690         (vect_gen_prolog_loop_niters): Likewise.
7691         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7693 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7695         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
7696         to a vec<stmt_vec_info>.
7697         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
7698         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
7699         vec<stmt_vec_info>.
7700         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
7701         to a vec<stmt_vec_info>.
7702         (bst_traits::value_type, bst_traits::value_type): Likewise.
7703         (bst_traits::hash): Update accordingly.
7704         (vect_get_and_check_slp_defs): Change the stmts parameter from
7705         a vec<gimple *> to a vec<stmt_vec_info>.
7706         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
7707         (vect_build_slp_tree): Likewise.
7708         (vect_build_slp_tree_2): Likewise.  Update uses of
7709         SLP_TREE_SCALAR_STMTS.
7710         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
7711         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
7712         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
7713         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
7714         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
7715         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
7716         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
7717         (vect_get_constant_vectors, vect_get_slp_defs)
7718         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
7719         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
7720         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
7721         instead of gimple stmts.
7722         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
7723         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
7724         (vect_slp_analyze_instance_dependence): Update uses of
7725         SLP_TREE_SCALAR_STMTS.
7726         (vect_slp_analyze_and_verify_node_alignment): Likewise.
7727         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
7728         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
7729         (get_initial_defs_for_reduction): Likewise.
7730         (vect_create_epilog_for_reduction): Likewise.
7731         (vectorize_fold_left_reduction): Likewise.
7732         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
7733         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
7734         (can_vectorize_live_stmts): Likewise.
7736 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7738         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
7739         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7740         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
7741         than gimple stmts.
7742         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
7743         of a gimple stmt.
7744         (gather_scalar_reductions): Update after above interface changes.
7745         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
7746         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
7747         than gimple stmts.
7748         (vect_force_simple_reduction): Likewise.
7749         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
7750         LOOP_VINFO_REDUCTIONS.
7751         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7753 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7755         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
7756         a gimple stmt to a stmt_vec_info.
7757         * tree-vect-loop.c (vect_active_double_reduction_p)
7758         (vect_force_simple_reduction, vectorizable_reduction): Update
7759         accordingly.
7761 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7763         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
7764         vec<gimple *> to a vec<stmt_vec_info>.
7765         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
7766         the reduction_phis argument from a vec<gimple *> to a
7767         vec<stmt_vec_info>.
7768         (vectorizable_reduction): Likewise the phis local variable that
7769         is passed to vect_create_epilog_for_reduction.  Update for new type
7770         of SLP_TREE_VEC_STMTS.
7771         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
7772         (vectorizable_live_operation): Likewise.
7773         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
7774         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
7776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7778         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
7779         a gimple stmt to a stmt_vec_info.
7780         (vectorizable_condition, vectorizable_live_operation)
7781         (vectorizable_reduction, vectorizable_induction): Pass back the
7782         vectorized statement as a stmt_vec_info.
7783         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
7784         use of STMT_VINFO_VEC_STMT.
7785         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
7786         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
7787         as stmt_vec_infos rather than gimple stmts.
7788         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
7789         to a stmt_vec_info.
7790         (vectorizable_live_operation): Likewise.
7791         (vectorizable_reduction, vectorizable_induction): Likewise,
7792         updating use of STMT_VINFO_VEC_STMT.
7793         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
7794         of STMT_VINFO_VEC_STMT.
7795         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
7796         (vectorizable_simd_clone_call, vectorizable_conversion)
7797         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7798         (vectorizable_store, vectorizable_load, vectorizable_condition)
7799         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
7800         from a gimple stmt to a stmt_vec_info.
7801         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
7802         pointer to a stmt_vec_info to the vectorizable_* routines.
7804 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7806         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7807         a gimple stmt to a stmt_vec_info.
7808         (is_pattern_stmt_p): Update accordingly.
7809         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7810         (vect_record_grouped_load_vectors): Likewise.
7811         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7812         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7813         (vect_model_reduction_cost): Likewise.
7814         (vect_create_epilog_for_reduction): Likewise.
7815         (vectorizable_reduction, vectorizable_induction): Likewise.
7816         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7817         Return the stmt_vec_info for the pattern statement.
7818         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7819         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7820         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7821         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7822         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7823         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7824         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7825         (free_stmt_vec_info, vect_is_simple_use): Likewise.
7827 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7829         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7830         (vect_finish_stmt_generation): Likewise.
7831         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7832         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7833         (vect_build_gather_load_calls): Use the return value of the above
7834         functions instead of a separate call to vinfo_for_stmt.  Use narrow
7835         scopes for the input gimple stmt and wider scopes for the associated
7836         stmt_vec_info.  Use vec_info::lookup_def when setting these
7837         stmt_vec_infos from an SSA_NAME definition.
7838         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7839         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7840         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7841         (vectorizable_store, vectorizable_load, vectorizable_condition)
7842         (vectorizable_comparison): Likewise.
7843         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7844         (vectorizable_reduction): Likewise.
7846 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7848         * tree-vectorizer.h (vect_is_simple_use): Add an optional
7849         stmt_vec_info * parameter before the optional gimple **.
7850         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7851         (process_use, vect_get_vec_def_for_operand_1): Update callers.
7852         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7853         * tree-vect-loop.c (vectorizable_reduction): Likewise.
7854         (vectorizable_live_operation): Likewise.
7855         * tree-vect-patterns.c (type_conversion_p): Likewise.
7856         (vect_look_through_possible_promotion): Likewise.
7857         (vect_recog_rotate_pattern): Likewise.
7858         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7860 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7862         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7863         a typedef to a wrapper class.
7864         (NULL_STMT_VEC_INFO): New macro.
7865         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7866         (stmt_vec_info::operator*): New function.
7867         (stmt_vec_info::operator gimple *): Likewise.
7868         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7869         (add_stmt_costs): Likewise.
7870         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7871         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7872         (vect_get_known_peeling_cost): Likewise.
7873         (vect_estimate_min_profitable_iters): Likewise.
7874         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7875         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7876         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7877         (vectorizable_store, free_stmt_vec_infos): Likewise.
7878         (new_stmt_vec_info): Change return type of xcalloc to
7879         _stmt_vec_info *.
7881 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7883         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7884         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7885         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7886         a single_imm_use-based sequence.
7887         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7889 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7891         * tree-vectorizer.h (vec_info::lookup_def): Declare.
7892         * tree-vectorizer.c (vec_info::lookup_def): New function.
7893         * tree-vect-patterns.c (vect_get_internal_def): Use it.
7894         (vect_widened_op_tree): Likewise.
7895         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7896         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
7897         (vectorizable_reduction): Likewise.
7898         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
7899         of a gimple *.
7900         (vect_is_slp_reduction): Update calls accordingly.  Use
7901         vec_info::lookup_def.
7902         (vect_is_simple_reduction): Likewise
7903         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
7905 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7907         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
7908         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
7909         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
7910         of vinfo_for_stmt.
7911         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
7912         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
7913         (vect_update_vf_for_slp, vect_analyze_loop_operations)
7914         (vect_is_slp_reduction, vectorizable_induction)
7915         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
7916         * tree-vect-patterns.c (vect_init_pattern_stmt):
7917         (vect_determine_min_output_precision_1, vect_determine_precisions)
7918         (vect_pattern_recog): Likewise.
7919         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
7920         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
7921         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
7922         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7923         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
7924         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
7925         info field from a loop to a loop_vec_info.
7927 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7929         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7930         (vec_info::add_stmt): Declare.
7931         * tree-vectorizer.c (vec_info::add_stmt): New function.
7932         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7933         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7934         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7935         (vectorizable_induction): Likewise.
7936         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7937         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7938         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7939         (vectorizable_load): Likewise.
7940         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7941         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7942         (vect_recog_gather_scatter_pattern): Likewise.
7943         (append_pattern_def_seq): Likewise.  Remove a check that is
7944         performed by add_stmt itself.
7946 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7948         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7949         which make_ssa_name was called with new_stmt before new_stmt
7950         had been created.
7952 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7954         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7955         split out from...
7956         (vect_is_slp_reduction): ...here...
7957         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
7958         that are already known to be false.
7960 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7962         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7963         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
7964         STMT_VINFO_NUM_SLP_USES when it's true.
7965         (vect_free_slp_instance): Add a final_p parameter and pass it to
7966         vect_free_slp_tree.
7967         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7968         (vect_analyze_slp_instance): Likewise.
7969         (vect_slp_analyze_operations): Likewise.
7970         (vect_slp_analyze_bb_1): Likewise.
7971         * tree-vectorizer.c (vec_info): Likewise.
7972         * tree-vect-loop.c (vect_transform_loop): Likewise.
7974 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7976         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7977         function is not called for second and subsequent members of
7978         a reduction group.
7980 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7982         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7983         cases for nested loops from here to ...
7984         (vect_create_epilog_for_reduction): ...here.  Only call
7985         vect_is_simple_use for inner-loop reductions.
7987 2018-07-31  Martin Liska  <mliska@suse.cz>
7989         PR gcov-profile/85338
7990         PR gcov-profile/85350
7991         PR gcov-profile/85372
7992         * profile.c (struct location_triplet): New.
7993         (struct location_triplet_hash): Likewise.
7994         (output_location): Do not output a BB that
7995         is already recorded for a line.
7996         (branch_prob): Use streamed_locations.
7998 2018-07-31  Martin Liska  <mliska@suse.cz>
8000         PR gcov-profile/85370
8001         * coverage.c (coverage_begin_function): Do not mark target
8002         clones as artificial functions.
8004 2018-07-31  Martin Liska  <mliska@suse.cz>
8006         PR gcov-profile/83813
8007         PR gcov-profile/84758
8008         PR gcov-profile/85217
8009         PR gcov-profile/85332
8010         * profile.c (branch_prob): Do not record GOTO expressions
8011         for GIMPLE statements which locations are already streamed.
8013 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8015         * gcc.c (handle_spec_function): Accept a soft_matched_part
8016         argument, as do_spec_1.  Pass it down to ...
8017         (eval_spec_function): Accept a soft_matched_part argument,
8018         and pass it down to ...
8019         (do_spec_2): Accept a soft_matched_part argument, and pass
8020         it down to do_spec_1.
8021         (do_spec_1): Pass soft_matched_part to handle_spec_function.
8022         (handle_braces): Update call to handle_spec_function.
8023         (driver::set_up_specs): Update calls to do_spec_2.
8024         (compare_debug_dump_opt_spec_function): Likewise.
8025         (compare_debug_self_opt_spec_function): Likewise.
8027 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8029         * common.opt (nolibc): New option.
8030         * doc/invoke.texi (Link Options): Document it.
8031         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8032         * config/alpha/linux.h: Likewise.
8033         * config/arc/elf.h: Likewise.
8034         * config/arm/uclinux-elf.h: Likewise.
8035         * config/arm/unknown-elf.h: Likewise.
8036         * config/avr/avrlibc.h: Likewise.
8037         * config/bfin/bfin.h: Likewise.
8038         * config/bfin/linux.h: Likewise.
8039         * config/bfin/uclinux.h: Likewise.
8040         * config/darwin.h: Likewise.
8041         * config/darwin10.h: Likewise.
8042         * config/darwin12.h: Likewise.
8043         * config/gnu-user.h: Likewise.
8044         * config/lm32/uclinux-elf.h: Likewise.
8045         * config/pa/pa-hpux11.h: Likewise.
8046         * config/pa/pa64-hpux.h: Likewise.
8047         * config/sparc/sparc.h: Likewise.
8049 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8051         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
8052         undefined variables.
8054 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8056         PR target/86640
8057         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8058         instead of GEN_INT.
8060 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8062         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8063         terminated string literal.
8065 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8067         PR rtl-optimization/85160
8068         * combine.c (is_just_move): New function.
8069         (try_combine): Allow combining two instructions into two if neither of
8070         the original instructions was a move.
8072 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
8074         PR target/86673
8075         * doc/extend.texi (Global Register Variables): Discourage use of type
8076         qualifiers.
8077         (Local Register Variables): Likewise.
8079 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
8081         PR tree-optimization/86506
8082         * hwint.h (ceil_log2): Resync with hwint.c implementation.
8084 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
8086         PR target/86547
8087         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8088         hard_regno, make sure no insn between `from` and `to` clobbers it.
8090 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
8091             Tom de Vries  <tdevries@suse.de>
8093         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8094         (PTX_DEFAULT_RUNTIME_DIM): ... this.
8095         (nvptx_goacc_validate_dims): Set default worker and gang dims to
8096         PTX_DEFAULT_RUNTIME_DIM.
8097         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8099 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
8101         * config/pa/pa.c (pa_output_addr_vec): Align address table.
8102         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8103         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
8105 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
8107         * config/rs6000/constraints.md (wG constraint): Delete, no longer
8108         used.
8109         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
8110         predicate to reflect toc fusion has been deleted.
8111         (toc_fusion_mem_raw): Delete, no longer used.
8112         (toc_fusion_mem_wrapped): Likewise.
8113         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
8114         fusion mask bit.
8115         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
8116         Delete, no longer used.
8117         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
8118         meant to be used for toc fusion.
8119         (rs6000_debug_print_mode): Delete toc fusion debugging.
8120         (rs6000_debug_reg_global): Likewise.
8121         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
8122         fusion and secondary reload support that were never used.
8123         (rs6000_option_override_internal): Delete TOC fusion, that was only
8124         partially defined, and it did not work unless you also used the
8125         -mcmodel= switch.
8126         (rs6000_legitimate_address_p): Delete TOC fusion support.
8127         (rs6000_opt_masks): Likewise.
8128         (fusion_wrap_memory_address): Delete function, no longer used.
8129         (fusion_split_address); Delete TOC fusion support.
8130         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
8131         longer used with toc fusion being deleted.
8132         (TARGET_TOC_FUSION_FP): Likewise.
8133         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
8134         UNSPEC.
8135         (toc fusion spliter): Delete TOC fusion support.
8136         (toc_fusionload_<mode>): Likewise.
8137         (toc_fusionload_di): Likewise.
8138         (fusion_gpr_load_<mode>): Delete generator function, this insn no
8139         longer needs to be named.  Rename predicate to delete TOC fusion.
8140         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8141         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8142         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8143         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8144         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
8146 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8148         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
8149         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
8150         __int128 in built-in function prototypes.
8151         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
8152         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
8154 2018-07-27  Martin Sebor  <msebor@redhat.com>
8156         PR tree-optimization/86696
8157         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
8158         types, including enums.
8159         (handle_char_store): Be prepared for the above function to fail.
8161 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
8163         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
8164         when optimization level is lower than 2 or optimize for size.
8166 2018-07-26  Martin Sebor  <msebor@redhat.com>
8168         PR tree-optimization/86043
8169         PR tree-optimization/86042
8170         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
8171         (get_string_cst_length): Rename...
8172         (get_min_string_length): ...to this.  Add argument.
8173         (handle_char_store): Extend to handle multi-character stores by
8174         MEM_REF.
8175         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
8176         * tree.h (initializer_zerop): Add argument.
8178 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
8180         PR middle-end/86660
8181         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
8182         declare target to variables if they have always,{to,from,tofrom} map
8183         kinds.
8185 2018-07-26  Martin Liska  <mliska@suse.cz>
8187         PR lto/86548
8188         * lto-wrapper.c: Add linker_output as prefix
8189         for ltrans_output_file.
8191 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
8193         PR rtl-optimization/85805
8194         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
8195         value for hard registers if that was written in the same mode.
8197 2018-07-26  Martin Liska  <mliska@suse.cz>
8199         PR gcov-profile/86536
8200         * gcov.c (format_gcov): Use printf format %.*f directly
8201         and do not handle special values.
8203 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8205         * common/config/arc/arc-common.c (arc_option_optimization_table):
8206         Update default optimizations for size.
8208 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8210         * config/arc/arc.md (movsf_insn): Add short instruction selection.
8211         * config/arc/constraints.md (CfZ): New constraint.
8212         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
8213         (subsf3_fpu): Likewise.
8214         (cmpsf_fpu): Likewise.
8215         (cmpsf_fpu_uneq): Likewise.
8217 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8219         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
8220         accesses as well.
8221         (arc_is_uncached_mem_p): uncached applies to both the variable and
8222         the pointer.
8224 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8226         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
8227         register names.
8229 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
8231         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
8232         field "m_scopes" from vec to auto_vec.
8234 2018-07-25  Martin Liska  <mliska@suse.cz>
8236         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
8237         return type.
8239 2018-07-25  Richard Biener  <rguenther@suse.de>
8241         PR debug/86654
8242         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
8243         special wrt context_die late.
8244         (gen_subprogram_die): Re-use DIEs in local scope.
8246 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
8248         PR tree-optimization/86644
8249         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
8251 2018-07-25  Martin Liska  <mliska@suse.cz>
8253         PR middle-end/86645
8254         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
8255         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
8257 2018-07-25  Martin Liska  <mliska@suse.cz>
8259         PR sanitizer/79635
8260         * params.def: Explain ASan abbreviation and provide
8261         a documentation link.
8263 2018-07-24  Martin Sebor  <msebor@redhat.com>
8265         PR tree-optimization/86622
8266         PR tree-optimization/86532
8267         * builtins.h (string_length): Declare.
8268         * builtins.c (c_strlen): Correct handling of non-constant offsets.
8269         (check_access): Be prepared for non-constant length ranges.
8270         (string_length): Make extern.
8271         * expr.c (string_constant): Only handle the minor non-constant
8272         array index.  Use string_constant to compute the length of
8273         a generic string constant.
8275 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
8277         PR tree-optimization/86618
8278         * tree-vect-stmts.c (vectorizable_call): Don't take the address
8279         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
8281 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
8283         PR tree-optimization/86636
8284         * json.cc (json::object::set): Fix comment.  Add assertions.
8285         (json::array::append): Move here from json.h.  Add comment and an
8286         assertion.
8287         (json::string::string): Likewise.
8288         * json.h (json::array::append): Move to json.cc.
8289         (json::string::string): Likewise.
8290         * optinfo-emit-json.cc
8291         (optrecord_json_writer::impl_location_to_json): Assert that we
8292         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
8293         wrapper around it.  Expand the location once, rather than three
8294         times.
8295         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
8296         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
8297         wrappers.
8298         (optrecord_json_writer::optinfo_to_json): Likewise, in four
8299         places.  Fix some overlong lines.
8301 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
8303         * config/aarch64/aarch64-simd.md
8304         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
8305         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
8306         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
8307         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
8308         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
8309         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
8310         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
8311         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
8312         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
8314 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
8316         PR middle-end/86627
8317         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
8318         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
8319         and abs_d == d, do the power of two handling if profitable.
8321 2018-07-24  Richard Biener  <rguenther@suse.de>
8323         * match.pd: Add BIT_FIELD_REF canonicalizations.
8325 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8327         PR c/86617
8328         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
8330 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8332         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
8333         terminated STRING_CST object.
8335 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8337         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
8339 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
8341         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
8342         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
8343         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
8344         the elements into a register.
8345         (rs6000_split_v4si_init_di_reg): Delete.
8346         (rs6000_split_v4si_init): Delete.
8347         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
8348         (vsx_init_v4si): Rewrite as a define_expand.
8350 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
8352         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
8353         zero_extend argument from memory): New.
8355 2018-07-22  Martin Sebor  <msebor@redhat.com>
8357         PR bootstrap/86621
8358         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
8359         diagnosing calls with unknown arguments unless -Walloca-larger-than
8360         is restricted to less than PTRDIFF_MAX bytes.
8362 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
8364         * doc/gcov.texi (Invoking Gcov): Editorial changes.
8366 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
8368         * pretty-print.c (text_info::set_location): Remove redundant
8369         "line_table" parameter from call to rich_location::set_range.
8371 2018-07-20  Martin Sebor  <msebor@redhat.com>
8373         PR middle-end/82063
8374         * builtins.c (expand_builtin_alloca): Adjust.
8375         * calls.c (alloc_max_size): Simplify.
8376         * cgraphunit.c (cgraph_node::expand): Adjust.
8377         * common.opt (larger_than_size, warn_frame_larger_than): Remove
8378         variables.
8379         (frame_larger_than_size): Same.
8380         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
8381         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
8382         Initialize.
8383         * doc/invoke.texi (GCC Command Options): Document option arguments.
8384         Explain byte-size arguments and suffixes.
8385         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
8386         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
8387         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
8388         * doc/options.texi (UInteger): Expand.
8389         (Host_Wide_Int, ByteSize): Document new properties.
8390         * final.c (final_start_function_1): Include sizes in an error message.
8391         * function.c (frame_offset_overflow): Same.
8392         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
8393         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
8394         Diagnose unbounded alloca calls only for limits of less than
8395         PTRDIFF_MAX.
8396         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
8397         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
8398         for alloca(0).
8399         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
8400         only for limits of less than PTRDIFF_MAX.
8401         * langhooks-def.h (lhd_handle_option): Change function argument
8402         to HOST_WIDE_INT.
8403         * langhooks.c (lhd_handle_option): Same.
8404         * langhooks.h (handle_option): Same.
8405         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
8406         ByteSize flags.
8407         (var_type, var_type_struct): Same.
8408         (var_set): Handle ByteSize flag.
8409         * optc-gen.awk: Add comments to output to ease debugging.  Make
8410         use of HOST_WIDE_INT where appropriate.
8411         * opts-gen-save.awk:  Use %lx to format unsigned long.
8412         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
8413         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
8414         arguments.  Parse bytes-size suffixes.
8415         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
8416         (enum_value_to_arg): Same.
8417         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
8418         (handle_option): Adjust.
8419         (generate_option): Change function argument to HOST_WIDE_INT.
8420         (cmdline_handle_error): Adjust.
8421         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
8422         (set_option): Change function argument to HOST_WIDE_INT.
8423         (option_enabled): Handle cl_host_wide_int.
8424         (get_option_state): Handle CLVC_SIZE.
8425         (control_warning_option): Same.
8426         * opts.c (common_handle_option): Change function argument to
8427         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
8428         OPT_Wvla_larger_than_.
8429         * opts.h (enum cl_var_type): Add an enumerator.
8430         * stor-layout.c (layout_decl): Print a more meaningful warning.
8431         * toplev.c (output_stack_usage): Adjust.
8433 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
8435         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
8436         call to inline_expand_builtin_string_cmp.
8437         (expand_builtin_strcmp): Likewise.
8438         (expand_builtin_strncmp): Likewise.
8439         (inline_string_cmp): Delete the last parameter, change char_type_node
8440         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
8441         two operands.
8442         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
8443         the inlining expansion on target where the type of the call has same or
8444         narrower precision than unsigned char.
8446 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
8448         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
8449         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
8450         * common.opt (fsave-optimization-record): New option.
8451         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
8452         * doc/invoke.texi (-fsave-optimization-record): New option.
8453         * dumpfile.c: Include "optinfo-emit-json.h".
8454         (struct kv_pair): Move to coretypes.h.
8455         (optgroup_options): Make non-static.
8456         (dump_context::end_scope): Call
8457         optimization_records_maybe_pop_dump_scope.
8458         * dumpfile.h (optgroup_options): New decl.
8459         * json.cc: New file.
8460         * json.h: New file.
8461         * optinfo-emit-json.cc: New file.
8462         * optinfo-emit-json.h: New file.
8463         * optinfo.cc: Include "optinfo-emit-json.h".
8464         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
8465         (optinfo_enabled_p): Check optimization_records_enabled_p.
8466         (optinfo_wants_inlining_info_p): Likewise.
8467         * optinfo.h: Update comment.
8468         * profile-count.c (profile_quality_as_string): New function.
8469         * profile-count.h (profile_quality_as_string): New decl.
8470         (profile_count::quality): New accessor.
8471         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
8472         and optinfo_emit_json_cc_tests.
8473         * selftest.h (selftest::json_cc_tests): New decl.
8474         (selftest::optinfo_emit_json_cc_tests): New decl.
8475         * toplev.c: Include "optinfo-emit-json.h".
8476         (compile_file): Call optimization_records_finish.
8477         (do_compile): Call optimization_records_start.
8478         * tree-ssa-live.c: Include optinfo.h.
8479         (remove_unused_scope_block_p): Retain inlining information if
8480         optinfo_wants_inlining_info_p returns true.
8482 2018-07-20  Richard Biener  <rguenther@suse.de>
8484         PR debug/86585
8485         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
8486         to cover -flto-partition=none.
8488 2018-07-20  Martin Liska  <mliska@suse.cz>
8490         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
8491         (get_decl_source_range): Remove unused function.
8493 2018-07-20  Richard Biener  <rguenther@suse.de>
8495         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
8496         (struct vn_phi_s): Likewise.
8497         (struct vn_reference_s): Likewise.
8498         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
8499         for searching the slot of an entry known to be in the hash itself.
8500         (vn_phi_hasher::equal): Likewise.
8501         (vn_reference_hasher::equal): Likewise.
8502         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
8503         globals.
8504         (optimistic_info, current_info): Remove, keeping only valid_info.
8505         (vn_reference_lookup_1): Remove fallback lookup.
8506         (vn_reference_lookup_2): Likewise.
8507         (vn_nary_op_lookup_1): Likewise.
8508         (vn_phi_lookup): Likewise.
8509         (vn_nary_build_or_lookup_1): Make sure to not chain the built
8510         hash element.
8511         (vn_reference_insert): Adjust, chain the inserted hash element
8512         at last_inserted_ref.
8513         (vn_reference_insert_pieces): Likewise.
8514         (visit_reference_op_call): Likewise.
8515         (vn_nary_op_insert_into): Chain the inserted hash element at
8516         last_inserted_nary.
8517         (vn_nary_op_insert_pieces): Adjust.
8518         (vn_nary_op_insert): Likewise.
8519         (vn_nary_op_insert_stmt): Likewise.
8520         (vn_phi_insert): Adjust, chain the inserted hash element at
8521         last_inserted_phi.
8522         (process_scc): Remove clearing and copying the optimistic
8523         table.  Instead remove elements inserted during an optimistic
8524         iteration from the single table we maintain.
8525         (init_scc_vn): Adjust.
8526         (free_scc_vn): Likewise.
8527         (sccvn_dom_walker::record_cond): Likewise.
8528         (sccvn_dom_walker::after_dom_children): Likewise.
8530 2018-07-19  Martin Sebor  <msebor@redhat.com>
8532         PR tree-optimization/84047
8533         PR tree-optimization/83776
8534         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
8535         (check_array_bounds): Call it.
8537 2018-07-19  Martin Sebor  <msebor@redhat.com>
8539         * align.h (align_flags): Use member initialization.
8541 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
8543         * Makefile.in (OBJS): Add optinfo.o.
8544         * coretypes.h (class symtab_node): New forward decl.
8545         (struct cgraph_node): New forward decl.
8546         (class varpool_node): New forward decl.
8547         * dump-context.h: New file.
8548         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
8549         "tree-pass.h".
8550         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
8551         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
8552         (set_alt_dump_file): Likewise.
8553         (dump_context::~dump_context): New dtor.
8554         (dump_gimple_stmt): Move implementation to...
8555         (dump_context::dump_gimple_stmt): ...this new member function.
8556         Add the stmt to any pending optinfo, creating one if need be.
8557         (dump_gimple_stmt_loc): Move implementation to...
8558         (dump_context::dump_gimple_stmt_loc): ...this new member function.
8559         Start a new optinfo and add the stmt to it.
8560         (dump_gimple_expr): Move implementation to...
8561         (dump_context::dump_gimple_expr): ...this new member function.
8562         Add the stmt to any pending optinfo, creating one if need be.
8563         (dump_gimple_expr_loc): Move implementation to...
8564         (dump_context::dump_gimple_expr_loc): ...this new member function.
8565         Start a new optinfo and add the stmt to it.
8566         (dump_generic_expr): Move implementation to...
8567         (dump_context::dump_generic_expr): ...this new member function.
8568         Add the tree to any pending optinfo, creating one if need be.
8569         (dump_generic_expr_loc): Move implementation to...
8570         (dump_context::dump_generic_expr_loc): ...this new member
8571         function.  Add the tree to any pending optinfo, creating one if
8572         need be.
8573         (dump_printf): Move implementation to...
8574         (dump_context::dump_printf_va): ...this new member function.  Add
8575         the text to any pending optinfo, creating one if need be.
8576         (dump_printf_loc): Move implementation to...
8577         (dump_context::dump_printf_loc_va): ...this new member function.
8578         Start a new optinfo and add the stmt to it.
8579         (dump_dec): Move implementation to...
8580         (dump_context::dump_dec): ...this new member function.  Add the
8581         value to any pending optinfo, creating one if need be.
8582         (dump_context::dump_symtab_node): New member function.
8583         (dump_context::get_scope_depth): New member function.
8584         (dump_context::begin_scope): New member function.
8585         (dump_context::end_scope): New member function.
8586         (dump_context::ensure_pending_optinfo): New member function.
8587         (dump_context::begin_next_optinfo): New member function.
8588         (dump_context::end_any_optinfo): New member function.
8589         (dump_context::s_current): New global.
8590         (dump_context::s_default): New global.
8591         (dump_scope_depth): Delete global.
8592         (dumpfile_ensure_any_optinfo_are_flushed): New function.
8593         (dump_symtab_node): New function.
8594         (get_dump_scope_depth): Reimplement in terms of dump_context.
8595         (dump_begin_scope): Likewise.
8596         (dump_end_scope): Likewise.
8597         (selftest::temp_dump_context::temp_dump_context): New ctor.
8598         (selftest::temp_dump_context::~temp_dump_context): New dtor.
8599         (selftest::verify_item): New function.
8600         (ASSERT_IS_TEXT): New macro.
8601         (ASSERT_IS_TREE): New macro.
8602         (ASSERT_IS_GIMPLE): New macro.
8603         (selftest::test_capture_of_dump_calls): New test.
8604         (selftest::dumpfile_c_tests): Call it.
8605         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
8606         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
8607         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
8608         descriptive comment.
8609         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
8610         (dump_node, dump_bb): Move these unrelated decls.
8611         (class dump_manager): Add leading comment.
8612         * optinfo.cc: New file.
8613         * optinfo.h: New file.
8615 2018-07-19  Michael Collison  <michael.collison@arm.com>
8616             Richard Henderson <rth@redhat.com>
8618         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
8619         (subti): Handle op1 zero.
8620         (subvti4, usub4ti4): New.
8621         (*sub<GPI>3_compare1_imm): New.
8622         (sub<GPI>3_carryinCV): New.
8623         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
8624         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
8626 2018-07-19  Michael Collison  <michael.collison@arm.com>
8627             Richard Henderson <rth@redhat.com>
8629         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
8630         (addti3): Create simpler code if low part is already known to be 0.
8631         (addvti4, uaddvti4): New.
8632         (*add<GPI>3_compareC_cconly_imm): New.
8633         (*add<GPI>3_compareC_cconly): New.
8634         (*add<GPI>3_compareC_imm): New.
8635         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
8636         handle constants within this pattern..
8637         (*add<GPI>3_compareV_cconly_imm): New.
8638         (*add<GPI>3_compareV_cconly): New.
8639         (*add<GPI>3_compareV_imm): New.
8640         (add<GPI>3_compareV): New.
8641         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
8642         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
8643         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
8644         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
8645         with 'comparison' operator.
8646         (*add<GPI>3_compareV_cconly_imm): Ditto.
8647         (*add<GPI>3_compareV_cconly): Ditto.
8648         (*add<GPI>3_compareV_imm): Ditto.
8649         (add<GPI>3_compareV): Ditto.
8650         (add<mode>3_carryinC): Ditto.
8651         (*add<mode>3_carryinC_zero): Ditto.
8652         (*add<mode>3_carryinC): Ditto.
8653         (add<mode>3_carryinV): Ditto.
8654         (*add<mode>3_carryinV_zero): Ditto.
8655         (*add<mode>3_carryinV): Ditto.
8657 2018-07-19  Michael Collison  <michael.collison@arm.com>
8658             Richard Henderson <rth@redhat.com>
8660         * config/aarch64/aarch64-modes.def (CC_V): New.
8661         * config/aarch64/aarch64-protos.h
8662         (aarch64_addti_scratch_regs): Declare
8663         (aarch64_subvti_scratch_regs): Declare.
8664         (aarch64_expand_subvti): Declare.
8665         (aarch64_gen_unlikely_cbranch): Declare
8666         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
8667         for signed overflow using CC_Vmode.
8668         (aarch64_get_condition_code_1): Handle CC_Vmode.
8669         (aarch64_gen_unlikely_cbranch): New function.
8670         (aarch64_addti_scratch_regs): New function.
8671         (aarch64_subvti_scratch_regs): New function.
8672         (aarch64_expand_subvti): New function.
8674 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8676         * config/aarch64/aarch64-option-extensions.def: New entry for profile
8677         extension.
8678         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
8679         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
8680         extension.
8682 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8684         PR target/83009
8685         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8686         address check not strict.
8688 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8690         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
8691         Umq with Umn.
8692         (store_pair_lanes<mode>): Likewise.
8693         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
8694         enum value 'ADDR_QUERY_LDP_STP_N'.
8695         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
8696         (aarch64_print_address_internal): Add declaration.
8697         (aarch64_print_ldpstp_address): Remove.
8698         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
8699         (aarch64_print_operand): Change printing of 'y'.
8700         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
8701         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
8702         'true' rather than '1'.
8703         * config/aarch64/constraints.md (Uml): Likewise.
8704         (Uml): Rename to Umn.
8705         (Umq): Remove.
8707 2018-07-19  Richard Biener  <rguenther@suse.de>
8709         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
8710         a trailing array.
8711         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
8712         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
8713         (vn_reference_hasher): Likewise.
8714         (struct vn_tables_s): Remove obstack and alloc-pool members.
8715         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
8716         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
8717         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
8718         (vn_reference_insert_pieces): Likewise.
8719         (alloc_vn_nary_op_noinit): Adjust.
8720         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
8721         (vn_phi_eq): Adjust.
8722         (shared_lookup_phiargs): Remove.
8723         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
8724         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
8725         (visit_reference_op_call): Likewise.
8726         (copy_nary, copy_phi, copy_reference): Remove.
8727         (process_scc): Rewind the obstack when iterating.  Do not
8728         copy the elements to valid_info but just move them from one
8729         hashtable to the other.
8730         (allocate_vn_table): Adjust.
8731         (free_vn_table): Likewise.
8732         (init_scc_vn): Likewise.
8733         (free_scc_vn): Likewise.
8735 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
8737         PR target/86560
8738         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
8739         indirect_return as function type attribute.
8740         (ix86_attribute_table): Change indirect_return to function
8741         type attribute.
8742         * doc/extend.texi: Update indirect_return attribute.
8744 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
8746         * wide-int.h (widest2_int): New.
8747         * gimple-fold.c (arith_overflowed_p): Use it.
8748         * tree.h (widest2_int_cst): New.
8749         * tree-vrp.c (wide_int_binop_overflow): Rename from
8750         vrp_int_const_binop.
8751         Rewrite to work on trees.
8752         (extract_range_from_multiplicative_op_1): Abstract code to...
8753         (wide_int_range_min_max): ...here.
8754         (wide_int_range_cross_product): ...and here.
8755         (extract_range_from_binary_expr_1): Abstract overflow code to...
8756         (wide_int_range_mult_wrapping): ...here.
8757         * tree-vrp.h (wide_int_range_cross_product): New.
8758         (wide_int_range_mult_wrapping): New.
8760 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
8761             Julia Koval  <julia.koval@intel.com>
8763         * config/i386/x86-tune-costs.h (skylake_memcpy,
8764         skylake_memset): Replace rep_prefix with unrolling for size 512.
8766 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
8768         PR middle-end/86544
8769         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
8770         comparision with EQ_EXPR in last stmt.
8772 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8774         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
8775         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
8776         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
8777         previously known as "PowerPC AltiVec Built-in Functions".  Move
8778         some material to new subsubsections "PowerPC AltiVec Built-in
8779         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
8780         ISA 2.07".
8781         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
8782         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
8783         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
8784         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
8786 2018-07-18  Richard Biener  <rguenther@suse.de>
8788         PR tree-optimization/86557
8789         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
8790         EXACT_DIV_EXPR.
8792 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
8794         * config/s390/s390.c (s390_function_profiler): Generate CFI.
8796 2018-07-17  Jeff Law  <law@redhat.com>
8798         * config/arm/arm.c (get_label_padding): Update for recent
8799         changes to label_to_alignment.
8801         PR tree-optimization/86010
8802         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
8804         * config/mips/mips.c (vr4130_align_insns): Update for recent
8805         changes to label_to_alignment.
8807         * config/frv/frv.c (frv_label_align): Update for recent changes
8808         to label_to_alignment.
8810         * config/nios2/nios2.c (nios2_label_align): Update for recent
8811         changes which dropped ALIGN_LABELS_LOG.
8813 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
8815         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8816         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8818 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
8820         * config/arc/arc.c (arc_label_align): Use align_labels instead of
8821         deprecated align_labels_log.
8823 2018-07-17  Richard Biener  <rguenther@suse.de>
8825         PR lto/86456
8826         * dwarf2out.c (init_sections_and_labels): Always generate
8827         a debug_line_str_section for early LTO debug.
8828         (dwarf2out_finish): Reset debug_line_str_hash output early.
8829         Bump counter for extra dwarf5 .debug_loc labels to not conflict
8830         with fat LTO part.
8831         (dwarf2out_early_finish): Output debug_line_str.
8833 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
8835         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8836         index register on z196 or later.
8838 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
8840         * config/s390/s390.c (s390_default_align): Set default function
8841         alignment to 16.
8842         (s390_override_options_after_change): Call s390_default align.
8843         (s390_option_override_internal): Call s390_default_align.
8844         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8846 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8848         PR middle-end/86542
8849         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8850         to _looptemp_ clauses, other than the first two.
8852 2018-07-17  Martin Liska  <mliska@suse.cz>
8854         * opts.c: Do not enable OPT_falign_* for -Os.
8856 2018-07-17  Martin Liska  <mliska@suse.cz>
8858         * align.h (MAX_CODE_ALIGN): New.
8859         (MAX_CODE_ALIGN_VALUE): New.
8860         * common/config/i386/i386-common.c (ix86_handle_option):
8861         (MAX_CODE_ALIGN): Moved to align.h.
8862         * final.c (MAX_CODE_ALIGN): Likewise.
8863         * opts.c (parse_and_check_align_values):
8864         (MAX_CODE_ALIGN): Likewise.
8865         (MAX_CODE_ALIGN_VALUE): Likewise.
8867 2018-07-17  Martin Liska  <mliska@suse.cz>
8869         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8870         in order to fulfil coding style.
8871         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8872         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8873         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8874         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8875         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8876         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8877         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8878         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8879         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8881 2018-07-17  Martin Liska  <mliska@suse.cz>
8883         * align.h: New file.
8884         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8885         directly.
8886         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8887         align_flags of label_to_alignment.
8888         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8889         align_flags class.
8890         * config/m68k/m68k.c: Do not use removed align_labels_value and
8891         align_loops_value.
8892         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
8893         (LOOP_ALIGN): Likewise.
8894         (LABEL_ALIGN): Likewise.
8895         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8896         Remove not used macro.
8897         (rs6000_loop_align): Change return type to align_flags.
8898         (rs6000_loop_align_max_skip): Remove.
8899         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
8900         Change return type to align_flags.
8901         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8902         Remove not used macro.
8903         (rs6000_loop_align):  Change return type to align_flags.
8904         (rs6000_loop_align_max_skip): Remove.
8905         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
8906         * config/rx/rx-protos.h (rx_align_for_label): Make it
8907         static function.
8908         * config/rx/rx.c (rx_align_for_label): Change return type
8909         to align_flags.
8910         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8911         macro definitions.
8912         into align_flags class.
8913         (LABEL_ALIGN): Likewise.
8914         (LOOP_ALIGN): Likewise.
8915         * config/s390/s390.c (s390_label_align): Use align_flags
8916         class member.
8917         (s390_asm_output_function_label): Likewise.
8918         * config/sh/sh.c (sh_override_options_after_change):
8919         Use align_flags class directly without macros.
8920         (find_barrier): Likewise.
8921         (barrier_align): Likewise.
8922         (sh_loop_align): Likewise.
8923         * config/spu/spu.c (spu_option_override):
8924         Use align_flags_tuple::get_value instead of removed macros.
8925         (spu_sched_init): Likewise.
8926         * config/spu/spu.h (GTY): Likewise.
8927         * config/visium/visium.c (visium_option_override):
8928         Set "8" as default secondary alignment.
8929         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8930         in order to guarantee secondary alignment of 8.
8931         * coretypes.h: Include align.h header file.
8932         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8933         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8934         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8935         * doc/tm.texi.in: Likewise.
8936         * final.c (struct label_alignment): Remove not used structure.
8937         (LABEL_ALIGN): Change type to align_flags.
8938         (LOOP_ALIGN): Likewise.
8939         (JUMP_ALIGN): Likewise.
8940         (default_loop_align_max_skip): Remove.
8941         (default_label_align_max_skip): Likewise.
8942         (default_jump_align_max_skip): Likewise.
8943         (default_label_align_after_barrier_max_skip):
8944         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8945         (LABEL_TO_MAX_SKIP): Remove.
8946         (label_to_alignment): Return align_flags type instead of integer.
8947         (label_to_max_skip): Remove.
8948         (align_fuzz): Use align_flags type.
8949         (compute_alignments): Use align_flags type and use align_flags::max
8950         to combine multiple alignments.
8951         (grow_label_align): Grow vec instead of C array.
8952         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8953         (shorten_branches):  Use align_flags type and use align_flags::max
8954         to combine multiple alignments.
8955         (final_scan_insn_1): Remove usage of secondary alignment that comes
8956         from label alignment, but instead use proper secondary alignment
8957         which is computed in grow_label_align.
8958         * flags.h (struct align_flags_tuple): Move to align.h.
8959         (struct align_flags): Likewise.
8960         (state_align_loops): Rename to align_loops.
8961         (state_align_jumps): Rename to align_jumps.
8962         (state_align_labels): Rename to align_labels.
8963         (state_align_functions): Rename to align_functions.
8964         (align_loops_log): Remove.
8965         (align_jumps_log): Remove.
8966         (align_labels_log): Remove.
8967         (align_functions_log): Remove.
8968         (align_loops_max_skip): Remove.
8969         (align_jumps_max_skip): Remove.
8970         (align_labels_max_skip): Remove.
8971         (align_functions_max_skip): Remove.
8972         (align_loops_value): Remove.
8973         (align_jumps_value): Remove.
8974         (align_labels_value): Remove.
8975         (align_functions_value): Remove.
8976         * output.h (label_to_alignment): Change return type to align_flags.
8977         (label_to_max_skip): Remove.
8978         * target.def: Remove loop_align_max_skip, label_align_max_skip,
8979         jump_align_max_skip macros.
8980         * targhooks.h (default_loop_align_max_skip): Remove.
8981         (default_label_align_max_skip): Likewise.
8982         (default_jump_align_max_skip): Likewise.
8983         (default_label_align_after_barrier_max_skip): Remove.
8984         * toplev.c (read_log_maxskip): Use ::normalize function.
8985         (parse_N_M): Remove not used argument and also call ::normalize.
8986         (parse_alignment_opts): Do not pass unused arguments.
8987         * varasm.c (assemble_start_function): Use directly align_functions
8988         instead of removed macros.
8989         * system.h: Do not poison removed macros.
8991 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
8993         PR middle-end/86539
8994         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8995         and cond temporaries don't have reference type if iterator has
8996         pointer type.  For init use &for_pre_body instead of pre_p if
8997         for_pre_body is non-empty.
8999 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9001         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9002         double-double modes to SFmode directly directly.
9003         (trunc<mode>sf2_fprs): Delete.
9005 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9007         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9008         for conversions between IFmode and the decimal floating point modes.
9009         (init_float128_ieee): Use the correct names for conversions between
9010         KFmode and the decimal floating point modes.
9012 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9014         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9015         for the conversions between TDmode and IFmode.
9016         (init_float128_ieee): Use more correct names for the conversions
9017         between TDmode and KFmode.
9019 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
9021         PR tree-optimization/86526
9022         * builtins.c (expand_builtin_memcmp): Formatting fixes.
9023         (inline_expand_builtin_string_cmp): Likewise.
9024         (inline_string_cmp): Likewise.  Use c_readstr instead of
9025         builtin_memcpy_read_str.  Add unit_mode temporary.
9027 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9029         PR middle-end/86528
9030         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9031         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9033 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9035         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9036         Alphabetize prototypes of built-in functions, separating out
9037         built-in functions that are listed in this section but should be
9038         described elsewhere.
9040 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
9042         PR target/86511
9043         * expmed.c (emit_store_flag): Do not emit setcc followed by a
9044         conditional move when trapping comparison was split to a
9045         non-trapping one (and vice versa).
9047 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9049         * config/s390/s390.c (s390_function_profiler): Generate nops
9050         instead of profiler call sequences.
9051         * config/s390/s390.opt: Add the new option.
9053 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9055         * config/s390/s390.c (s390_function_profiler): Generate
9056         __mcount_loc section.
9057         * config/s390/s390.opt: Add the new option.
9059 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9061         * common.opt: Add the new warning.
9062         * config/s390/s390.c (s390_function_profiler): Emit "brasl
9063         %r0,__fentry__" when -mfentry is specified.
9064         (s390_option_override_internal): Disallow -mfentry for 31-bit
9065         CPUs.
9066         * config/s390/s390.opt: Add the new option.
9068 2018-07-16  Richard Biener  <rguenther@suse.de>
9070         PR lto/86523
9071         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9072         for function-local FUNCTION_DECL and RESULT_DECL immediately.
9074 2018-07-16  Martin Liska  <mliska@suse.cz>
9076         PR ipa/86529
9077         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9078         to ::get_create.
9080 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9082         * config/arc/arcHS.md: Update ARCHS scheduling rules.
9084 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9086         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9087         for ARCHS4x.
9088         * config/arc/arc-cpus.def (hs4x): New cpu.
9089         (hs4xd): Likewise.
9090         * config/arc/arc-tables.opt: Regenerate.
9091         * config/arc/arc.c (arc_sched_issue_rate): New function.
9092         (TARGET_SCHED_ISSUE_RATE): Define.
9093         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9094         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9095         fpu_cvt.
9096         (attr tune): Add ARCHS4x tune values.
9097         (attr tune_dspmpy): Define.
9098         (*tst): Correct instruction type.
9099         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9100         * config/arc/arcHS4x.md: New file.
9101         * config/arc/fpu.md: Update instruction type attributes.
9102         * config/arc/t-multilib: Regenerate.
9104 2018-07-16  Tom de Vries  <tdevries@suse.de>
9106         PR debug/86455
9107         * var-tracking.c (vt_initialize): Fix pre_dec handling.
9109 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9111         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
9112         early clobber.
9114 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
9116         PR tree-optimization/86514
9117         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
9118         conversion to a boolean type from a type with greater precision.
9120 2018-07-16  Tom de Vries  <tdevries@suse.de>
9122         * var-tracking.c (vt_initialize): Print adjusted insn slim if
9123         dump_flags request TDF_SLIM.
9125 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
9127         * fold-const.c (int_const_binop_1): Abstract...
9128         (wide_int_binop): ...wide int code here.
9129         (poly_int_binop): ...poly int code here.
9130         Abstract the rest of int_const_binop_1 into int_const_binop.
9131         * fold-const.h (wide_int_binop): New.
9132         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
9133         Remove useless PLUS/MINUS_EXPR case.
9134         (zero_nonzero_bits_from_vr): Move wide int code...
9135         (zero_nonzero_bits_from_bounds): ...here.
9136         (extract_range_from_binary_expr_1): Move mask optimization code...
9137         (range_easy_mask_min_max): ...here.
9138         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
9139         (range_easy_mask_min_max): New.
9141 2018-07-15  Jeff Law  <law@redhat.com>
9143         PR target/85993
9144         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
9145         block.
9147 2018-07-14  Jim Wilson  <jimw@sifive.com>
9149         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
9151 2018-07-14  Paul Koning  <ni1d@arrl.net>
9153         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
9155 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
9157         * lto-streamer-out.c (copy_function_or_variable): Dump info about
9158         copying section.
9160 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
9161             Steve Munroe  <munroesj52@gmail.com>
9163         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
9164         (_mm_andnot_si128): Likewise.
9165         (_mm_or_si128): Likewise.
9166         (_mm_xor_si128): Likewise.
9168 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
9170         PR middle-end/78809
9171         * builtins.c (expand_builtin_memcmp): Inline the calls first
9172         when result_eq is false.
9173         (expand_builtin_strcmp): Inline the calls first.
9174         (expand_builtin_strncmp): Likewise.
9175         (inline_string_cmp): New routine. Expand a string compare
9176         call by using a sequence of char comparison.
9177         (inline_expand_builtin_string_cmp): New routine. Inline expansion
9178         a call to str(n)cmp/memcmp.
9179         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
9180         New option.
9181         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
9183 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
9185         * config/arm/driver-arm.c: Include arm-native.h.
9186         (host_detect_local_cpu): Use auto-generated data tables.
9187         (vendors, arm_cpu_table): Delete.  Move part information to ...
9188         * config/arm/arm-cpus.in: ... here.
9189         * config/arm/parsecpu.awk (gen_native): New function.
9190         (vendor, part): New CPU fields.
9191         (END): Add support for building the native CPU detection tables.
9192         * config/arm/t-arm (arm-native.h): Add build rule.
9193         (driver-arm.o): Add dependency on arm-native.h.
9195 2018-07-13  Richard Biener  <rguenther@suse.de>
9197         PR middle-end/85974
9198         * match.pd (addr1 - addr2): Allow either of the operand to
9199         have a conversion.
9201 2018-07-13  Tom de Vries  <tdevries@suse.de>
9203         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
9204         in remap_ssa_name.
9206 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
9208         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
9209         arrays instead of numbered variables.
9211 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
9213         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
9214         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
9216 2018-07-13  Richard Biener  <rguenther@suse.de>
9218         PR debug/86452
9219         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
9220         instead of get_context_die.
9222 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
9223             Richard Biener  <rguenther@suse.de>
9225         PR middle-end/86489
9226         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
9227         that the loop latch destination where phi is defined.
9229 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
9231         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
9232         (riscv_expand_epilogue): Add assertion to check interrupt mode.
9233         (riscv_set_current_function): Extract getting interrupt type to new
9234         function.
9235         (riscv_get_interrupt_type): New function.
9236         (riscv_merge_decl_attributes): New function, checking interrupt type is
9237         same.
9238         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
9240 2018-07-12  Paul Koning  <ni1d@arrl.net>
9242         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
9243         directive.
9245 2018-07-12  Paul Koning  <ni1d@arrl.net>
9247         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
9248         zero reference, add doloop_end instead.
9249         * doc/md.texi (decrement_and_branch_until_zero): Remove.
9250         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
9251         detail for doloop_end.
9253 2018-07-12  Martin Sebor  <msebor@redhat.com>
9255         PR c/86453
9256         * attribs.c (decl_attributes): Reject conflicting attributes before
9257         calling attribute handlers.
9259 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9261         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
9262          parameter.
9263         (gcc::dump_manager::get_dump_file_name): likewise.
9264         (dump_begin): Likewise.
9265         * dumpfile.h (dump_begin): Update prototype.
9266         (gcc::dump_manager::get_dump_file_name,
9267         gcc::dump_manager::get_dump_file_name): Update prototype.
9269 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9271         * internal-fn.h (vectorizable_internal_fn_p): New function.
9272         * tree-vect-slp.c (compatible_calls_p): Likewise.
9273         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
9274         to internal functions.
9275         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
9277 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9279         * fold-const.h (inverse_conditions_p): Declare.
9280         * fold-const.c (inverse_conditions_p): New function.
9281         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
9282         that test the inverse condition of a conditional internal function.
9283         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
9284         * internal-fn.c (internal_fn_mask_index): Handle conditional
9285         internal functions.
9286         (vectorized_internal_fn_supported_p): New function.
9287         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
9288         (any_pred_load_store): Replace with...
9289         (need_to_predicate): ...this new variable.
9290         (redundant_ssa_names): New variable.
9291         (ifcvt_can_use_mask_load_store): Move initial checks to...
9292         (ifcvt_can_predicate): ...this new function.  Handle tree codes
9293         for which a conditional internal function exists.
9294         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
9295         instead of ifcvt_can_use_mask_load_store.  Update after variable
9296         name change.
9297         (predicate_load_or_store): New function, split out from
9298         predicate_mem_writes.
9299         (check_redundant_cond_expr): New function.
9300         (value_available_p): Likewise.
9301         (predicate_rhs_code): Likewise.
9302         (predicate_mem_writes): Rename to...
9303         (predicate_statements): ...this.  Use predicate_load_or_store
9304         and predicate_rhs_code.
9305         (combine_blocks, tree_if_conversion): Update after above name changes.
9306         (ifcvt_local_dce): Handle redundant_ssa_names.
9307         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
9308         general conditional functions.
9309         * tree-vect-stmts.c (vectorizable_call): Likewise.
9311 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9312             Alan Hayward  <alan.hayward@arm.com>
9313             David Sherwood  <david.sherwood@arm.com>
9315         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
9316         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
9317         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
9318         plus and minus and convert them into IFN_COND_FMA-based sequences.
9319         (convert_mult_to_fma): Handle conditional plus and minus.
9321 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9323         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
9324         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
9325         (cond_fnms_optab): New optabs.
9326         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
9327         internal functions.
9328         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
9329         * internal-fn.h (get_conditional_internal_fn): Declare.
9330         (get_unconditional_internal_fn): Likewise.
9331         * internal-fn.c (cond_ternary_direct): New macro.
9332         (expand_cond_ternary_optab_fn): Likewise.
9333         (direct_cond_ternary_optab_supported_p): Likewise.
9334         (FOR_EACH_COND_FN_PAIR): Likewise.
9335         (get_conditional_internal_fn): New function.
9336         (get_unconditional_internal_fn): Likewise.
9337         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
9338         (gimple_match_op::gimple_match_op): Add a new overload for 5
9339         operands.
9340         (gimple_match_op::set_op): Likewise.
9341         (gimple_resimplify5): Declare.
9342         * genmatch.c (decision_tree::gen): Generate simplifications for
9343         5 operands.
9344         * gimple-match-head.c (gimple_simplify): Define an overload for
9345         5 operands.  Handle calls with 5 arguments in the top-level overload.
9346         (convert_conditional_op): Handle conversions from unconditional
9347         internal functions to conditional ones.
9348         (gimple_resimplify5): New function.
9349         (build_call_internal): Pass a fifth operand.
9350         (maybe_push_res_to_seq): Likewise.
9351         (try_conditional_simplification): Try converting conditional
9352         internal functions to unconditional internal functions.
9353         Handle 3-operand unconditional forms.
9354         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
9355         Define ternary equivalents of the current rules for binary conditional
9356         internal functions.
9357         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
9358         ternary operations.
9359         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
9360         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
9361         (optab): Handle them.
9362         (SVE_COND_FP_TERNARY): New int iterator.
9363         (sve_fmla_op, sve_fmad_op): New int attributes.
9364         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
9365         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
9366         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
9368 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9370         * target.def (preferred_else_value): New target hook.
9371         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
9372         * doc/tm.texi: Regenerate.
9373         * targhooks.h (default_preferred_else_value): Declare.
9374         * targhooks.c (default_preferred_else_value): New function.
9375         * internal-fn.h (conditional_internal_fn_code): Declare.
9376         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
9377         (get_conditional_internal_fn): Use it.
9378         (conditional_internal_fn_code): New function.
9379         * gimple-match.h (gimple_match_cond): New struct.
9380         (gimple_match_op): Add a cond member function.
9381         (gimple_match_op::gimple_match_op): Update all forms to take a
9382         gimple_match_cond.
9383         * genmatch.c (expr::gen_transform): Use the same condition as res_op
9384         for the suboperation, but don't specify a particular else_value.
9385         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
9386         (visit_nary_op, visit_reference_op_load): Pass
9387         gimple_match_cond::UNCOND to the gimple_match_op constructor.
9388         * gimple-match-head.c: Include tree-eh.h
9389         (convert_conditional_op): New function.
9390         (maybe_resimplify_conditional_op): Likewise.
9391         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
9392         (gimple_resimplify2): Likewise.
9393         (gimple_resimplify3): Likewise.
9394         (gimple_resimplify4): Likewise.
9395         (maybe_push_res_to_seq): Return null for conditional operations.
9396         (try_conditional_simplification): New function.
9397         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
9398         constructor.
9399         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
9400         IFN_COND_* call.
9401         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
9402         function.
9403         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
9405 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9407         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9408         DECL_FCONTEXT
9409         (hash_tree): Do not hash DECL_FCONTEXT
9410         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
9411         Do not stream DECL_FCONTEXT.
9412         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
9413         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
9415 2018-07-12  Richard Biener  <rguenther@suse.de>
9417         PR debug/86462
9418         * dwarf2out.c (gen_block_die): Only output blocks when they have
9419         at least one !DECL_IGNORED_P variable.
9421 2018-07-12  Richard Biener  <rguenther@suse.de>
9423         PR target/84829
9424         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
9425         Remove -mieee-fp handling.
9427 2018-07-12  Richard Biener  <rguenther@suse.de>
9429         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
9430         left-over from last patch.
9432 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
9434         PR tree-optimization/86492
9435         * gimple-ssa-store-merging.c
9436         (imm_store_chain_info::coalesce_immediate_stores): Call
9437         check_no_overlap even for the merge_overlapping case.  Formatting fix.
9439 2018-07-12  Richard Biener  <rguenther@suse.de>
9441         PR middle-end/86479
9442         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
9443         move possibly trapping operations into the conditional.
9445 2018-07-12  Richard Biener  <rguenther@suse.de>
9447         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
9448         (vn_lookup_simplify_result): Remove recursion limit applied
9449         here.
9450         (vn_nary_build_or_lookup_1): Adjust.
9451         (try_to_simplify): Likewise.
9452         * gimple-match-head.c (gimple_resimplify1): Instead apply one
9453         here.
9454         (gimple_resimplify2): Likewise.
9455         (gimple_resimplify3): Likewise.
9456         (gimple_resimplify4): Likewise.
9458 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
9460         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
9461         Use __mmask64 type instead of __mmask8 for __M argument.
9462         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
9463         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
9464         __U argument.
9465         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
9466         __mmask16 for __M argument.
9467         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
9468         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
9469         to __mmask16 instead of __mmask8.
9470         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
9471         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
9472         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
9473         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
9474         instead of __mmask16 for __U argument.
9475         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
9476         __mmask16 instead of __mmask8 for __U argument.
9477         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
9478         __U argument.
9479         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
9480         __mmask16.
9481         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
9482         argument.
9483         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
9484         __U argument.
9485         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
9486         __mmask16.
9487         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
9488         of __mmask16.
9489         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
9490         __U argument.
9491         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
9492         __U argument.
9493         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
9494         __U argument.
9495         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
9496         __U argument.
9497         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
9498         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
9499         return type as well as __M argument type and all casts from __mmask8
9500         to __mmask32.
9501         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
9502         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
9503         return type as well as __M argument type and all casts from __mmask8
9504         to __mmask16.
9505         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
9506         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
9507         return type as well as __M argument type and all casts from __mmask8
9508         to __mmask32.
9509         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
9510         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
9511         return type as well as __M argument type and all casts from __mmask8
9512         to __mmask16.
9513         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
9514         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
9515         __mmask16.
9517 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
9519         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
9520         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
9521         for __U argument.
9523 2018-07-11  Paul Koning  <ni1d@arrl.net>
9525         * doc/md.texi (define_subst): Document how multiple occurrences of
9526         the same argument in the replacement pattern are handled.
9528 2018-07-11  Paul Koning  <ni1d@arrl.net>
9530         * doc/extend.texi (Common Variable Attributes): Move "mode" into
9531         alphabetical order.
9532         (Common Type Attributes): Add "mode" attribute.
9534 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
9536         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
9537         stream DECL_ORIGINAL_TYPE.
9538         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
9539         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
9540         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
9541         Do not walk original type.
9542         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
9543         external decls.
9544         (write_ts_decl_non_common_tree_pointers): Do not stream
9545         DECL_ORIGINAL_TYPE
9546         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
9547         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
9549 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
9551         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
9552         thread twice from the same starting edge.
9554 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
9556         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
9557         * gimple.c (gimple_call_nonnull_result_p): ...here...
9558         (gimple_call_nonnull_arg): ...and here.
9559         * gimple.h (gimple_call_nonnull_result_p): New.
9560         (gimple_call_nonnull_arg): New.
9562 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
9564         * config/arm/arm-cpus.in: Move information from fpu field of each
9565         cpu definition to the isa field.
9566         * config/arm/parsecpu.awk (fpu): Delete match rule.
9567         (gen_comm_data): Don't add bits from the CPU's FPU entry.
9569 2018-07-11  Richard Biener  <rguenther@suse.de>
9571         PR debug/86457
9572         * dwarf2out.c (init_sections_and_labels): Use
9573         output_asm_line_debug_info consistently.
9574         (dwarf2out_early_finish): Likewise.
9575         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
9576         type units.
9578 2018-07-11  Richard Biener  <rguenther@suse.de>
9580         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
9581         Rework father_bb setting in a way to avoid propagating constants
9582         multiple times on a loop body.
9584 2018-07-10  Mark Wielaard  <mark@klomp.org>
9586         PR debug/86459
9587         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
9589 2018-07-10  Richard Biener  <rguenther@suse.de>
9591         * hash-map.h (hash_map::iterator::operator*): Return
9592         references to key and value.
9594 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
9596         PR c++/86443
9597         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
9598         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
9599         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
9600         (gimplify_omp_for): For composite loops, move outer
9601         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
9602         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
9603         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
9604         TREE_LIST for both the original class iterator and the "last" helper
9605         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
9606         loop, remember has_decl_expr from outer composite loops for the
9607         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
9609 2018-07-09  Martin Sebor  <msebor@redhat.com>
9611         PR middle-end/77357
9612         PR middle-end/86428
9613         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
9614         accessing implicitly initialized array elements.
9615         * expr.c (string_constant): Handle string initializers of
9616         character arrays within aggregates.
9617         * gimple-fold.c (fold_array_ctor_reference): Add argument.
9618         Store element offset.  As a special case, handle zero size.
9619         (fold_nonarray_ctor_reference): Same.
9620         (fold_ctor_reference): Add argument.  Store subobject offset.
9621         * gimple-fold.h (fold_ctor_reference): Add argument.
9623 2018-07-09  Paul Koning  <ni1d@arrl.net>
9625         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
9626         (pdp11_insn_cost): New function.
9627         (pdp11_md_asm_adjust): New function.
9628         (TARGET_INVALID_WITHIN_DOLOOP): Define.
9629         (pdp11_rtx_costs): Update to match machine better.
9630         (output_addr_const_pdp11): Correct format mismatch warnings.
9631         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
9632         * config/pdp11/pdp11.md: General change to add base_cost and/or
9633         length attributes for use by new pdp11_insn_cost function.
9634         (MIN_BRANCH): Correct definition.
9635         (MIN_SOB): Ditto.
9636         (doloop_end): Use standard pattern name for looping pattern.
9637         (doloop_end_nocc): New.
9638         (movsf): Add another constraint alternative.
9639         (zero_extendqihi2): Add constraint alternatives for not in place
9640         extend.
9641         (zero_extendhisi2): Remove.
9642         (shift patterns): Add CC handling variants.
9643         (bswaphi2): New.
9644         (bswapsi2): New.
9645         (rothi3): New.
9646         (define_peephole2): New peephole to recognize mov that sets CC for
9647         subsequent test.
9649 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9651         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
9652         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
9653         wi::add.
9655 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
9657         PR c/86420
9658         * real.c (real_nextafter): Return true if result is denormal.
9660 2018-07-09  Martin Liska  <mliska@suse.cz>
9662         * common.opt: Add back wrongly removed attribute.
9664 2018-07-09  Richard Biener  <rguenther@suse.de>
9666         PR debug/86413
9667         * dwarf2out.c (gen_block_die): For an early generated DIE
9668         always output high/low PC attributes.
9670 2018-07-09  Tom de Vries  <tdevries@suse.de>
9672         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
9673         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
9674         onto VAR_DECL with abstract origin.
9676 2018-07-07  Jim Wilson  <jimw@sifive.com>
9678         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
9680 2018-07-07  Tom de Vries  <tdevries@suse.de>
9682         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
9684 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
9686         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
9687         overflow_type.
9688         (combine_bound): Use wide-int overflow calculation instead of
9689         rolling our own.
9690         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
9691         overflow_type.
9692         * fold-const.c (int_const_binop_2): Same.
9693         (extract_muldiv_1): Same.
9694         (fold_div_compare): Same.
9695         (fold_abs_const): Same.
9696         * match.pd: Same.
9697         * poly-int.h (add): Same.
9698         (sub): Same.
9699         (neg): Same.
9700         (mul): Same.
9701         * predict.c (predict_iv_comparison): Same.
9702         * profile-count.c (slow_safe_scale_64bit): Same.
9703         * simplify-rtx.c (simplify_const_binary_operation): Same.
9704         * tree-chrec.c (tree_fold_binomial): Same.
9705         * tree-data-ref.c (split_constant_offset_1): Same.
9706         * tree-if-conv.c (idx_within_array_bound): Same.
9707         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
9708         * tree-ssa-phiopt.c (minmax_replacement): Same.
9709         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
9710         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
9711         * vr-values.c (vr_values::adjust_range_with_scev): Same.
9712         * wide-int.cc (wi::add_large): Same.
9713         (wi::mul_internal): Same.
9714         (wi::sub_large): Same.
9715         (wi::divmod_internal): Same.
9716         * wide-int.h: Change overflow type to overflow_type for neg, add,
9717         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
9718         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
9719         mul_internal, divmod_internal.
9720         (overflow_type): New enum.
9721         (accumulate_overflow): New.
9723 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
9725         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
9726         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
9728 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
9730         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
9731         argument is checked for zero before entering loop, avoid checking again.
9733 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
9735         * gimplify.h (generic_expr_could_trap_p): Set as global function.
9736         * gimplify.h (generic_expr_could_trap_p): Likwise.
9737         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
9739 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
9741         PR tree-optimization/86401
9742         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
9743         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
9744         (fold_bit_and_mask): ... here.  New helper function for match.pd.
9745         * fold-const.h (fold_bit_and_mask): Declare.
9746         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
9748 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
9750         PR target/86324
9751         * target.def (translate_mode_attribute): New hook.
9752         * targhooks.h (default_translate_mode_attribute): Declare.
9753         * targhooks.c (default_translate_mode_attribute): New function.
9754         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
9755         * doc/tm.texi: Regenerate.
9756         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
9757         (rs6000_translate_mode_attribute): New function.
9759 2018-07-06  Paul Koning  <ni1d@arrl.net>
9761         * doc/md.texi (define_split): Document DONE and FAIL.
9762         (define_peephole2): Ditto.
9764 2018-07-05  Jeff Law  <law@redhat.com>
9766         PR tree-optimization/86010
9767         * tree-ssa-dse.c (compute_trims): More aggressively trim at
9768         both the head and tail of mem* and str* calls.
9770 2018-07-05  Jim Wilson  <jimw@sifive.com>
9772         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
9774 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
9776         * config/aarch64/aarch64-simd.md: correct flags text for
9777         MIN_EXPR replacement.
9779 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
9781         * configure: Regenerated.
9783 2018-07-05  Carl Love  <cel@us.ibm.com>
9785         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
9786         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9787         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
9788         VSX_BUILTIN_DOUBLEL_V4SF.
9790 2018-07-05  Martin Sebor  <msebor@redhat.com>
9792         PR c++/86400
9793         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
9794         than its domain to compute its the upper bound of a char array.
9796 2018-07-05  Nathan Sidwell  <nathan@acm.org>
9798         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
9799         * doc/cpp.texi: Update comment.
9800         * doc/tm.texi: Rebuilt.
9801         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
9802         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
9803         * doc/extend.texi (Backwards Compatibility): Clarify it is system
9804         headers affected by extern "C".
9805         * system.h: Poison NO_IMPLICIT_EXTERN_C.
9806         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9807         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9808         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9809         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9810         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9811         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9812         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9813         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9814         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9815         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9816         config/sparc/sp64-elf.h, config/spu/spu.h,
9817         config/stormy16/stormy16.h, config/v850/v850.h,
9818         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9819         define NO_IMPLICIT_EXTERN_C.
9820         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9822 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9824         PR target/84711
9825         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9826         instead of GET_MODE_SIZE when comparing Units.
9828 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9830         PR target/84711
9831         * rtlanal.c (set_noop_p): Constrain on mode change,
9832         include hard-reg-set.h
9834 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
9836         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9838 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
9840         Revert
9841         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9843         PR sanitizer/84250
9844         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9845         libasan.
9846         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9848 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
9850         PR sanitizer/84250
9851         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9852         libasan.
9853         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9855 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
9857         PR middle-end/86380
9858         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9860 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
9862         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9863         neg_*_op* variables.
9865 2018-07-04  Martin Liska  <mliska@suse.cz>
9867         * tree-switch-conversion.c: Define
9868         max_ratio_for_speed and max_ratio_for_size constants.
9870 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9871             Martin Liska  <mliska@suse.cz>
9873         PR middle-end/66240
9874         PR target/45996
9875         PR c/84100
9876         * common.opt: Rename align options with 'str_' prefix.
9877         * common/config/i386/i386-common.c (set_malign_value): New
9878         function.
9879         (ix86_handle_option): Use it to set -falign-* options/
9880         * config/aarch64/aarch64-protos.h (struct tune_params): Change
9881         type from int to string.
9882         * config/aarch64/aarch64.c: Update default values from int
9883         to string.
9884         * config/alpha/alpha.c (alpha_override_options_after_change):
9885         Likewise.
9886         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9887         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9888         max skip conditionally.
9889         * config/i386/freebsd.h (SUBALIGN_LOG): New.
9890         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9891         max skip conditionally.
9892         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9893         max skip conditionally.
9894         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
9895         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9896         max skip conditionally.
9897         * config/i386/i386.c (struct ptt): Change type from int to
9898         string.
9899         (ix86_default_align): Set default values.
9900         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
9901         max skip conditionally.
9902         * config/i386/iamcu.h (SUBALIGN_LOG): New.
9903         (ASM_OUTPUT_MAX_SKIP_ALIGN):
9904         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
9905         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9906         max skip conditionally.
9907         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
9908         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
9909         * config/i386/x86-64.h (SUBALIGN_LOG): New.
9910         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9911         max skip conditionally.
9912         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9913         * config/ia64/ia64.c (ia64_option_override): Set default values
9914         for alignment options.
9915         * config/m68k/m68k.c: Handle new str_align_* options.
9916         * config/mips/mips.c (mips_set_compression_mode): Change
9917         type of constants.
9918         (mips_option_override): Set default values for options.
9919         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9920         Likewise.
9921         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9922         Likewise.
9923         * config/rx/rx.c (rx_option_override): Likewise.
9924         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
9925         (LABEL_ALIGN): Use align_labels_log.
9926         (LOOP_ALIGN): Use align_loops_align.
9927         * config/s390/s390.c (s390_asm_output_function_label): Use new
9928         macros.
9929         * config/sh/sh.c (sh_override_options_after_change):
9930         Change type of constants.
9931         * config/spu/spu.c (spu_sched_init): Likewise.
9932         * config/sparc/sparc.c (sparc_option_override): Set default
9933         values for options.
9934         * config/visium/visium.c (visium_option_override): Likewise.
9935         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9936         emit p2align format with last argument if it's not needed.
9937         * doc/invoke.texi: Document extended format of -falign-*.
9938         * final.c: Use align_labels alignment.
9939         * flags.h (struct target_flag_state): Change type to use
9940         align_flags.
9941         (struct align_flags_tuple): New.
9942         (struct align_flags): Likewise.
9943         (align_loops_log): Redefine macro to use new types.
9944         (align_loops_max_skip): Redefine macro to use new types.
9945         (align_jumps_log): Redefine macro to use new types.
9946         (align_jumps_max_skip): Redefine macro to use new types.
9947         (align_labels_log): Redefine macro to use new types.
9948         (align_labels_max_skip): Redefine macro to use new types.
9949         (align_functions_log): Redefine macro to use new types.
9950         (align_loops): Redefine macro to use new types.
9951         (align_jumps): Redefine macro to use new types.
9952         (align_labels): Redefine macro to use new types.
9953         (align_functions): Redefine macro to use new types.
9954         (align_functions_max_skip): Redefine macro to use new types.
9955         (align_loops_value): New macro.
9956         (align_jumps_value): New macro.
9957         (align_labels_value): New macro.
9958         (align_functions_value): New macro.
9959         * function.c (invoke_set_current_function_hook): Propagate
9960         alignment values from flags to global variables default in
9961         topleev.h.
9962         * ipa-icf.c (sem_function::equals_wpa): Use
9963         cl_optimization_option_eq instead of memcmp.
9964         * lto-streamer.h (cl_optimization_stream_out): Support streaming
9965         of string types.
9966         (cl_optimization_stream_in): Likewise.
9967         * optc-save-gen.awk: Support strings in cl_optimization.
9968         * opth-gen.awk: Likewise.
9969         * opts.c (finish_options): Remove error checking of invalid
9970         value ranges.
9971         (MAX_CODE_ALIGN): Remove.
9972         (MAX_CODE_ALIGN_VALUE): Likewise.
9973         (parse_and_check_align_values): New function.
9974         (check_alignment_argument): Likewise.
9975         (common_handle_option): Use check_alignment_argument.
9976         * opts.h (parse_and_check_align_values): Declare.
9977         * toplev.c (init_alignments): Remove.
9978         (read_log_maxskip): New.
9979         (parse_N_M): Likewise.
9980         (parse_alignment_opts): Likewise.
9981         (backend_init_target): Remove usage of init_alignments.
9982         * toplev.h (parse_alignment_opts): Declare.
9983         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9984         argument.
9985         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9986         * tree.c (cl_option_hasher::equal): New.
9987         * varasm.c: Use new global macros.
9989 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
9991         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9992         Use a simpler align directive also if MAXSKIP = ALIGN-1.
9993         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9994         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9995         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9996         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9997         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9998         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9999         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10000         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10001         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10002         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10003         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10005 2018-07-04  Martin Liska  <mliska@suse.cz>
10006             Jonathan Wakely  <jwakely@redhat.com>
10008         * coverage.c: Use correct type.
10009         * doc/invoke.texi: Language correction.
10011 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
10013         PR target/85620
10014         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10015         ENDBRANCH for non-tail call which may return via indirect branch.
10016         * doc/extend.texi: Document indirect_return attribute.
10018 2018-07-03  Martin Sebor  <msebor@redhat.com>
10020         PR tree-optimization/86274
10021         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10022         precondition.
10023         (format_floating): Correct handling of infinities and NaNs.
10025 2018-07-03  Martin Sebor  <msebor@redhat.com>
10027         * print-tree.c (print_real_cst): New function.
10028         (print_node_brief): Call it.
10029         (print_node): Ditto.
10031 2018-07-03  Jeff Law  <law@redhat.com>
10033         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10034         into a single pattern.
10036         * config/h8300/h8300.md (ors code_iterator): New.
10037         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10038         a single pattern and single splitter.
10039         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10040         (iorqi3_1, xorqi3_1): Likewise.
10041         (iorqi3, xorqi3 expanders): Similarly.
10043         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10044         (movmd_internal) into a single pattern using the P mode iterator.
10045         (movmd splitters): Similarly.
10046         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
10047         (movsd splitters): Similarly.
10049         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
10050         ADDB, ADDW and ADDL into a single ADD attribute which selects the
10051         right table based on the size of the operand.
10052         * config/h8300/h8300.md (length_table): Corresponding changes. All
10053         references to "addb", "addw" and "addl" changed to "add".
10054         (btst patterns): Merge two variants into a single pattern.
10055         (tstqi, tsthi): Likewise.
10056         (addhi3_incdec, addsi3_incdec): Likewise.
10057         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
10058         (mulhi3, mulsi3): Likewise.
10059         (udivhi3, udivsi3): Likewise.
10060         (divhi3, divsi3): Likewise.
10061         (andorqi3, andorhi3, andorsi3): Likewise.
10063 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
10065         PR target/85694
10066         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10067         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10069 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10071         PR tree-optimization/85694
10072         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10073         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10074         UNSPEC_URHADD.
10075         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10076         (<u>avg<mode>3_ceil): New patterns.
10078 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
10080         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10081         scan-tree-dump directive.
10082         * gcc.dg/vect/slp-perm-2.c: Likewise.
10083         * gcc.dg/vect/slp-perm-3.c: Likewise.
10084         * gcc.dg/vect/slp-perm-5.c: Likewise.
10085         * gcc.dg/vect/slp-perm-6.c: Likewise.
10086         * gcc.dg/vect/slp-perm-7.c: Likewise.
10087         * gcc.dg/vect/slp-perm-8.c: Likewise.
10089 2018-07-03  Marek Polacek  <polacek@redhat.com>
10091         PR middle-end/86202
10092         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10094 2018-07-03  Richard Biener  <rguenther@suse.de>
10096         PR ipa/86389
10097         * tree-ssa-structalias.c (find_func_clobbers): Properly
10098         handle indirect calls.
10100 2018-07-03  Jeff Law  <law@redhat.com>
10102         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10103         (shifts): New code iterator.
10104         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
10105         expander.  Fix HImode handling on H8/SX.
10106         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
10107         (subqi3, subhi3, subsi3 expanders): Likewise.
10108         (andqi3, andhi3, andsi3 expanders): Likewise.
10109         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
10110         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
10111         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
10112         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
10113         (zero_extendqihi2, zero_extendqisi2): Likewise.
10114         (extendqihi2, extendqisi2): Likewise.
10115         (rotlqi3, rotlhi3, rotlsi3): Likewise.
10116         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
10117         (rotlqi3_1, rotlhi3_1): Likewise.
10118         (logicalhi3_sn, logicalsi3_sn): Likewise.
10119         (logicalhi3, logicalsi3): Likewise.
10121 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10123         * tree-vect-patterns.c (vect_recog_rotate_pattern)
10124         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10125         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
10126         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
10127         type to append_pattern_def_seq instead of creating a stmt_vec_info
10128         directly.
10129         (build_mask_conversion): Likewise.  Remove vinfo argument.
10130         (vect_add_conversion_to_patterm): Likewise, renaming to...
10131         (vect_add_conversion_to_pattern): ...this.
10132         (vect_recog_mask_conversion_pattern): Update call to
10133         build_mask_conversion.  Pass the vector type to
10134         append_pattern_def_seq here too.
10135         (vect_recog_gather_scatter_pattern): Update call to
10136         vect_add_conversion_to_pattern.
10138 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10140         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
10141         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10142         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
10143         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
10144         STMT_VINFO_PATTERN_DEF_SEQ to null here.
10145         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
10146         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10147         append_pattern_def_seq instead of new_pattern_def_seq.
10148         (vect_recog_divmod_pattern): Do both of the above.
10149         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
10150         is null.
10152 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10154         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10155         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
10156         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
10157         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
10158         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
10159         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10160         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
10161         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10162         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10163         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
10164         parameter with a single stmt_vec_info.
10165         (vect_recog_func_ptr): Likewise.
10166         (vect_recog_gather_scatter_pattern): Likewise, folding in...
10167         (vect_try_gather_scatter_pattern): ...this.
10168         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
10169         the stmt_vec_info of the statement to be matched.  Don't clear
10170         STMT_VINFO_RELATED_STMT.
10171         (vect_pattern_recog): Update call accordingly.
10173 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10175         PR tree-optimization/85694
10176         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
10177         (uavgM3_ceil): Document new optabs.
10178         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
10179         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
10180         functions.
10181         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
10182         (savg_ceil_optab): New optabs.
10183         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
10184         (vect_vect_recog_func_ptrs): Add it.
10185         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
10186         constant directly from the associated lhs.
10188 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10190         * tree-vect-patterns.c (vect_split_statement): New function.
10191         (vect_convert_input): Use it to try to split an existing cast.
10193 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10195         * poly-int.h (print_hex): New function.
10196         * dumpfile.h (dump_dec, dump_hex): Declare.
10197         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
10198         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
10199         min_input_precision, operation_precision and operation_sign.
10200         * tree-vect-patterns.c (vect_get_range_info): New function.
10201         (vect_same_loop_or_bb_p, vect_single_imm_use)
10202         (vect_operation_fits_smaller_type): Delete.
10203         (vect_look_through_possible_promotion): Add an optional
10204         single_use_p parameter.
10205         (vect_recog_over_widening_pattern): Rewrite to use new
10206         stmt_vec_info infomration.  Handle one operation at a time.
10207         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
10208         (vect_truncatable_operation_p, vect_set_operation_type)
10209         (vect_set_min_input_precision): New functions.
10210         (vect_determine_min_output_precision_1): Likewise.
10211         (vect_determine_min_output_precision): Likewise.
10212         (vect_determine_precisions_from_range): Likewise.
10213         (vect_determine_precisions_from_users): Likewise.
10214         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
10215         (vect_vect_recog_func_ptrs): Put over_widening first.
10216         Add cast_forwprop.
10217         (vect_pattern_recog): Call vect_determine_precisions.
10219 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10221         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
10222         statements that have been replaced by further pattern statements.
10223         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
10225 2018-07-03  Richard Biener  <rguenther@suse.de>
10227         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10228         always set *dt.  Dump vectype in vectype overload.
10229         * dumpfile.h (dump_gimple_expr): New function.
10230         (dump_gimple_expr_loc): Likewise.
10231         * dumpfile.c (dump_gimple_expr): New function.
10232         (dump_gimple_expr_loc): Likewise.
10234 2018-07-02  Jeff Law  <law@redhat.com>
10236         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
10237         the H8/300, H8/300H and H8/S variants into a single pattern.
10238         (movhi_h8300, movqi_h8300hs): Similarly.
10239         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
10240         (QHI mode iterator): New.
10242         * config/h8300/h8300.md: Remove trailing whitespace.
10244 2018-07-02  Jim Wilson  <jimw@sifive.com>
10246         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
10247         instead of emit_insn for interrupt returns.
10248         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
10249         (riscv_sret, riscv_uret): Likewise.
10251 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10253         * pretty-print.c (selftest::test_pp_format): Move save and restore
10254         of quotes to class auto_fix_quotes, and add an instance.
10255         * selftest.c: Include "intl.h".
10256         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
10257         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
10258         * selftest.h (selftest::auto_fix_quotes): New class.
10260 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
10262         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
10263         (aarch64_sve_prepare_conditional_op): Remove.
10264         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
10265         Allow aarch64_simd_reg_or_zero as select operand; remove
10266         the aarch64_sve_prepare_conditional_op call.
10267         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
10268         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10269         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
10270         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
10271         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
10272         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
10273         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
10274         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
10275         and a splitters to match all of the *_any patterns.
10276         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
10278         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
10279         (SVE_COND_FP_BINARY_REV): Remove.
10280         (sve_int_op_rev, sve_fp_op_rev): New.
10281         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
10282         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
10283         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
10284         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
10285         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
10286         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
10287         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
10288         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
10289         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
10291         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
10292         Remove match_dup 1 from the inner unspec.
10293         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10295         * config/aarch64/aarch64.md (movprfx): New attr.
10296         (length): Default movprfx to 8.
10297         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
10298         (*madd<SVE_I>, *msub<SVE_I): Likewise.
10299         (*<su>mul<SVE_I>3_highpart): Likewise.
10300         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
10301         (*v<ASHIFT><SVE_I>3): Likewise.
10302         (*<su><MAXMIN><SVE_I>3): Likewise.
10303         (*<su><MAXMIN><SVE_F>3): Likewise.
10304         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
10305         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
10306         (*div<SVE_F>4): Likewise.
10308 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
10310         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
10311         in dump string.
10313 2018-07-02  Richard Biener  <rguenther@suse.de>
10315         PR tree-optimization/86363
10316         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
10317         memset argument refers to a non-variable address.
10319 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
10321         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
10322         {PLUS,MINUS}_EXPR code to...
10323         (adjust_symbolic_bound): ...here,
10324         (combine_bound): ...here,
10325         (set_value_range_with_overflow): ...and here.
10327 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
10329         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
10330         code...
10331         (extract_range_from_abs_expr): ...here.
10333 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
10335         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
10336         -fno-omit-frame-pointer when not optimizing.
10338 2018-07-02  Martin Liska  <mliska@suse.cz>
10340         PR ipa/86279
10341         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
10342         (propagate_nothrow): Likewise.
10344 2018-07-02  Martin Liska  <mliska@suse.cz>
10346         PR ipa/86323
10347         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
10349 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10351         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
10352         function in r262149, changing "loc" param from source_location to
10353         const dump_location_t &.
10354         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
10355         declaration, as above.
10357 2018-07-01  Paul Koning  <ni1d@arrl.net>
10359         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10360         -munit-asm, -mgnu-asm, -mdec-asm.
10361         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
10362         (pdp11_output_labelref): New.
10363         (pdp11_output_def): New.
10364         (pdp11_output_addr_vec_elt): New.
10365         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
10366         %# and %@ format codes.
10367         (pdp11_option_override): New.
10368         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
10369         (pdp11_output_ident): New.
10370         (pdp11_asm_named_section): New.
10371         (pdp11_asm_init_sections): New.
10372         (pdp11_file_start): New.
10373         (pdp11_file_end): New.
10374         (output_ascii): Use .ascii/.asciz for -mdec-asm.
10375         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
10376         %o, like %c but octal.
10377         (pdp11_option_override): New.
10378         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
10379         -mdec-asm.
10380         (DATA_SECTION_ASM_OP): Ditto.
10381         (READONLY_DATA_SECTION_ASM_OP): New.
10382         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
10383         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
10384         (ASM_OUTPUT_LABELREF): Ditto.
10385         (ASM_OUTPUT_DEF): Ditto.
10386         (ASM_OUTPUT_EXTERNAL): New.
10387         (ASM_OUTPUT_SOURCE_FILENAME): New.
10388         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
10389         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
10390         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
10391         %# and %@ format codes.
10392         * config/pdp11/pdp11.opt (mgnu-asm): New.
10393         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
10394         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
10395         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
10397 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
10399         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
10400         dereferencing path[] beyond its length.
10401         (debug_path): New.
10402         (debug_all_paths): New.
10403         (rewire_first_differing_edge): New.
10404         (adjust_paths_after_duplication): New.
10405         (duplicate_thread_path): Call adjust_paths_after_duplication.
10406         Add new argument.
10407         (thread_through_all_blocks): Add new argument to
10408         duplicate_thread_path.
10410 2018-06-30  Jim Wilson  <jimw@sifive.com>
10412         * config/riscv/predicates.md (p2m1_shift_operand): New.
10413         (high_mask_shift_operand): New.
10414         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
10415         pattern using p2m1_shift_operand.
10416         (lshsi3_zero_extend_3+2): New combiner pattern using
10417         high_mask_shift_operand.
10419 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10421         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
10422         split out from...
10423         (vect_recog_rotate_pattern): ...here.
10424         (vect_convert_input): Try to insert casts of invariants in the
10425         preheader.
10426         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
10427         preheader to be empty.
10429 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10431         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
10432         vector type.  If given, install it in the new statement's
10433         STMT_VINFO_VECTYPE.
10434         (vect_element_precision): New function.
10435         (vect_unpromoted_value): New struct.
10436         (vect_unpromoted_value::vect_unpromoted_value): New function.
10437         (vect_unpromoted_value::set_op): Likewise.
10438         (vect_look_through_possible_promotion): Likewise.
10439         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
10440         (vect_widened_op_tree, vect_convert_input): Likewise.
10441         (vect_convert_inputs, vect_convert_output): Likewise.
10442         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
10443         to handle the optional cast of the multiplication result and
10444         vect_widened_op_tree to detect the widened multiplication itself.
10445         Do not require the input and output of promotion casts to have
10446         the same sign, but base the signedness of the operation on the
10447         input rather than the result.  If the pattern includes two
10448         promotions, check that those promotions have the same sign.
10449         Do not restrict the MULT_EXPR handling to a double-width result;
10450         handle quadruple-width results and wider.  Use vect_convert_inputs
10451         to convert the inputs to the common type.
10452         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
10453         to handle the optional cast of the ABS result.  Also allow a sign
10454         change or a sign extension between the ABS and MINUS.
10455         Use vect_widened_op_tree to detect the widened subtraction and use
10456         vect_convert_inputs to convert the inputs to the common type.
10457         (vect_handle_widen_op_by_const): Delete.
10458         (vect_recog_widen_op_pattern): New function.
10459         (vect_recog_widen_mult_pattern): Use it.
10460         (vect_recog_widen_shift_pattern): Likewise.
10461         (vect_recog_widen_sum_pattern): Use
10462         vect_look_through_possible_promotion to handle the promoted
10463         PLUS_EXPR operand.
10465 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10467         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
10468         the containing gimple_seq *.
10469         * gimple-iterator.h (gsi_for_stmt): Declare it.
10470         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
10471         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
10472         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10473         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10474         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
10475         checks.
10476         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
10477         split out from...
10478         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
10479         statement being replaced is part of an existing pattern
10480         definition sequence, inserting the new pattern statements before
10481         the original one.
10482         (vect_pattern_recog_1): Don't return a bool.  If the statement
10483         is already part of a pattern, instead apply pattern matching
10484         to the pattern definition statements.  Don't clear the
10485         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
10486         (vect_pattern_recog): Don't break after the first match;
10487         continue processing the pattern definition statements instead.
10488         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
10490 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10492         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10493         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10494         (vect_recog_widen_sum_pattern): Use it.
10496 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10498         * tree-vect-loop.c (vectorizable_reduction): Assert that the
10499         phi is not a pattern statement and has not been replaced by
10500         a pattern statement.
10501         * tree-vect-patterns.c (type_conversion_p): Don't check
10502         STMT_VINFO_IN_PATTERN_P.
10503         (vect_recog_vector_vector_shift_pattern): Likewise.
10504         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
10505         the pattern statement rather than the original statement; check
10506         directly for a WIDEN_MULT_EXPR here.
10507         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
10508         vect_is_simple_use to return the pattern statement rather
10509         than the original statement; use is_pattern_stmt_p to check
10510         for such a pattern statement.
10511         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
10512         to return the pattern statement rather than the original statement;
10513         don't do the same transformation here.
10514         (vect_is_simple_use): If the defining statement has been replaced
10515         by a pattern statement, return the pattern statement instead.
10516         Remove the corresponding (local) transformation from the vectype
10517         overload.
10519 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
10521         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
10522         end and default to null.
10523         * tree-vect-loop.c (vect_create_epilog_for_reduction)
10524         (vectorizable_reduction): Update calls accordingly, dropping the
10525         gimple ** argument if the passed-back statement isn't needed.
10526         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
10527         (vect_recog_rotate_pattern): Likewise.
10528         (vect_recog_mask_conversion_pattern): Likewise.
10529         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10530         (vect_mask_constant_operand_p): Likewise.
10531         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
10532         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
10533         (get_group_load_store_type, get_load_store_type): Likewise.
10534         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
10535         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
10536         (vectorizable_conversion, vectorizable_assignment): Likewise.
10537         (vectorizable_shift, vectorizable_operation): Likewise.
10538         (vectorizable_store, vect_is_simple_cond): Likewise.
10539         (vectorizable_condition, vectorizable_comparison): Likewise.
10540         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
10541         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
10542         and move it to the end.  Cope with null def_stmt_outs.
10544 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10546         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
10548 2018-06-29  Jeff Law  <law@redhat.com>
10550         * config/v850/v850.c (v850_legitimate_address_p): Handle large
10551         displacements for TARGET_V850E2V3 and newer.
10552         (TARGET_LRA_P): Remove.  Defaults to LRA now.
10553         * config/v850/v850.md (sign23byte_load): Remove.
10554         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
10555         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
10557 2018-06-29  Martin Liska  <mliska@suse.cz>
10559         PR lto/85759
10560         * coverage.c (coverage_init): Mangle full path name.
10561         * doc/invoke.texi: Document the change.
10562         * gcov-io.c (mangle_path): New.
10563         * gcov-io.h (mangle_path): Likewise.
10564         * gcov.c (mangle_name): Use mangle_path for path mangling.
10566 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10568         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
10569         if starting source register is not even.
10571 2018-06-29  Martin Liska  <mliska@suse.cz>
10573         PR tree-optimization/86263
10574         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
10575         Make edge redirection.
10577 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
10579         * dumpfile.c (dump_loc): Add indentation based on scope depth.
10580         (dump_scope_depth): New variable.
10581         (get_dump_scope_depth): New function.
10582         (dump_begin_scope): New function.
10583         (dump_end_scope): New function.
10584         * dumpfile.h (get_dump_scope_depth): New declaration.
10585         (dump_begin_scope): New declaration.
10586         (dump_end_scope): New declaration.
10587         (class auto_dump_scope): New class.
10588         (AUTO_DUMP_SCOPE): New macro.
10589         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
10590         AUTO_DUMP_SCOPE.
10592 2018-06-29  Richard Biener  <rguenther@suse.de>
10594         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
10595         compute_all_dependences succeeds.
10596         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
10597         exceed --param loop-max-datarefs-for-datadeps.
10599 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
10601         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
10603 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
10605         PR target/86348
10606         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
10607         alternative 0 in preferred_for_speed attribute.
10609 2018-06-28  Paul Koning  <ni1d@arrl.net>
10611         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
10612         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
10613         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
10614         * config/pdp11/pdp11.md: Correct "length" attribute calculation
10615         for shift insn patterns.
10617 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
10619         * cgraph.c (cgraph_node::get_body): Replace assignments to
10620         "dump_file" with calls to set_dump_file.
10621         * dumpfile.c (alt_dump_file): Make static, and group with...
10622         (alt_flags): ...this definition.
10623         (dumps_are_enabled): New variable.
10624         (refresh_dumps_are_enabled): New function.
10625         (set_dump_file): New function.
10626         (set_alt_dump_file): New function.
10627         (gcc::dump_manager::dump_start): Replace assignments to
10628         "dump_file" and "alt_dump_file" with calls to set_dump_file and
10629         set_alt_dump_file.
10630         (gcc::dump_manager::dump_finish): Likewise.
10631         * dumpfile.h (alt_dump_file): Delete decl.
10632         (dumps_are_enabled): New variable decl.
10633         (set_dump_file): New function decl.
10634         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
10635         global.
10636         * tree-nested.c (lower_nested_functions): Replace assignments to
10637         "dump_file" with calls to set_dump_file.
10639 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
10641         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
10642         goto_locus of each outgoing edge of each basic block.
10644 2018-06-28  Richard Biener  <rguenther@suse.de>
10646         * dwarf2out.c (decl_scope_table): Remove.
10647         (push_decl_scope): Likewise.
10648         (pop_decl_scope): Likewise.
10649         (gen_type_die_for_member): Do not call push/pop_decl_scope.
10650         (gen_struct_or_union_type_die): Likewise.
10651         (gen_tagged_type_die): Likewise.
10652         (dwarf2out_init): Do not initialize decl_scope_table.
10653         (dwarf2out_c_finalize): Do not free it.
10655 2018-06-28  Richard Biener  <rguenther@suse.de>
10657         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
10658         deciding whether to not re-use a DIE.
10660 2018-06-28  Richard Biener  <rguenther@suse.de>
10662         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
10663         DW_AT_abstract_origin attribute.
10665 2018-06-28  Martin Liska  <mliska@suse.cz>
10667         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10668         Use newly introduced constants.
10669         * tree-switch-conversion.h (struct jump_table_cluster):
10670         Define max_ratio_for_size and max_ratio_for_speed.
10672 2018-06-28  Martin Liska  <mliska@suse.cz>
10674         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10675         Add new checking assert to catch invalid state.
10676         (jump_table_cluster::can_be_handled): Handle single case
10677         clusters.
10678         (jump_table_cluster::is_beneficial): Bail out for such case.
10679         (bit_test_cluster::find_bit_tests):
10680         Add new checking assert to catch invalid state.
10681         (bit_test_cluster::can_be_handled): Handle single case
10682         clusters.
10683         (bit_test_cluster::is_beneficial): Bail out for such case.
10684         (switch_decision_tree::analyze_switch_statement):
10685         Fix comment.
10687 2018-06-28  Martin Liska  <mliska@suse.cz>
10689         * common.opt: Introduce -completion option.
10690         * gcc.c (driver_handle_option): Handle it.
10691         (driver::main): Print completions if completion
10692         is set.
10693         * opt-suggestions.c (option_proposer::get_completions):
10694         New function.
10695         (option_proposer::suggest_completion): Likewise.
10696         (option_proposer::find_param_completions): Likewise.
10697         (verify_autocompletions): Likewise.
10698         (test_completion_valid_options): Likewise.
10699         (test_completion_valid_params): Likewise.
10700         (in_completion_p): Likewise.
10701         (empty_completion_p): Likewise.
10702         (test_completion_partial_match): Likewise.
10703         (test_completion_garbage): Likewise.
10704         (opt_proposer_c_tests): Likewise.
10705         * opt-suggestions.h: Declare new functions.
10706         * opts.c (common_handle_option): Handle OPT__completion_.
10707         * selftest-run-tests.c (selftest::run_tests): Add
10708         opt_proposer_c_tests.
10709         * selftest.c (assert_str_startswith): New.
10710         * selftest.h (assert_str_startswith): Likewise.
10711         (opt_proposer_c_tests): New.
10712         (ASSERT_STR_STARTSWITH): Likewise.
10714 2018-06-28  Martin Liska  <mliska@suse.cz>
10716         * Makefile.in: Add opt-suggestions.o.
10717         * gcc-main.c: Include opt-suggestions.h.
10718         * gcc.c (driver::driver): Likewise.
10719         (driver::~driver): Remove m_option_suggestions.
10720         (driver::build_option_suggestions): Moved to option_proposer.
10721         (driver::suggest_option): Likewise.
10722         (driver::handle_unrecognized_options): Use option_proposer.
10723         * gcc.h (class driver): Add new memver m_option_proposer.
10724         * opt-suggestions.c: New file.
10725         * opt-suggestions.h: New file.
10727 2018-06-28  Martin Liska  <mliska@suse.cz>
10729         * vec.h (class auto_string_vec): New (moved from auto_argvec).
10730         (auto_string_vec::~auto_string_vec): Likewise.
10732 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
10734         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
10735         prevent_decl_creation_for_types fields up and add reset_location field.
10736         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
10737         statement if id->reset_location is true.
10738         (copy_edges_for_bb): Do not set goto_locus on the new edges if
10739         id->reset_location is true.
10740         (copy_phis_for_bb): Force input_location on the arguments if
10741         id->reset_location is true.
10742         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10743         is set on the function to be inlined.
10745 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
10747         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
10749 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
10751         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
10752         registers for Pmode.
10753         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
10754         hard registers for the clobbered pseudo.
10756 2018-06-27  Paul Koning  <ni1d@arrl.net>
10758         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10759         mutually exclusive options.
10760         * config/pdp11/constraints.md (h): New constraint.
10761         (O): Update definition to match shift code generation.
10762         (D): New constraint.
10763         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
10764         (CCFP): Remove.
10765         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
10766         function.
10767         (output_jump): Change arguments.
10768         (pdp11_fixed_cc_regs): New function.
10769         (pdp11_cc_mode): Ditto.
10770         (pdp11_expand_shift): Ditto.
10771         (pdp11_assemble_shift): Ditto.
10772         (pdp11_small_shift): Ditto.
10773         (pdp11_branch_cost): Remove.
10774         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
10775         from output.
10776         (pdp11_register_move_cost): Update for CC registers.
10777         (pdp11_rtx_costs): Add case for LSHIFTRT.
10778         (pdp11_output_jump): Add CCNZ mode conditional branches.
10779         (notice_update_cc_on_set): Remove.
10780         (pdp11_cc_mode): New function.
10781         (simple_memory_operand): Correct pre/post decrement case.
10782         (no_side_effect_operand): New function.
10783         (pdp11_regno_reg_class): Add CC_REGS class.
10784         (pdp11_fixed_cc_regs): New function.
10785         (pdp11_small_shift): New function.
10786         (pdp11_expand_shift): New function to expand shift insns.
10787         (pdp11_assemble_shift): New function to output shifts.
10788         (pdp11_branch_cost): Remove.
10789         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
10790         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
10791         (WCHAR_TYPE): Ditto.
10792         (PTRDIFF_TYPE): Ditto.
10793         (ADJUST_INSN_LENGTH): New macro.
10794         (FIXED_REGISTERS): Add CC registers.
10795         (CALL_USED_REGISTERS): Ditto.
10796         (reg_class): Ditto.
10797         (REG_CLASS_NAMES): Ditto.
10798         (REG_CLASS_CONTENTS): Ditto.
10799         (SELECT_CC_MODE): Use new function.
10800         (TARGET_FLAGS_REGNUM): New macro.
10801         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
10802         (cc0_reg_rtx): Remove.
10803         (CC_STATUS_MDEP): Remove.
10804         (CC_STATUS_MDEFP_INIT): Remove.
10805         (CC_IN_FPU): Remove.
10806         (NOTICE_UPDATE_CC): Remove.
10807         (REGISTER_NAMES): Add CC registers.
10808         (BRANCH_COST): Change to constant 1.
10809         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10810         handling.
10811         * config/pdp11/pdp11.opt (mbcopy): Remove.
10812         (mbcopy-builtin): Remove.
10813         (mbranch-cheap): Remove.
10814         (mbranch-expensive): Remove.
10815         * config/pdp11/predicates.md (expand_shift_operand): Update to
10816         match shift code generation.
10817         (ccnz_operator): New predicate.
10818         * doc/invoke.texi (PDP-11 Options): Remove deleted options
10819         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10820         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
10821         exclusive options.
10822         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
10823         description of O constraint.
10825 2018-06-27  Jeff Law  <law@redhat.com>
10826             Austin Law  <austinklaw@gmail.com>
10828         * config/v850/v850.md (addsi3_set_flags): New pattern.
10829         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10830         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10831         (zero_extendhisi2_v850_set_flags): Likewise.
10832         (zero_extendqisi2_v850_set_flags): Likewise.
10833         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10834         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10835         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10837         * config/v850/v850-protos.h (notice_update_cc): Remove.
10838         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10839         (v850_print_operand): Handle 'D' and "d".
10840         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10841         Add handling of arithmetic/logical operations compared against zero.
10842         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10843         Do not look at v850_compare_op, instead get mode from last argument.
10844         (v850_gen_compare): Remove
10845         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10846         after reload for prologue insns.
10847         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10848         patterns.
10849         (construct_save_jarl): Likewise.
10850         (TARGET_FLAGS_REGNUM): Define.
10851         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10852         (NOTICE_UPDATE_CC): Remove.
10853         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10854         than cc0.  Conditionalize on reload_completed.
10855         (cmpsi_insn, setfcc_insn): Likewise.
10856         (tst1 splitter): Turn into define_and_split which sets the flags
10857         after reload.
10858         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10859         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10860         (cstoresf4, cstoredf4): Clobber the flags.
10861         (cmpsi, cmpsf, cmpdf): Remove expanders.
10862         (setf_insn): Remove pattern.
10863         (addsi3): Turn into define_and_split which clobbers the flags after
10864         reload and a suitable pattern (addsi3_clobber_flags) for use after
10865         reload.
10866         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10867         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10868         (ashrsi3, ashrsi3_v850e2): Likewise.
10869         (bins): Clobber the flags.
10870         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10871         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10872         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10873         (call_value_internal_short, call_value_internal_long): Likewise.
10874         (callt_save_interrupt, callt_return_interrupt): Likewise.
10875         (save_interrupt, return_interrupt): Likewise.
10876         (callt_save_all_interrupt, save_all_interrupt): Likewise.
10877         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10878         (restore_all_interrupt, _restore_all_interrupt): Likewise.
10879         (All FP comparisons): Only allow after reload has completed.
10880         (trfsr): Likewise.
10881         (divh, divhu): Tweak output template.
10882         (branch_z_normal, branch_z_invert): Remove
10883         (branch_nz_normal, branch_nz_invert): Likewise.
10884         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10886         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10887         * config/v850/v850.c (notice_update_cc): Remove.
10888         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
10889         (CC_NO_CARRY): Likewise.
10890         (NOTICE_UPDATE_CC): Define to nothing.
10891         * config/v850/v850.md: Remove block comment on cc0 handling
10892         Remove "cc" attribute from all patterns.  Remove cc_status handling
10893         from all patterns.  Minor formatting fixes.
10895 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10897         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
10898         (cortex-a76.cortex-a55): Likewise.
10899         * config/aarch64/aarch64-tune.md: Regenerate.
10900         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
10901         cortex-a76.cortex-a55.
10903 2018-06-27  Jeff Law  <law@redhat.com>
10905         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
10906         (MULTILIB_DIRNAMES): Similarly.
10908 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
10910         * gimple.h (gimple_return_retbnd): Delete.
10911         (gimple_return_set_retbnd): Likewise.
10912         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
10913         gimple_return_set_retbnd.
10914         * gimple-pretty-print.c (dump_gimple_return): Remove call to
10915         gimple_return_retbnd and adjust.
10916         * tree-inline.h (struct copy_body_data): Remove retbnd field.
10917         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
10918         Explicitly return NULL in a couple more cases.  Move assertion
10919         on debug statements and remove unreachable code.
10920         (reset_debug_binding): Do not test id->retbnd.
10921         (expand_call_inline): Do not set it.
10923 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
10925         * configure.ac: Add --disable-gcov option.
10926         * configure: Regenerate.
10927         * Makefile.in: Honour @enable_gcov@.
10928         * doc/install.texi: Document --disable-gcov.
10930 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10932         * config/arm/arm-cpus.in (cortex-a76): New entry.
10933         (cortex-a76.cortex-a55): Likewise.
10934         * config/arm/arm-tables.opt: Regenerate.
10935         * config/arm/arm-tune.md: Likewise.
10936         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10937         * doc/invoke.texi (ARM Options): Document cortex-a76 and
10938         cortex-a76.cortex-a55.
10940 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
10942         PR target/85769
10943         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10945 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10947         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10948         comment.
10949         (EPILOGUE_USES): Likewise.
10951 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
10953         * tree-inline.c (remap_location): New function extracted from...
10954         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
10955         (copy_phis_for_bb): ...here.  Call remap_location.
10956         (copy_cfg_body): Adjust call to copy_edges_for_bb.
10958 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
10960         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10961         unaligned vsx for 16B memset.
10963 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
10965         PR target/86285
10966         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10967         ieee128_float_type_node to long_double_type_node unless
10968         TARGET_LONG_DOUBLE_128 is set.
10970 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
10972         * cfgloop.c (get_loop_location): Convert return type from
10973         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10974         by implicit construction from rtx_insn *, and using
10975         dump_user_location_t::from_function_decl for the fallback case.
10976         * cfgloop.h (get_loop_location): Convert return type from
10977         location_t to dump_user_location_t.
10978         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10979         dump_printf_loc to pass in a dump_location_t rather than a
10980         location_t, via the gimple stmt.
10981         * coverage.c (get_coverage_counts): Update calls to
10982         dump_printf_loc to pass in dump_location_t rather than a
10983         location_t.
10984         * doc/optinfo.texi (Dump types): Convert example of
10985         dump_printf_loc from taking "locus" to taking "insn".  Update
10986         description of the "_loc" calls to cover dump_location_t.
10987         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10988         "selftest.h".
10989         (dump_user_location_t::dump_user_location_t): New constructors,
10990         from gimple *stmt and rtx_insn *.
10991         (dump_user_location_t::from_function_decl): New function.
10992         (dump_loc): Make static.
10993         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10994         const dump_location_t &.
10995         (dump_generic_expr_loc): Delete.
10996         (dump_printf_loc): Convert param "loc" from location_t to
10997         const dump_location_t &.
10998         (selftest::test_impl_location): New function.
10999         (selftest::dumpfile_c_tests): New function.
11000         * dumpfile.h: Include "profile-count.h".
11001         (class dump_user_location_t): New class.
11002         (struct dump_impl_location_t): New struct.
11003         (class dump_location_t): New class.
11004         (dump_printf_loc): Convert 2nd param from source_location to
11005         const dump_location_t &.
11006         (dump_generic_expr_loc): Delete.
11007         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11008         const dump_location_t &.
11009         * gimple-fold.c (fold_gimple_assign): Update call to
11010         dump_printf_loc to pass in a dump_location_t rather than a
11011         location_t, via the gimple stmt.
11012         (gimple_fold_call): Likewise.
11013         * gimple-loop-interchange.cc
11014         (loop_cand::analyze_iloop_reduction_var): Update for change to
11015         check_reduction_path.
11016         (tree_loop_interchange::interchange): Update for change to
11017         find_loop_location.
11018         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11019         change in return-type of find_loop_location.
11020         (graphite_regenerate_ast_isl): Likewise.
11021         * graphite-optimize-isl.c (optimize_isl): Likewise.
11022         * graphite.c (graphite_transform_loops): Likewise.
11023         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11024         pass in a dump_location_t rather than a location_t, via the
11025         gimple stmt.
11026         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11027         * ipa.c (walk_polymorphic_call_targets): Likewise.
11028         * loop-unroll.c (report_unroll): Convert "locus" param from
11029         location_t to dump_location_t.
11030         (decide_unrolling): Update for change to get_loop_location's
11031         return type.
11032         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11033         location_t to dump_user_location_t.
11034         (grid_find_single_omp_among_assignments_1): Updates calls to
11035         dump_printf_loc to pass in a dump_location_t rather than a
11036         location_t, via the gimple stmt.
11037         (grid_parallel_clauses_gridifiable): Convert "tloc" from
11038         location_t to dump_location_t.  Updates calls to dump_printf_loc
11039         to pass in a dump_location_t rather than a location_t, via the
11040         gimple stmt.
11041         (grid_inner_loop_gridifiable_p): Likewise.
11042         (grid_dist_follows_simple_pattern): Likewise.
11043         (grid_gfor_follows_tiling_pattern): Likewise.
11044         (grid_target_follows_gridifiable_pattern): Likewise.
11045         (grid_attempt_target_gridification): Convert initialization
11046         of local "grid" from memset to zero-initialization; FIXME: does
11047         this require C++11?  Update call to dump_printf_loc to pass in a
11048         optinfo_location rather than a location_t, via the gimple stmt.
11049         * profile.c (read_profile_edge_counts): Updates call to
11050         dump_printf_loc to pass in a dump_location_t rather than a
11051         location_t
11052         (compute_branch_probabilities): Likewise.
11053         * selftest-run-tests.c (selftest::run_tests): Call
11054         dumpfile_c_tests.
11055         * selftest.h (dumpfile_c_tests): New decl.
11056         * tree-loop-distribution.c (pass_loop_distribution::execute):
11057         Update for change in return type of find_loop_location.
11058         * tree-parloops.c (parallelize_loops): Likewise.
11059         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11060         "locus" from location_t to dump_user_location_t.
11061         (canonicalize_loop_induction_variables): Likewise.
11062         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11063         for change in return type of find_loop_location.
11064         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11065         to dump_printf_loc to pass in a dump_location_t rather than a
11066         location_t, via the stmt.
11067         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11068         Likewise.
11069         * tree-vect-loop-manip.c (find_loop_location): Convert return
11070         type from source_location to dump_user_location_t.
11071         (vect_do_peeling): Update for above change.
11072         (vect_loop_versioning): Update for change in type of
11073         vect_location.
11074         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11075         from location_t to dump_user_location_t.
11076         (vect_estimate_min_profitable_iters): Update for change in type
11077         of vect_location.
11078         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11079         location_t to dump_location_t.
11080         (vect_slp_bb): Update for change in type of vect_location.
11081         * tree-vectorizer.c (vect_location): Convert from source_location
11082         to dump_user_location_t.
11083         (try_vectorize_loop_1): Update for change in vect_location's type.
11084         (vectorize_loops): Likewise.
11085         (increase_alignment): Likewise.
11086         * tree-vectorizer.h (vect_location): Convert from source_location
11087         to dump_user_location_t.
11088         (find_loop_location): Convert return type from source_location to
11089         dump_user_location_t.
11090         (check_reduction_path): Convert 1st param from location_t to
11091         dump_user_location_t.
11092         * value-prof.c (check_counter): Update call to dump_printf_loc to
11093         pass in a dump_user_location_t rather than a location_t; update
11094         call to error_at for change in type of "locus".
11095         (check_ic_target): Update call to dump_printf_loc to
11096         pass in a dump_user_location_t rather than a location_t, via the
11097         call_stmt.
11099 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11101         * config/s390/s390.h (enum processor_flags): Do not use
11102         default tune parameter when -march was specified.
11104 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
11106         PR target/86314
11107         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
11108         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
11109         operands.
11111 2018-06-26  Richard Biener  <rguenther@suse.de>
11113         PR tree-optimization/86287
11114         PR bootstrap/86316
11115         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
11116         (vect_analyze_loop): Initialize n_stmts.
11118 2018-06-26  Richard Biener  <rguenther@suse.de>
11120         PR middle-end/86271
11121         * fold-const.c (fold_convertible_p): Pointer extension
11122         isn't valid.
11124 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
11126         PR debug/86064
11127         * dwarf2out.c (loc_list_has_views): Adjust comments.
11128         (dw_loc_list): Split single cross-partition range with
11129         nonzero locview.
11131 2018-06-25  Jeff Law  <law@redhat.com>
11133         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
11134         on -mbig-switch by default.
11136         * config/v850/predicates.md (const_float_1_operand): Fix match_code
11137         test.
11138         (const_float_0_operand): Remove unused predicate.
11139         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
11140         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
11141         (recipsf2): New expander.  Original pattern now called
11142         (recipsf2_insn).
11143         (recipdf2, recipdf2_insn): Similarly.
11144         (rsqrtsf2, rsqrtsf2_insn): Similarly
11145         (rsqrtdf2, rsqrtdf2_insn): Similarly
11147 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
11149         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
11150         Simplify logic for FreeBSD (twice).
11152 2018-06-25  Martin Sebor  <msebor@redhat.com>
11154         PR tree-optimization/86204
11155         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
11156         a strnlen result if it's less than the length of the string.
11158 2018-06-25  Martin Sebor  <msebor@redhat.com>
11160         PR tree-optimization/85700
11161         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
11162         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
11163         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
11165 2018-06-25  Martin Sebor  <msebor@redhat.com>
11167         * doc/extend.texi (Zero-length arrays): Update and clarify.
11169 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
11171         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
11172         added IEEE/IBM long double multilib support on PowerPC little
11173         endian Linux systems.
11174         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11175         (MULTILIB_DEFAULTS): Likewise.
11176         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11177         Likewise.
11178         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
11179         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
11180         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
11182 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
11184         PR middle-end/86311
11185         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
11186         (REORDER_45): Likewise.
11188 2018-06-25  Jeff Law  <law@redhat.com>
11190         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
11191         dividend to 32 bits.  Adjust length.
11192         (udivmodhi4): Cleanup output template.  Fix length.
11194 2018-06-25  Carl Love  <cel@us.ibm.com>
11196         * config/rs6000/vsx.md: Change word selector to prefered location.
11198 2018-06-25  Richard Biener  <rguenther@suse.de>
11200         PR tree-optimization/86304
11201         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
11202         epilogue-if-converted loops as well.
11204 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
11206         * lto-section-out.c (lto_begin_section): Do not print section
11207         name for noaddr and unnumbered dumps.
11209 2018-06-25  Richard Biener  <rguenther@suse.de>
11211         * tree-vectorizer.h (struct vec_info_shared): New structure
11212         with parts split out from struct vec_info and loop_nest from
11213         struct _loop_vec_info.
11214         (struct vec_info): Adjust accordingly.
11215         (struct _loop_vec_info): Likewise.
11216         (LOOP_VINFO_LOOP_NEST): Adjust.
11217         (LOOP_VINFO_DATAREFS): Likewise.
11218         (LOOP_VINFO_DDRS): Likewise.
11219         (struct _bb_vec_info): Likewise.
11220         (BB_VINFO_DATAREFS): Likewise.
11221         (BB_VINFO_DDRS): Likewise.
11222         (struct _stmt_vec_info): Add dr_aux member.
11223         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
11224         (DR_MISALIGNMENT_UNINITIALIZED): New.
11225         (set_dr_misalignment): Adjust.
11226         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
11227         (vect_analyze_loop): Adjust prototype.
11228         (vect_analyze_loop_form): Likewise.
11229         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11230         Compute dependences lazily.
11231         (vect_record_base_alignments): Use shared datarefs/ddrs.
11232         (vect_verify_datarefs_alignment): Likewise.
11233         (vect_analyze_data_refs_alignment): Likewise.
11234         (vect_analyze_data_ref_accesses): Likewise.
11235         (vect_analyze_data_refs): Likewise.
11236         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
11237         constructor parameter for shared part.
11238         (vect_analyze_loop_form): Pass in shared part and adjust.
11239         (vect_analyze_loop_2): Pass in storage for the number of
11240         stmts.  Move loop nest finding to the caller.  Compute
11241         datarefs lazily.
11242         (vect_analyze_loop): Pass in shared part.
11243         (vect_transform_loop): Verify shared datarefs are unchanged.
11244         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
11245         constructor parameter for shared part.
11246         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
11247         (vect_slp_bb): Verify shared datarefs are unchanged before
11248         transform.
11249         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
11250         change.
11251         (new_stmt_vec_info): Initialize DR_AUX misalignment to
11252         DR_MISALIGNMENT_UNINITIALIZED.
11253         * tree-vectorizer.c (vec_info::vec_info): Add constructor
11254         parameter for shared part.
11255         (vec_info::~vec_info): Adjust.
11256         (vec_info_shared::vec_info_shared): New.
11257         (vec_info_shared::~vec_info_shared): Likewise.
11258         (vec_info_shared::save_datarefs): Likewise.
11259         (vec_info_shared::check_datarefs): Likewise.
11260         (try_vectorize_loop_1): Construct shared part live for analyses
11261         of a single loop for multiple vector sizes.
11262         * tree-parloops.c (gather_scalar_reductions): Adjust.
11264 2018-06-25  Richard Biener  <rguenther@suse.de>
11266         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
11267         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
11268         (vect_analyze_data_refs): Remove similar code from here and
11269         simplify accordingly.
11271 2018-06-25  Richard Biener  <rguenther@suse.de>
11273         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
11274         for reverse storage order accesses rather than asserting
11275         they cannot happen here.
11277 2018-06-25  Tom de Vries  <tdevries@suse.de>
11279         PR debug/86257
11280         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
11281         Use data16 instead of .byte for insn prefix.
11283 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
11285         PR C++/86082
11286         * parser.c (make_char_string_pack): Pass this literal chars
11287         through cpp_interpret_string.
11288         (cp_parser_userdef_numeric_literal): Check the result of
11289         make_char_string_pack.
11291 2018-06-24  Maya Rashish  <coypu@sdf.org>
11293         * ginclude/stddef.h: Simplify conditions around avoiding
11294         re-definition of __size_t.
11296 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11298         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
11299         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
11301 2018-06-22  Maya Rashish  <coypu@sdf.org>
11303         * doc/invoke.texi (mno-fancy-math-387): Update for changes
11304         made to OpenBSD and NetBSD through the years.
11306 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11308         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11309         behavior of vec_pack (vector double, vector double) to match
11310         behavior of vec_float2 (vector double, vector double).
11312 2018-06-22  Olivier Hainque  <hainque@adacore.com>
11314         * gimplify.c (gimplify_function_tree): Prevent creation
11315         of a trampoline for the address of the current function
11316         passed to entry/exit instrumentation hooks.
11318 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
11320         PR target/86222
11321         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
11322         correctly.
11324 2018-06-22  Martin Liska  <mliska@suse.cz>
11326         PR tree-optimization/86263
11327         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11328         Bail out if is_enabled is false.
11329         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
11330         New declaration.
11331         (jump_table_cluster::is_enabled): New function.
11333 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11335         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11336         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
11337         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
11338         (lto_input_ts_binfo_tree_pointers): Likewise.
11339         * tree-streamer-out.c (streamer_write_tree_bitfields,
11340         write_ts_binfo_tree_pointers): Likewise.
11341         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
11343 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11345         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
11347 2018-06-22  Martin Liska  <mliska@suse.cz>
11349         * symbol-summary.h (get): Make it pure and inline move
11350         functionality from ::get function.
11351         (get): Remove and inline into ::get and ::get_create.
11352         (get_create): Move code from ::get function.
11354 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11356         PR target/85994
11357         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
11358         -x assembler-with-cpp.
11360 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11362         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
11363         _FILE_OFFSET_BITS=64 for C++.
11365 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
11367         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
11368         conversion insn that shows up when pr85657-3.c is compiled using
11369         IEEE 128-bit long double.
11370         (neg<mode>2_internal): Use the correct mode to check whether the
11371         mode is IBM extended.
11372         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
11373         multiply and divide external functions from being created more
11374         than once.
11376 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11378         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
11379         functions.
11380         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
11381         the edge can be forwarded.
11382         (cfg_layout_merge_blocks): Likewise.
11384 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11386         * except.c (finish_eh_generation): Commit edge insertions only after
11387         the EH edges have been redirected from post-landing to landing pads.
11389 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11391         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
11392         create_tmp_var_for to create the FRAME decl.
11393         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
11395 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
11397         * tree-inline.c (copy_edges_for_bb): Minor tweak.
11398         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
11399         debug statement when resetting its value.
11400         (expand_call_inline): Copy the locus of the call onto the assignment
11401         of the return value, if any.  Use local variable in more cases.
11403 2018-06-21  Martin Liska  <mliska@suse.cz>
11405         * ipa-pure-const.c (propagate_nothrow): Use
11406         funct_state_summaries->get.
11407         (dump_malloc_lattice): Likewise.
11408         (propagate_malloc): Likewise.
11410 2018-06-21  Richard Biener  <rguenther@suse.de>
11412         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
11413         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
11414         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
11415         comment.
11416         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
11417         BLOCK_ABSTRACT_ORIGIN unconditionally.
11419 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
11421         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
11422         deleting it.
11423         * ipa-reference.c (ipa_reference_c_finalize): Delete
11424         ipa_ref_opt_sum_summaries and set it to NULL.
11426 2018-06-21  Tom de Vries  <tdevries@suse.de>
11428         PR tree-optimization/85859
11429         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
11430         test with comment from bb_no_side_effects_p.
11432 2018-06-21  Richard Biener  <rguenther@suse.de>
11434         PR tree-optimization/86232
11435         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
11436         max for constant niter.
11438 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11440         * config/aarch64/aarch64-simd.md
11441         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
11443 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11445         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
11446         Make opernads of the unspec commutative.
11448 2018-06-21  Richard Biener  <rguenther@suse.de>
11450         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
11451         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
11452         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
11453         (vect_analyze_data_ref_dependence): Re-order checks to deal with
11454         NULL DR_STEP.
11455         (vect_record_base_alignments): Do not record base alignment
11456         for gathers or scatters.
11457         (vect_compute_data_ref_alignment): Drop return value that is always
11458         true.  Bail out early for gathers or scatters.
11459         (vect_enhance_data_refs_alignment): Bail out early for gathers
11460         or scatters.
11461         (vect_find_same_alignment_drs): Likewise.
11462         (vect_analyze_data_refs_alignment): Remove dead code.
11463         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11464         (vect_analyze_data_refs): For possible gathers or scatters do
11465         not create an alternate DR, just check their possible validity
11466         and mark them.  Adjust DECL_NONALIASED handling to not rely
11467         on DR_BASE_ADDRESS.
11468         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
11469         update inits of gathers or scatters.
11470         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
11471         Also copy gather/scatter flag to pattern vinfo.
11473 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11475         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11476         behavior of vec_packsu (vector unsigned long long, vector unsigned
11477         long long) to match behavior of vec_packs with same signature.
11479 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
11480             Thomas Schwinge <thomas@codesourcery.com>
11481             Cesar Philippidis  <cesar@codesourcery.com>
11483         * gimplify.c (gimplify_scan_omp_clauses): Add support for
11484         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11485         (gimplify_adjust_omp_clauses): Likewise.
11486         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
11487         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
11488         (gimplify_omp_target_update): Update handling of acc update and
11489         enter/exit data.
11490         * omp-low.c (install_var_field): Remove unused parameter
11491         base_pointers_restrict.
11492         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
11493         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
11494         FINALIZE}
11495         (omp_target_base_pointers_restrict_p): Delete.
11496         (scan_omp_target): Update call to scan_sharing_clauses.
11497         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
11498         FINALIZE}.
11499         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
11500         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11501         (convert_local_omp_clauses): Likewise.
11502         * tree-pretty-print.c (dump_omp_clause): Likewise.
11503         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
11504         FINALIZE}.
11505         (omp_clause_code_name): Likewise.
11507 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
11509         PR debug/86194
11510         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
11511         be narrowed.
11513         PR tree-optimization/86231
11514         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
11515         anti-range don't overwrite *vr0min before using it to compute *vr0max.
11517 2018-06-20  Tom de Vries  <tdevries@suse.de>
11519         PR tree-optimization/86097
11520         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
11521         iv type if signedness of iv type is not the same as that of *nit.
11523 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
11525         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
11526         EDGE_EH edges, verify they are all EDGE_EH.
11528 2018-06-20  Maya Rashish  <coypu@sdf.org>
11530         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
11532 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11534         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
11535         * config/aarch64/aarch64.c (xgene1_tunings): Add
11536         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
11537         (aarch64_mode_valid_for_sched_fusion_p):
11538         Allow 16-byte modes.
11539         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
11540         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
11541         128-bit modes.
11542         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
11543         New pattern.
11544         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
11545         * config/aarch64/iterators.md (VQ2): New mode iterator.
11547 2018-06-20  Martin Liska  <mliska@suse.cz>
11549         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11550         Change default ratio from 10 to 8.
11552 2018-06-20  Martin Liska  <mliska@suse.cz>
11554         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11555         New.
11556         (bit_test_cluster::find_bit_tests): Likewise.
11557         (switch_decision_tree::analyze_switch_statement): Find clusters.
11558         * tree-switch-conversion.h (struct jump_table_cluster): Document
11559         hierarchy.
11561 2018-06-20  Martin Liska  <mliska@suse.cz>
11563         * tree-switch-conversion.c (switch_conversion::collect):
11564         Record m_uniq property.
11565         (switch_conversion::expand): Bail out for special conditions.
11566         (group_cluster::~group_cluster): New.
11567         (group_cluster::group_cluster): Likewise.
11568         (group_cluster::dump): Likewise.
11569         (jump_table_cluster::emit): New.
11570         (switch_decision_tree::fix_phi_operands_for_edges): New.
11571         (struct case_node): Remove struct.
11572         (jump_table_cluster::can_be_handled): New.
11573         (case_values_threshold): Moved to header.
11574         (reset_out_edges_aux): Likewise.
11575         (jump_table_cluster::is_beneficial): New.
11576         (bit_test_cluster::can_be_handled): Likewise.
11577         (add_case_node): Remove.
11578         (bit_test_cluster::is_beneficial): New.
11579         (case_bit_test::cmp): New.
11580         (bit_test_cluster::emit): New.
11581         (expand_switch_as_decision_tree_p): Remove.
11582         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
11583         (fix_phi_operands_for_edge): Likewise.
11584         (switch_decision_tree::analyze_switch_statement): New.
11585         (compute_cases_per_edge): Move ...
11586         (switch_decision_tree::compute_cases_per_edge): ... here.
11587         (try_switch_expansion): Likewise.
11588         (switch_decision_tree::try_switch_expansion): Likewise.
11589         (record_phi_operand_mapping): Likewise.
11590         (switch_decision_tree::record_phi_operand_mapping): Likewise.
11591         (emit_case_decision_tree): Likewise.
11592         (switch_decision_tree::emit): Likewise.
11593         (balance_case_nodes): Likewise.
11594         (switch_decision_tree::balance_case_nodes): Likewise.
11595         (dump_case_nodes): Likewise.
11596         (switch_decision_tree::dump_case_nodes): Likewise.
11597         (emit_jump): Likewise.
11598         (switch_decision_tree::emit_jump): Likewise.
11599         (emit_cmp_and_jump_insns): Likewise.
11600         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
11601         (emit_case_nodes): Likewise.
11602         (switch_decision_tree::emit_case_nodes): Likewise.
11603         (conditional_probability): Remove.
11604         * tree-switch-conversion.h (enum cluster_type): New.
11605         (PRINT_CASE): New.
11606         (struct cluster): Likewise.
11607         (cluster::cluster): Likewise.
11608         (struct simple_cluster): Likewise.
11609         (simple_cluster::simple_cluster): Likewise.
11610         (struct group_cluster): Likewise.
11611         (struct jump_table_cluster): Likewise.
11612         (struct bit_test_cluster): Likewise.
11613         (struct min_cluster_item): Likewise.
11614         (struct case_tree_node): Likewise.
11615         (case_tree_node::case_tree_node): Likewise.
11616         (jump_table_cluster::case_values_threshold): Likewise.
11617         (struct case_bit_test): Likewise.
11618         (struct switch_decision_tree): Likewise.
11619         (struct switch_conversion): Likewise.
11620         (switch_decision_tree::reset_out_edges_aux): Likewise.
11622 2018-06-20  Martin Liska  <mliska@suse.cz>
11624         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
11625         (hoist_edge_and_branch_if_true): Likewise.
11626         (expand_switch_using_bit_tests_p): Likewise.
11627         (struct case_bit_test): Likewise.
11628         (case_bit_test_cmp): Likewise.
11629         (emit_case_bit_tests): Likewise.
11630         (switch_conversion::switch_conversion): New class.
11631         (struct switch_conv_info): Remove old struct.
11632         (collect_switch_conv_info): More to ...
11633         (switch_conversion::collect): ... this.
11634         (check_range): Likewise.
11635         (switch_conversion::check_range): Likewise.
11636         (check_all_empty_except_final): Likewise.
11637         (switch_conversion::check_all_empty_except_final): Likewise.
11638         (check_final_bb): Likewise.
11639         (switch_conversion::check_final_bb): Likewise.
11640         (create_temp_arrays): Likewise.
11641         (switch_conversion::create_temp_arrays): Likewise.
11642         (free_temp_arrays): Likewise.
11643         (gather_default_values): Likewise.
11644         (switch_conversion::gather_default_values): Likewise.
11645         (build_constructors): Likewise.
11646         (switch_conversion::build_constructors): Likewise.
11647         (constructor_contains_same_values_p): Likewise.
11648         (switch_conversion::contains_same_values_p): Likewise.
11649         (array_value_type): Likewise.
11650         (switch_conversion::array_value_type): Likewise.
11651         (build_one_array): Likewise.
11652         (switch_conversion::build_one_array): Likewise.
11653         (build_arrays): Likewise.
11654         (switch_conversion::build_arrays): Likewise.
11655         (gen_def_assigns): Likewise.
11656         (switch_conversion::gen_def_assigns): Likewise.
11657         (prune_bbs): Likewise.
11658         (switch_conversion::prune_bbs): Likewise.
11659         (fix_phi_nodes): Likewise.
11660         (switch_conversion::fix_phi_nodes): Likewise.
11661         (gen_inbound_check): Likewise.
11662         (switch_conversion::gen_inbound_check): Likewise.
11663         (process_switch): Use the newly created class.
11664         (switch_conversion::expand): New.
11665         (switch_conversion::~switch_conversion): New.
11666         * tree-switch-conversion.h: New file.
11668 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11670         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
11671         tree-vect-patterns.c.
11672         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
11673         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
11674         (vect_recog_sad_pattern): Likewise.
11675         (vect_recog_widen_sum_pattern): Likewise.
11676         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
11677         (vect_recog_widen_shift_pattern): Remove the type_in argument.
11678         (vect_recog_rotate_pattern): Likewise.
11679         (vect_recog_mult_pattern): Likewise.
11680         (vect_recog_vector_vector_shift_pattern): Likewise.
11681         (vect_recog_divmod_pattern): Likewise.
11682         (vect_recog_mixed_size_cond_pattern): Likewise.
11683         (vect_recog_bool_pattern): Likewise.
11684         (vect_recog_mask_conversion_pattern): Likewise.
11685         (vect_try_gather_scatter_pattern): Likewise.
11686         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
11687         (vect_recog_over_widening_pattern): Likewise.
11688         (vect_recog_gather_scatter_pattern): Likewise.
11689         (vect_recog_func_ptr): Move from tree-vectorizer.h
11690         (vect_vect_recog_func_ptrs): Move further down the file.
11691         (vect_recog_func): Likewise.  Remove the third argument.
11692         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
11693         (vect_pattern_recog_1): Expect the pattern function to do any
11694         necessary target tests.  Also expect it to provide a vector type.
11695         Remove the type_in handling.
11697 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11699         * tree-vect-patterns.c (vect_pattern_detected): New function.
11700         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
11701         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
11702         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
11703         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
11704         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11705         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11706         (vect_recog_mask_conversion_pattern)
11707         (vect_try_gather_scatter_pattern): Likewise.
11709 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11711         * tree-vect-patterns.c (vect_get_internal_def): New function.
11712         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11713         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
11714         (search_type_for_mask_1): Use it.
11716 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11718         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11719         redundant WIDEN_SUM_EXPR handling.
11720         (vect_recog_sad_pattern): Likewise.
11722 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11724         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11725         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
11726         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
11727         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
11728         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
11730 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11732         * tree-vect-stmts.c (vectorizable_call): Make sure that we
11733         use the stmt_vec_info of the original bb statement for the
11734         new zero assignment, even if the call is part of a pattern.
11736 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11738         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
11739         that the sequence is attached to the original statement rather
11740         than the pattern statement.
11741         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
11742         PATTERN_DEF_SEQ from the original statement rather than
11743         the main pattern statement.
11744         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
11745         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
11746         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
11748 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
11750         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
11751         definition statements before the early exit for statements that aren't
11752         live or relevant.
11753         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
11754         split out from...
11755         (vect_transform_loop): ...here.  Process pattern definition
11756         statements without first checking whether the main pattern
11757         statement is live or relevant.
11759 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11761         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
11762         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
11764 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
11766         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
11767         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
11768         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
11769         (expand_block_compare): Change select_block_compare_mode call.
11770         (expand_strncmp_align_check): Use new functions, fix comment.
11771         (emit_final_str_compare_gpr): New function.
11772         (expand_strn_compare): Refactor and clean up code.
11773         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
11775 2018-06-19  Tony Reix  <tony.reix@atos.com>
11776             Damien Bergamini  <damien.bergamini@atos.com>
11777             David Edelsohn  <dje.gcc@gmail.com>
11779         * collect2.c (static_obj): New variable.
11780         (static_libs): New variable.
11781         (is_in_list): Uncomment declaration.
11782         (main): Track AIX libraries linked statically.
11783         (is_in_list): Uncomment definition.
11784         (scan_prog_file): Don't add AIX shared libraries initializer
11785         to constructor list if linking statically.
11787 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
11789         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
11790         constant.
11791         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
11793 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
11795         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
11796         blocks.
11798 2018-06-19  Martin Liska  <mliska@suse.cz>
11800         * config/i386/i386.c (ix86_can_inline_p): Do not use
11801         ipa_fn_summaries::get_create.
11802         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
11803         get.
11804         (devirtualization_time_bonus): Likewise.
11805         (ipcp_propagate_stage): Likewise.
11806         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11807         (edge_set_predicate): Likewise.
11808         (evaluate_conditions_for_known_args): Likewise.
11809         (evaluate_properties_for_edge): Likewise.
11810         (ipa_call_summary::reset): Tranform to ...
11811         (ipa_call_summary::~ipa_call_summary): ... this.
11812         (ipa_fn_summary::reset): Transform to ...
11813         (ipa_fn_summary::~ipa_fn_summary): ... this.
11814         (ipa_fn_summary_t::remove): Rename to ...
11815         (ipa_fn_summary_t::remove_callees): ... this.
11816         (ipa_fn_summary_t::duplicate): Use placement new
11817         instead of memory copy.
11818         (ipa_call_summary_t::duplicate): Likewise.
11819         (ipa_call_summary_t::remove): Remove.
11820         (dump_ipa_call_summary): Change get_create to get.
11821         (ipa_dump_fn_summary): Dump only when summary exists.
11822         (analyze_function_body): Use symbol_summary::get instead
11823         of get_create.
11824         (compute_fn_summary): Likewise.
11825         (estimate_edge_devirt_benefit): Likewise.
11826         (estimate_edge_size_and_time): Likewise.
11827         (inline_update_callee_summaries): Likewise.
11828         (remap_edge_change_prob): Likewise.
11829         (remap_edge_summaries): Likewise.
11830         (ipa_merge_fn_summary_after_inlining): Likewise.
11831         (write_ipa_call_summary): Likewise.
11832         (ipa_fn_summary_write): Likewise.
11833         (ipa_free_fn_summary): Likewise.
11834         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11835         (struct ipa_call_summary): Likewise.
11836         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11837         of get_create.
11838         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11839         (estimate_size_after_inlining): Likewise.
11840         (estimate_growth): Likewise.
11841         (growth_likely_positive): Likewise.
11842         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11843         (inline_call): Likewise.
11844         * ipa-inline.c (caller_growth_limits): Likewise.
11845         (can_inline_edge_p): Likewise.
11846         (can_inline_edge_by_limits_p): Likewise.
11847         (compute_uninlined_call_time): Likewise.
11848         (compute_inlined_call_time): Likewise.
11849         (want_inline_small_function_p): Likewise.
11850         (edge_badness): Likewise.
11851         (update_caller_keys): Likewise.
11852         (update_callee_keys): Likewise.
11853         (inline_small_functions): Likewise.
11854         (inline_to_all_callers_1): Likewise.
11855         (dump_overall_stats): Likewise.
11856         (early_inline_small_functions): Likewise.
11857         (early_inliner): Likewise.
11858         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11859         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11860         * ipa-pure-const.c (malloc_candidate_p): Likewise.
11861         * ipa-split.c (execute_split_functions): Likewise.
11862         * symbol-summary.h: Likewise.
11863         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11865 2018-06-19  Richard Biener  <rguenther@suse.de>
11867         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11868         (vectorize_loops): ... here.  Fix dbgcnt handling.
11869         (try_vectorize_loop): Wrap try_vectorize_loop_1.
11871 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
11873         PR target/86197
11874         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11875         ieee128 argument takes up only one (vector) register, not two (floating
11876         point) registers.
11878 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
11880         * gimplify.c (gimplify_init_constructor): Really never clear for an
11881         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11883 2018-06-19  Richard Biener  <rguenther@suse.de>
11885         PR tree-optimization/86179
11886         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11887         after failed recognition.
11889 2018-06-18  Martin Sebor  <msebor@redhat.com>
11891         PR middle-end/85602
11892         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
11893         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
11894         Handle integer subtraction.
11895         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
11896         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
11898 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11900         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
11901         param from rtx to rtx_insn *.
11902         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
11903         param.
11904         (frv_ifcvt_modify_insn): Likwise.
11905         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
11906         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
11907         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
11908         as_a <rtx_insn *> cast to local "unprotected_region" once
11909         it's been established that it's not NULL or pc_rtx.
11910         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
11911         param "sethi" from rtx to rtx_insn *.
11912         (nds32_group_float_insns): Likewise for param "insn".
11913         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
11914         param.
11915         (vax_output_int_subtract): Likewise.
11916         * config/vax/vax.c (vax_output_int_add): Likewise for param
11917         "insn".
11918         (vax_output_int_subtract): Likewise.
11919         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
11920         (emit_pattern_after): Likewise for param "after".
11921         (emit_insn_after): Likewise.
11922         (emit_jump_insn_after): Likewise.
11923         (emit_call_insn_after): Likewise.
11924         (emit_debug_insn_after): Likewise.
11925         (emit_pattern_before): Likewise for param "before".
11926         (emit_insn_before): Likewise.
11927         (emit_jump_insn_before): Likewise.
11928         * final.c (get_insn_template): Likewise for param "insn", removing
11929         a cast.
11930         * output.h (get_insn_template): Likewise for 2nd param.
11931         * rtl.h (emit_insn_before): Likewise.
11932         (emit_jump_insn_before): Likewise.
11933         (emit_debug_insn_before_noloc): Likewise.
11934         (emit_insn_after): Likewise.
11935         (emit_jump_insn_after): Likewise.
11936         (emit_call_insn_after): Likewise.
11937         (emit_debug_insn_after): Likewise.
11938         (set_insn_deleted): Likewise for param.
11940 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
11942         PR target/85358
11943         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11944         floating point modes, so that IFmode is numerically greater than
11945         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11946         to declare the ordering.  This prevents IFmode from being
11947         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11948         machine.  Include rs6000-modes.h to share the fractional values
11949         between genmodes* and the rest of the compiler.
11950         (IFmode): Likewise.
11951         (KFmode): Likewise.
11952         (TFmode): Likewise.
11953         * config/rs6000/rs6000-modes.h: New file.
11954         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11955         meaning of rs6000_long_double_size so that 126..128 selects an
11956         appropriate 128-bit floating point type.
11957         (rs6000_option_override_internal): Likewise.
11958         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11959         (TARGET_LONG_DOUBLE_128): Change the meaning of
11960         rs6000_long_double_size so that 126..128 selects an appropriate
11961         128-bit floating point type.
11962         (LONG_DOUBLE_TYPE_SIZE): Update comment.
11963         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11964         source and destination to match the standard usage.
11965         (truncifkf2): Likewise.
11966         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11967         ISA 2.07 to use an explicit clobber, instead of passing in a
11968         temporary.
11969         (copysign<mode>3_soft): Likewise.
11971 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
11973         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11974         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11975         (vect_slp_analyze_instance_dependence): Likewise.
11976         (vect_enhance_data_refs_alignment): Likewise.
11977         (vect_analyze_data_refs_alignment): Likewise.
11978         (vect_slp_analyze_and_verify_instance_alignment
11979         (vect_analyze_data_ref_accesses): Likewise.
11980         (vect_prune_runtime_alias_test_list): Likewise.
11981         (vect_analyze_data_refs): Likewise.
11982         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11983         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11984         (vect_analyze_scalar_cycles_1): Likewise.
11985         (vect_get_loop_niters): Likewise.
11986         (vect_analyze_loop_form_1): Likewise.
11987         (vect_update_vf_for_slp): Likewise.
11988         (vect_analyze_loop_operations): Likewise.
11989         (vect_analyze_loop): Likewise.
11990         (vectorizable_induction): Likewise.
11991         (vect_transform_loop): Likewise.
11992         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11993         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11994         (vect_make_slp_decision): Likewise.
11995         (vect_detect_hybrid_slp): Likewise.
11996         (vect_slp_analyze_operations): Likewise.
11997         (vect_slp_bb): Likewise.
11998         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11999         (vectorizable_bswap): Likewise.
12000         (vectorizable_call): Likewise.
12001         (vectorizable_simd_clone_call): Likewise.
12002         (vectorizable_conversion): Likewise.
12003         (vectorizable_assignment): Likewise.
12004         (vectorizable_shift): Likewise.
12005         (vectorizable_operation): Likewise.
12006         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12008 2018-06-18  Martin Sebor  <msebor@redhat.com>
12010         PR tree-optimization/81384
12011         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12012         * builtins.c (expand_builtin_strnlen): New function.
12013         (expand_builtin): Call it.
12014         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12015         * builtins.def (BUILT_IN_STRNLEN): New.
12016         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12017         Warn for bounds in excess of maximum object size.
12018         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12019         single-value ranges.  Handle strnlen.
12020         (handle_builtin_strlen): Handle strnlen.
12021         (strlen_check_and_optimize_stmt): Same.
12022         * doc/extend.texi (Other Builtins): Document strnlen.
12024 2018-06-18  Maya Rashish  <coypu@sdf.org>
12026         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12027         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12028         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12030         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12031         here to ...
12032         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12034 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12036         * tree.c (escaped_string::escape): Replace cast to char * by
12037         const_cast<char *> (unescaped).
12039 2018-06-18  Nick Clifton  <nickc@redhat.com>
12041         PR 84195
12042         * tree.c (escaped_string): New class.  Converts an unescaped
12043         string into its escaped equivalent.
12044         (warn_deprecated_use): Use the new class to convert the
12045         deprecation message, if present.
12046         (test_escaped_strings): New self test.
12047         (test_c_tests): Add test_escaped_strings.
12048         * doc/extend.texi (deprecated): Add a note that the
12049         deprecation message is affected by the -fmessage-length
12050         option, and that control characters will be escaped.
12051         (#pragma GCC error): Document this pragma.
12052         (#pragma GCC warning): Likewise.
12053         * doc/invoke.texi (-fmessage-length): Document this option's
12054         effect on the #warning and #error preprocessor directives and
12055         the deprecated attribute.
12057 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
12059         * tree.c (decl_value_expr_lookup): Revert latest change.
12060         (decl_value_expr_insert): Likewise.
12062 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
12064         * gimplify.c (nonlocal_vlas): Delete.
12065         (nonlocal_vla_vars): Likewise.
12066         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12067         referenced VLAs.
12068         (gimplify_body): Do not create and destroy nonlocal_vlas.
12069         * tree-nested.c: Include diagnostic.h.
12070         (use_pointer_in_frame): Tweak.
12071         (lookup_field_for_decl): Add assertion and declare the transformation.
12072         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12073         internal error when the reference is in a wrong context.  Do not
12074         create a debug decl by default.
12075         (note_nonlocal_block_vlas): Delete.
12076         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12077         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
12078         create a debug decl by default.
12079         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12080         call is in a wrong context.
12081         (fixup_vla_decls): New function.
12082         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
12083         debug variables were created.
12084         * tree.c (decl_value_expr_lookup): Add checking assertion.
12085         (decl_value_expr_insert): Likewise.
12087 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12089         PR middle-end/82479
12090         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12091         * tree-scalar-evolution.c (interpret_expr): Likewise.
12092         (expression_expensive_p): Likewise.
12093         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12094         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12095         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12096         (ssa_defined_by_minus_one_stmt_p): New.
12098 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12100         PR middle-end/64946
12101         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12102         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12103         * dojump.c (do_jump): Likewise.
12104         * expr.c (expand_expr_real_2): Check operand type's sign.
12105         * fold-const.c (const_unop): Handle ABSU_EXPR.
12106         (fold_abs_const): Likewise.
12107         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
12108         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12109         (strip_sign_op_1): Likesise.
12110         * match.pd: Add new pattern to generate ABSU_EXPR.
12111         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
12112         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12113         * tree-eh.c (operation_could_trap_helper_p): Likewise.
12114         * tree-inline.c (estimate_operator_cost): Likewise.
12115         * tree-pretty-print.c (dump_generic_node): Likewise.
12116         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
12117         * tree.def (ABSU_EXPR): New.
12119 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
12121         PR middle-end/86095
12122         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
12123         documented as preserved for backward compatibility only.
12124         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
12126         PR rtl-optimization/86108
12127         * bb-reorder.c (create_forwarder_block): Renamed to ...
12128         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
12129         jump from new landing pad to the second part.
12130         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
12131         Adjust callers.
12133 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
12135         PR middle-end/85878
12136         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
12137         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
12138         Only call store_expr for halves if the mode is the same.
12140         PR middle-end/86123
12141         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
12142         Fix up comment formatting.
12144 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12146         * typed-splay-tree.h (typed_splay_tree::remove): New function.
12147         (typed_splay_tree::closure,
12148         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
12149         (typed_splay_tree::typed_splay_tree,
12150         typed_splay_tree::operator =): Declared private.
12151         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
12152         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
12153         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
12154         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
12155         typed_splay_tree::splay_tree_splay,
12156         typed_splay_tree::splay_tree_foreach_helper,
12157         typed_splay_tree::splay_tree_insert,
12158         typed_splay_tree::splay_tree_remove,
12159         typed_splay_tree::splay_tree_lookup,
12160         typed_splay_tree::splay_tree_predecessor,
12161         typed_splay_tree::splay_tree_successor,
12162         typed_splay_tree::splay_tree_min,
12163         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
12164         (typed_splay_tree::root, typed_splay_tree::comp,
12165         typed_splay_tree::delete_key,
12166         typed_splay_tree::delete_value): New data members.
12167         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
12168         typed_splay_tree::remove.
12170 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
12172         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
12173         -mginv and -mno-ginv to the assembler.
12174         * config/mips/mips.opt (-mcrc): New option.
12175         (-mginv): Likewise.
12176         * doc/invoke.text (-mcrc): Document.
12177         (-mginv): Likewise.
12179 2018-06-15  Nick Clifton  <nickc@redhat.com>
12181         PR 84195
12182         * tree.c (escaped_string): New class.  Converts an unescaped
12183         string into its escaped equivalent.
12184         (warn_deprecated_use): Use the new class to convert the
12185         deprecation message, if present.
12186         (test_escaped_strings): New self test.
12187         (test_c_tests): Add test_escaped_strings.
12188         * doc/extend.texi (deprecated): Add a note that the
12189         deprecation message is affected by the -fmessage-length
12190         option, and that control characters will be escaped.
12191         (#pragma GCC error): Document this pragma.
12192         (#pragma GCC warning): Likewise.
12193         * doc/invoke.texi (-fmessage-length): Document this option's
12194         effect on the #warning and #error preprocessor directives and
12195         the deprecated attribute.
12197 2018-06-15  Richard Biener  <rguenther@suse.de>
12199         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
12200         here, also noting vector size used.
12201         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
12202         size used in MSG_OPTIMIZED_LOCATIONS dump.
12203         (pass_slp_vectorize::execute): Adjust.
12205 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
12207         PR target/85968
12208         * config/arc/arc.c (arc_return_address_register): Fix
12209         if-condition.
12211 2018-06-15  Richard Biener  <rguenther@suse.de>
12213         PR middle-end/86159
12214         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
12215         leave useless conversion stripping to force_gimple_operand_gsi.
12216         (gimplify_build2): Likewise.
12217         (gimplify_build1): Likewise.
12219 2018-06-15  Richard Biener  <rguenther@suse.de>
12221         PR middle-end/86076
12222         * tree-cfg.c (move_stmt_op): unshare invariant addresses
12223         before adjusting their block.
12225 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12227         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
12228         multilibs for *-*-rtems*.
12229         * config/riscv/t-rtems: New file.
12231 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
12233         PR middle-end/86122
12234         * match.pd ((A +- CST1) +- CST2): Punt if last resort
12235         unsigned_type_for returns NULL.
12237         PR target/85945
12238         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
12239         subregs of multi-word pseudos unless the float mode has word size.
12241 2018-06-14  Richard Biener  <rguenther@suse.de>
12243         PR middle-end/86139
12244         * tree-vect-generic.c (build_word_mode_vector_type): Remove
12245         duplicate and harmful type_hash_canon.
12246         * tree.c (type_hash_canon): Assert we didn't find ourselves.
12248 2018-06-14  Richard Biener  <rguenther@suse.de>
12250         PR ipa/86124
12251         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
12252         NULL cgraph_node.
12254 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12256         * config/rtems.h (STDINT_LONG32): Define.
12258 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
12259             Prachi Godbole  <prachi.godbole@imgtec.com>
12261         * config/mips/mips-cpus.def: Define P6600.
12262         * config/mips/mips-tables.opt: Regenerate.
12263         * config/mips/mips.c (mips_ucbranch_type): New enum.
12264         (mips_rtx_cost_data): Add support for P6600.
12265         (mips_issue_rate): Likewise.
12266         (mips_multipass_dfa_lookahead): Likewise.
12267         (mips_avoid_hazard): Likewise.
12268         (mips_reorg_process_insns): Likewise.
12269         (mips_classify_branch_p6600): New function.
12270         * config/mips/mips.h (TUNE_P6600): New define.
12271         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
12272         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
12273         * config/mips/mips.md: Include p6600.md.
12274         (processor): Add p6600.
12275         * config/mips/p6600.md: New file.
12276         * doc/invoke.texi: Add p6600 to supported architectures.
12278 2018-06-13  Martin Sebor  <msebor@redhat.com>
12280         PR tree-optimization/86114
12281         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
12282         of integer types.
12283         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
12285 2018-06-13  Richard Biener  <rguenther@suse.de>
12287         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
12288         Properly set vector type of the intermediate stmt.
12289         * tree-vect-stmts.c (vectorizable_operation): The destination
12290         var always has vectype_out type.
12292 2018-06-13  Jeff Law  <law@redhat.com>
12294         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
12295         integer 0 for argument to print_rtl_with_bb.
12296         (rl78_reorg): Likewise.
12298 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
12300         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
12301         from rtx to rtx_insn *.
12302         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
12303         "label".
12304         (add_sched_insns_for_speculation): Likewise for local "target",
12305         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12306         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
12307         from rtx_insn ** to rtx_code_label **.
12308         (reorg_emit_nops): Likewise.
12309         (c6x_reorg): Likewise for local "call_labels".
12310         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
12311         rtx to rtx_insn *.
12312         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
12313         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
12314         the loops over LABEL_REFS.
12315         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
12316         braf_label.
12317         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12318         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
12319         (split_branches): Strengthen local "olabel" from rtx to
12320         rtx_insn *, adding a safe_as_a cast.
12321         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
12322         to "rtx_insn *".
12323         (add_insn_after): Likewise for first two params.
12324         (add_insn_before): Likewise.
12325         (remove_insn): Likewise for param.
12326         (emit_pattern_before_noloc): Likewise for second and third params.
12327         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
12328         (emit_call_insn_before_noloc): Likewise.
12329         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
12330         to "rtx_insn *".
12331         (emit_barrier_before): Likewise.
12332         (emit_label_before): Strengthen "label" param from "rtx" to
12333         "rtx_code_label *".  Strengthen "before" param from "rtx" to
12334         "rtx_insn *".
12335         (emit_insn_after_1): Strengthen "after" param from "rtx" to
12336         "rtx_insn *".
12337         (emit_pattern_after_noloc): Likewise.
12338         (emit_insn_after_noloc): Likewise.
12339         (emit_jump_insn_after_noloc): Likewise.
12340         (emit_call_insn_after_noloc): Likewise.
12341         (emit_debug_insn_after_noloc): Likewise.
12342         (emit_barrier_after): Likewise.
12343         (emit_label_after): Likewise for both params.
12344         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
12345         "loc" param from "int" to "location_t".
12346         (emit_insn_after_setloc): Likewise.
12347         (emit_jump_insn_after_setloc): Likewise.
12348         (emit_call_insn_after_setloc): Likewise.
12349         (emit_debug_insn_after_setloc): Likewise.
12350         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
12351         "loc" param from "int" to "location_t".
12352         (emit_pattern_before): Convert NULL_RTX to NULL.
12353         (emit_insn_before_setloc): Convert "loc" param from "int" to
12354         "location_t".
12355         (emit_jump_insn_before_setloc): Likewise.
12356         (emit_call_insn_before_setloc): Likewise.
12357         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
12358         rtx_insn *.  Convert "loc" param from "int" to "location_t".
12359         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
12360         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
12361         Convert 3rd param from "int" to "location_t".
12362         (emit_barrier_before, emit_barrier_after, next_real_insn):
12363         Strengthen param from rtx to rtx_insn *.
12364         (emit_label_before): Strengthen 1st param from "rtx" to
12365         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
12366         "rtx_insn *".
12367         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
12368         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
12369         Strengthen 2nd param from "rtx" to "rtx_insn *".
12370         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
12371         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
12372         Likewise. Convert 3rd param from "int" to "location_t".
12373         (emit_label_after): Strengthen 1st param from "rtx" to
12374         "rtx_code_label *".
12375         (next_real_insn, remove_insn): Strengthen param from "rtx" to
12376         "rtx_insn *".
12377         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
12378         from "rtx" to "rtx_insn *".
12380 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
12382         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
12383         bodies streamed in with -Q.
12384         * dumpfile.c (dump_files): Add lto-stream-out dump file.
12385         * dumpfile.h (tree_dump_index): Add lto_stream_out.
12386         * gimple-streamer-out.c: Include gimple-pretty-print.h
12387         (output_bb): Dump stmts streamed.
12388         * lto-section-out.c: Include print-tree.h
12389         (lto_begin_section): Dump sections created.
12390         (lto_output_decl_index): Dump decl encoded.
12391         * lto-streamer-out.c: Include print-tree.h
12392         (create_output_block): Dump output block created.
12393         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
12394         (output_function): Dump function output.
12395         (output_constructor): Dump constructor streamed.
12396         (write_global_stream): Output indexes encoded.
12397         (produce_asm_for_decls): Dump streams encoded.
12398         * lto-streamer.c (streamer_dump_file): New global var.
12399         * lto-streamer.h (streamer_dump_file): Declare.
12400         * passes.c (ipa_write_summaries): Initialize streamer dump.
12401         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
12402         in.
12404 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
12406         PR target/86048
12407         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
12408         offsets for register save directives.  Emit a second batch of save
12409         directives, if need be, when the function accesses prior frames.
12411 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12413         * config/arc/fpu.md (fmasf4): Force operand to register.
12414         (fnmasf4): Likewise.
12416 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12418         * config/arc/arc-protos.h (arc_pad_return): Remove.
12419         * config/arc/arc.c (machine_function): Remove force_short_suffix
12420         and size_reason.
12421         (arc_print_operand): Adjust printing of '&'.
12422         (arc_verify_short): Remove conditional printing of short suffix.
12423         (arc_final_prescan_insn): Remove reference to size_reason.
12424         (pad_return): New function.
12425         (arc_reorg): Call pad_return.
12426         (arc_pad_return): Remove.
12427         (arc_init_machine_status): Remove reference to force_short_suffix.
12428         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
12429         (attr length): When attribute iscompact is true force to 2
12430         regardless; in the case of maybe check if we want to force the
12431         instruction to have 4 bytes length.
12432         (nopv): Change it to generate 4 byte long nop as well.
12433         (blockage): New pattern.
12434         (simple_return): Remove call to arc_pad_return.
12435         (p_return_i): Likewise.
12437 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12439         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
12441 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12443         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
12444         ARC cores.
12446 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12448         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
12449         for ARC700 and ARCv2.
12451 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
12453         PR target/86076
12454         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
12455         operands[2] instead of operands[1].
12458 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
12460         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
12461         case, check whether the outer register overlaps an unallocatable
12462         register, not just whether it fits the required class.
12464 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
12466         * poly-int.h (can_div_trunc_p): Add new overload in which all values
12467         are poly_ints.
12468         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
12469         (memrefs_conflict_p): Likewise.
12470         (init_alias_analysis): Likewise.
12471         * cfgexpand.c (expand_debug_expr): Likewise.
12472         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
12473         * cse.c (fold_rtx): Likewise.
12474         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
12475         * expr.c (emit_block_move_hints): Likewise.
12476         (clear_storage_hints, push_block, emit_push_insn): Likewise.
12477         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
12478         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
12479         (emit_group_store): Likewise.
12480         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
12481         to read the PRE/POST_MODIFY increment.
12482         * calls.c (store_one_arg): Use strip_offset.
12483         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
12484         poly_int_rtx_p.
12485         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
12486         by a VEC_SELECT.
12487         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
12488         (simplify_binary_operation_1): Extend CONST_INT handling to
12489         poly_int_rtx_p.
12490         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
12491         than a HOST_WIDE_INT.
12492         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
12493         poly_int64.
12494         (adjust_mems, add_stores): Update accodingly.
12495         (vt_canonicalize_addr): Track polynomial offsets.
12496         (emit_note_insn_var_location): Likewise.
12497         (vt_add_function_parameter): Likewise.
12498         (vt_initialize): Likewise.
12500 2018-06-12  Jeff Law  <law@redhat.com>
12502         * config.gcc (alpha*-*-freebsd*): Remove.
12503         * config/alpha/freebsd.h: Remove.
12505 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
12507         PR other/69968
12508         * spellcheck-tree.c (levenshtein_distance): Rename to...
12509         (get_edit_distance): ...this, and update for underlying renaming.
12510         * spellcheck-tree.h (levenshtein_distance): Rename to...
12511         (get_edit_distance): ...this.
12512         * spellcheck.c (levenshtein_distance): Rename to...
12513         (get_edit_distance): ...this.  Convert from Levenshtein distance
12514         to Damerau-Levenshtein distance by supporting transpositions of
12515         adjacent characters.  Rename "v1" to "v_next" and "v0" to
12516         "v_one_ago".
12517         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
12518         (selftest::test_edit_distance_unit_test_oneway): ...this, and
12519         update for underlying renaming.
12520         (selftest::levenshtein_distance_unit_test): Rename to...
12521         (selftest::test_get_edit_distance_unit): ...this, and update for
12522         underlying renaming.
12523         (selftest::test_find_closest_string): Add example from PR 69968
12524         where transposition helps
12525         (selftest::test_metric_conditions): Update for renaming.
12526         (selftest::test_metric_conditions): Likewise.
12527         (selftest::spellcheck_c_tests): Likewise.
12528         * spellcheck.h (levenshtein_distance): Rename both overloads to...
12529         (get_edit_distance): ...this.
12530         (best_match::consider): Update for renaming.
12532 2018-06-12  Martin Sebor  <msebor@redhat.com>
12534         PR tree-optimization/85259
12535         * builtins.c (compute_objsize): Handle constant offsets.
12536         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
12537         true iff a warning has been issued.
12538         * gimple.h (gimple_nonartificial_location): New function.
12539         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
12540         gimple_nonartificial_location and handle -Wno-system-headers.
12541         (handle_builtin_stxncpy): Same.
12543 2018-06-12  Martin Sebor  <msebor@redhat.com>
12545         PR c/85931
12546         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
12548 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
12550         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
12551         BUILTIN_VEC_XST entries for pointer to double and long long.
12553 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
12555         PR target/85990
12556         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12557         Update comments.
12558         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12559         Likewise.
12561 2018-06-12  Martin Liska  <mliska@suse.cz>
12563         * doc/options.texi: Document IntegerRange.
12565 2018-06-12  Martin Liska  <mliska@suse.cz>
12567         * config/i386/i386.opt: Make MPX-related options as Deprecated.
12568         * opt-functions.awk: Handle Deprecated flag.
12569         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
12570         and report error.
12571         (read_cmdline_option): Report warning for a deprecated option.
12572         * opts.h (struct cl_option): Add new field cl_deprecated.
12573         (CL_ERR_DEPRECATED): New.
12575 2018-06-12  Martin Liska  <mliska@suse.cz>
12577         * doc/options.texi: Document Deprecated option flag.
12579 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12581         * config/arc/arc-arch.h (arc_extras): New enum.
12582         (arc_cpu_t):Add field extra.
12583         (arc_cpu_types): Consider the extras.
12584         * config/arc/arc-cpus.def: Add extras info.
12585         * config/arc/arc-opts.h (processor_type): Consider extra field.
12586         * config/arc/arc.c (arc_override_options): Handle extra field.
12588 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
12590         * config/arc/arc-arch.h: Update ARC_OPTX macro.
12591         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
12592         field.
12593         * config/arc/arc.c (arc_init): Update pic warning.
12594         (irq_range): Update irq range parsing warnings.
12595         (arc_override_options): Update various warning messages.
12596         (arc_handle_aux_attribute): Likewise.
12598 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
12600         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
12602 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12604         * doc/sourcebuild.texi: Document usage of line number 0 in verify
12605         compiler messages directives.
12607 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
12609         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
12610         * config/mips/mips-tables.opt: Regenerate.
12611         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
12612         mips64r6.
12613         * doc/invoke.texi: Document -march=i6500.
12615 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
12617         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
12618         (i6400_gpmul): Add cpu_unit.
12619         (i6400_gpdiv): Likewise.
12620         (i6400_msa_add_d): Update reservations.
12621         (i6400_msa_int_add) Likewise.
12622         (i6400_msa_short_logic3) Likewise.
12623         (i6400_msa_short_logic2) Likewise.
12624         (i6400_msa_short_logic) Likewise.
12625         (i6400_msa_move) Likewise.
12626         (i6400_msa_cmp) Likewise.
12627         (i6400_msa_short_float2) Likewise.
12628         (i6400_msa_div_d) Likewise.
12629         (i6400_msa_long_logic1) Likewise.
12630         (i6400_msa_long_logic2) Likewise.
12631         (i6400_msa_mult) Likewise.
12632         (i6400_msa_long_float2) Likewise.
12633         (i6400_msa_long_float4) Likewise.
12634         (i6400_msa_long_float5) Likewise.
12635         (i6400_msa_long_float8) Likewise.
12636         (i6400_fpu_fadd): Include frint type.
12637         (i6400_fpu_store): New define_insn_reservation.
12638         (i6400_fpu_load): Likewise.
12639         (i6400_fpu_move): Likewise.
12640         (i6400_fpu_fcmp): Likewise.
12641         (i6400_fpu_fmadd): Likewise.
12642         (i6400_int_mult): Include imul3nc type and update reservation.
12643         (i6400_int_div): Include idiv3 type and update reservation.
12644         (i6400_int_load): Update to check type not move_type.
12645         (i6400_int_store): Likewise.
12646         (i6400_int_prefetch): Set zero latency.
12648 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
12650         * gcc.c: Document new %@{...} sequence.
12651         (LINK_COMMAND_SPEC): Use it for the -L switches.
12652         (cpp_unique_options): Use it for the -I switches.
12653         (at_file_argbuf): New global variable.
12654         (in_at_file): Likewise.
12655         (alloc_args): Create at_file_argbuf.
12656         (clear_args): Truncate at_file_argbuf.
12657         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
12658         (open_at_file): New function.
12659         (close_at_file): Likewise.
12660         (create_at_file): Delete.
12661         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
12662         <'o'>: Likewise.
12663         <'@'>: New case.
12664         (validate_switches_from_spec): Deal with %@{...} sequence.
12665         (validate_switches): Likewise.
12666         (driver::finalize): Call clear_args.
12668 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
12670         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
12672 2018-06-11  Martin Sebor  <msebor@redhat.com>
12674         * doc/invoke.texi (-Wall): List -Wc++17-compat.
12675         (Wno-class-memaccess): Add @opindex.
12676         (Wno-templates, Wno-multiple-inheritance): Same.
12677         (Wno-virtual-inheritance, Wno-namespaces): Same.
12678         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
12679         (Wno-format-overflow, Wno-format-truncation): Same.
12680         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
12681         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
12682         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
12683         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
12684         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
12685         (Wno-misspelled-isr): Same.
12687 2018-06-11  Martin Sebor  <msebor@redhat.com>
12689         * PR tree-optimization/86083
12690         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
12692 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
12694         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
12696 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
12698         PR target/85755
12699         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
12700         on the correct operand.
12701         (*movdi_internal64): Ditto.
12703 2018-06-11  Martin Liska  <mliska@suse.cz>
12705         PR tree-optimization/86089
12706         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
12708 2018-06-11  Julia Koval  <julia.koval@intel.com>
12710         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
12711         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
12712         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
12714 2018-06-11  Olivier Hainque  <hainque@adacore.com>
12716         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
12717         for Ada with strict dwarf2.
12719 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
12721         PR target/85755
12722         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
12723         addresses.
12725 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
12727         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
12729 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
12731         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
12732         TARGET_ELF.
12734 2018-06-08  Martin Liska  <mliska@suse.cz>
12736         * tree-cfg.h (debug_function): Fix argument type to match
12737         implementation.
12739 2018-06-08  Martin Liska  <mliska@suse.cz>
12741         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
12742         Remove usage of MPX-related (and removed) fields.
12743         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
12745 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
12747         * cfg.c (debug): Use TDF_NONE rather than 0.
12748         * cfghooks.c (debug): Likewise.
12749         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
12750         (struct dump_option_value_info): Convert to...
12751         (struct kv_pair): ...this template type.
12752         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
12753         rather than 0.
12754         (optinfo_verbosity_options): Likewise.
12755         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
12756         OPTGROUP_NONE.
12757         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
12758         than int for "optgroup_flags" param.
12759         (dump_generic_expr_loc): Use dump_flags_t rather than int for
12760         "dump_kind" param.
12761         (dump_dec): Likewise.
12762         (dump_finish): Use TDF_NONE rather than 0.
12763         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
12764         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
12765         than 0.  Update for change to option_ptr.
12766         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
12767         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
12768         0.  Update for changes to optinfo_verbosity_options and
12769         optgroup_options.
12770         (opt_info_switch_p): Convert optgroup_flags from int to
12771         optgroup_flags_t.
12772         (dump_basic_block): Use dump_flags_t rather than int
12773         for "dump_kind" param.
12774         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
12775         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
12776         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
12777         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
12778         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
12779         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
12780         TDF_NONE): Convert from macros to...
12781         (enum dump_flag): ...this new enum.
12782         (dump_flags_t): Update to use enum.
12783         (operator|, operator&, operator~, operator|=, operator&=):
12784         Implement for dump_flags_t.
12785         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
12786         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
12787         Convert from macros to...
12788         (enum optgroup_flag): ...this new enum.
12789         (optgroup_flags_t): New typedef.
12790         (operator|, operator|=): Implement for optgroup_flags_t.
12791         (struct dump_file_info): Convert field "alt_flags" to
12792         dump_flags_t.  Convert field "optgroup_flags" to
12793         optgroup_flags_t.
12794         (dump_basic_block): Use dump_flags_t rather than int for param.
12795         (dump_generic_expr_loc): Likewise.
12796         (dump_dec): Likewise.
12797         (dump_register): Convert param "optgroup_flags" to
12798         optgroup_flags_t.
12799         (opt_info_enable_passes): Likewise.
12800         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
12801         than 0.
12802         * gimple-pretty-print.c (debug): Likewise.
12803         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
12804         (merged_store_group::apply_stores): Likewise.
12805         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12806         * gimple.c (verify_gimple_pp): Likewise.
12807         * graphite-poly.c (print_pbb_body): Likewise.
12808         * passes.c (pass_manager::register_one_dump_file): Convert
12809         local "optgroup_flags" to optgroup_flags_t.
12810         * print-tree.c (print_node): Use TDF_NONE rather than 0.
12811         (debug): Likewise.
12812         (debug_body): Likewise.
12813         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12814         to optgroup_flags_t.
12815         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12816         than 0.
12817         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12818         (convert_mult_to_fma): Likewise.
12819         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12820         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12821         * tree-vect-data-refs.c (dump_lower_bound): Convert param
12822         "dump_kind" to dump_flags_t.
12824 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
12826         * config/rs6000/rs6000.c (min, max): Delete.
12828 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
12830         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12831         -mabi=spe and -mabi=no-spe.
12833 2018-06-08  Martin Liska  <mliska@suse.cz>
12835         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12836         where we expect an existing summary.
12838 2018-06-08  Martin Liska  <mliska@suse.cz>
12840         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12841         * ipa-inline.h (estimate_edge_growth): Likewise.
12843 2018-06-08  Martin Liska  <mliska@suse.cz>
12845         * cgraph.c (function_version_hasher::hash): Use
12846         cgraph_node::get_uid ().
12847         (function_version_hasher::equal):
12848         * cgraph.h (cgraph_node::get_uid): New method.
12849         * ipa-inline.c (update_caller_keys): Use
12850         cgraph_node::get_uid ().
12851         (update_callee_keys): Likewise.
12852         * ipa-utils.c (searchc): Likewise.
12853         (ipa_reduced_postorder): Likewise.
12854         * lto-cgraph.c (input_node): Likewise.
12855         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12856         * symbol-summary.h (symtab_insertion): Likewise.
12857         (symtab_removal): Likewise.
12858         (symtab_duplication): Likewise.
12859         * tree-pretty-print.c (dump_function_header): Likewise.
12860         * tree-sra.c (convert_callers_for_node): Likewise.
12862 2018-06-08  Martin Liska  <mliska@suse.cz>
12864         * cgraph.c (symbol_table::create_edge): Always assign a new
12865         unique number.
12866         (symbol_table::free_edge): Do not recycle numbers.
12867         * cgraph.h (cgraph_edge::get): New method.
12868         * symbol-summary.h (symtab_removal): Use it.
12869         (symtab_duplication): Likewise.
12870         (call_summary::hashable_uid): Remove.
12872 2018-06-08  Martin Liska  <mliska@suse.cz>
12874         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12875         (initialize_growth_caches): Remove.
12876         (free_growth_caches): Likewise.
12877         (do_estimate_edge_time): Use edge_growth_cache.
12878         (do_estimate_edge_size): Likewise.
12879         (do_estimate_edge_hints): Likewise.
12880         * ipa-inline.c (reset_edge_caches): Likewise.
12881         (recursive_inlining): Likewise.
12882         (inline_small_functions): Likewise.
12883         * ipa-inline.h (initialize_growth_caches): Remove.
12884         (estimate_edge_size): Likewise.
12885         (estimate_edge_time): Likewise.
12886         (estimate_edge_hints): Likewise.
12887         (reset_edge_growth_cache): Likewise.
12888         * symbol-summary.h (call_summary::remove): New method.
12890 2018-06-08  Martin Liska  <mliska@suse.cz>
12892         * ipa-cp.c (class edge_clone_summary): New summary.
12893         (grow_edge_clone_vectors): Remove.
12894         (ipcp_edge_duplication_hook): Remove.
12895         (class edge_clone_summary_t): New call_summary class.
12896         (ipcp_edge_removal_hook): Remove.
12897         (edge_clone_summary_t::duplicate): New function.
12898         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
12899         (create_specialized_node): Likewise.
12900         (ipcp_driver): Initialize edge_clone_summaries and do not
12901         register hooks.
12903 2018-06-08  Martin Liska  <mliska@suse.cz>
12905         * symbol-summary.h (get): New function.
12906         (call_summary::m_initialize_when_cloning): New class member.
12908 2018-06-08  Martin Liska  <mliska@suse.cz>
12910         * cgraph.c (cgraph_node::remove): Do not recycle uid.
12911         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
12912         (symbol_table::allocate_cgraph_symbol): Do not set uid.
12913         * passes.c (uid_hash_t): Record removed_nodes by their uids.
12914         (remove_cgraph_node_from_order): Use the removed_nodes set.
12915         (do_per_function_toporder): Likwise.
12916         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
12917         instead of summary_uid.
12918         (symtab_removal): Likewise.
12919         (symtab_duplication): Likewise.
12921 2018-06-08  Martin Liska  <mliska@suse.cz>
12923         * ipa-cp.c (ipcp_store_bits_results): Use
12924         ipcp_transformation_sum.
12925         (ipcp_store_vr_results): Likewise.
12926         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12927         to ...
12928         (ipcp_transformation_initialize): ... this.
12929         (ipa_set_node_agg_value_chain):
12930         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12931         (write_ipcp_transformation_info): Likewise.
12932         (read_ipcp_transformation_info): Likewise.
12933         (ipcp_update_bits): Likewise.
12934         (ipcp_update_vr): Likewise.
12935         (ipcp_transform_function): Likewise.
12936         * ipa-prop.h: Rename ipcp_transformation_summary to
12937         ipcp_transformation.
12938         (class ipcp_transformation_t): New function summary.
12939         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12940         (ipa_get_agg_replacements_for_node): Likewise.
12942 2018-06-08  Martin Liska  <mliska@suse.cz>
12944         * ipa-pure-const.c (struct funct_state_d): Do it class instead
12945         of struct.
12946         (class funct_state_summary_t): New function_summary class.
12947         (has_function_state): Remove.
12948         (get_function_state): Likewise.
12949         (set_function_state): Likewise.
12950         (add_new_function): Likewise.
12951         (funct_state_summary_t::insert): New function.
12952         (duplicate_node_data): Remove.
12953         (remove_node_data): Remove.
12954         (funct_state_summary_t::duplicate): New function.
12955         (register_hooks): Create new funct_state_summaries.
12956         (pure_const_generate_summary): Use it.
12957         (pure_const_write_summary): Likewise.
12958         (pure_const_read_summary): Likewise.
12959         (propagate_pure_const): Likewise.
12960         (propagate_nothrow): Likewise.
12961         (dump_malloc_lattice): Likewise.
12962         (propagate_malloc): Likewise.
12963         (execute): Do not register hooks, just remove summary
12964         instead.
12965         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12966         constructor.
12968 2018-06-08  Martin Liska  <mliska@suse.cz>
12970         * ipa-reference.c (remove_node_data): Remove.
12971         (duplicate_node_data): Likewise.
12972         (class ipa_ref_var_info_summary_t): New class.
12973         (class ipa_ref_opt_summary_t): Likewise.
12974         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12975         (get_reference_optimization_summary): Use
12976         ipa_ref_opt_sum_summaries.
12977         (set_reference_vars_info): Remove.
12978         (set_reference_optimization_summary): Likewise.
12979         (ipa_init): Create summaries.
12980         (init_function_info): Use function summary.
12981         (ipa_ref_opt_summary_t::duplicate): New function.
12982         (ipa_ref_opt_summary_t::remove): New function.
12983         (get_read_write_all_from_node): Fix GNU coding style.
12984         (propagate): Use function summary.
12985         (write_node_summary_p): Fix GNU coding style.
12986         (stream_out_bitmap): Likewise.
12987         (ipa_reference_read_optimization_summary): Use function summary.
12988         (ipa_reference_c_finalize): Do not release hooks.
12990 2018-06-08  Martin Liska  <mliska@suse.cz>
12992         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12993         (analyze_function_body): Extract multiple calls of get_create.
12994         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12995         * ipa-inline.c (recursive_inlining): Use ::get method.
12996         * ipa-inline.h (estimate_edge_growth): Likewise.
12998 2018-06-08  Martin Liska  <mliska@suse.cz>
13000         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13001         HSA_INVALID.
13002         (hsa_function_summary::hsa_function_summary): Use the new enum
13003         value.
13004         (hsa_gpu_implementation_p): Use hsa_summaries::get.
13005         * hsa-gen.c (hsa_get_host_function): Likewise.
13006         (get_brig_function_name): Likewise.
13007         * ipa-hsa.c (process_hsa_functions): Likewise.
13008         (ipa_hsa_write_summary): Likewise.
13009         * symbol-summary.h (symtab_duplication): Use ::get function/
13010         (get): New function.
13012 2018-06-08  Martin Liska  <mliska@suse.cz>
13014         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13015         of get.
13016         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13017         (hsa_register_kernel): Likewise.
13018         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13019         * hsa-gen.c (hsa_get_host_function): Likewise.
13020         (get_brig_function_name): Likewise.
13021         (generate_hsa): Likewise.
13022         (pass_gen_hsail::execute): Likewise.
13023         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13024         (devirtualization_time_bonus): Likewise.
13025         (ipcp_propagate_stage): Likewise.
13026         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13027         (edge_set_predicate): Likewise.
13028         (evaluate_conditions_for_known_args): Likewise.
13029         (evaluate_properties_for_edge): Likewise.
13030         (ipa_fn_summary::reset): Likewise.
13031         (ipa_fn_summary_t::duplicate): Likewise.
13032         (dump_ipa_call_summary): Likewise.
13033         (ipa_dump_fn_summary): Likewise.
13034         (analyze_function_body): Likewise.
13035         (compute_fn_summary): Likewise.
13036         (estimate_edge_devirt_benefit): Likewise.
13037         (estimate_edge_size_and_time): Likewise.
13038         (estimate_calls_size_and_time): Likewise.
13039         (estimate_node_size_and_time): Likewise.
13040         (inline_update_callee_summaries): Likewise.
13041         (remap_edge_change_prob): Likewise.
13042         (remap_edge_summaries): Likewise.
13043         (ipa_merge_fn_summary_after_inlining): Likewise.
13044         (ipa_update_overall_fn_summary): Likewise.
13045         (read_ipa_call_summary): Likewise.
13046         (inline_read_section): Likewise.
13047         (write_ipa_call_summary): Likewise.
13048         (ipa_fn_summary_write): Likewise.
13049         (ipa_free_fn_summary): Likewise.
13050         * ipa-hsa.c (process_hsa_functions): Likewise.
13051         (ipa_hsa_write_summary): Likewise.
13052         (ipa_hsa_read_section): Likewise.
13053         * ipa-icf.c (sem_function::merge): Likewise.
13054         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13055         (do_estimate_edge_time): Likewise.
13056         (estimate_size_after_inlining): Likewise.
13057         (estimate_growth): Likewise.
13058         (growth_likely_positive): Likewise.
13059         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13060         (inline_call): Likewise.
13061         * ipa-inline.c (caller_growth_limits): Likewise.
13062         (can_inline_edge_p): Likewise.
13063         (can_inline_edge_by_limits_p): Likewise.
13064         (compute_uninlined_call_time): Likewise.
13065         (compute_inlined_call_time): Likewise.
13066         (want_inline_small_function_p): Likewise.
13067         (edge_badness): Likewise.
13068         (update_caller_keys): Likewise.
13069         (update_callee_keys): Likewise.
13070         (recursive_inlining): Likewise.
13071         (inline_small_functions): Likewise.
13072         (inline_to_all_callers_1): Likewise.
13073         (dump_overall_stats): Likewise.
13074         (early_inline_small_functions): Likewise.
13075         (early_inliner): Likewise.
13076         * ipa-inline.h (estimate_edge_growth): Likewise.
13077         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13078         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13079         * ipa-prop.h (IPA_NODE_REF): Likewise.
13080         (IPA_EDGE_REF): Likewise.
13081         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13082         (propagate_malloc): Likewise.
13083         * ipa-split.c (execute_split_functions): Likewise.
13084         * symbol-summary.h: Rename get to get_create.
13085         (get): Likewise.
13086         (get_create): Likewise.
13087         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13089 2018-06-08  Martin Liska  <mliska@suse.cz>
13091         * symbol-summary.h (release): Move definition out of class
13092         declaration.
13093         (symtab_removal): Likewise.
13094         (symtab_duplication): Likewise.
13096 2018-06-08  Martin Liska  <mliska@suse.cz>
13098         * symbol-summary.h (function_summary): Move constructor
13099         implementation out of class declaration.
13100         (release): Likewise.
13101         (symtab_insertion): Likewise.
13102         (symtab_removal): Likewise.
13103         (symtab_duplication): Likewise.
13104         (get): Likewise.
13106 2018-06-08  Martin Liska  <mliska@suse.cz>
13108         * Makefile.in: Remove support for MPX (macros, related functions,
13109         fields in cgraph_node, ...).
13110         * builtin-types.def (BT_BND): Likewise.
13111         (BT_FN_BND_CONST_PTR): Likewise.
13112         (BT_FN_CONST_PTR_BND): Likewise.
13113         (BT_FN_VOID_PTR_BND): Likewise.
13114         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
13115         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
13116         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
13117         (expand_builtin_mempcpy_with_bounds): Likewise.
13118         (expand_builtin_memset_with_bounds): Likewise.
13119         (expand_builtin_memset_args): Likewise.
13120         (std_expand_builtin_va_start): Likewise.
13121         (expand_builtin): Likewise.
13122         (expand_builtin_with_bounds): Likewise.
13123         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
13124         (DEF_LIB_BUILTIN_CHKP): Likewise.
13125         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
13126         (DEF_CHKP_BUILTIN): Likewise.
13127         (BUILT_IN_MEMCPY): Likewise.
13128         (BUILT_IN_MEMMOVE): Likewise.
13129         (BUILT_IN_MEMPCPY): Likewise.
13130         (BUILT_IN_MEMSET): Likewise.
13131         (BUILT_IN_STPCPY): Likewise.
13132         (BUILT_IN_STRCAT): Likewise.
13133         (BUILT_IN_STRCHR): Likewise.
13134         (BUILT_IN_STRCPY): Likewise.
13135         (BUILT_IN_STRLEN): Likewise.
13136         (BUILT_IN_MEMCPY_CHK): Likewise.
13137         (BUILT_IN_MEMMOVE_CHK): Likewise.
13138         (BUILT_IN_MEMPCPY_CHK): Likewise.
13139         (BUILT_IN_MEMSET_CHK): Likewise.
13140         (BUILT_IN_STPCPY_CHK): Likewise.
13141         (BUILT_IN_STRCAT_CHK): Likewise.
13142         (BUILT_IN_STRCPY_CHK): Likewise.
13143         * calls.c (store_bounds): Likewise.
13144         (emit_call_1): Likewise.
13145         (special_function_p): Likewise.
13146         (maybe_warn_nonstring_arg): Likewise.
13147         (initialize_argument_information): Likewise.
13148         (finalize_must_preallocate): Likewise.
13149         (compute_argument_addresses): Likewise.
13150         (expand_call): Likewise.
13151         * cfgexpand.c (expand_call_stmt): Likewise.
13152         (expand_return): Likewise.
13153         (expand_gimple_stmt_1): Likewise.
13154         (pass_expand::execute): Likewise.
13155         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
13156         (cgraph_node::remove): Likewise.
13157         (cgraph_node::dump): Likewise.
13158         (cgraph_node::verify_node): Likewise.
13159         * cgraph.h (chkp_function_instrumented_p): Likewise.
13160         (symtab_node::get_alias_target): Likewise.
13161         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
13162         (cgraph_local_p): Likewise.
13163         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
13164         (cgraph_edge::rebuild_references): Likewise.
13165         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
13166         (walk_polymorphic_call_targets): Likewise.
13167         (cgraph_node::expand_thunk): Likewise.
13168         (symbol_table::output_weakrefs): Likewise.
13169         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
13170         (ix86_handle_option): Likewise.
13171         * config/i386/constraints.md: Likewise.
13172         * config/i386/i386-builtin-types.def (BND): Likewise.
13173         (VOID): Likewise.
13174         (PVOID): Likewise.
13175         (ULONG): Likewise.
13176         * config/i386/i386-builtin.def (BDESC_END): Likewise.
13177         (BDESC_FIRST): Likewise.
13178         (BDESC): Likewise.
13179         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
13180         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
13181         * config/i386/i386.c (enum reg_class): Likewise.
13182         (ix86_target_string): Likewise.
13183         (ix86_option_override_internal): Likewise.
13184         (ix86_conditional_register_usage): Likewise.
13185         (ix86_valid_target_attribute_inner_p): Likewise.
13186         (ix86_set_indirect_branch_type): Likewise.
13187         (ix86_set_current_function): Likewise.
13188         (ix86_function_arg_regno_p): Likewise.
13189         (init_cumulative_args): Likewise.
13190         (ix86_function_arg_advance): Likewise.
13191         (ix86_function_arg): Likewise.
13192         (ix86_pass_by_reference): Likewise.
13193         (ix86_function_value_regno_p): Likewise.
13194         (ix86_function_value_1): Likewise.
13195         (ix86_function_value_bounds): Likewise.
13196         (ix86_return_in_memory): Likewise.
13197         (ix86_setup_incoming_vararg_bounds): Likewise.
13198         (ix86_va_start): Likewise.
13199         (indirect_thunk_need_prefix): Likewise.
13200         (print_reg): Likewise.
13201         (ix86_print_operand): Likewise.
13202         (ix86_expand_call): Likewise.
13203         (ix86_output_function_return): Likewise.
13204         (reg_encoded_number): Likewise.
13205         (BDESC_VERIFYS): Likewise.
13206         (ix86_init_mpx_builtins): Likewise.
13207         (ix86_init_builtins): Likewise.
13208         (ix86_emit_cmove): Likewise.
13209         (ix86_emit_move_max): Likewise.
13210         (ix86_expand_builtin): Likewise.
13211         (ix86_builtin_mpx_function): Likewise.
13212         (ix86_get_arg_address_for_bt): Likewise.
13213         (ix86_load_bounds): Likewise.
13214         (ix86_store_bounds): Likewise.
13215         (ix86_load_returned_bounds): Likewise.
13216         (ix86_store_returned_bounds): Likewise.
13217         (ix86_class_likely_spilled_p): Likewise.
13218         (ix86_hard_regno_mode_ok): Likewise.
13219         (x86_order_regs_for_local_alloc): Likewise.
13220         (ix86_mitigate_rop): Likewise.
13221         (ix86_bnd_prefixed_insn_p): Likewise.
13222         (ix86_mpx_bound_mode): Likewise.
13223         (ix86_make_bounds_constant): Likewise.
13224         (ix86_initialize_bounds): Likewise.
13225         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
13226         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
13227         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
13228         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
13229         (TARGET_CHKP_BOUND_MODE): Likewise.
13230         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
13231         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
13232         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
13233         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
13234         * config/i386/i386.h (TARGET_MPX): Likewise.
13235         (TARGET_MPX_P): Likewise.
13236         (VALID_BND_REG_MODE): Likewise.
13237         (FIRST_BND_REG): Likewise.
13238         (LAST_BND_REG): Likewise.
13239         (enum reg_class): Likewise.
13240         (BND_REG_P): Likewise.
13241         (BND_REGNO_P): Likewise.
13242         (BNDmode): Likewise.
13243         (ADJUST_INSN_LENGTH): Likewise.
13244         * config/i386/i386.md: Likewise.
13245         * config/i386/i386.opt: Likewise.
13246         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
13247         (defined): Likewise.
13248         (LINK_MPX): Likewise.
13249         (MPX_SPEC): Likewise.
13250         (LIBMPX_SPEC): Likewise.
13251         (LIBMPXWRAPPERS_SPEC): Likewise.
13252         (CHKP_SPEC): Likewise.
13253         * config/i386/predicates.md: Likewise.
13254         * dbxout.c (dbxout_type): Likewise.
13255         * doc/extend.texi: Likewise.
13256         * doc/invoke.texi: Likewise.
13257         * doc/md.texi: Likewise.
13258         * doc/tm.texi: Likewise.
13259         * doc/tm.texi.in: Likewise.
13260         * dwarf2out.c (is_base_type): Likewise.
13261         (gen_formal_types_die): Likewise.
13262         (gen_subprogram_die): Likewise.
13263         (gen_type_die_with_usage): Likewise.
13264         (gen_decl_die): Likewise.
13265         (dwarf2out_late_global_decl): Likewise.
13266         * expr.c (expand_assignment): Likewise.
13267         (emit_storent_insn): Likewise.
13268         (store_expr_with_bounds): Likewise.
13269         (store_expr): Likewise.
13270         (expand_expr_real_1): Likewise.
13271         * expr.h (store_expr_with_bounds): Likewise.
13272         * function.c (use_register_for_decl): Likewise.
13273         (struct bounds_parm_data): Likewise.
13274         (assign_parms_augmented_arg_list): Likewise.
13275         (assign_parm_find_entry_rtl): Likewise.
13276         (assign_parm_is_stack_parm): Likewise.
13277         (assign_parm_load_bounds): Likewise.
13278         (assign_bounds): Likewise.
13279         (assign_parms): Likewise.
13280         (expand_function_start): Likewise.
13281         * gcc.c (CHKP_SPEC): Likewise.
13282         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
13283         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
13284         (wrestrict_dom_walker::check_call): Likewise.
13285         * gimple.c (gimple_build_call_from_tree): Likewise.
13286         * gimple.h (enum gf_mask): Likewise.
13287         (gimple_call_with_bounds_p): Likewise.
13288         (gimple_call_set_with_bounds): Likewise.
13289         * gimplify.c (gimplify_init_constructor): Likewise.
13290         * ipa-cp.c (initialize_node_lattices): Likewise.
13291         (propagate_constants_across_call): Likewise.
13292         (find_more_scalar_values_for_callers_subset): Likewise.
13293         * ipa-hsa.c (process_hsa_functions): Likewise.
13294         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
13295         * ipa-icf.c (sem_function::merge): Likewise.
13296         * ipa-inline.c (early_inliner): Likewise.
13297         * ipa-pure-const.c (warn_function_noreturn): Likewise.
13298         (warn_function_cold): Likewise.
13299         (propagate_pure_const): Likewise.
13300         * ipa-ref.h (enum GTY): Likewise.
13301         * ipa-split.c (find_retbnd): Likewise.
13302         (consider_split): Likewise.
13303         (split_function): Likewise.
13304         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
13305         * ipa.c (walk_polymorphic_call_targets): Likewise.
13306         (symbol_table::remove_unreachable_nodes): Likewise.
13307         (process_references): Likewise.
13308         (cgraph_build_static_cdtor_1): Likewise.
13309         * lto-cgraph.c (lto_output_node): Likewise.
13310         (output_refs): Likewise.
13311         (compute_ltrans_boundary): Likewise.
13312         (input_overwrite_node): Likewise.
13313         (input_node): Likewise.
13314         (input_cgraph_1): Likewise.
13315         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
13316         * passes.c (pass_manager::execute_early_local_passes): Likewise.
13317         (class pass_chkp_instrumentation_passes): Likewise.
13318         (make_pass_chkp_instrumentation_passes): Likewise.
13319         * passes.def: Likewise.
13320         * rtl.h (struct GTY): Likewise.
13321         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
13322         * stor-layout.c (layout_type): Likewise.
13323         * symtab.c: Likewise.
13324         * target.def: Likewise.
13325         * targhooks.c (default_chkp_bound_type): Likewise.
13326         (default_chkp_bound_mode): Likewise.
13327         (default_builtin_chkp_function): Likewise.
13328         (default_chkp_function_value_bounds): Likewise.
13329         (default_chkp_make_bounds_constant): Likewise.
13330         (default_chkp_initialize_bounds): Likewise.
13331         * targhooks.h (default_chkp_bound_type): Likewise.
13332         (default_chkp_bound_mode): Likewise.
13333         (default_builtin_chkp_function): Likewise.
13334         (default_chkp_function_value_bounds): Likewise.
13335         (default_chkp_make_bounds_constant): Likewise.
13336         (default_chkp_initialize_bounds): Likewise.
13337         * toplev.c (compile_file): Likewise.
13338         (process_options): Likewise.
13339         * tree-core.h (DEF_BUILTIN): Likewise.
13340         (DEF_BUILTIN_CHKP): Likewise.
13341         * tree-inline.c (declare_return_variable): Likewise.
13342         (remap_gimple_stmt): Likewise.
13343         (copy_bb): Likewise.
13344         (initialize_inlined_parameters): Likewise.
13345         (expand_call_inline): Likewise.
13346         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
13347         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
13348         (make_pass_ipa_chkp_produce_thunks): Likewise.
13349         (make_pass_chkp): Likewise.
13350         (make_pass_chkp_opt): Likewise.
13351         (make_pass_chkp_instrumentation_passes): Likewise.
13352         * tree-pretty-print.c (dump_generic_node): Likewise.
13353         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
13354         * tree-ssa-dce.c (propagate_necessity): Likewise.
13355         (eliminate_unnecessary_stmts): Likewise.
13356         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
13357         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
13358         * tree-ssa-sccvn.h: Likewise.
13359         * tree-ssa-strlen.c (get_string_length): Likewise.
13360         (valid_builtin_call): Likewise.
13361         (adjust_last_stmt): Likewise.
13362         (handle_builtin_strchr): Likewise.
13363         (handle_builtin_strcpy): Likewise.
13364         (handle_builtin_stxncpy): Likewise.
13365         (handle_builtin_memcpy): Likewise.
13366         (handle_builtin_strcat): Likewise.
13367         (strlen_check_and_optimize_stmt): Likewise.
13368         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
13369         * tree-streamer-in.c: Likewise.
13370         * tree-streamer.c (record_common_node): Likewise.
13371         * tree.c (tree_code_size): Likewise.
13372         (wide_int_to_tree_1): Likewise.
13373         (type_contains_placeholder_1): Likewise.
13374         (build_common_tree_nodes): Likewise.
13375         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
13376         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
13377         (POINTER_BOUNDS_P): Likewise.
13378         (BOUNDED_TYPE_P): Likewise.
13379         (BOUNDED_P): Likewise.
13380         (CALL_WITH_BOUNDS_P): Likewise.
13381         (pointer_bounds_type_node): Likewise.
13382         * value-prof.c (gimple_ic): Likewise.
13383         * var-tracking.c (vt_add_function_parameters): Likewise.
13384         * varasm.c (make_decl_rtl): Likewise.
13385         (assemble_start_function): Likewise.
13386         (output_constant): Likewise.
13387         (maybe_assemble_visibility): Likewise.
13388         * varpool.c (ctor_for_folding): Likewise.
13389         * chkp-builtins.def: Remove.
13390         * ipa-chkp.c: Remove.
13391         * ipa-chkp.h: Remove.
13392         * rtl-chkp.c: Remove.
13393         * rtl-chkp.h: Remove.
13394         * tree-chkp-opt.c: Remove.
13395         * tree-chkp.c: Remove.
13396         * tree-chkp.h: Remove.
13398 2018-06-07  Carl Love  <cel@us.ibm.com>
13400         * config/rs6000/vsx.md (vextract_fp_from_shorth,
13401         vextract_fp_from_shortl): Add BE support.
13403 2018-06-07  Paul Koning  <ni1d@arrl.net>
13405         * compare-elim.c (try_merge_compare): Don't merge compare if
13406         address contains a side effect.
13407         (try_eliminate_compare): Likewise.
13409 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
13411         * config.gcc: Support "tremont".
13412         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
13413         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13414         PROCESSOR_TREMONT.
13415         * config/i386/i386.c (m_TREMONT): Define.
13416         (processor_target_table): Add "tremont".
13417         (PTA_TREMONT): Define.
13418         (ix86_lea_outperforms): Add TARGET_TREMONT.
13419         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
13420         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
13421         and M_INTEL_GOLDMONT_PLUS.
13422         (fold_builtin_cpu): Add "tremont".
13423         (ix86_add_stmt_cost): Add TARGET_TREMONT.
13424         (ix86_option_override_internal): Add "tremont".
13425         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
13426         (processor_type): Add PROCESSOR_TREMONT.
13427         * config/i386/x86-tune.def: Add m_TREMONT.
13428         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
13430 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13432         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
13433         symbol defined for msp430i* devices to be lower case.
13435 2018-06-07  Richard Biener  <rguenther@suse.de>
13437         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
13438         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
13439         Properly wrap signed arithmetic if overflow wraps.
13441 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
13443         PR tree-optimization/69615
13444         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
13445         of a cast from a same precision integral SSA_NAME in a bb dominated
13446         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
13447         cast to utype if rhs2 has already a compatible type.
13449 2018-06-07  Richard Biener  <rguenther@suse.de>
13451         PR tree-optimization/85935
13452         * graphite-scop-detection.c (find_params_in_bb): Analyze
13453         condition operands with respect to the correct loop.  Assert
13454         the analysis doesn't fail.
13456 2018-06-04  Carl Love  <cel@us.ibm.com>
13458         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
13459         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
13460         as it is slightly cheaper.
13461         (first_match_or_eos_index_<mode>):
13462         Calculate index using natural element order.
13463         (first_match_index_<mode>):
13464         Calculate index using natural element order.
13465         (first_match_or_eos_index_<mode>):
13466         Calculate index using natural order.
13467         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
13468         for BE and LE modes.
13469         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
13470         P9V_BUILTIN_VCLZLSBB_V16QI.
13471         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
13472         specific.
13474 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13476         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
13477         indentation and line wrap for many prototypes.  Add missing
13478         @smallexample directives around block of prototypes for vec_xl and
13479         vec_xst.
13481 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
13483         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
13484         track if we pass or return IEEE 128-bit floating point.
13485         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
13486         C++ mangling that is compatible with GCC 8.1.
13487         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
13488         (init_cumulative_args): Note if we pass or return IEEE 128-bit
13489         floating point types.
13490         (rs6000_function_arg_advance_1): Likewise.
13491         (rs6000_mangle_type): Optionally generate mangled names that match
13492         what GCC 8.1 generated for IEEE 128-bit floating point types.
13493         (rs6000_globalize_decl_name): If we have an external function that
13494         passes or returns IEEE 128-bit types, generate a weak reference
13495         from the mangled name used in GCC 8.1 to the current mangled
13496         name.
13497         (rs6000_init_builtins): Make __ibm128 use the long double type if
13498         long double is IBM extended double.  Make __float128 use the long
13499         double type if long double is IEEE 128-bit.
13501         PR target/85657
13502         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
13503         macro for __ibm128 built-in functions.
13504         (PACK_IF): Add __ibm128 pack/unpack functions.
13505         (UNPACK_IF): Likewise.
13506         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
13507         enable long double built-in functions if long double is IEEE
13508         128-bit floating point.
13509         (rs6000_invalid_builtin): Update long double built-in function
13510         error message.
13511         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
13512         functions, adjust the built-in function to use the long double
13513         built-in function if __ibm128 and long double are the same type.
13514         * doc/extend.texi (PowerPC builtins): Update documention for
13515         __builtin_{,un}pack_longdouble.  Add documentation for
13516         __builtin_{,un}pack_ibm128.
13518 2018-06-06  Jim Wilson  <jimw@sifive.com>
13520         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
13521         (struct machine_function): New field interrupt_mode.
13522         (riscv_handle_type_attribute): New function.  Add forward declaration.
13523         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
13524         (riscv_expand_epilogue): Check interrupt_mode field.
13525         (riscv_set_current_function): Check interrupt attribute args and
13526         set interrupt_mode field.
13527         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
13528         (riscv_sret, riscv_uret): New.
13529         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
13530         new arguments to interrupt attribute.
13532 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
13534         PR target/63177
13535         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
13536         Don't handle -mcpu=power8 if -mpower9-vector is also used.
13538 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13540         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
13541         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
13542         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13543         several redundant entries.
13545 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
13547         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
13548         type from "rtx" to "rtx_insn *".
13549         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
13550         for local "call_insn", removing cast.
13551         (ix86_expand_call): Likewise, introducing a "call_insn" local.
13553 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
13555         PR tree-optimization/86066
13556         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
13557         for BIT_INSERT_EXPR stores.
13559 2018-06-06  Richard Biener  <rguenther@suse.de>
13561         PR tree-optimization/86062
13562         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13563         component refs ontop
13564         of to be offsetted base.
13566 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13568         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
13569         to be static and remove check on interrupt attribute name.
13571 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13573         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13574         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
13576 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
13578         PR target/79924
13579         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
13580         second argument.
13581         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
13582         Remove second argument, change how error is called.
13583         (aarch64_layout_arg): Remove second argument from
13584         aarch64_err_no_fpadvsimd call.
13585         (aarch64_init_cumulative_args): Ditto.
13586         (aarch64_gimplify_va_arg_expr): Ditto.
13587         * config/aarch64/aarch64.md (mov<mode>): Ditto.
13589 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
13591         * config/i386/i386.md (simple_return_indirect_internal): New expander.
13592         (*simple_return_indirect_internal<mode>): Rename from
13593         simple_return_indirect_internal.  Use W mode iterator.
13594         (rstorssp): New expander.
13595         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
13596         (clrssbsy): New expander.
13597         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
13599 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13601         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
13602         __typeof__.
13603         (cmse_check_pointed_object): Likewise.
13605 2018-06-05  Martin Liska  <mliska@suse.cz>
13607         PR gcov-profile/47618
13608         * doc/invoke.texi: Document how -fprofile-dir format
13609         is extended.
13611 2018-06-05  Richard Biener  <rguenther@suse.de>
13613         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
13614         removal pretend DOM info isn't available so we do not update
13615         it and only remove edges, not dominated blocks.  Actually free
13616         DOM info in case we removed something.  Remove unreachable blocks.
13617         (mfb_keep_latches): Work with either DOM info or marked backedges.
13618         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
13619         first.  Mark backedges if DOM info isn't available.
13620         (Re-)compute DOM info after cleanup_control_flow_pre.
13622 2018-06-05  Richard Biener  <rguenther@suse.de>
13624         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
13625         (locus_discrim_hasher::hash): Adjust.
13626         (locus_discrim_hasher::equal): Likewise.
13627         (next_discriminator_for_locus): Work on line directly.
13628         (same_line_p): Pass in expanded locus1 as well.
13629         (assign_discriminators): Avoid redundant location expansions.
13631 2018-06-05  Richard Biener  <rguenther@suse.de>
13633         PR tree-optimization/86046
13634         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
13635         if required after clearing TREE_ADDRESSABLE.
13637 2018-06-05  Richard Biener  <rguenther@suse.de>
13639         PR tree-optimization/86047
13640         * tree-ssa-loop.c (for_each_index): Glob handling of all
13641         decls and constants and really handle all of them.
13643 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13645         PR target/81497
13646         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
13647         qualifier_void_pointer and qualifier_const_void_pointer.
13648         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
13649         (arm_init_builtins): Handle the above.
13650         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
13651         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
13652         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
13653         void intrinsics.
13655 2018-06-05  Martin Liska  <mliska@suse.cz>
13657         * auto-profile.c (read_autofdo_file): Do not use
13658         gcov_ctr_summary struct.
13659         (afdo_callsite_hot_enough_for_early_inline): Likewise.
13660         * coverage.c (struct counts_entry): Likewise.
13661         (read_counts_file): Read just single summary entry.
13662         (get_coverage_counts): Use gcov_summary struct.
13663         * coverage.h (get_coverage_counts): Likewise.
13664         * gcov-dump.c (dump_working_sets): Likewise.
13665         (tag_summary): Dump just single summary.
13666         * gcov-io.c (gcov_write_summary): Write just histogram
13667         summary.
13668         (gcov_read_summary): Read just single summary.
13669         (compute_working_sets): Use gcov_summary struct.
13670         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
13671         of GCOV_COUNTERS_SUMMABLE.
13672         (GCOV_COUNTERS_SUMMABLE): Remove.
13673         (GCOV_FIRST_VALUE_COUNTER): Replace with
13674         GCOV_COUNTER_V_INTERVAL.
13675         (struct gcov_ctr_summary): Remove.
13676         (struct gcov_summary): Directly use fields of former
13677         gcov_ctr_summary.
13678         (compute_working_sets): Use gcov_summary struct.
13679         * gcov.c (read_count_file): Do not use ctrs fields.
13680         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
13681         struct.
13682         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
13683         struct.
13684         * profile.c: Likewise.
13685         * profile.h: Likewise.
13687 2018-06-05  Martin Liska  <mliska@suse.cz>
13689         PR gcov-profile/84846
13690         * gcov.c (output_lines): Print working directory only
13691         in intermediate format.
13693 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
13695         * config/s390/s390-builtin-types.def: Add void function type.
13696         * config/s390/s390-builtins.def: Use the function type for the
13697         tbeginc builtin.
13699 2018-06-04  Jim Wilson  <jimw@sifive.com>
13701         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
13702         to int.
13703         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
13704         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
13705         handle EH_RETURN_DATA_REGNO registers properly.
13706         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
13707         (riscv_expand_epilogue): Update comment.  Change argument name and
13708         type.  Update code to use new name and type.  Pass new args to
13709         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
13710         EXCEPTION_RETURN.
13711         * config/riscv/riscv.md (NORMAL_RETURN): New.
13712         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
13713         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
13714         (eh_return): Call gen_eh_return_internal and emit barrier.
13715         (eh_return_internal): Call riscv_expand_epilogue.
13717 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
13719         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
13720         bit_insertion field and declare can_be_merged_into method.
13721         (merged_store_group::can_be_merged_into): New method.
13722         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
13723         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
13724         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
13726 2018-06-04  Richard Biener  <rguenther@suse.de>
13728         PR tree-optimization/85955
13729         * builtins.c (fold_builtin_sincos): Convert pointers to
13730         destination to appropriate type before dereferencing.
13732 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
13734         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
13736 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
13738         * expr.c (expand_expr_real_1): Force the operand into memory if
13739         its TYPE_MODE is BLKmode and if there is no integer mode for
13740         the number of bits being extracted.
13742 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
13744         PR target/85832
13745         PR target/86036
13746         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13747         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
13749 2018-06-04  Richard Biener  <rguenther@suse.de>
13751         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
13752         (cleanup_tree_cfg_noloop): ... single caller.  Do
13753         start_recording_case_labels later.
13755 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
13757         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
13758         to _IMMINTRIN_H_INCLUDED.
13759         * config/i386/pconfigintrin.h: Ditto.
13760         * config/i386/waitpkgintrin.h: Ditto.
13761         * config/i386/immintrin.h: Add includes for sgxintrin.h,
13762         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
13763         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
13764         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
13765         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
13766         waitpkgintrin.h and cldemoteintrin.h.
13768 2018-06-04  Richard Biener  <rguenther@suse.de>
13770         PR tree-optimization/86038
13771         * tracer.c (find_best_successor): Check probability for
13772         being initialized, bail out if not.
13774 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
13776         PR target/86003
13777         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
13778         of bits to ignore when comparing architectures.
13780 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
13782         PR tree-optimization/69615
13783         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
13784         maximum or minimum of the type, try to merge it also as if
13785         range1 is + [-, x - 1] or + [x + 1, -].
13787         PR c++/86025
13788         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
13790 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
13792         PR tree-optimization/86034
13793         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
13794         the unsigned bitfield type in a bit insertion sequence if it does not
13795         have a larger precision than the bitfield size.
13796         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
13798 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
13800         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
13802 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
13804         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
13805         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13806         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13807         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13809 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
13811         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13812         Disable -fdelete-null-pointer-checks for ELF toolchain.
13814 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13815             Kito Cheng  <kito.cheng@gmail.com>
13817         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13818         (nds32le-*-*, nds32be-*-*): Integrate checking process.
13819         (nds32*-*-*): Add glibc and uclibc conditions.
13820         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13821         (TARGET_EXCEPT_UNWIND_INFO): Define.
13822         * config/nds32/elf.h: New file.
13823         * config/nds32/linux.h: New file.
13824         * config/nds32/nds32-elf.opt: New file.
13825         * config/nds32/nds32-linux.opt: New file.
13826         * config/nds32/nds32-fp-as-gp.c
13827         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13828         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13829         TARGET_LINUX_ABI.
13830         (nds32_asm_file_end): Ditto.
13831         (nds32_print_operand): Ditto.
13832         (nds32_insert_attributes): Ditto.
13833         (nds32_init_libfuncs): New function.
13834         (TARGET_HAVE_TLS): Define.
13835         (TARGET_INIT_LIBFUNCS): Define.
13836         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13837         spec content.
13838         (TARGET_ELF): Apply different mcmodel setting.
13839         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13840         been migrated into elf.h and linux.h files.
13841         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13842         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13843         (mcmodel): The content has been migrated into nds32-elf.opt and
13844         nds32-linux.opt files.
13845         * config/nds32/t-elf: New file.
13846         * config/nds32/t-linux: New file.
13848 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
13849             Shiva Chen  <shiva0217@gmail.com>
13851         * config/nds32/constants.md (unspec_volatile_element): Add
13852         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13853         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13854         optimization.
13855         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13856         (make_pass_nds32_fp_as_gp): Declare.
13857         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13858         optmization pass.
13859         (nds32_asm_function_end_prologue): Remove unused asm output.
13860         (nds32_asm_function_begin_epilogue): Remove unused asm output.
13861         (nds32_asm_file_start): Output necessary fp_as_gp information.
13862         (nds32_option_override): Adjust register usage.
13863         (nds32_expand_prologue): Consider fp_as_gp situation.
13864         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13865         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13866         (epilogue): Ditto.
13867         (return): Ditto.
13868         (simple_return): Ditto.
13869         (omit_fp_begin): Output special directive for fp_as_gp.
13870         (omit_fp_end): Output special directive for fp_as_gp.
13871         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13872         mforbid-fp-as-gp): New options.
13874 2018-06-01  Mark Wielaard  <mark@klomp.org>
13876         * dwarf2out.c (dwarf2out_finish): Remove generation of
13877         DW_AT_loclists_base.
13879 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
13881         * gimple-ssa-store-merging.c: Include gimple-fold.h.
13882         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13883         (struct merged_store_group): Add bit_insertion field.
13884         (dump_char_array): Use standard hexadecimal format.
13885         (merged_store_group::merged_store_group): Set bit_insertion to false.
13886         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
13887         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
13888         also print the mask in the dump file.
13889         (pass_store_merging::gate): Minor tweak.
13890         (imm_store_chain_info::coalesce_immediate): Fix wrong association
13891         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
13892         stores with INTEGER_CST stores.
13893         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
13894         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
13895         and use it throughout.  Generate bit insertion sequences if need be.
13896         (pass_store_merging::process_store): Remove redundant condition.
13897         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
13899 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
13901         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
13902         the 128-bit floating point types.  Fix function comment.
13904 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13906         * config/aarch64/aarch64-simd.md
13907         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
13908         mnemonics.
13909         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13910         mnemonics.
13912 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13914         PR tree-optimization/85989
13915         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
13916         variable.
13917         (backprop::intersect_uses): Check it when deciding whether this
13918         is a backedge reference.
13919         (backprop::process_block): Add each phi to m_visited_phis
13920         after visiting it, then clear it at the end.
13922 2018-06-01  Richard Biener  <rguenther@suse.de>
13924         * tree-vectorizer.h (vect_dr_stmt): New function.
13925         (vect_get_load_cost): Adjust.
13926         (vect_get_store_cost): Likewise.
13927         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13928         Use vect_dr_stmt instead of DR_SMTT.
13929         (vect_record_base_alignments): Likewise.
13930         (vect_calculate_target_alignment): Likewise.
13931         (vect_compute_data_ref_alignment): Likewise and make static.
13932         (vect_update_misalignment_for_peel): Likewise.
13933         (vect_verify_datarefs_alignment): Likewise.
13934         (vector_alignment_reachable_p): Likewise.
13935         (vect_get_data_access_cost): Likewise.  Pass down
13936         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13937         (vect_get_peeling_costs_all_drs): Likewise.
13938         (vect_peeling_hash_get_lowest_cost): Likewise.
13939         (vect_enhance_data_refs_alignment): Likewise.
13940         (vect_find_same_alignment_drs): Likewise.
13941         (vect_analyze_data_refs_alignment): Likewise.
13942         (vect_analyze_group_access_1): Likewise.
13943         (vect_analyze_group_access): Likewise.
13944         (vect_analyze_data_ref_access): Likewise.
13945         (vect_analyze_data_ref_accesses): Likewise.
13946         (vect_vfa_segment_size): Likewise.
13947         (vect_small_gap_p): Likewise.
13948         (vectorizable_with_step_bound_p): Likewise.
13949         (vect_prune_runtime_alias_test_list): Likewise.
13950         (vect_analyze_data_refs): Likewise.
13951         (vect_supportable_dr_alignment): Likewise.
13952         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13953         (vect_gen_prolog_loop_niters): Likewise.
13954         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13955         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13956         modify DR_STMT.
13957         (vect_recog_mask_conversion_pattern): Likewise.
13958         (vect_try_gather_scatter_pattern): Likewise.
13959         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13960         to vect_get_store_cost.
13961         (vect_get_store_cost): Get stmt_info instead of DR.
13962         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13963         (vect_get_load_cost): Get stmt_info instead of DR.
13965 2018-06-01  Richard Biener  <rguenther@suse.de>
13967         PR middle-end/86017
13968         * gimple-fold.c (var_decl_component_p): Also allow offsetted
13969         vars wrapped in MEM_REFs.
13971 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
13973         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13974         Fix subreg tests so that we only return a choice between
13975         GENERAL_REGS and FP_REGS if the original classes included both.
13977 2018-06-01  Richard Biener  <rguenther@suse.de>
13979         PR ipa/85960
13980         * tree-ssa-structalias.c (get_function_part_constraint):
13981         Handle NULL fi->decl.
13982         (find_func_aliases_for_call): Properly handle indirect
13983         fi from direct call.
13984         (find_func_clobbers): Likewise.
13985         (ipa_pta_execute): Likewise.
13986         (create_variable_info_for): For functions that are ifunc_resolver
13987         resolve to a varinfo that contains the result of the resolver call.
13988         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13989         aliases.
13991 2018-05-31  Michael Collison  <michael.collison@arm.com>
13993         * config/aarch64/aarch64.md:
13994         (*fix_to_zero_extenddfdi2): New pattern.
13995         * gcc.target/aarch64/fix_extend1.c: New testcase.
13997 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
13999         PR middle-end/78809
14000         PR middle-end/83026
14001         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14002         and BUILT_IN_STRNCMP_EQ.
14003         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14004         BUILT_IN_STRNCMP_EQ.
14005         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14006         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14007         (gimple_fold_builtin): Likewise.
14008         * tree-ssa-strlen.c (compute_string_length): New function.
14009         (determine_min_obsize): New function.
14010         (handle_builtin_string_cmp): New function to handle calls to
14011         string compare functions.
14012         (strlen_optimize_stmt): Add handling to builtin string compare
14013         calls.
14014         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14015         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14016         * tree.c (build_common_builtin_nodes): Add new defines of
14017         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14019 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
14021         PR target/85984
14022         * bb-reorder.c (pass_partition_blocks::gate): Return false for
14023         functions with naked attribute.
14025 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
14027         * config/i386/sse.md (avx_vec_concat<mode>):
14028         Substitute concat_tg_mode mode attribute with xtg_mode.
14029         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14030         (concat_tg_mode): Remove mode attribute.
14032 2018-05-31  Martin Sebor  <msebor@redhat.com>
14034         PR c/82063
14035         * calls.c (alloc_max_size): Correct a logic error/typo.
14036         Treat excessive arguments as infinite.  Warn for invalid arguments.
14037         * doc/invoke.texi (-Walloc-size-larger-than): Update.
14039 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
14041         PR target/85829
14042         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14043         and movx for Haswell.
14045 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
14046             Cesar Philippidis  <cesar@codesourcery.com>
14048         PR middle-end/85879
14049         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
14050         when emitting error on private/firstprivate reductions.
14051         * omp-low.c (lower_omp_target): Avoid reference-type processing
14052         on pointers for firstprivate clause.
14054 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14056         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14057         (st1x2): Likewise.
14058         (st1x3): Likewise.
14059         * config/aarch64/aarch64-simd.md
14060         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14061         (aarch64_ld1_x3_<mode>): Likewise
14062         (aarch64_st1x2<VALLDIF:mode>): Likewise
14063         (aarch64_st1_x2_<mode>): Likewise
14064         (aarch64_st1x3<VALLDIF:mode>): Likewise
14065         (aarch64_st1_x3_<mode>): Likewise
14066         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14067         (vld1_s8_x3): Likewise.
14068         (vld1_u16_x3): Likewise.
14069         (vld1_s16_x3): Likewise.
14070         (vld1_u32_x3): Likewise.
14071         (vld1_s32_x3): Likewise.
14072         (vld1_u64_x3): Likewise.
14073         (vld1_s64_x3): Likewise.
14074         (vld1_f16_x3): Likewise.
14075         (vld1_f32_x3): Likewise.
14076         (vld1_f64_x3): Likewise.
14077         (vld1_p8_x3): Likewise.
14078         (vld1_p16_x3): Likewise.
14079         (vld1_p64_x3): Likewise.
14080         (vld1q_u8_x3): Likewise.
14081         (vld1q_s8_x3): Likewise.
14082         (vld1q_u16_x3): Likewise.
14083         (vld1q_s16_x3): Likewise.
14084         (vld1q_u32_x3): Likewise.
14085         (vld1q_s32_x3): Likewise.
14086         (vld1q_u64_x3): Likewise.
14087         (vld1q_s64_x3): Likewise.
14088         (vld1q_f16_x3): Likewise.
14089         (vld1q_f32_x3): Likewise.
14090         (vld1q_f64_x3): Likewise.
14091         (vld1q_p8_x3): Likewise.
14092         (vld1q_p16_x3): Likewise.
14093         (vld1q_p64_x3): Likewise.
14094         (vst1_s64_x2): Likewise.
14095         (vst1_u64_x2): Likewise.
14096         (vst1_f64_x2): Likewise.
14097         (vst1_s8_x2): Likewise.
14098         (vst1_p8_x2): Likewise.
14099         (vst1_s16_x2): Likewise.
14100         (vst1_p16_x2): Likewise.
14101         (vst1_s32_x2): Likewise.
14102         (vst1_u8_x2): Likewise.
14103         (vst1_u16_x2): Likewise.
14104         (vst1_u32_x2): Likewise.
14105         (vst1_f16_x2): Likewise.
14106         (vst1_f32_x2): Likewise.
14107         (vst1_p64_x2): Likewise.
14108         (vst1q_s8_x2): Likewise.
14109         (vst1q_p8_x2): Likewise.
14110         (vst1q_s16_x2): Likewise.
14111         (vst1q_p16_x2): Likewise.
14112         (vst1q_s32_x2): Likewise.
14113         (vst1q_s64_x2): Likewise.
14114         (vst1q_u8_x2): Likewise.
14115         (vst1q_u16_x2): Likewise.
14116         (vst1q_u32_x2): Likewise.
14117         (vst1q_u64_x2): Likewise.
14118         (vst1q_f16_x2): Likewise.
14119         (vst1q_f32_x2): Likewise.
14120         (vst1q_f64_x2): Likewise.
14121         (vst1q_p64_x2): Likewise.
14122         (vst1_s64_x3): Likewise.
14123         (vst1_u64_x3): Likewise.
14124         (vst1_f64_x3): Likewise.
14125         (vst1_s8_x3): Likewise.
14126         (vst1_p8_x3): Likewise.
14127         (vst1_s16_x3): Likewise.
14128         (vst1_p16_x3): Likewise.
14129         (vst1_s32_x3): Likewise.
14130         (vst1_u8_x3): Likewise.
14131         (vst1_u16_x3): Likewise.
14132         (vst1_u32_x3): Likewise.
14133         (vst1_f16_x3): Likewise.
14134         (vst1_f32_x3): Likewise.
14135         (vst1_p64_x3): Likewise.
14136         (vst1q_s8_x3): Likewise.
14137         (vst1q_p8_x3): Likewise.
14138         (vst1q_s16_x3): Likewise.
14139         (vst1q_p16_x3): Likewise.
14140         (vst1q_s32_x3): Likewise.
14141         (vst1q_s64_x3): Likewise.
14142         (vst1q_u8_x3): Likewise.
14143         (vst1q_u16_x3): Likewise.
14144         (vst1q_u32_x3): Likewise.
14145         (vst1q_u64_x3): Likewise.
14146         (vst1q_f16_x3): Likewise.
14147         (vst1q_f32_x3): Likewise.
14148         (vst1q_f64_x3): Likewise.
14149         (vst1q_p64_x3): Likewise.
14151 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14153         * config/msp430/msp430.c (msp430_output_labelref): Prepend
14154         user_label_prefix to name.
14156         * tree-core.h: Update comment about the format of NAME string
14157         passed to handler in attribute_spec.
14159         * config/msp430/msp430.md: Remove erroneous subreg expression from
14160         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
14161         zero_extend{q,h}isi2.
14163 2018-05-30  Borislav Petkov  <bp@suse.de>
14165         * doc/extend.texi: Document some architecture specific
14166         constraints and sort entries.
14168 2018-05-30  Martin Sebor  <msebor@redhat.com>
14170         PR middle-end/85369
14171         * builtins.c (expand_builtin_stpcpy_1): New function.
14172         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
14173         only if the former succeeds.
14175 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14177         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
14178         in saphira.
14180 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14182         * doc/invoke.texi (-flinker-output): Document
14184 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14186         * passes.c (ipa_write_summaries): Only modify statements if body
14187         is in memory.
14188         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
14189         incrementally linking.
14190         (ipa_passes): Likewise.
14191         * lto-cgraph.c (lto_output_node): When incrementally linking do not
14192         pass down resolution info.
14193         * common.opt (flag_incremental_link): Update info.
14194         * gcc.c (plugin specs): Turn flinker-output=* to
14195         -plugin-opt=-linker-output-known
14196         * toplev.c (compile_file): Also cut compilation when doing incremental
14197         link.
14198         * flag-types. (enum lto_partition_model): Add
14199         LTO_LINKER_OUTPUT_NOLTOREL.
14200         (invoke.texi): Add -flinker-output docs.
14201         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
14202         link same way as WPA; do not stream in dead initializers.
14204         * dwarf2out.c (dwarf2out_die_ref_for_decl,
14205         darf2out_register_external_decl): Support incremental link.
14207 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14209         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
14211 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14213         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
14214         it down to simple_object_copy_lto_debug_sections.
14215         (run_gcc): Determine incremental LTO link time and configure
14216         lto1 into non-wpa mode, disable renaming of debug sections.
14218 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14220         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14221         descriptions of various incorrectly documented functions.
14223 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14225         Revert:
14226         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14227         address check not strict.
14229 2018-05-30  Richard Biener  <rguenther@suse.de>
14231         PR tree-optimization/85964
14232         * tracer.c (better_p): Drop initialized count check, we only
14233         call the function with initialized counts now.
14234         (find_best_successor): Do find a best edge if one
14235         has uninitialized count.
14236         (find_best_predecessor): Likewise.  Do BB frequency check only
14237         if count is initialized.
14239 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
14241         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14242         (aarch64_ldrstr_offset_compare): New.
14243         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
14244         load/store orderings.
14245         (aarch64_gen_adjusted_ldpstp): Likewise.
14247 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
14249         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14250         Check for subset of GENERAL_REGS and FP_REGS.
14251         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14252         r=w alternative.
14254 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
14256         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
14257         and wi::to_poly_offset.  Add the current offset and then check
14258         whether the sum fits, rather than using an unchecked addition of
14259         a checked term.  Check for a shwi rather than a uhwi.
14260         * expr.c (get_bit_range): Use tree_to_poly_uint64.
14261         (store_constructor): Use poly_int_tree_p.
14262         (expand_expr_real_1): Likewise.
14263         * function.c (assign_temp): Likewise.
14264         * fold-const.c (const_binop): Use poly_int_tree_p and
14265         wi::to_poly_offset.
14266         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
14267         division.
14268         * ipa-icf-gimple.c (func_checker::compare_operand): Use
14269         to_poly_offset for MEM offsets.
14270         * ipa-icf.c (sem_variable::equals): Likewise.
14271         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
14272         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
14273         wi::to_poly_offset for BIT_FIELD_REF offsets.
14274         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
14275         wi::to_poly_offset.
14276         * var-tracking.c (emit_note_insn_var_location): Use
14277         tree_to_poly_uint64.
14279 2018-05-29  Jim Wilson  <jimw@sifive.com>
14281         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
14283 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
14285         PR target/85950
14286         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
14287         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
14288         sequence.
14289         (sse4_1_round<mode>2): Use nonimmediate_operand
14290         for operand 1 predicate.
14292 2018-05-29  Martin Sebor  <msebor@redhat.com>
14293             Richard Biener  <rguenther@suse.de>
14295         PR testsuite/85888
14296         * calls.c (get_size_range): Call determine_value_range instead
14297         of get_value_range..
14298         * tree-vrp.h (determine_value_range): Declared new function.
14299         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
14301 2018-05-29  Richard Biener  <rguenther@suse.de>
14303         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
14304         sure to use non-pattern stmts for get_earlier_stmt arguments.
14305         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
14306         called on pattern stmts.
14307         (get_later_stmt): Likewise.
14309 2018-05-29  Martin Liska  <mliska@suse.cz>
14311         PR gcov-profile/85759
14312         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
14313         env variables.
14315 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
14317         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
14318         VEC_UNPACK_*_EXPR.
14319         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
14320         VEC_PACK_*_EXPR.
14322         PR target/85918
14323         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14324         VEC_PACK_FLOAT_EXPR): New tree codes.
14325         * tree-pretty-print.c (op_code_prio): Handle
14326         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14327         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14328         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14329         * tree-inline.c (estimate_operator_cost): Likewise.
14330         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
14331         * fold-const.c (const_binop): Likewise.
14332         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
14333         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14334         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14335         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
14336         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14337         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14338         * expr.c (expand_expr_real_2): Likewise.
14339         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
14340         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
14341         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
14342         optabs.
14343         * optabs.c (expand_widen_pattern_expr): For
14344         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
14345         sign from result type rather than operand's type.
14346         (expand_binop_directly): For vec_packu_float_optab and
14347         vec_packs_float_optab allow result type to be different from operand's
14348         type.
14349         * optabs-tree.c (optab_for_tree_code): Handle
14350         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14351         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
14352         * tree-vect-generic.c (expand_vector_operations_1):  Handle
14353         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14354         VEC_PACK_FLOAT_EXPR.
14355         * tree-vect-stmts.c (supportable_widening_operation): Handle
14356         FIX_TRUNC_EXPR.
14357         (supportable_narrowing_operation): Handle FLOAT_EXPR.
14358         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
14359         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
14360         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
14361         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
14362         mode attributes.
14363         (vec_pack<floatprefix>_float_<mode>): New expander.
14364         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
14365         attributes.
14366         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
14367         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
14368         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
14369         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
14370         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
14371         Document.
14372         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
14373         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
14374         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14375         VEC_PACK_FLOAT_EXPR): Document.
14377 2018-05-29  Richard Biener  <rguenther@suse.de>
14379         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
14380         member.
14381         (stmt_vec_info_vec): Make pointer.
14382         (init_stmt_vec_info_vec): Remove.
14383         (free_stmt_vec_info_vec): Likewise.
14384         (set_stmt_vec_info_vec): New function.
14385         (free_stmt_vec_infos): Likewise.
14386         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
14387         (set_vinfo_for_stmt): Likewise.
14388         (get_earlier_stmt): Likewise.
14389         (get_later_stmt): Likewise.
14390         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
14391         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
14392         (vec_info::~vec_info): Free stmt_vec_infos.
14393         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
14394         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
14395         (pass_slp_vectorize::execute): Likewise.
14396         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
14397         (free_stmt_vec_info_vec): Likewise.
14398         (set_stmt_vec_info_vec): New function.
14399         (free_stmt_vec_infos): Likewise.
14400         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
14401         the global stmt_vec_info_vec.
14402         * tree-parloops.c (gather_scalar_reductions): Use
14403         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
14404         vector.
14406 2018-05-29  Richard Biener  <rguenther@suse.de>
14408         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
14410 2018-05-29  Martin Liska  <mliska@suse.cz>
14411             David Malcolm  <dmalcolm@redhat.com>
14413         * vec.c (test_reverse): New.
14414         (vec_c_tests): Add new test.
14415         * vec.h (vl_ptr>::reverse): New function.
14417 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
14419         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
14421         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
14422         and later.
14424 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14426         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
14428 2018-05-28  Richard Biener  <rguenther@suse.de>
14430         PR tree-optimization/85933
14431         * tree-vect-data-refs.c (vect_record_base_alignments): Only
14432         look at stmts marked as vectorizable.
14434 2018-05-28  Richard Biener  <rguenther@suse.de>
14436         PR tree-optimization/85934
14437         * tree-vect-generic.c (expand_vector_operations_1): Hoist
14438         vector boolean check before scalar optimization.
14440 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
14442         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
14443         for armv5te.
14445 2018-05-28  Mark Wielaard  <mark@klomp.org>
14447         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
14448         if it is an expression containing a minus sign.
14450 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
14452         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
14454 2018-05-27  Paul Koning  <ni1d@arrl.net>
14456         * config/pdp11/pdp11.md (truncsihi2): Remove.
14458 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
14459             Chung-Ju Wu  <jasonwucj@gmail.com>
14461         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
14462         implementation.
14463         (unaligned_store_dw): Ditto.
14464         * config/nds32/nds32-memory-manipulation.c
14465         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
14466         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
14467         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
14468         (emit_setmem_word_loop): Rename to ...
14469         (emit_setmem_doubleword_loop): ... this.
14470         (nds32_gen_dup_4_byte_to_word_value): New function.
14471         (nds32_gen_dup_8_byte_to_double_word_value): New function.
14472         (nds32_expand_setmem_loop): Refine implementation.
14473         (nds32_expand_setmem_loop_v3m): Ditto.
14474         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
14475         pattern.
14477 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
14479         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
14481 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
14483         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
14484         (nds32_init_machine_status): Initialize machine->attr_naked_p and
14485         machine->attr_no_prologue_p.
14486         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
14487         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
14488         (nds32_expand_epilogue): Consider attr_naked_p.
14489         (nds32_expand_epilogue_v3pop): Likewise.
14490         (nds32_can_use_return_insn): Likewise.
14491         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
14492         attr_no_prologue_p fields.
14493         * config/nds32/nds32.opt (mret-in-naked-func): New option.
14495 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
14497         PR target/85918
14498         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
14499         attributes.
14500         * config/i386/sse.md
14501         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
14502         Rename to ...
14503         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
14504         ... this.
14505         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
14506         Rename to ...
14507         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
14508         ... this.
14509         (*<floatsuffix>floatv2div2sf2): Rename to ...
14510         (*float<floatunssuffix>v2div2sf2): ... this.
14511         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
14512         (float<floatunssuffix>v2div2sf2_mask): ... this.
14513         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
14514         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
14515         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
14516         to ...
14517         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
14518         ... this.
14519         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14520         Rename to ...
14521         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14522         ... this.
14523         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14524         Rename to ...
14525         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14526         ... this.
14527         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
14528         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
14529         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
14530         gen_ufix_truncv8dfv8si2.
14531         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
14532         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
14533         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
14534         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
14535         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
14536         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
14537         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
14538         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
14540 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
14542         PR target/85900
14543         PR target/85345
14544         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
14546 2018-05-25  Jim Wilson  <jimw@sifive.com>
14548         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
14549         * config/riscv/riscv.c (struct machine_function): Add
14550         interrupt_handler_p and attribute_checked_p fields.
14551         (riscv_attribute_table): Add interrupt.
14552         (riscv_interrupt_type_p): New.
14553         (riscv_save_reg_p): Save extra regs for interrupt handler.
14554         (riscv_use_save_libcall): Return false  for interrupt handler.
14555         (riscv_first_stack_step): Add forward declaration.
14556         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
14557         for interrupt handler with large frame.  Use it for saved reg list.
14558         (riscv_expand_prologue): Move flag_stack_usage_info support to
14559         eliminate duplication.
14560         (riscv_expand_epilogue): Generate mret for interrupt handler.
14561         (riscv_epilogue_uses): New.
14562         (riscv_can_use_return_insn): Return false for interrupt handler.
14563         (riscv_function_ok_for_sibcall): Likewise.
14564         (riscv_set_current_function): Add interrupt handler support.
14565         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
14566         * config/riscv/riscv.md (UNSPECV_MRET): New.
14567         (GP_REGNUM): New.
14568         (riscv_frflags, riscv_fsflags): Use tab after opcode.
14569         (riscv_mret): New.
14570         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
14572 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
14574         PR tree-optimization/85712
14575         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
14576         this candidate has already been replaced in-situ by a copy.
14578 2018-05-25  Jason Merrill  <jason@redhat.com>
14580         PR c++/80485 - inline function non-zero address.
14581         * symtab.c (nonzero_address): Check DECL_COMDAT.
14583 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
14585         PR target/83628
14586         * config/alpha/alpha.md (ashlsi3): New insn pattern.
14587         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
14588         extension of SImode operation.  Use const123_operand predicate.
14589         (*saddsi_1): Remove.
14590         (*saddl_se_1): Ditto.
14591         (*ssubsi_1): Ditto.
14592         (*ssubl_se_1): Ditto.
14593         * config/alpha/predicates.md (const123_operand): New predicate.
14594         * config/alpha/constraints.md (P): Use IN_RANGE.
14596 2018-05-25  Richard Biener  <rguenther@suse.de>
14598         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
14599         defaulted to true.
14600         (ref_maybe_used_by_stmt_p): Likewise.
14601         (stmt_may_clobber_ref_p): Likewise.
14602         (stmt_may_clobber_ref_p_1): Likewise.
14603         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
14604         and pass it along.
14605         (ref_maybe_used_by_stmt_p): Likewise.
14606         (stmt_may_clobber_ref_p): Likewise.
14607         (stmt_may_clobber_ref_p_1): Likewise.
14608         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
14609         the alias oracle to disambiguate DRs with stmts DR analysis
14610         couldn't handle.
14611         (vect_analyze_data_refs): Do not give up on not analyzable
14612         DRs for BB vectorization.  Remove code truncating the dataref
14613         vector.
14615 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
14617         PR target/85832
14618         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14619         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
14620         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
14622 2018-05-25  Richard Biener  <rguenther@suse.de>
14624         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
14625         function, combining stmt data ref gathering and fatal analysis
14626         parts.
14627         (vect_analyze_data_refs): Remove now redudnant code and simplify.
14628         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
14629         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
14630         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
14631         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
14633 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14635         PR tree-optimization/85720
14636         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
14637         SCC if all partitions are builtins.
14638         (version_loop_by_alias_check): New parameter.  Generate cancelable
14639         runtime alias check if all partitions are builtins.
14640         (distribute_loop): Update call to above function.
14642 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14644         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
14645         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
14646         (parm_default_def_partition_arg): Ditto.
14647         (set_parm_default_def_partition): Ditto.
14648         (get_parm_default_def_partitions): Ditto and make it static.
14649         (get_undefined_value_partitions): Ditto and make it static.
14650         (remove_ssa_form): Refactor call to init_var_map here.
14651         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
14652         computation for loop region.
14653         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
14654         (register_default_def): Delete.
14655         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
14656         (parm_default_def_partition_arg): Ditto.
14657         (set_parm_default_def_partition): Ditto.
14658         (get_parm_default_def_partitions): Ditto and make it static.
14659         (get_undefined_value_partitions): Ditto and make it static.
14660         (coalesce_with_default, coalesce_with_default): Update comment.
14661         (create_coalesce_list_for_region): New func factored out from
14662         create_outofssa_var_map.
14663         (populate_coalesce_list_for_outofssa): New func factored out from
14664         create_outofssa_var_map and coalesce_ssa_name.
14665         (create_outofssa_var_map): Delete.
14666         (coalesce_ssa_name): Refactor to support live range computation.
14667         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
14668         (get_parm_default_def_partitions): Delete.
14669         (get_undefined_value_partitions): Ditto.
14670         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
14671         computation for loop region.
14672         (new_tree_live_info, loe_visit_block): Ditto.
14673         (live_worklist, set_var_live_on_entry): Ditto.
14674         (calculate_live_on_exit, verify_live_on_entry): Ditto.
14675         * tree-ssa-live.h (struct _var_map): New fields.
14676         (init_var_map): Change decl.
14677         (region_contains_p): New.
14679 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
14681         * tree-ssa-live.h (live_merge_and_clear): Delete.
14683 2018-05-25  Richard Biener  <rguenther@suse.de>
14685         PR c++/85912
14686         * tree-dump.c (dequeue_and_dump): Remove access to removed
14687         operand 2 of a SWITCH_EXPR.
14689 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14691         * doc/sourcebuild.texi (vect_double_cond_arith): Include
14692         multiplication and division.
14693         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
14694         (cond_udiv@var{m}, cond_umod@var{m}): Document.
14695         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
14696         (cond_udiv_optab, cond_umod_optab): New optabs.
14697         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
14698         (IFN_COND_RDIV): New internal functions.
14699         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
14700         TRUNC_MOD_EXPR and RDIV_EXPR.
14701         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
14702         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
14703         New unspecs.
14704         (SVE_INT_BINARY): Include mult.
14705         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
14706         (optab, sve_int_op): Handle mult.
14707         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
14708         UNSPEC_COND_DIV.
14709         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
14710         for SVE_INT_BINARY_SD.
14712 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14714         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
14715         (optab, sve_int_op): Handle div and udiv.
14716         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
14717         for SVE_INT_BINARY_SD.
14718         (*<optab><mode>3): New insn for the same.
14720 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14722         * tree-vect-patterns.c: Include predict.h.
14723         (vect_recog_divmod_pattern): Restrict check for division support
14724         to when optimizing for size.
14726 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14728         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
14729         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
14730         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
14731         (gimple_match_op::set_op): Likewise.
14732         (gimple_resimplify4): Declare.
14733         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
14734         (expr::gen_transform): Likewise.
14735         (decision_tree::gen): Generate a simplification routine for 4 operands.
14736         * gimple-match-head.c (gimple_simplify): Add an overload for
14737         4 operands.  In the top-level function, handle up to 4 call
14738         arguments and call gimple_resimplify4.
14739         (gimple_resimplify4): New function.
14740         (build_call_internal): Pass a fourth operand.
14741         (maybe_push_to_seq): Likewise.
14742         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
14743         Fold VEC_COND_EXPRs of an operation and a default value into
14744         an IFN_COND_* function if possible.
14745         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
14746         New unspecs.
14747         (SVE_COND_FP_BINARY): Include them.
14748         (optab, sve_fp_op): Handle them.
14749         (SVE_INT_BINARY_REV): New code iterator.
14750         (SVE_COND_FP_BINARY_REV): New int iterator.
14751         (commutative): New int attribute.
14752         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
14753         Declare.
14754         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
14755         function.
14756         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
14757         (*cond_<optab><mode>): New patterns for reversed operands.
14759 2018-05-25  Richard Biener  <rguenther@suse.de>
14761         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
14762         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
14763         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
14764         (STMT_VINFO_GROUPED_ACCESS): Adjust.
14765         * tree-vect-data-refs.c (everywhere): Adjust users.
14766         * tree-vect-loop.c (everywhere): Likewise.
14767         * tree-vect-slp.c (everywhere): Likewise.
14768         * tree-vect-stmts.c (everywhere): Likewise.
14769         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14771 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14773         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
14774         Rename to...
14775         (gcc_cv_as_section_exclude): ... this.
14776         Try Solaris as #exclude syntax.
14777         * configure: Regenerate.
14778         * config.in: Regenerate.
14779         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
14780         SECTION_EXCLUDE.
14781         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
14782         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
14784         * varasm.c (default_elf_asm_named_section): Don't check if
14785         HAVE_GAS_SECTION_EXCLUDE is defined.
14787 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14789         * doc/md.texi: Update the documentation of the cond_* optabs
14790         to mention the new final operand.  Fix GET_MODE_NUNITS call.
14791         Describe the scalar case too.
14792         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
14793         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
14794         instead of 2.
14795         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
14796         (get_conditional_internal_fn): Update comment.
14797         * tree-vect-loop.c (vectorizable_reduction): Pass the original
14798         accumulator value as a final argument to conditional functions.
14799         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
14800         a define_expand and add an "else" operand.  Assert for now that
14801         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
14802         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
14803         (*cond_<optab><mode>): New patterns.
14804         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
14805         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14806         (UNSPEC_COND_EOR): Delete.
14807         (optab): Remove associated mappings.
14808         (SVE_INT_BINARY): New code iterator.
14809         (sve_int_op): Remove int attribute and add "minus" to the code
14810         attribute.
14811         (SVE_COND_INT_OP): Delete.
14812         (SVE_COND_FP_OP): Rename to...
14813         (SVE_COND_FP_BINARY): ...this.
14815 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
14817         * optabs.c (can_reuse_operands_p): New function.
14818         (maybe_legitimize_operands): Try to reuse the results for
14819         earlier operands.
14821 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14823         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14824         Add {q} suffix to insn mnemonic.
14826 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14828         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14829         (msp430_warn_func_return): New.
14831 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
14833         * fold-const.c (tree_nonzero_bits): New function.
14834         * fold-const.h (tree_nonzero_bits): Likewise.
14835         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14836         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14838 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
14840         PR target/85900
14841         PR target/85345
14842         * varasm.c (assemble_alias): Check ifunc_resolver only on
14843         FUNCTION_DECL.
14845 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
14847         PR target/85903
14848         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14849         when memory input operand is handled.
14851 2018-05-24  Luis Machado  <luis.machado@linaro.org>
14853         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14854         global.
14855         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14857 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14859         * match.pd: Delay FMA folds until after vectorization.
14861 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14863         PR target/83009
14864         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14865         address check not strict.
14867 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
14869         * gimple-match.h (gimple_match_op): New class.
14870         (mprts_hook): Replace parameters with a gimple_match_op *.
14871         (maybe_build_generic_op): Likewise.
14872         (gimple_simplified_result_is_gimple_val): Replace parameters with
14873         a const gimple_match_op *.
14874         (gimple_simplify): Replace code_helper * and tree * parameters with
14875         a gimple_match_op * parameter.
14876         (gimple_resimplify1): Replace code_helper *, tree and tree *
14877         parameters with a gimple_match_op * parameter.
14878         (gimple_resimplify2): Likewise.
14879         (gimple_resimplify3): Likewise.
14880         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14881         parameters with a gimple_match_op * parameter.
14882         * gimple-match-head.c (gimple_simplify): Change prototypes of
14883         auto-generated functions to take a gimple_match_op * instead of
14884         separate code_helper * and tree * parameters.  Make the same
14885         change in the top-level overload and update calls to the
14886         gimple_resimplify routines.  Update calls to the auto-generated
14887         functions and to maybe_push_res_to_seq in the publicly-facing
14888         operation-specific gimple_simplify overloads.
14889         (gimple_match_op::MAX_NUM_OPS): Define.
14890         (gimple_resimplify1): Replace rcode and ops with a single res_op
14891         parameter.  Update call to gimple_simplify.
14892         (gimple_resimplify2): Likewise.
14893         (gimple_resimplify3): Likewise.
14894         (mprts_hook): Replace parameters with a gimple_match_op *.
14895         (maybe_build_generic_op): Likewise.
14896         (build_call_internal): Replace type, nargs and ops with
14897         a gimple_match_op *.
14898         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
14899         with a single gimple_match_op *.  Update calls to mprts_hook,
14900         build_call_internal and gimple_simplified_result_is_gimple_val.
14901         Factor out code that is common to the tree_code and combined_fn cases.
14902         * genmatch.c (expr::gen_transform): Replace tem_code and
14903         tem_ops with a gimple_match_op called tem_op.  Update calls
14904         to the gimple_resimplify functions and maybe_push_res_to_seq.
14905         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
14906         res_ops.  Update call to the gimple_resimplify functions.
14907         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
14908         (decision_tree::gen): Make the functions take a gimple_match_op *
14909         called res_op instead of separate res_code and res_ops parameters.
14910         Update call accordingly.
14911         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
14912         and ops with a single res_op parameter.  Update calls to
14913         maybe_build_generic_op and maybe_push_res_to_seq.
14914         (fold_stmt_1): Update calls to gimple_simplify and
14915         replace_stmt_with_simplification.
14916         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
14917         and gimple_simplified_result_is_gimple_val.
14918         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
14919         gimple_simplify.
14920         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
14921         with a gimple_match_op *.
14922         (vn_nary_build_or_lookup): Likewise.  Update call to
14923         vn_nary_build_or_lookup_1.
14924         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
14925         gimple_match_op *.  Update calls to the gimple_resimplify routines
14926         and to gimple_simplified_result_is_gimple_val.
14927         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14928         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14929         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14930         (visit_nary_op): Likewise.
14931         (visit_reference_op_load): Likewise.
14933 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14935         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14936         modifier for printing the step amount.
14938 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
14940         PR target/78849
14941         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14942         types.
14944 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
14946         * doc/sourcebuild.texi (Endianness): New subsubsection.
14948 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14950         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14951         <prefetch_dynamic_strides>: New const bool field.
14952         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14953         prefetch_dynamic_strides.
14954         (exynosm1_prefetch_tune): Likewise.
14955         (thunderxt88_prefetch_tune): Likewise.
14956         (thunderx_prefetch_tune): Likewise.
14957         (thunderx2t99_prefetch_tune): Likewise.
14958         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14959         false.
14960         (aarch64_override_options_internal): Update to set
14961         PARAM_PREFETCH_DYNAMIC_STRIDES.
14962         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14963         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14964         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14965         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14966         prefetch-dynamic-strides setting.
14968 2018-05-23  Luis Machado  <luis.machado@linaro.org>
14970         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14971         <minimum_stride>: New const int field.
14972         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14973         minimum_stride field defaulting to -1.
14974         (exynosm1_prefetch_tune): Likewise.
14975         (thunderxt88_prefetch_tune): Likewise.
14976         (thunderx_prefetch_tune): Likewise.
14977         (thunderx2t99_prefetch_tune): Likewise.
14978         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14979         <default_opt_level>: Set to 3.
14980         (aarch64_override_options_internal): Update to set
14981         PARAM_PREFETCH_MINIMUM_STRIDE.
14982         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14983         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14984         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14985         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14986         stride is constant and is below the minimum stride threshold.
14988 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14990         * config/arm/arm-cpus.in (mode26): Delete.
14991         (armv4): Delete mode26 reference.
14992         * config/arm/arm.c (arm_configure_build_target): Delete use of
14993         isa_bit_mode26.
14995 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
14997         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14998         New insn pattern.
14999         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15000         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
15001         for non-SSE modes.
15002         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
15003         (floatunsdidf2): Ditto.
15005 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15007         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15008         (fixuns_trunc<mode>si2_avx512f): Ditto.
15009         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15010         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15011         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15013 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15015         PR rtl-optimization/79985
15016         * df-scan.c (df_insn_refs_collect): Remove special case for
15017         global registers and asm statements.
15019 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15021         * extend.texi (Global Register Variables): Rewrite the bullet list.
15022         Note that the register is available for allocation. Note that access
15023         via inline asm must use constraints. Add note about async-signal
15024         handlers. Remove paragraph about automagic register selection.
15026 2018-05-23  Richard Biener  <rguenther@suse.de>
15028         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15029         of fixed offset from memset VN.
15031 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
15033         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15034         first_interp field.
15035         (alloc_cand_and_find_basis): Initialize first_interp field.
15036         (slsr_process_mul): Modify first_interp field.
15037         (slsr_process_add): Likewise.
15038         (slsr_process_cast): Modify first_interp field for each new
15039         interpretation.
15040         (slsr_process_copy): Likewise.
15041         (dump_candidate): Dump first_interp field.
15042         (replace_mult_candidate): Process all interpretations, not just
15043         subsequent ones.
15044         (replace_rhs_if_not_dup): Likewise.
15045         (replace_one_candidate): Likewise.
15047 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
15049         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
15050         Add new boolean.
15051         (aarch64_needs_frame_chain): New function.
15052         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
15054 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
15056         PR target/84882
15057         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
15058         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15059         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15060         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15061         as true for strict-align.
15062         (aarch64_can_inline_p): Perform checks even when callee has no
15063         attributes to check for strict alignment.
15064         * doc/extend.texi (AArch64 Function Attributes): Document
15065         no-strict-align.
15066         * doc/invoke.texi: (AArch64 Options): Likewise.
15068 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
15070         PR tree-optimization/85853
15071         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15072         the handling of the root of the node to...
15073         (vect_slp_analyze_node_operations_1): ...this new function,
15074         and run the whole thing with the child nodes' def types
15075         set according to their SLP node's def type.
15077 2018-05-23  Richard Biener  <rguenther@suse.de>
15079         PR middle-end/85874
15080         * tree-data-ref.c (create_runtime_alias_checks): Defer
15081         and ignore overflow warnings.
15083 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
15085         PR tree-optimization/85822
15086         * tree-vrp.c (is_masked_range_test): Fix handling of negative
15087         constants.
15089 2018-05-23  Richard Biener  <rguenther@suse.de>
15091         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15092         memset constants via native_interpret_expr.
15094 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
15096         PR target/85345
15097         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15098         attribute.
15099         (cgraph_node::create_alias): Likewise.
15100         (cgraph_node::get_availability): Check ifunc_resolver instead
15101         of looking up ifunc attribute.
15102         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15103         * varasm.c (do_assemble_alias): Likewise.
15104         (assemble_alias): Likewise.
15105         (default_binds_local_p_3): Likewise.
15106         * cgraph.h (cgraph_node): Add ifunc_resolver.
15107         (cgraph_node::only_called_directly_or_aliased_p): Return false
15108         for IFUNC resolver.
15109         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
15110         attribute.
15111         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
15112         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
15113         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
15114         instead of looking up ifunc attribute.
15116 2018-05-22  Luis Machado  <luis.machado@linaro.org>
15118         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
15120 2018-05-22  Martin Sebor  <msebor@redhat.com>
15122         PR middle-end/85359
15123         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
15124         only when expasion succeeds.
15125         (expand_builtin_strcmp): Same.
15126         (expand_builtin_strncmp): Same.
15128 2018-05-22  Martin Sebor  <msebor@redhat.com>
15130         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
15132 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15133             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15135         * config/aarch64/aarch64-ldpstp.md: Replace uses of
15136         aarch64_mem_pair_operand with memory_operand and delete operand swapping
15137         code.
15138         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15139         Add check for legitimate_address.
15140         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
15141         (aarch64_swap_ldrstr_operands): New.
15142         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
15143         Define prototype.
15145 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15146             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15148         * config/aarch64/aarch64.md: New patterns to generate stp
15149         and ldp.
15150         (store_pair_sw, store_pair_dw): New patterns to generate stp for
15151         single words and double words.
15152         (load_pair_sw, load_pair_dw): Likewise.
15153         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
15154         Delete.
15155         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
15156         Delete.
15157         * config/aarch64/aarch64-ldpstp.md: Modify peephole
15158         for different mode ldpstp and add peephole for merged zero stores.
15159         Likewise for loads.
15160         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15161         Add size check.
15162         (aarch64_gen_store_pair): Rename calls to match new patterns.
15163         (aarch64_gen_load_pair): Rename calls to match new patterns.
15164         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
15165         (load_pair<DREG:mode><DREG2:mode>): ... This.
15166         (store_pair<mode>): Rename to...
15167         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
15168         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
15169         New mode iterators.
15170         (V_INT_EQUIV): Handle SImode.
15171         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
15172         New predicate.
15174 2018-05-22  Martin Sebor  <msebor@redhat.com>
15176         PR c/85623
15177         * calls.c (maybe_warn_nonstring_arg): Use string length to set
15178         or ajust the presumed bound on an operation to avoid unnecessary
15179         warnings.
15181 2018-05-22  Martin Sebor  <msebor@redhat.com>
15183         PR tree-optimization/85826
15184         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
15185         assuming that a DECL necesarily has a constant size.
15187 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
15189         PR middle-end/85862
15190         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
15192 2018-05-22  Richard Biener  <rguenther@suse.de>
15194         PR tree-optimization/85834
15195         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
15196         non-constant and non-zero memset arguments.
15198 2018-05-22  Martin Liska  <mliska@suse.cz>
15200         PR ipa/85607
15201         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
15203 2018-05-22  Richard Biener  <rguenther@suse.de>
15205         PR tree-optimization/85863
15206         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
15207         comparisons when vectype is specified.
15208         (vectorizable_condition): Do not specify vectype for
15209         vect_is_simple_cond when SLP vectorizing.
15211 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
15213         PR target/85657
15214         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
15215         define __ibm128 as long double.
15216         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
15217         as a distinct type when IEEE 128-bit support is enabled.
15218         (init_float128_ieee): Fix up conversions between IFmode and IEEE
15219         128-bit types to use the correct functions.
15220         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
15221         convert between 128-bit floating point types that have different
15222         modes but the same representation, instead of using gen_lowpart to
15223         makean alias.
15224         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
15225         KFmode.
15226         (IFKF_reg): New attributes to give the register constraints for
15227         IFmode and KFmode.
15228         (extend<mode>tf2_internal): New insns to mark an explicit
15229         conversion between 128-bit floating point types that have a
15230         different mode but share the same representation.
15232 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
15234         PR tree-optimization/85814
15235         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
15236         a null return from get_strinfo when unsharing the next
15237         strinfo in the chain.
15239 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
15241         PR gcc/84923
15242         * varasm.c (weak_finish): Clean up weak_decls.
15244 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15246         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
15247         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
15248         UNSPEC_UADALP values.
15249         * config/aarch64/iterators.md (ABAL): New int iterator.
15250         (ABDL2): Likewise.
15251         (ADALP): Likewise.
15252         (sur): Add mappings for the above.
15253         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
15254         New define_insn.
15255         (aarch64_<sur>abal<mode>_4): Likewise.
15256         (aarch64_<sur>adalp<mode>_3): Likewise.
15257         (<sur>sadv16qi): New define_expand.
15259 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
15261         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
15262         (*movdf_internal): Ditto.
15263         (*rcpsf2_sse): Ditto.
15264         (*rsqrtsf2_sse): Ditto.
15265         (*sqrt<mode>2_sse): Ditto.
15267 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
15269         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
15270         eor3q<mode>4.
15271         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
15272         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15273         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
15274         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
15275         vbcaxq_s64): New.
15276         * config/aarch64/arm_neon.h: Likewise.
15277         * config/aarch64/iterators.md (VQ_I): New.
15279 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
15281         * config.gcc: Add arc/t-multilib-linux to tmake_file for
15282         arc*-*-linux*.
15283         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
15284         MULTILIB_DIRNAMES
15286 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15288         * config/nds32/constraints.md (S): New constraint.
15289         * config/nds32/nds32.md (call_internal): Use constraint S.
15290         (call_value_internal): Likewise.
15291         (sibcall_internal): Likewise.
15292         (sibcall_value_internal): Likewise.
15294 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15295             Chung-Ju Wu  <jasonwucj@gmail.com>
15297         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15298         into consideration.
15300 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15301             Chung-Ju Wu  <jasonwucj@gmail.com>
15303         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
15304         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
15305         (nds32_rtx_costs_impl): Simplify.
15306         (nds32_address_cost_impl): Simplify.
15307         (nds32_init_rtx_costs): New function.
15308         (nds32_rtx_costs_speed_prefer): Likewise.
15309         (nds32_rtx_costs_size_prefer): Likewise.
15310         (nds32_address_cost_speed_prefer): Likewise.
15311         (nds32_address_cost_speed_fwprop): Likewise.
15312         (nds32_address_cost_size_prefer): Likewise.
15313         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
15314         * config/nds32/nds32.c (nds32_option_override): Use
15315         nds32_init_rtx_costs function.
15317 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15319         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
15320         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
15321         (TARGET_PIPELINE_N8): Likewise.
15322         (TARGET_PIPELINE_N10): Likewise.
15323         (TARGET_PIPELINE_N13): Likewise.
15324         (TARGET_PIPELINE_GRAYWOLF): Likewise.
15326 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
15328         * config/nds32/nds32-fpu.md: Update copyright year.
15330 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15332         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
15334 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15336         * config/nds32/nds32.c
15337         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
15338         * config/nds32/nds32.opt (minline-asm-r15): New option.
15340 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15342         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
15343         MASK_HW_ABS.
15344         * config/nds32/nds32.md (abssi2): New pattern.
15346 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
15348         * config/i386/i386.md (rex64namesuffix): New mode attribute.
15349         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
15350         Merge insn pattern from sse_cvtsi2ss<round_name> and
15351         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
15352         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
15353         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
15354         using SWI48 mode iterator.
15355         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
15356         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
15357         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
15358         pattern from sse_cvttss2si<round_saeonly_name>
15359         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
15360         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
15361         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
15362         using SWI48 mode iterator.
15363         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
15364         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
15365         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
15366         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
15367         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
15368         using SWI48 mode iterator.
15369         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
15370         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
15371         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
15372         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
15373         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
15374         SWI48 mode iterator.
15375         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
15376         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
15377         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
15378         pattern from sse_cvttsd2si<round_saeonly_name>
15379         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
15381 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15383         * config/nds32/nds32-md-auxiliary.c
15384         (nds32_valid_smw_lwm_base_p): Refine.
15385         (nds32_output_smw_single_word): Refine.
15386         (nds32_output_smw_double_word): New.
15387         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
15389 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15391         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
15392         (nds32_output_stack_pop): Refine.
15393         (nds32_expand_unaligned_load): Refine.
15394         (nds32_expand_unaligned_store): Refine.
15396 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
15397             Chung-Ju Wu  <jasonwucj@gmail.com>
15399         * config/nds32/constants.md: Add TP_REGNUM constant.
15400         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
15401         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
15402         UNSPEC_ADD32.
15403         * config/nds32/nds32-doubleword.md: Consider flag_pic.
15404         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
15405         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
15406         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
15407         and PIC code generation.
15408         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
15409         code generation.
15410         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
15411         optimization.
15412         * config/nds32/nds32.md: Support TLS and PIC.
15413         * config/nds32/nds32.c: Support TLS and PIC.
15414         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
15415         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
15416         predicate.
15418 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
15420         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
15421         mode with E_ prefix.
15423 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
15424             Chung-Ju Wu  <jasonwucj@gmail.com>
15426         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
15427         * config/nds32/nds32-md-auxiliary.c
15428         (symbolic_reference_mentioned_p): New.
15429         (nds32_legitimize_ict_address): New.
15430         (nds32_expand_ict_move): New.
15431         (nds32_indirect_call_referenced_p): New.
15432         (nds32_symbol_binds_local_p): Delete.
15433         (nds32_long_call_p): Modify.
15434         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
15435         * config/nds32/nds32-protos.h
15436         (symbolic_reference_mentioned_p): Declare.
15437         (nds32_legitimize_ict_address): Declare.
15438         (nds32_expand_ict_move): Declare.
15439         (nds32_indirect_call_referenced_p): Declare.
15440         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
15441         (nds32_relax_group): Use nds32_ict_const_p as condition.
15442         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
15443         (nds32_asm_file_start): Output ict_model directive in asm code.
15444         (nds32_legitimate_address_p): Consider indirect call.
15445         (nds32_print_operand): Consider indirect call.
15446         (nds32_print_operand_address): Consider indirect call.
15447         (nds32_insert_attributes): Handle "indirect_call" attribute.
15448         (TARGET_LEGITIMATE_ADDRESS_P): Define.
15449         (TARGET_LEGITIMATE_CONSTANT_P): Define.
15450         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
15451         (TARGET_DELEGITIMIZE_ADDRESS): Define.
15452         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
15453         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
15454         (TARGET_ICT_MODEL_SMALL): Define.
15455         (TARGET_ICT_MODEL_LARGE): Define.
15456         * config/nds32/nds32.md (movsi): Consider ict model.
15457         (call, call_value): Consider ict model.
15458         (sibcall, sibcall_value): Consider ict model.
15459         * config/nds32/nds32.opt (mict-model): New option.
15460         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
15461         model.
15463 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
15464             Monk Chiang  <sh.chiang04@gmail.com>
15465             Jim Wilson <jimw@sifive.com>
15467         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
15468         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
15469         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
15470         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
15471         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
15472         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
15473         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
15474         compute save_libcall_adjustment properly.
15475         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
15476         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
15477         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
15478         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
15479         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
15480         (ABI_SPEC): Handle mabi=ilp32e.
15481         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
15482         (RVE): Add RVE mask.
15483         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
15484         <-march>: Add rv32e as an example.
15486 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
15488         PR c++/82899
15489         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
15490         (intra_create_variable_infos): Handle C++ constructors.
15492 2018-05-18  Martin Liska  <mliska@suse.cz>
15494         * passes.def: Remove a redundant pass.
15496 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
15498         PR bootstrap/85838
15499         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
15501 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15503         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
15504         (ARMv4): Update.
15505         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
15506         (ARMv6m): Update.
15507         (armv2, armv2a, armv3, armv3m): Delete architectures.
15508         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
15509         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
15510         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
15511         Delete cpus.
15512         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
15513         (*mulsidi3adddi): Likewise.
15514         (mulsidi3): Likewise.
15515         (*mulsidi3_nov6): Likewise.
15516         (umulsidi3): Likewise.
15517         (umulsidi3_nov6): Likewise.
15518         (umaddsidi4): Likewise.
15519         (*umulsidi3adddi): Likewise.
15520         (smulsi3_highpart): Likewise.
15521         (*smulsi3_highpart_nov6): Likewise.
15522         (umulsi3_highpart): Likewise.
15523         (*umulsi3_highpart_nov6): Likewise.
15524         * config/arm/arm.h (arm_arch3m): Delete.
15525         * config/arm/arm.c (arm_arch3m): Delete.
15526         (arm_option_override_internal): Update armv3-related comment.
15527         (arm_configure_build_target): Delete use of isa_bit_mode32.
15528         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
15529         (arm_rtx_costs_internal): Delete check of arm_arch3m.
15530         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
15531         (mulsa3): Likewise.
15532         (mulusa3): Likewise.
15533         * config/arm/arm-protos.h (arm_arch3m): Delete.
15534         * config/arm/arm-tables.opt: Regenerate.
15535         * config/arm/arm-tune.md: Likewise.
15536         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
15537         deleted architectures.
15539 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15541         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
15542         (armv5t, armv5te): New features.
15543         (ARMv5, ARMv5e): Delete fgroups.
15544         (ARMv5t, ARMv5te): Adjust for above changes.
15545         (ARMv6m): Likewise.
15546         (armv5, armv5e): Delete arches.
15547         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
15548         arm_arch5.
15549         (*call_reg_arm): Likewise.
15550         (*call_value_reg_armv5): Likewise.
15551         (*call_value_reg_arm): Likewise.
15552         (*call_symbol): Likewise.
15553         (*call_value_symbol): Likewise.
15554         (*sibcall_insn): Likewise.
15555         (*sibcall_value_insn): Likewise.
15556         (clzsi2): Likewise.
15557         (prefetch): Likewise.
15558         (define_split and define_peephole2 dependent on arm_arch5):
15559         Likewise.
15560         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
15561         arm_arch5e.
15562         (TARGET_ARM_QBIT): Likewise.
15563         (TARGET_DSP_MULTIPLY): Likewise.
15564         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
15565         (arm_arch5, arm_arch5e): Delete.
15566         (arm_arch5t, arm_arch5te): Declare.
15567         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
15568         (arm_arch5t): Declare.
15569         (arm_option_reconfigure_globals): Update for the above.
15570         (arm_options_perform_arch_sanity_checks): Update comment, replace
15571         use of arm_arch5 with arm_arch5t.
15572         (use_return_insn): Likewise.
15573         (arm_emit_call_insn): Likewise.
15574         (output_return_instruction): Likewise.
15575         (arm_final_prescan_insn): Likewise.
15576         (arm_coproc_builtin_available): Likewise.
15577         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
15578         arm_arch5e with arm_arch5t and arm_arch5te.
15579         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
15580         (arm_arch5t, arm_arch5te): Declare.
15581         * config/arm/arm-tables.opt: Regenerate.
15582         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
15583         * config/arm/t-multilib: Likewise.
15584         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
15585         instead of arm_arch5.
15586         (*call_reg_thumb1): Likewise.
15587         (*call_value_reg_thumb1_v5): Likewise.
15588         (*call_value_reg_thumb1): Likewise.
15589         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
15590         unreachable path.
15591         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
15593 2018-05-18  Martin Liska  <mliska@suse.cz>
15595         PR gcov-profile/84846
15596         * doc/gcov.texi: Document -t option of gcov tool.
15598 2018-05-18  Martin Liska  <mliska@suse.cz>
15600         PR gcov-profile/84846
15601         * gcov.c (print_usage): Add new -t option.
15602         (process_args): Handle the option.
15603         (generate_results): Use stdout as output when requested by
15604         the option.
15606 2018-05-18  Martin Liska  <mliska@suse.cz>
15608         PR gcov-profile/84846
15609         * coverage.c (coverage_init): Write PWD to .gcno file.
15610         * doc/gcov.texi: Document how working directory is printed.
15611         * gcov-dump.c (dump_gcov_file): Print PWD.
15612         * gcov.c (output_intermediate_file): Likewise.
15613         (read_graph_file): Read PWD string.
15614         (output_lines): Print PWD.
15616 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15618         PR middle-end/85817
15619         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
15620         for retval and return false if all args to phi are zero.
15622 2018-05-18  Richard Biener  <rguenther@suse.de>
15624         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
15625         method.
15626         (evrp_dom_walker::before_dom_children): Call it.
15628 2018-05-18  Richard Biener  <rguenther@suse.de>
15630         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
15631         results when processing array refs with variable index.
15633 2018-05-18  Toon Moene  <toon@moene.org>
15635         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
15636         directly after that of -floop-interchange. Indicate that both
15637         options are enabled by default when specifying -O3.
15639 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15641         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
15642         iterator.  Delete separate integer-mode vec_set<mode> expander.
15643         (aarch64_simd_vec_setv2di): Delete.
15644         (vec_setv2di): Delete.
15645         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
15646         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
15647         the "w, r" alternative.
15649 2018-05-18  Martin Liska  <mliska@suse.cz>
15651         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
15652         * tree-pass.h (make_pass_lower_switch_O0): New function.
15653         * tree-switch-conversion.c (node_has_low_bound): Remove.
15654         (node_has_high_bound): Likewise.
15655         (node_is_bounded): Likewise.
15656         (class pass_lower_switch): Make it a template type and create
15657         two instances.
15658         (pass_lower_switch::execute): Add template argument.
15659         (make_pass_lower_switch): New function.
15660         (make_pass_lower_switch_O0): New function.
15661         (do_jump_if_equal): Remove.
15662         (emit_case_nodes): Simplify to just handle all 3 cases and leave
15663         all the hard work to tree optimization passes.
15665 2018-05-18  Martin Liska  <mliska@suse.cz>
15667         * dbgcnt.c (limit_low): Renamed from limit.
15668         (limit_high): New variable.
15669         (dbg_cnt_is_enabled): Check for upper limit.
15670         (dbg_cnt): Adjust dumping.
15671         (dbg_cnt_set_limit_by_index): Add new argument for high
15672         value.
15673         (dbg_cnt_set_limit_by_name): Likewise.
15674         (dbg_cnt_process_single_pair): Parse new format.
15675         (dbg_cnt_process_opt): Use strtok.
15676         (dbg_cnt_list_all_counters): Remove 'value' and add
15677         'limit_high'.
15678         * doc/invoke.texi: Document changes.
15680 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
15682         * doc/sourcebuild.texi (scalar_all_fma): Document.
15683         * tree.def (FMA_EXPR): Delete.
15684         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
15685         * internal-fn.c (ternary_direct): New macro.
15686         (expand_ternary_optab_fn): Likewise.
15687         (direct_ternary_optab_supported_p): Likewise.
15688         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
15689         * builtins.c (fold_builtin_fma): Delete.
15690         (fold_builtin_3): Don't call it.
15691         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
15692         * expr.c (expand_expr_real_2): Likewise.
15693         * fold-const.c (operand_equal_p): Likewise.
15694         (fold_ternary_loc): Likewise.
15695         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
15696         * gimple.c (DEFTREECODE): Likewise.
15697         * gimplify.c (gimplify_expr): Likewise.
15698         * optabs-tree.c (optab_for_tree_code): Likewise.
15699         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
15700         * tree-eh.c (operation_could_trap_p): Likewise.
15701         (stmt_could_throw_1_p): Likewise.
15702         * tree-inline.c (estimate_operator_cost): Likewise.
15703         * tree-pretty-print.c (dump_generic_node): Likewise.
15704         (op_code_prio): Likewise.
15705         * tree-ssa-loop-im.c (stmt_cost): Likewise.
15706         * tree-ssa-operands.c (get_expr_operands): Likewise.
15707         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
15708         * fold-const-call.h (fold_fma): Delete.
15709         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
15710         CFN_FNMA and CFN_FNMS.
15711         (fold_fma): Delete.
15712         * genmatch.c (combined_fn): New enum.
15713         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
15714         (commutative_op): New function.
15715         (commutate): Use it.  Handle more than 2 operands.
15716         (dt_operand::gen_gimple_expr): Use commutative_op.
15717         (parser::parse_expr): Allow :c to be used with non-binary
15718         operators if the commutative operand is known.
15719         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
15720         CFN_FMS, CFN_FNMA and CFN_FNMS.
15721         (backprop::process_assign_use): Remove FMA_EXPR handling.
15722         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
15723         (gen_hsa_fma): New function.
15724         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
15725         IFN_FNMA and IFN_FNMS.
15726         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
15727         * gimple-fold.h (follow_all_ssa_edges): Declare.
15728         * gimple-fold.c (follow_all_ssa_edges): New function.
15729         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
15730         gimple_build interface and use follow_all_ssa_edges to fold the result.
15731         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
15732         instead of checking for optabs directly.
15733         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
15734         rather than FMA_EXPRs.
15735         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
15736         call to IFN_FMA instead of an FMA_EXPR.
15738 2018-05-17  Jim Wilson  <jimw@sifive.com>
15740         * expr.c (do_tablejump): When converting index to Pmode, if we have a
15741         sign extended promoted subreg, and the range does not have the sign bit
15742         set, then do a sign extend.
15744         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
15745         test, check for sign extended subreg and/or constant operands, and
15746         do a sign extend in that case.
15748 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
15750         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
15751         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
15752         Add untyped.
15753         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
15754         Change logics_shift_reg to logics_shift_imm.
15755         (thunderx2t99_fp_loadpair_basic): Delete.
15756         (thunderx2t99_fp_storepair_basic): Delete.
15757         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
15758         (thunderx2t99_asimd_polynomial): Delete.
15759         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
15760         and neon_fp_mul_d_scalar_q.
15761         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15762         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
15763         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
15764         (thunderx2t99_asimd_lut): Add missing tbl types.
15765         (thunderx2t99_asimd_ext): Delete.
15766         (thunderx2t99_asimd_load1_1_mult): Delete.
15767         (thunderx2t99_asimd_load1_2_mult): Delete.
15768         (thunderx2t99_asimd_load1_ldp): New.
15769         (thunderx2t99_asimd_load1): New.
15770         (thunderx2t99_asimd_load2): Add missing *load2* types.
15771         (thunderx2t99_asimd_load3): New.
15772         (thunderx2t99_asimd_load4): New.
15773         (thunderx2t99_asimd_store1_1_mult): Delete.
15774         (thunderx2t99_asimd_store1_2_mult): Delete.
15775         (thunderx2t99_asimd_store2_mult): Delete.
15776         (thunderx2t99_asimd_store2_onelane): Delete.
15777         (thunderx2t99_asimd_store_stp): New.
15778         (thunderx2t99_asimd_store1): New.
15779         (thunderx2t99_asimd_store2): New.
15780         (thunderx2t99_asimd_store3): New.
15781         (thunderx2t99_asimd_store4): New.
15783 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
15785         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
15786         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
15788 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
15789             Segher Boessenkool  <segher@kernel.crashing.org>
15791         PR target/85698
15792         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
15793         operand.
15795 2018-05-17  Richard Biener  <rguenther@suse.de>
15797         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
15798         for pruning loop and prune defs feeding only already visited PHIs.
15800 2018-05-17  Richard Biener  <rguenther@suse.de>
15802         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
15804 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
15805             Richard Biener  <rguenther@suse.de>
15807         PR tree-optimization/85793
15808         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15809         for VMAT_ELEMENTWISE.
15811 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
15813         * internal-fn.h (lookup_internal_fn): Declare
15814         * internal-fn.c (lookup_internal_fn): New function.
15815         * gimple.c (gimple_build_call_from_tree): Handle calls to
15816         internal functions.
15817         * gimple-pretty-print.c (dump_gimple_call): Print "." before
15818         internal function names.
15819         * tree-pretty-print.c (dump_generic_node): Likewise.
15820         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15822 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
15824         * gimple-fold.h (gimple_build): Make the function forms take
15825         combined_fn rather than built_in_function.
15826         (gimple_simplify): Likewise.
15827         * gimple-match-head.c (gimple_simplify): Likewise.
15828         * gimple-fold.c (gimple_build): Likewise.
15829         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15830         rather than gimple_build_call_internal.
15831         (get_initial_defs_for_reduction): Likewise.
15832         (vect_create_epilog_for_reduction): Likewise.
15833         (vectorizable_live_operation): Likewise.
15835 2018-05-17  Martin Liska  <mliska@suse.cz>
15837         * gimple-ssa-sprintf.c (format_directive): Do not use
15838         space in between 'G_' and '('.
15840 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
15842         PR target/85323
15843         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15844         even if the mask is not all ones.
15846         PR target/85323
15847         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15848         vector.
15849         (ix86_gimple_fold_builtin): Likewise.
15851         PR target/85323
15852         * config/i386/i386.c: Include tree-vector-builder.h.
15853         (ix86_vector_shift_count): New function.
15854         (ix86_fold_builtin): Fold shift builtins by scalar count.
15855         (ix86_gimple_fold_builtin): Likewise.
15857         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15858         _mm512_setzero): New intrinsics.
15860 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
15861             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15863         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15864         code generation for cases where splatting a value is not useful.
15865         * simplify-rtx.c (simplify_ternary_operation): Simplify
15866         vec_merge across a vec_duplicate and a paradoxical subreg forming
15867         a vector mode to a vec_concat.
15869 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
15871         * config.gcc: Support "goldmont-plus".
15872         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15873         "goldmont-plus".
15874         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15875         PROCESSOR_GOLDMONT_PLUS.
15876         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15877         (processor_target_table): Add "goldmont-plus".
15878         (PTA_GOLDMONT_PLUS): Define.
15879         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15880         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15881         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15882         (fold_builtin_cpu): Add "goldmont-plus".
15883         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15884         (ix86_option_override_internal): Add "goldmont-plus".
15885         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15886         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15887         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15888         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15890 2018-05-17  Richard Biener  <rguenther@suse.de>
15892         PR tree-optimization/85757
15893         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
15894         remove defs that only feed that PHI from further processing.
15896 2018-05-16  Jim Wilson  <jimw@sifive.com>
15898         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
15899         asterisk to name.
15900         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
15902 2018-05-16  Mark Wielaard  <mark@klomp.org>
15904         * dwarf2out.c (count_index_strings): New function.
15905         (output_indirect_strings): Call count_index_strings and generate
15906         header for dwarf_version >= 5.
15908 2018-05-16  Mark Wielaard  <mark@klomp.org>
15910         * dwarf2out.c (dwarf_FORM): New function.
15911         (set_indirect_string): Use dwarf_FORM.
15912         (reset_indirect_string): Likewise.
15913         (size_of_die): Likewise.
15914         (value_format): Likewise.
15915         (output_die): Likewise.
15916         (add_skeleton_AT_string): Likewise.
15917         (output_macinfo_op): Likewise.
15918         (index_string): Likewise.
15919         (output_index_string_offset): Likewise.
15920         (output_index_string): Likewise.
15921         (count_index_strings): Likewise.
15923 2018-05-16  Carl Love  <cel@us.ibm.com>
15925         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15926         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15928 2018-05-16  Martin Jambor  <mjambor@suse.cz>
15930         * ipa-prop.c (ipa_free_all_edge_args): Remove.
15931         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15933 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
15935         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15936         (fnma<mode>4): Likewise.
15937         (fms<mode>4): Likewise.
15938         (fnms<mode>4): Likewise.
15939         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15940         (aarch64_fnma<mode>4): Likewise.
15941         (aarch64_fms<mode>4): Likewise.
15942         (aarch64_fnms<mode>4): Likewise.
15943         (aarch64_fnmadd<mode>4): Likewise.
15945 2018-05-16  Jason Merrill  <jason@redhat.com>
15947         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
15949 2018-05-16  Richard Biener  <rguenther@suse.de>
15951         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15952         (dump_stmt_cost): Declare.
15953         (add_stmt_cost): Dump cost we add.
15954         (add_stmt_costs): New function.
15955         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15956         No longer exported.
15957         (vect_analyze_stmt): Adjust prototype.
15958         (vectorizable_condition): Likewise.
15959         (vectorizable_live_operation): Likewise.
15960         (vectorizable_reduction): Likewise.
15961         (vectorizable_induction): Likewise.
15962         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15963         cost vector to pass to vectorizable_ and record afterwards.
15964         (vect_model_reduction_cost): Take cost vector argument and adjust.
15965         (vect_model_induction_cost): Likewise.
15966         (vectorizable_reduction): Likewise.
15967         (vectorizable_induction): Likewise.
15968         (vectorizable_live_operation): Likewise.
15969         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15970         SLP_TREE_NUMBER_OF_VEC_STMTS.
15971         (vect_analyze_slp_cost_1): Remove.
15972         (vect_analyze_slp_cost): Likewise.
15973         (vect_slp_analyze_node_operations): Take visited args and
15974         a target cost vector.  Avoid processing already visited stmt sets.
15975         (vect_slp_analyze_operations): Use a local cost vector to gather
15976         costs and register those of non-discarded instances.
15977         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15978         (vect_schedule_slp_instance): Remove copying of
15979         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
15980         zero.
15981         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15982         adding cost.  Record cost entry location.
15983         (vect_prologue_cost_for_slp_op): Function to compute cost of
15984         a constant or invariant generated for SLP vect in the prologue,
15985         split out from vect_analyze_slp_cost_1.
15986         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
15987         (vect_model_promotion_demotion_cost): Likewise.
15988         (vect_model_store_cost): Likewise, make static.
15989         (vect_model_load_cost): Likewise.
15990         (vectorizable_bswap): Add cost vector arg and adjust.
15991         (vectorizable_call): Likewise.
15992         (vectorizable_simd_clone_call): Likewise.
15993         (vectorizable_conversion): Likewise.
15994         (vectorizable_assignment): Likewise.
15995         (vectorizable_shift): Likewise.
15996         (vectorizable_operation): Likewise.
15997         (vectorizable_store): Likewise.
15998         (vectorizable_load): Likewise.
15999         (vectorizable_condition): Likewise.
16000         (vectorizable_comparison): Likewise.
16001         (can_vectorize_live_stmts): Likewise.
16002         (vect_analyze_stmt): Likewise.
16003         (vect_transform_stmt): Adjust calls to vectorizable_*.
16004         * tree-vectorizer.c: Include gimple-pretty-print.h.
16005         (dump_stmt_cost): New function.
16007 2018-05-16  Richard Biener  <rguenther@suse.de>
16009         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16010         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16011         * tree-ssa-dse.c: Include tree-ssa-loop.h.
16012         (check_name): New callback.
16013         (dse_classify_store): Track cycles via a visited bitmap of PHI
16014         defs and simplify handling of in-loop and across loop dead stores
16015         and properly fail for loop-variant refs.  Handle byte-tracking with
16016         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16017         limiting the walk.
16019 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
16021         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16022         (vect_get_mask_type_for_stmt): Likewise.
16023         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16024         split out from...
16025         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
16026         to determine the statement's vector type and the vector type that
16027         should be used for calculating nunits.  Deal with cases in which
16028         the type has to be deferred.
16029         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16030         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16031         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16032         (vect_determine_vf_for_stmt): New functions, split out from...
16033         (vect_determine_vectorization_factor): ...here.
16034         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16035         (vect_get_mask_type_for_stmt): New functions, split out from
16036         vect_determine_vectorization_factor.
16038 2018-05-16  Richard Biener  <rguenther@suse.de>
16040         * tree-cfg.c (verify_gimple_assign_ternary): Properly
16041         verify the [VEC_]COND_EXPR embedded comparison.
16043 2018-05-15  Martin Sebor  <msebor@redhat.com>
16045         PR tree-optimization/85753
16046         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
16047         RECORD_TYPE in addition to ARRAY_TYPE.
16049 2018-05-15  Martin Sebor  <msebor@redhat.com>
16051         PR middle-end/85643
16052         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
16054 2018-05-15  Richard Biener  <rguenther@suse.de>
16056         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
16057         add by_clobber_p one.  Change algorithm to collect all defs
16058         representing uses we need to walk and try reducing them to
16059         a single one before failing.
16060         (dse_dom_walker::dse_optimize_stmt): Adjust.
16062 2018-05-13  Mark Wielaard  <mark@klomp.org>
16064         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16065         (size_of_loc_descr): Likewise.
16066         (output_loc_operands): Likewise.
16067         (output_loc_operands_raw): Likewise.
16068         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16069         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16070         (hash_loc_operands): Likewise.
16071         (compare_loc_operands): Likewise.
16073 2018-05-14  Mark Wielaard  <mark@klomp.org>
16075         * dwarf2out.c (count_index_addrs): New function.
16076         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16078 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16080         PR tree-optimization/83648
16081         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16082         return value as malloc candidate.
16084 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16086         PR ipa/85734
16087         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16088         param as true in call to suggest_attribute.
16090 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
16092         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16093         -mreadonly-in-sdata.
16095 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16097         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16098         New pattern.
16099         (aarch64_crypto_aesd_fused): Likewise.
16101 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
16103         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
16104         (movsi_aarch64): Likewise.
16105         (load_pairsi): Likewise.
16106         (load_pairdi): Likewise.
16107         (store_pairsi): Likewise.
16108         (store_pairdi): Likewise.
16109         (load_pairsf): Likewise.
16110         (load_pairdf): Likewise.
16111         (store_pairsf): Likewise.
16112         (store_pairdf): Likewise.
16113         (zero_extend): Likewise.
16114         (trunc): Swap alternatives.
16115         (fcvt_target): Add '?' to prefer w over r.
16117 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
16119         PR target/85756
16120         * config/i386/i386.md: Disallow non-commutative arithmetics in
16121         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
16122         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
16123         in the peephole2 before it.
16125 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
16127         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
16128         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
16129         (ix86_handle_option): Handle -mcldemote.
16130         * config.gcc: New header.
16131         * config/i386/cldemoteintrin.h: New file.
16132         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
16133         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16134         -mcldemote.
16135         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16136         OPTION_MASK_ISA_CLDEMOTE.
16137         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
16138         (ix86_valid_target_attribute_inner_p): Ditto.
16139         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
16140         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
16141         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
16142         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
16143         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
16144         (cldemote): New.
16145         * config/i386/i386.opt: Add -mcldemote.
16146         * config/i386/x86intrin.h: New header.
16147         * doc/invoke.texi: Add -mcldemote.
16149 2018-05-14  Richard Biener  <rguenther@suse.de>
16151         * doc/match-and-simplify.texi: Adjust :s documentation.
16153 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
16155         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
16156         intended memcpy size.
16157         (REORDER_45): Likewise.
16159 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
16161         * sort.cc: New file.
16162         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
16163         * vec.c (qsort_chk): Use gcc_qsort.
16164         * Makefile.in (OBJS-libcommon): Add sort.o.
16165         (build/sort.o): New target.  Use it...
16166         (BUILD_RTL): ... here, and...
16167         (build/gencfn-macros): ... here, and...
16168         (build/genmatch): ... here.
16170 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16171             Chung-Ju Wu  <jasonwucj@gmail.com>
16173         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
16174         * config/nds32/nds32-graywolf.md: New file.
16175         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
16176         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
16177         pipeline.
16178         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
16179         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
16180         * config/nds32/nds32.md (pipeline_model): Add graywolf.
16181         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
16182         * config/nds32/pipelines.md: Include n15 settings.
16184 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16185             Chung-Ju Wu  <jasonwucj@gmail.com>
16187         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
16188         * config/nds32/nds32-n13.md: New file.
16189         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
16190         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
16191         pipeline.
16192         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
16193         * config/nds32/nds32.md (pipeline_model): Add n13.
16194         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
16195         * config/nds32/pipelines.md: Include n13 settings.
16197 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16198             Chung-Ju Wu  <jasonwucj@gmail.com>
16200         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
16201         * config/nds32/nds32-n10.md: New file.
16202         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
16203         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
16204         pipeline.
16205         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
16206         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
16207         * config/nds32/nds32.md (pipeline_model): Add n10.
16208         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
16209         * config/nds32/pipelines.md: Include n10 settings.
16211 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
16212             Kito Cheng  <kito.cheng@gmail.com>
16213             Chung-Ju Wu  <jasonwucj@gmail.com>
16215         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
16216         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16217         Add enum values for DSP extension instructions.
16218         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
16219         New constraints.
16220         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
16221         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
16222         New code iterators.
16223         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
16224         * config/nds32/nds32-dspext.md: New file for DSP implementation.
16225         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
16226         * config/nds32/nds32-intrinsic.md: Likewise.
16227         * config/nds32/nds32_intrinsic.h: Likewise.
16228         * config/nds32/nds32-md-auxiliary.c: Likewise.
16229         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
16230         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
16231         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
16232         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
16233         * config/nds32/nds32-protos.h: New declarations for DSP extension.
16234         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
16235         TYPE_DMAC in switch statement.
16236         * config/nds32/nds32.c: New checking and implementation for DSP
16237         extension instructions.
16238         * config/nds32/nds32.h: Likewise.
16239         * config/nds32/nds32.md: Likewise.
16240         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
16241         * config/nds32/predicates.md: Implement new predicates for DSP
16242         extension.
16244 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
16246         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
16247         Reformat alternatives and attributes so it is easier to identify
16248         which constraints/attributes go with which instruction.
16249         (mov<mode>_hardfloat32, FMOVE64): Likewise.
16250         (mov<mode>_softfloat32, FMOVE64): Likewise.
16251         (mov<mode>_hardfloat64, FMOVE64): Likewise.
16252         (mov<mode>_softfloat64, FMOVE64): Likewise.
16254 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16256         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16257         subsection.
16258         (Basic PowerPC Built-in Functions): The new name of the
16259         subsection previously known as "PowerPC Built-in Functions".
16260         (Basic PowerPC Built-in Functions Available on all Configurations):
16261         New subsubsection.
16262         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
16263         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
16264         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
16265         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
16267 2018-05-11  Martin Jambor  <mjambor@suse.cz>
16269         PR ipa/85655
16270         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
16271         single const.
16273 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16275         PR target/85733
16276         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
16278 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
16280         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
16281         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
16282         (ix86_handle_option): Handle -mwaitpkg.
16283         * config.gcc: New header.
16284         * config/i386/cpuid.h (bit_WAITPKG): New bit.
16285         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
16286         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
16287         function type.
16288         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16289         OPTION_MASK_ISA_WAITPKG.
16290         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
16291         (ix86_option_override_internal): Add PTA_WAITPKG.
16292         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
16293         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
16294         IX86_BUILTIN_TPAUSE.
16295         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
16296         __builtin_ia32_umwait and __builtin_ia32_tpause.
16297         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
16298         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
16299         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
16300         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
16301         UNSPECV_TPAUSE): New.
16302         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
16303         * config/i386/i386.opt: Add -mwaitpkg.
16304         * config/i386/waitpkgintrin.h: New file.
16305         * config/i386/x86intrin.h: New header.
16306         * doc/invoke.texi: Add -mwaitpkg.
16308 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16310         PR target/85606
16311         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
16312         equivalent.
16313         (cortex-m0): Use armv6s-m isa.
16314         (cortex-m0plus): Likewise.
16315         (cortex-m1): Likewise.
16316         (cortex-m0.small-multiply): Likewise.
16317         (cortex-m0plus.small-multiply): Likewise.
16318         (cortex-m1.small-multiply): Likewise.
16320 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
16321             Jakub Jelinek  <jakub@redhat.com>
16323         PR tree-optimization/85692
16324         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
16325         source permute as well.
16327 2018-05-11  Martin Liska  <mliska@suse.cz>
16329         PR sanitizer/85556
16330         * doc/extend.texi: Document LLVM style format for no_sanitize
16331         attribute.
16333 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
16335         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
16336         mode_supports_vsx_dform_quad to mode_supports_dq_form.
16337         (mode_supports_vsx_dform_quad): Likewise.
16338         (mode_supports_vmx_dform): Move these functions to be next to the
16339         other mode_supports functions.
16340         (mode_supports_dq_form): Likewise.
16341         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
16342         mode_supports_dq_form.
16343         (reg_offset_addressing_ok_p): Likewise.
16344         (offsettable_ok_by_alignment): Likewise.
16345         (rs6000_legitimate_offset_address_p): Likewise.
16346         (legitimate_lo_sum_address_p): Likewise.
16347         (rs6000_legitimize_address): Likewise.
16348         (rs6000_legitimize_reload_address): Likewise.
16349         (rs6000_secondary_reload_inner): Likewise.
16350         (rs6000_preferred_reload_class): Likewise.
16351         (rs6000_output_move_128bit): Likewise.
16353 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
16355         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
16356         Generate SImode target register for null target.
16357         <case IX86_BUILTIN_XGETBV>: Ditto.
16358         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
16359         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
16361 2018-05-10  Carl Love  <cel@us.ibm.com>
16363         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16364         dcbtt and dcbtstt if operands[2] is 0.
16366 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
16368         PR target/85693
16369         * config/i386/sse.md (usadv64qi): New expander.
16371 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
16373         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
16374         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
16375         -maltivec=be support.
16376         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
16377         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
16378         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
16379         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
16380         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
16381         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
16382         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
16383         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
16384         altivec_vsumsws): Adjust.
16385         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
16386         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
16387         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
16388         support.
16389         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
16390         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
16391         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
16392         (altivec_lve<VI_char>x): Delete expand.
16393         (*altivec_lve<VI_char>x_internal): Rename to...
16394         (altivec_lve<VI_char>x): ... this.
16395         (altivec_lvxl_<mode>): Delete expand.
16396         (*altivec_lvxl_<mode>_internal): Rename to ...
16397         (altivec_lvxl_<mode>): ... this.
16398         (altivec_stvxl_<mode>): Delete expand.
16399         (*altivec_stvxl_<mode>_internal): Rename to ...
16400         (altivec_stvxl_<mode>): ... this.
16401         (altivec_stve<VI_char>x): Delete expand.
16402         (*altivec_stve<VI_char>x_internal): Rename to ...
16403         (altivec_stve<VI_char>x): ... this.
16404         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
16405         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
16406         reduc_plus_scal_<mode>): Adjust.
16407         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
16408         comment.
16409         (rs6000_cpu_cpp_builtins): Adjust.
16410         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
16411         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
16412         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
16413         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
16414         -maltivec=be support.
16415         (rs6000_split_vec_extract_var): Adjust.
16416         (rs6000_split_v4si_init): Adjust.
16417         (swap_selector_for_mode): Delete.
16418         (altivec_expand_lvx_be, altivec_expand_stvx_be,
16419         altivec_expand_stvex_be): Delete.
16420         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
16421         -maltivec=be support.
16422         (rs6000_gimple_fold_builtin): Ditto.
16423         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
16424         Adjust.
16425         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
16426         (TARGET_DIRECT_MOVE_64BIT): Adjust.
16427         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
16428         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
16429         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
16430         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
16431         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
16432         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
16433         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
16434         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
16435         anonymous split): Adjust.
16436         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
16437         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
16439 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
16441         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
16442         when --with-gxx-include-dir is also specified.
16443         * configure: Regenerate.
16445 2018-05-09  Jim Wilson  <jimw@sifive.com>
16447         PR target/84797
16448         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
16449         * config/riscv/t-withmultilib: New.
16450         * config/riscv/withmultilib.h: New.
16451         * doc/install.texi: Document RISC-V --with-multilib-list support.
16453 2018-05-09  Richard Biener  <rguenther@suse.de>
16455         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
16456         vector.
16457         (vect_bb_vectorization_profitable_p): Adjust.  Compute
16458         actual scalar cost using the cost vector and the add_stmt_cost
16459         machinery.
16461 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16463         PR rtl-optimization/85645
16464         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
16465         in the REG_CFA_REGISTER note for LR, don't leave it empty.
16467 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16469         PR rtl-optimization/85645
16470         * shrink-wrap.c (spread_components): Return a boolean saying if
16471         anything was changed.
16472         (try_shrink_wrapping_separate): Iterate spread_components until
16473         nothing changes anymore.
16475 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16477         PR rtl-optimization/85645
16478         * regrename.c (build_def_use): Also kill the chains that include the
16479         destination of a REG_CFA_REGISTER note.
16481 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
16483         PR rtl-optimization/85645
16484         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
16485         insn that has a REG_CFA_REGISTER note.
16487 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
16489         * cfgexpand.c (expand_clobber): New function.
16490         (expand_gimple_stmt_1): Use it.
16491         * tree-vect-stmts.c (vect_clobber_variable): New function,
16492         split out from...
16493         (vectorizable_simd_clone_call): ...here.
16494         (vectorizable_store): Emit a clobber either side of an
16495         IFN_STORE_LANES sequence.
16496         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
16498 2018-05-09  Tom de Vries  <tom@codesourcery.com>
16500         PR target/85626
16501         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
16502         (define_insn "trap_if_false"): Add exit after trap.
16504 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
16506         PR rtl-optimization/85638
16507         * bb-reorder.c: Include common/common-target.h.
16508         (create_forwarder_block): New function extracted from...
16509         (fix_up_crossing_landing_pad): ...here.  Rename into...
16510         (dw2_fix_up_crossing_landing_pad): ...this.
16511         (sjlj_fix_up_crossing_landing_pad): New function.
16512         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
16513         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
16514         from both partitions and exit the loop after one iteration.
16516 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16518         Revert:
16519         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16520         subsection.
16521         (Basic PowerPC Built-in Functions): The new name of the
16522         subsection previously known as "PowerPC Built-in Functions".
16523         (Basic PowerPC Built-in Functions Available on all Configurations):
16524         New subsubsection.
16525         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16526         subsubsection.
16527         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16528         subsubsection.
16529         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16530         subsubsection.
16531         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16532         subsubsection.
16534 2018-05-08  Jim Wilson  <jimw@sifive.com>
16536         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
16537         (LD_EMUL_SUFFIX): New.
16538         (LINK_SPEC): Use it.
16540 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16542         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16543         subsection.
16544         (Basic PowerPC Built-in Functions): The new name of the
16545         subsection previously known as "PowerPC Built-in Functions".
16546         (Basic PowerPC Built-in Functions Available on all Configurations):
16547         New subsubsection.
16548         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16549         subsubsection.
16550         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16551         subsubsection.
16552         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16553         subsubsection.
16554         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16555         subsubsection.
16557 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
16559         PR target/85683
16560         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
16561         after cmpelim optimization.
16563 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
16565         * config.gcc: Support "goldmont".
16566         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
16567         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16568         PROCESSOR_GOLDMONT.
16569         * config/i386/i386.c (m_GOLDMONT): Define.
16570         (processor_target_table): Add "goldmont".
16571         (PTA_GOLDMONT): Define.
16572         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
16573         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
16574         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
16575         (fold_builtin_cpu): Add "goldmont".
16576         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
16577         (ix86_option_override_internal): Add "goldmont".
16578         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
16579         (processor_type): Add PROCESSOR_GOLDMONT.
16580         * config/i386/i386.md: Add CPU "glm".
16581         * config/i386/glm.md: New file.
16582         * config/i386/x86-tune.def: Add m_GOLDMONT.
16583         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
16585 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
16587         PR target/85572
16588         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
16589         E_V4DImode.
16590         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
16591         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
16592         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
16594         PR target/85317
16595         * config/i386/i386.c (ix86_fold_builtin): Handle
16596         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
16598         PR target/85480
16599         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
16600         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
16602 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
16604         PR target/85658
16605         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
16606         (check_arch): Likewise.
16607         (check_fpu): Return the result rather than printing it.
16608         (end arch): Fix operator precedence.
16609         (end cpu): Likewise.
16610         (END): Print the result from check_fpu.
16612 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
16613             Alan Hayward  <alan.hayward@arm.com>
16614             David Sherwood  <david.sherwood@arm.com>
16616         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
16617         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
16618         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
16619         (*fcmuo<mode>_and): New patterns.
16621 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
16623         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
16624         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
16625         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
16626         (cmp_op, sve_imm_con): New code attributes.
16627         (SVE_COND_INT_CMP, imm_con): Delete.
16628         (cmp_op): Remove above unspecs from int attribute.
16629         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
16630         to...
16631         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
16632         comparison-specific unspecs.
16633         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
16634         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
16635         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
16636         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
16637         (*vec_fcm<cmp_op><mode>): Rename to...
16638         (*fcm<cmp_op><mode>): ...this and adjust likewise.
16639         (*vec_fcmuo<mode>): Rename to...
16640         (*fcmuo<mode>): ...this and adjust likewise.
16641         (*pred_fcm<cmp_op><mode>): New pattern.
16642         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
16643         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
16644         functions.
16645         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
16646         and UNORDERED.
16647         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
16648         (aarch64_emit_sve_predicated_cond): New function.
16649         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
16650         (aarch64_emit_unspec_cond_or): Replace with...
16651         (aarch64_emit_sve_or_conds): ...this new function.  Use
16652         aarch64_emit_sve_ptrue_op for the individual comparisons and
16653         aarch64_emit_binop to OR them together.
16654         (aarch64_emit_inverted_unspec_cond): Replace with...
16655         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
16656         aarch64_emit_sve_ptrue_op for the comparison and
16657         aarch64_emit_unop to invert the result.
16658         (aarch64_expand_sve_vec_cmp_float): Update after the above
16659         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
16661 2018-05-07  Nathan Sidwell  <nathan@acm.org>
16663         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
16664         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
16665         (Backwards Compatibility): Likewise.
16667 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16669         PR bootstrap/85681
16670         Revert:
16671         2018-05-07  Luis Machado  <luis.machado@linaro.org>
16673         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16674         <prefetch_dynamic_strides>: New const bool field.
16675         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16676         prefetch_dynamic_strides.
16677         (exynosm1_prefetch_tune): Likewise.
16678         (thunderxt88_prefetch_tune): Likewise.
16679         (thunderx_prefetch_tune): Likewise.
16680         (thunderx2t99_prefetch_tune): Likewise.
16681         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16682         to false.
16683         (aarch64_override_options_internal): Update to set
16684         PARAM_PREFETCH_DYNAMIC_STRIDES.
16685         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16686         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16687         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16688         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16689         prefetch-dynamic-strides setting.
16691         2018-05-07  Luis Machado  <luis.machado@linaro.org>
16693         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16694         <minimum_stride>: New const int field.
16695         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16696         minimum_stride field.
16697         (exynosm1_prefetch_tune): Likewise.
16698         (thunderxt88_prefetch_tune): Likewise.
16699         (thunderx_prefetch_tune): Likewise.
16700         (thunderx2t99_prefetch_tune): Likewise.
16701         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16702         (aarch64_override_options_internal): Update to set
16703         PARAM_PREFETCH_MINIMUM_STRIDE.
16704         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16705         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16706         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16707         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16708         stride is constant and is below the minimum stride threshold.
16710 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16712         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
16713         to 512.
16715 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16717         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16718         <prefetch_dynamic_strides>: New const bool field.
16719         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16720         prefetch_dynamic_strides.
16721         (exynosm1_prefetch_tune): Likewise.
16722         (thunderxt88_prefetch_tune): Likewise.
16723         (thunderx_prefetch_tune): Likewise.
16724         (thunderx2t99_prefetch_tune): Likewise.
16725         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16726         to false.
16727         (aarch64_override_options_internal): Update to set
16728         PARAM_PREFETCH_DYNAMIC_STRIDES.
16729         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16730         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16731         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16732         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16733         prefetch-dynamic-strides setting.
16735 2018-05-07  Luis Machado  <luis.machado@linaro.org>
16737         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16738         <minimum_stride>: New const int field.
16739         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16740         minimum_stride field.
16741         (exynosm1_prefetch_tune): Likewise.
16742         (thunderxt88_prefetch_tune): Likewise.
16743         (thunderx_prefetch_tune): Likewise.
16744         (thunderx2t99_prefetch_tune): Likewise.
16745         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16746         (aarch64_override_options_internal): Update to set
16747         PARAM_PREFETCH_MINIMUM_STRIDE.
16748         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16749         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16750         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16751         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16752         stride is constant and is below the minimum stride threshold.
16754 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
16756         PR c++/85659
16757         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
16758         the type is addressable.  Don't force op into register if it has
16759         BLKmode.
16761 2018-05-05  Roland McGrath  <mcgrathr@google.com>
16763         PR other/77609
16764         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
16765         any section for which we don't know a specific type it should have,
16766         regardless of name.  Previously this was done only for the exact
16767         names ".init_array", ".fini_array", and ".preinit_array".
16768         (default_elf_asm_named_section): Add comment about
16769         relationship with default_section_type_flags and SECTION_NOTYPE.
16770         (get_section): Don't consider it a type conflict if one side has
16771         SECTION_NOTYPE and the other doesn't, as long as neither has the
16772         SECTION_BSS et al used in the default_section_type_flags logic.
16774 2018-05-05  Tom de Vries  <tom@codesourcery.com>
16776         PR target/85653
16777         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
16778         (workaround_barsyncs): New function.
16779         (nvptx_reorg): Use workaround_barsyncs.
16780         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16781         (define_expand "nvptx_membar_cta"): New define_expand.
16782         (define_insn "*nvptx_membar_cta"): New insn.
16784 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
16786         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
16787         To improve optimization opportunities.
16788         * builtin-types.def: The new needed builtin types for the above.
16790 2018-05-04  Richard Biener  <rguenther@suse.de>
16792         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
16793         * gimple-ssa-store-merging.c
16794         (imm_store_chain_info::output_merged_store): Remove redundant create,
16795         release split_store vector contents on failure.
16796         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
16797         scalar stmt vector on cache hit.
16799 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
16801         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
16802         Xilinx FP support.
16803         * config.gcc (powerpc-xilinx-eabi*): Remove.
16804         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16805         support.
16806         (fusion_addis_mem_combo_load): Ditto.
16807         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16808         FP support.
16809         (rs6000_cpu_cpp_builtins): Ditto.
16810         * config/rs6000/rs6000-linux.c
16811         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16812         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16813         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16814         support.
16815         (rs6000_setup_reg_addr_masks): Ditto.
16816         (rs6000_init_hard_regno_mode_ok): Ditto.
16817         (rs6000_option_override_internal): Ditto.
16818         (legitimate_lo_sum_address_p): Ditto.
16819         (rs6000_legitimize_address): Ditto.
16820         (rs6000_legitimize_reload_address): Ditto.
16821         (rs6000_legitimate_address_p): Ditto.
16822         (abi_v4_pass_in_fpr): Ditto.
16823         (setup_incoming_varargs): Ditto.
16824         (rs6000_gimplify_va_arg): Ditto.
16825         (rs6000_split_multireg_move): Ditto.
16826         (rs6000_savres_strategy): Ditto.
16827         (rs6000_emit_prologue_components): Ditto.
16828         (rs6000_emit_epilogue_components): Ditto.
16829         (rs6000_emit_prologue): Ditto.
16830         (rs6000_emit_epilogue): Ditto.
16831         (rs6000_elf_file_end): Ditto.
16832         (rs6000_function_value): Ditto.
16833         (rs6000_libcall_value): Ditto.
16834         * config/rs6000/rs6000.h: Ditto.
16835         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16836         (TARGET_MINMAX): ... this.  New.
16837         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16838         * config/rs6000/rs6000.md: Remove Xilinx FP support.
16839         (*movsi_internal1_single): Delete.
16840         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16841         mfpu=, mxilinx-fpu): Delete.
16842         * config/rs6000/singlefp.h: Delete.
16843         * config/rs6000/sysv4.h: Remove Xilinx FP support.
16844         * config/rs6000/t-rs6000: Ditto.
16845         * config/rs6000/t-xilinx: Delete.
16846         * config/rs6000/titan.md: Adjust for fp_type removal.
16847         * config/rs6000/vsx.md: Remove Xilinx FP support.
16848         (VStype_simple): Delete.
16849         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16850         * config/rs6000/xfpu.h: Delete.
16851         * config/rs6000/xfpu.md: Delete.
16852         * config/rs6000/xilinx.h: Delete.
16853         * config/rs6000/xilinx.opt: Delete.
16854         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16855         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16857 2018-05-04  Tom de Vries  <tom@codesourcery.com>
16859         PR libgomp/85639
16860         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16861         if ignore == 0.
16863 2018-05-04  Richard Biener  <rguenther@suse.de>
16865         PR middle-end/85627
16866         * tree-complex.c (update_complex_assignment): We are always in SSA form.
16867         (expand_complex_div_wide): Likewise.
16868         (expand_complex_operations_1): Likewise.
16869         (expand_complex_libcall): Preserve EH info of the original stmt.
16870         (tree_lower_complex): Handle removed blocks.
16871         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16872         on complex multiplication and division libcall builtins.
16874 2018-05-04  Richard Biener  <rguenther@suse.de>
16876         PR middle-end/85574
16877         * fold-const.c (negate_expr_p): Restrict negation of operand
16878         zero of a division to when we know that can happen without
16879         overflow.
16880         (fold_negate_expr_1): Likewise.
16882 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
16884         PR libstdc++/85466
16885         * real.h (real_nextafter): Declare.
16886         * real.c (real_nextafter): New function.
16887         * fold-const-call.c (fold_const_nextafter): New function.
16888         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16889         CASE_CFN_NEXTTOWARD.
16890         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16891         even when arg1_mode is different from arg0_mode.
16893 2018-05-03  Nathan Sidwell  <nathan@acm.org>
16895         * doc/extend.texi (Deprecated Features): Remove
16896         -ffriend-injection.
16897         (Backwards Compatibility): Likewise.
16898         * doc/invoke.texi (C++ Language Options): Likewise.
16899         (C++ Dialect Options): Likewise.
16901 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
16903         PR target/85530
16904         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
16905         _mm512_mask_mullox_epi64): New intrinsics.
16907 2018-05-03  Tom de Vries  <tom@codesourcery.com>
16909         PR testsuite/85106
16910         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16911         dump files): Add offload-tree.
16913 2018-05-03  Richard Biener  <rguenther@suse.de>
16915         PR tree-optimization/85615
16916         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
16917         to loops not nested in BBs loop father to avoid creating multi-entry
16918         loops.
16920 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16922         PR tree-optimization/70291
16923         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
16924         arguments.  Change return type to tree.  Emit libcall as a new
16925         statement rather than replacing existing one when inplace_p is true.
16926         (expand_complex_multiplication_components): New function.
16927         (expand_complex_multiplication): Expand floating-point complex
16928         multiplication using the above.
16929         (expand_complex_division): Rename inner_type parameter to type.
16930         Update expand_complex_libcall call-site.
16931         (expand_complex_operations_1): Update expand_complex_multiplication
16932         and expand_complex_division call-sites.
16934 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
16936         PR target/85582
16937         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16938         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16939         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16940         the highest significant bit of the shift count mask is clear.  In
16941         check whether and[sq]i3 is needed verify that all significant bits
16942         of the shift count other than the highest are set.
16944 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16946         PR libgomp/82428
16947         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16948         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16949         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16950         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16951         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16952         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16953         __builtin_goacc_parlevel_size.
16955 2018-05-02  Richard Biener  <rguenther@suse.de>
16957         PR tree-optimization/85597
16958         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16959         do not use split vect_get_vec_defs call but call vect_get_slp_defs
16960         directly.
16962 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16964         PR testsuite/85106
16965         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16966         dump files): Add ltrans-tree.
16968 2018-05-02  Tom de Vries  <tom@codesourcery.com>
16970         PR testsuite/85106
16971         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16972         dump files): Add wpa-ipa.
16974 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
16976         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
16977         powerpc*-*-linux*paired* target.
16978         * config/rs6000/750cl.h: Delete.
16979         * config/rs6000/paired.h: Delete.
16980         * config/rs6000/paired.md: Delete.
16981         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16982         float support.
16983         * config/rs6000/rs6000-builtin.def: Remove paired float support.
16984         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16985         comment.  Remove paired float support.
16986         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16987         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16988         VECTOR_PAIRED.
16989         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16990         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16991         declarations.
16992         * config/rs6000/rs6000.c: Remove paired float support.
16993         (paired_expand_vector_init, paired_expand_vector_move,
16994         paired_emit_vector_compare, paired_emit_vector_cond_expr,
16995         (paired_expand_lv_builtin, paired_expand_stv_builtin,
16996         paired_expand_builtin, paired_expand_predicate_builtin,
16997         paired_init_builtins): Delete.
16998         * config/rs6000/rs6000.h: Remove paired float support.
16999         * config/rs6000/rs6000.md: Remove paired float support.
17000         (move_from_CR_ov_bit): Delete.
17001         * config/rs6000/rs6000.opt (mpaired): Delete.
17002         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17003         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17005 2018-05-02  Richard Biener  <rguenther@suse.de>
17007         PR middle-end/85567
17008         * gimplify.c (gimplify_save_expr): When in SSA form allow
17009         SAVE_EXPRs to compute to SSA vars.
17011 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17013         PR target/85582
17014         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17015         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17016         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17017         clobber operands[2], instead use a new pseudo.  Formatting fixes.
17019 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
17021         PR tree-optimization/85586
17022         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17023         exit early for statements in the same group if the accesses are
17024         not strided.
17026 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17028         PR lto/85451
17029         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17030         error message.
17032 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
17034         PR tree-optimization/85143
17035         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17037 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17039         PR lto/85451
17040         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17041         not found" error message.
17043 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17045         PR other/83786
17046         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
17047         * vec.c (test_ordered_remove_if): New function.
17048         (vec_c_tests): Call test_ordered_remove_if.
17049         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
17050         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
17051         * tree-vect-patterns.c (vect_pattern_recog_1): Use
17052         VEC_ORDERED_REMOVE_IF.
17054 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17056         PR tree-optimization/82665
17057         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
17058         pointer subtraction where arguments come from a memchr call.
17060 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
17062         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17063         --push-state --as-needed and --pop-state instead of --as-needed and
17064         --no-as-needed if ld supports it.
17065         * configure: Regenerated.
17067         PR web/85578
17068         * doc/install.texi2html: Replace _002d with - and _002a with * in
17069         generated html files using sed.
17071 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17073         PR c++/85523
17074         * gcc-rich-location.c (blank_line_before_p): New function.
17075         (use_new_line): New function.
17076         (gcc_rich_location::add_fixit_insert_formatted): New function.
17077         * gcc-rich-location.h
17078         (gcc_rich_location::add_fixit_insert_formatted): New function.
17080 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17082         * selftest.c (assert_streq): Rename "expected" and "actual" to
17083         "val1" and "val2".  Extend NULL-handling to cover both inputs
17084         symmetrically, while still requiring both to be non-NULL for a pass.
17085         * selftest.h (assert_streq): Rename "expected" and "actual" to
17086         "val1" and "val2".
17087         (ASSERT_EQ): Likewise.
17088         (ASSERT_EQ_AT): Likewise.
17089         (ASSERT_KNOWN_EQ): Likewise.
17090         (ASSERT_KNOWN_EQ_AT): Likewise.
17091         (ASSERT_NE): Likewise.
17092         (ASSERT_MAYBE_NE): Likewise.
17093         (ASSERT_MAYBE_NE_AT): Likewise.
17094         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
17095         the assertion to pass.
17096         (ASSERT_STREQ_AT): Likewise.
17098 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17100         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17101         interaction with -pie.
17103 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17105         * selftest.h: Fix alphabetization of per-source-file selftest
17106         declarations.
17108 2018-04-30  Jason Merrill  <jason@redhat.com>
17110         PR c++/61982 - dead stores to destroyed objects.
17111         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
17112         of clobber.
17114 2018-04-30  Jason Merrill  <jason@redhat.com>
17116         * tree.c (build_clobber): New.
17117         * tree.h: Declare it.
17118         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
17120 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17122         * diagnostic-show-locus.c (layout::layout): Update for
17123         location_get_source_line returning a char_span.
17124         (struct char_span): Move to input.h.
17125         (struct correction): Update for fields in char_span becoming
17126         private.
17127         (struct source_line): Update for location_get_source_line
17128         returning a char_span.
17129         (layout::print_line): Likewise.
17130         * edit-context.c (edited_file::print_content): Likewise.
17131         (edited_file::print_diff_hunk): Likewise.
17132         (edited_file::print_run_of_changed_lines): Likewise.
17133         (edited_file::get_num_lines): Likewise.
17134         (edited_line::edited_line): Likewise.
17135         * final.c (asm_show_source): Likewise.
17136         * input.c (location_get_source_line): Convert return type
17137         from const char * to char_span, losing the final "line_len"
17138         param.
17139         (dump_location_info): Update for the above.
17140         (get_substring_ranges_for_loc): Likewise.  Use a char_span
17141         when handling the literal within the line.
17142         (test_reading_source_line): Update for location_get_source_line
17143         returning a char_span.
17144         * input.h (class char_span): Move here from
17145         diagnostic-show-locus.c, converting from a struct to a class.
17146         Make data members private.
17147         (char_span::operator bool): New.
17148         (char_span::length): New.
17149         (char_span::get_buffer): New.
17150         (char_span::operator[]): New.
17151         (char_span::subspan): Make const.
17152         (char_span::xstrdup): New.
17153         (location_get_source_line): Convert return type from const char *
17154         to char_span, losing the final "line_size" param.
17156 2018-04-30  Jan Hubicka  <jh@suse.cz>
17158         * lto-wrapper.c (ltrans_priorities): New static var.
17159         (cmp_priority): New.
17160         (run_gcc): Read priorities and if doing parallel build order
17161         the Makefile by them.
17163 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17165         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
17167 2018-04-30  Richard Biener  <rguenther@suse.de>
17169         * tree-cfg.c (verify_address): Remove base argument, add
17170         flag whether to check TREE_ADDRESSABLE and do that.
17171         (verify_expr): Remove.
17172         (verify_types_in_gimple_reference): Add pieces from verify_expr.
17173         (verify_gimple_assign_single): Likewise.
17174         (verify_gimple_switch): Likewise.
17175         (verify_expr_location_1): Dereference tp once.  Add (disabled)
17176         piece from verify_expr.
17177         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
17179 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17181         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17183 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17185         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
17186         (small_data_pattern): Likewise.
17187         (arc_rewrite_small_data): Likewise.
17188         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
17189         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
17190         (get_symbol_alignment): New function.
17191         (legitimate_small_data_address_p): Likewise.
17192         (legitimate_scaled_address): Update, call
17193         legitimate_small_data_address_p.
17194         (output_sdata): New static variable.
17195         (arc_print_operand): Update how we handle small data operands.
17196         (arc_print_operand_address): Likewise.
17197         (arc_legitimate_address_p): Update, use
17198         legitimate_small_data_address_p.
17199         (arc_rewrite_small_data_p): Remove.
17200         (arc_rewrite_small_data_1): Likewise.
17201         (arc_rewrite_small_data): Likewise.
17202         (small_data_pattern): Likewise.
17203         (compact_sda_memory_operand): Update to use
17204         legitimate_small_data_address_p and get_symbol_alignment.
17205         (prepare_move_operands): Don't rewite sdata pattern.
17206         (prepare_extend_operands): Remove.
17207         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
17208         pattern.
17209         (zero_extendqisi2): Likewise.
17210         (zero_extendhisi2): Likewise.
17211         (extendqihi2): Likewise.
17212         (extendqisi2): Likewise.
17213         (extendhisi2): Likewise.
17214         (addsi3): Likewise.
17215         (subsi3): Likewise.
17216         (andsi3): Likewise.
17217         * config/arc/constraints.md (Usd): Change it to memory constraint.
17219 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17221         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
17222         as source of std instructions.
17223         * config/arc/arc.md (movsi_insn): Update pattern predicate to
17224         allow 6-bit constants as source for store instructions.
17225         (movdi_insn): Update instruction pattern to allow 6-bit constants
17226         as source for store instructions.
17228 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17230         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
17232 2018-04-30  Nathan Sidwell  <nathan@acm.org>
17233             Sandra Loosemore <sandra@codesourcery.com>
17235         * dumpfile.c (dump_open): Allow '-' for stdout.
17236         * doc/invoke.texi (Developer Options): Document dump filename
17237         determination early.  Document stdin/stdout selection.
17239 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
17241         Microblaze Target: PIC data text relative
17243         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
17244         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
17245         Add declaration.
17246         * config/microblaze/microblaze.h (microblaze_constant_address_p):
17247         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
17248         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
17249         New addressing mode for data-text relative position indepenedent code.
17250         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
17251         'ADDRESS_SYMBOLIC_TXT_REL'.
17252         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
17253         (microblaze_legitimate_pic_operand): Exclude function calls from
17254         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
17255         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
17256         addresses cases.
17257         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17258         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17259         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
17260         for 'address + offset'.
17261         (microblaze_expand_prologue): Add new function prologue call for
17262         'r20' assignation.
17263         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
17264         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
17265         table in case of TARGET_PIC_DATA_TEXT_REL.
17266         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
17267         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
17268         Add new macros 'UNSPEC_TEXT',
17269         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
17270         + exclude function calls from 'UNSPEC_PLT' in case of data text
17271         relative mode.
17272         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17273         new target hook for generating address diff vector tables in case of
17274         flag_pic.
17275         * doc/tm.texi : Regenerate.
17276         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
17277         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
17278         of addr diff vector generation.
17279         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17280         target hook definition.
17281         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
17282         Add default function for generate_pic_addr_diff_vec -> flag_pic.
17283         * doc/invoke.texi (Add new pic option): Add new microblaze pic
17284         option for data text relative.
17286 2018-04-30  Richard Biener  <rguenther@suse.de>
17288         * tree-chrec.h (evolution_function_is_constant_p): Remove
17289         redundant check.
17290         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
17292 2018-04-30  Richard Biener  <rguenther@suse.de>
17294         PR bootstrap/85571
17295         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
17297 2018-04-30  Richard Biener  <rguenther@suse.de>
17299         PR tree-optimization/28364
17300         PR tree-optimization/85275
17301         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17302         copying first exit test.
17304 2018-04-28  Mark Wielaard  <mark@klomp.org>
17306         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
17307         dwarf_version >= 5.
17308         (dwarf_AT): Handle DW_AT_addr_base.
17309         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
17311 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
17313         PR target/84431
17314         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
17315         (*ashl<dwi>3_doubleword_mask_1): Ditto.
17316         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
17317         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
17319 2018-04-28  Richard Biener  <rguenther@suse.de>
17321         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
17322         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
17323         to reflect use.  Only add interesting stmts.
17325 2018-04-27  Martin Jambor  <mjambor@suse.cz>
17327         PR ipa/85549
17328         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
17329         the jump function allows for passing through aggregate values.
17331 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
17333         * input.h (in_system_header_at): Convert from macro to inline
17334         function.
17335         (from_macro_expansion_at): Likewise.
17336         (from_macro_definition_at): Likewise.
17338 2018-04-27  Jeff Law  <law@redhat.com>
17340         * config.gcc: Mark tile* targets as deprecated/obsolete.
17342 2018-04-27  Richard Biener  <rguenther@suse.de>
17344         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
17345         fix for ILP32.
17347 2018-04-27  Richard Biener  <rguenther@suse.de>
17349         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
17351 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
17353         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
17354         with Yd constraint. Set "preferred_for_speed" attribute from
17355         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
17356         with Yd constraint.
17357         (*movdi_internal): Ditto.
17358         (movti_interunit splitters): Remove
17359         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
17360         (movdi_interunit splitters): Ditto.
17361         * config/i386/constraints.md (Ye): Remove.
17362         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
17364 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17366         PR target/85512
17367         * config/aarch64/constraints.md (Usg): Limit to 31.
17368         (Usj): Limit to 63.
17370 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
17372         PR tree-optimization/85529
17373         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
17374         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
17375         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
17376         zero extension or masking of the MSB bit.
17377         (optimize_range_tests): Add FIRST_BB argument, pass it through
17378         to optimize_range_tests_var_bound.
17379         (maybe_optimize_range_tests, reassociate_bb): Adjust
17380         optimize_range_tests callers.
17382 2018-04-26  Richard Biener  <rguenther@suse.de>
17383             Jakub Jelinek  <jakub@redhat.com>
17385         * cgraph.h (symbol_table): Just declare debug method here.
17386         * symtab.c (symbol_table::debug): Define.
17388 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
17390         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
17392 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
17394         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
17395         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
17396         (*movdi_internal): Substitute Yi and Yj constraint with x
17397         and Ym and Yn constraint with y constraint.  Update "isa"
17398         attribute and set "preferred_for_speed" attribute from
17399         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
17400         (*movsi_internal): Ditto.
17401         (*movdf_internal): Ditto.
17402         (*movsf_internal): Ditto.
17403         (*zero_extendsidi2): Ditto.
17404         * config/i386/sse.md (vec_set<mode>_0): Ditto.
17405         (sse2_loadld): Ditto.
17406         (*vec_extract<ssevecmodelower>_0): Ditto.
17407         (*vec_extractv4si_0_zext_sse4): Ditto.
17408         (vec_concatv2di): Ditto.
17409         (*vec_dup<mode>): Ditto.
17410         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
17411         * config/i386/constraints.md (Yi): Remove.
17412         (Yj): Remove.
17413         (Ym): Remove.
17414         (Yn): Remove.
17416 2018-04-26  Nathan Sidwell  <nathan@acm.org>
17418         * dumpfile.c (dump_open): New.
17419         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
17420         (dump_finish): Detect stdio/stderr by value not name.
17422 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
17424         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
17426 2018-04-26  Tom de Vries  <tom@codesourcery.com>
17428         PR target/84952
17429         * config/nvptx/nvptx.c (verify_neutering_jumps)
17430         (verify_neutering_labels): New function
17431         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
17433 2018-04-26  Tom de Vries  <tom@codesourcery.com>
17435         PR target/84025
17436         * config/nvptx/nvptx.c (needs_neutering_p): New function.
17437         (nvptx_single): Use needs_neutering_p to skip over insns that do not
17438         need neutering.
17440 2018-04-26  Richard Biener <rguenther@suse.de>
17441             Tom de Vries  <tom@codesourcery.com>
17443         PR lto/85422
17444         * lto-streamer-out.c (output_function): Fixup loops if required to match
17445         discovery done in the reader.
17447 2018-04-26  Richard Biener  <rguenther@suse.de>
17449         PR tree-optimization/85116
17450         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
17451         have a loop exit from the single latch predecessor.  Remove
17452         case of header with just condition.
17453         (ch_base::copy_headers): Exclude infinite loops from any
17454         processing.
17455         (pass_ch::execute): Record exits.
17457 2018-04-26  Richard Biener  <rguenther@suse.de>
17459         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
17460         prologue cost vector and pass it to vect_get_load_cost.
17461         (vect_get_peeling_costs_all_drs): Likewise.
17462         (vect_peeling_hash_get_lowest_cost): Likewise.
17463         (vect_enhance_data_refs_alignment): Likewise.
17465 2018-04-26  Richard Biener  <rguenther@suse.de>
17467         PR middle-end/85450
17468         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
17469         checking of integer<->pointer conversions.
17470         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
17471         sign-/zero-extending pointer types.
17472         (expand_omp_for_static_chunk): Likewise.
17474 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
17475             Jean Lee  <xiaoyur347@gmail.com>
17477         * config/mips/mips.c (mips_asan_shadow_offset): New function.
17478         (TARGET_ASAN_SHADOW_OFFSET): Define.
17479         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
17480         true for -fsanitize=address.
17482 2018-04-25  Mark Wielaard  <mark@klomp.org>
17484         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
17485         shorter ones.
17487 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
17489         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
17490         than "alu", remove explicit "memory" and "imm_disp" attributes.
17491         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
17493         PR middle-end/85414
17494         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
17495         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
17496         gen_lowpart_no_emit.
17498 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
17500         PR target/85473
17501         * config/i386/i386.c (ix86_expand_builtin): Change memory
17502         operand to XI, extend p0 to Pmode.
17503         * config/i386/i386.md: Change unspec volatile and operand
17504         1 mode to XI, change operand 0 mode to P.
17506 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17508         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
17509         GET_MODE_MASK before any checking.
17510         (nds32_can_use_bset_p): Likewise.
17511         (nds32_can_use_btgl_p): Likewise.
17513 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17515         * config/nds32/nds32-doubleword.md: New define_split pattern for
17516         illegal register number.
17518 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17520         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
17522 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
17524         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
17526 2018-04-25  Richard Biener  <rguenther@suse.de>
17528         * lto-streamer.h (LTO_major_version): Bump to 8.
17530 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
17532         * BASE-VER: Set to 9.0.0.
17534 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
17536         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
17537         in __abskf2 and __powikf2.
17539 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17541         PR target/85512
17542         * config/aarch64/constraints.md (Usg, Usj): New constraints.
17543         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
17544         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
17545         Use the above on operand 2.  Reindent.
17546         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
17548 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
17550         PR target/85485
17551         * common/config/i386/i386-common.c (ix86_handle_option): Don't
17552         handle OPT_mcet.
17553         * config/i386/i386.opt (mcet): Removed.
17554         * doc/install.texi: Remove -mcet documentation.
17555         * doc/invoke.texi: Likewise.
17557 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
17559         PR target/85485
17560         * doc/install.texi: Remove -mcet from bootstrap-cet.
17562 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17564         PR target/85511
17565         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
17566         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
17567         if TARGET_64BIT.
17569         PR target/85503
17570         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
17571         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
17572         containing a CONST_VECTOR.
17574 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
17576         * doc/install.texi: Update newlib dependency for nvptx.
17578 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17580         PR target/85508
17581         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
17582         instead of INTVAL when shifting x left.
17584 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
17586         PR tree-optimization/85478
17587         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
17588         vect_grouped_store_supported for single element vectors.
17590 2018-04-24  Richard Biener  <rguenther@suse.de>
17592         PR target/85491
17593         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
17594         load cost increase to the case of non-constant step.
17596 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
17598         PR target/84828
17599         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
17600         destination if any_malformed_asm.
17602 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
17604         PR middle-end/85496
17605         * expr.c (store_field): In the bitfield case, if the value comes from
17606         a function call and is returned in registers by means of a PARALLEL,
17607         do not change the mode of the temporary unless BLKmode and VOIDmode.
17609 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
17611         PR rtl-optimization/85423
17612         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
17613         dependencies to debug insns when the previous insn is non-debug.
17615 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
17617         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
17618         enums into a single definition.
17619         (fls): Fix predicates and printing.
17620         (seti): Likewise.
17622 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
17624         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
17625         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
17626         and short u6 immediate.
17627         (check_if_valid_sleep_operand): Remove.
17628         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
17630 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17632         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
17633         flag_always_save_lp condition.
17634         * config/nds32/nds32.opt (malways-save-lp): New option.
17636 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17638         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
17639         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
17640         * config/nds32/nds32.h
17641         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
17642         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
17644 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17646         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
17647         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
17649 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
17650             Chung-Ju Wu  <jasonwucj@gmail.com>
17652         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
17653         Declare.
17654         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
17655         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
17657 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17659         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
17661 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17663         * config/nds32/nds32-protos.h (nds32_data_alignment,
17664         nds32_local_alignment): Declare.
17665         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
17666         nds32_local_alignment): New functions.
17667         (TARGET_CONSTANT_ALIGNMENT): Define.
17668         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
17670 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17672         * config/nds32/nds32.c
17673         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
17674         (TARGET_MODES_TIEABLE_P): Likewise.
17676 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
17678         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
17679         level Ofast and Og.
17681 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
17682             Chung-Ju Wu  <jasonwucj@gmail.com>
17684         * config/nds32/constants.md (unspec_volatile_element): Add enum values
17685         for unaligned access.
17686         * config/nds32/nds32-intrinsic.c: Implementation of expanding
17687         unaligned access.
17688         * config/nds32/nds32-intrinsic.md: Likewise.
17689         * config/nds32/nds32_intrinsic.h: Likewise.
17690         * config/nds32/nds32.h (nds32_builtins): Likewise.
17691         * config/nds32/nds32.opt (munaligned-access): New option.
17692         * config/nds32/nds32.c (nds32_asm_file_start): Display
17693         flag_unaligned_access status.
17695 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
17697         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
17698         -mno-relax is present.
17699         * config/riscv/linux.h (LINK_SPEC): Ditto.
17701 2018-04-20  Martin Sebor  <msebor@redhat.com>
17703         PR c/85365
17704         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
17705         for null pointers.
17706         (gimple_fold_builtin_stxcpy_chk): Same.
17707         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
17709 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
17711         PR target/85456
17712         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
17713         __powikf2 when long double is IEEE 128-bit.
17715 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
17717         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
17718         step to make sure stack always aligned.
17720 2018-04-20  Carl Love  <cel@us.ibm.com>
17722         PR target/83402
17723         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
17724         size check for arg0.
17726 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
17727             Tom de Vries  <tom@codesourcery.com>
17729         PR target/85445
17730         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17731         Emit insns for calls too.
17732         (nvptx_find_par): Always look for worker-level predecessor insn.
17733         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
17734         calls.
17735         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
17736         (nvptx_process_pars): Propagate frames for calls.
17738 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
17740         PR target/85469
17741         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
17742         Removed.
17743         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
17744         (ix86_handle_option): Don't handle OPT_mibt.
17745         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
17746         __SHSTK__.
17747         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
17748         has_ibt and ibt.
17749         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
17750         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
17751         (ix86_target_macros): Define __CET__ with flag_cf_protection
17752         for -fcf-protection.
17753         * config/i386/i386.c (isa2_opts): Remove -mibt.
17754         * config/i386/i386.h (TARGET_IBT): Removed.
17755         (TARGET_IBT_P): Likewise.
17756         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
17757         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
17758         * config/i386/i386.opt (mcet): Update help message.
17759         (mshstk): Likewise.
17760         (mibt): Removed.
17761         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
17762         -mcet as an alias for -mshstk.
17764 2018-04-20  Richard Biener <rguenther@suse.de>
17766         PR middle-end/85475
17767         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
17768         complexity by forcing a single use of the multiply operand.
17770 2018-04-20  Martin Jambor  <mjambor@suse.cz>
17772         ipa/85449
17773         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
17774         recursion dependency to only apply to non-clones.
17776 2018-04-20  Martin Jambor  <mjambor@suse.cz>
17778         ipa/85447
17779         * ipa-cp.c (create_specialized_node): Check that clones of
17780         self-recursive edges exist during IPA-CP.
17782 2018-04-19  Toon Moene  <toon@moene.org>
17784         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
17785         by -O3.
17787 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
17789         PR tree-optimization/85467
17790         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
17791         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
17792         VECTOR_CST element to type.
17794 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17796         PR target/85397
17797         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
17798         * config/i386/i386.md (builtin_setjmp_setup): Removed.
17799         (builtin_longjmp): Likewise.
17800         (save_stack_nonlocal): New pattern.
17801         (restore_stack_nonlocal): Likewise.
17803 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17805         PR target/85404
17806         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17807         Replace ASM_OUTPUT_LABEL with fprintf.
17809 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
17811         PR target/85417
17812         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17813         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17814         * config/i386/i386-c.c (ix86_target_macros_internal): Also
17815         define __IBT__ and __SHSTK__ for -fcf-protection.
17816         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17817         TARGET_IBT.
17818         (ix86_trampoline_init): Likewise.
17819         (x86_output_mi_thunk): Likewise.
17820         (ix86_notrack_prefixed_insn_p): Likewise.
17821         (ix86_option_override_internal): Don't disallow -fcf-protection.
17822         * config/i386/i386.md (rdssp<mode>): Also enable for
17823         -fcf-protection.
17824         (incssp<mode>): Likewise.
17825         (nop_endbr): Likewise.
17826         * config/i386/i386.opt (mcet): Change help message to built-in
17827         functions only.
17828         (mibt): Likewise.
17829         (mshstk): Likewise.
17830         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17831         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
17832         enable CET built-in functions.
17834 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
17836         * common/config/i386/i386-common.c
17837         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17838         OPTION_MASK_ISA_MOVDIRI_UNSET,
17839         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17840         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17841         * config.gcc (movdirintrin.h): New header.
17842         * config/i386/cpuid.h (bit_MOVDIRI,
17843         bit_MOVDIR64B): New bits.
17844         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17845         and -mmvodir64b.
17846         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17847         (VOID, PVOID, PCVOID)): New function types.
17848         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17849         __builtin_ia32_directstoreu_u64,
17850         __builtin_ia32_movdir64b): New builtins.
17851         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17852         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17853         and -mmovdiri.
17854         (ix86_valid_target_attribute_inner_p): Ditto.
17855         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17856         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17857         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17858         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17859         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17860         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17861         (movdiri<mode>, movdir64b_<mode>): New.
17862         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17863         * config/i386/immintrin.h: Include movdirintrin.h.
17864         * config/i386/movdirintrin.h: New file.
17865         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17867 2018-04-19  Richard Biener  <rguenther@suse.de>
17869         PR middle-end/85455
17870         * cfg.c (clear_bb_flags): When loop state says we have
17871         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17873 2018-04-19  Richard Biener  <rguenther@suse.de>
17875         PR tree-optimization/84737
17876         * tree-vect-data-refs.c (vect_copy_ref_info): New function
17877         copying restrict info.
17878         (vect_setup_realignment): Use it.
17879         * tree-vectorizer.h (vect_copy_ref_info): Declare.
17880         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17881         the first DR to all generated stores.
17882         (vectorizable_load): Likewise for loads.
17884 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
17886         PR tree-optimization/85446
17887         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17888         the integral and pointer types to have the same precision.
17890         * doc/install.texi: Document --disable-cet being the default and
17891         --enable-cet=auto.
17893 2018-04-18  Martin Liska  <mliska@suse.cz>
17895         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17896         style.
17898 2018-04-18  Martin Liska  <mliska@suse.cz>
17900         Revert
17901         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
17903         PR ipa/83983
17904         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17905         arguments if they are comparable.
17907 2018-04-18  Martin Liska  <mliska@suse.cz>
17909         Revert
17910         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17912         PR lto/84805
17913         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17914         incomplete types.
17916 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
17918         PR target/85388
17919         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
17920         ENDBR after calling __morestack.
17922 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
17924         PR jit/85384
17925         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17926         by using gcc_base_ver to generate a gcc_driver_version, and use
17927         it when generating GCC_DRIVER_NAME.
17928         * configure: Regenerate.
17930 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
17932         PR target/81084
17933         * config.gcc: Obsolete powerpc*-*-*spe*.
17935 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17937         PR debug/84637
17938         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17939         (stabstr_D): Change type of unum from unsigned int to
17940         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
17941         type.
17943 2018-04-17  Jim Wilson  <jimw@sifive.com>
17945         PR 84856
17946         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17947         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17948         Set arg_pointer_offset after using pretend_args_size.
17950 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17952         PR rtl-optimization/85431
17953         * dse.c (record_store): Ignore zero width stores.
17955         PR sanitizer/85230
17956         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
17957         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17958         __builtin_stack_restore rather than after it.
17959         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17960         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17961         argument instead of virtual_dynamic_stack_rtx.
17963 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17965         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17966         New prototype.
17967         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17968         Add note to error message to explain internal mapping of overloaded
17969         built-in function name to non-overloaded built-in function name.
17970         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17971         function.
17973 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
17975         PR target/85424
17976         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17977         where the inputs overlap with the output.
17979 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
17981         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17982         (=v, v) alternative and explicit "memory" attribute.
17983         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
17984         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17985         attributes.
17986         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17987         "sselog1" type instead of "sselog".
17988         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17989         "sselog".  Remove explicit "memory" attribute.
17990         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17991         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17992         attributes.
17993         (vec_extract_hi_v32hi): Merge all alternatives into one, use
17994         "sselog1" type instead of "sselog".  Remove explicit "memory"
17995         attribute.
17996         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17997         use "sselog1" type instead of "sselog".  Remove explicit "memory"
17998         attribute.
17999         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18000         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18001         attributes.
18002         (vec_extract_hi_v64qi): Merge all alternatives into one, use
18003         "sselog1" type instead of "sselog".  Remove explicit "memory"
18004         attribute.
18005         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18006         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18007         attribute.
18009         PR target/85430
18010         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18012         PR middle-end/85414
18013         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18014         on a SUBREG.
18016 2018-04-17  Martin Jambor  <mjambor@suse.cz>
18018         PR ipa/85421
18019         * ipa-cp.c (create_specialized_node): Call
18020         expand_all_artificial_thunks if necessary.
18022 2018-04-17  Martin Liska  <mliska@suse.cz>
18024         PR lto/85405
18025         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18026         in message, remote space in between '_G' and '('.
18028 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18030         PR target/85281
18031         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18032         avx512f_vmcmp<mode>3<round_saeonly_name>,
18033         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18034         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18035         avx512f_rndscale<mode><round_saeonly_name>,
18036         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18037         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18038         Use %<iptr>2 instead of %2 for -masm=intel.
18039         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18040         avx512f_vcvttss2usi<round_saeonly_name>,
18041         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18042         -masm=intel.
18043         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18044         avx512f_vcvttsd2usi<round_saeonly_name>,
18045         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18046         Use %q1 instead of %1 for -masm=intel.
18047         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
18048         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
18049         of %3 for -masm=intel.
18050         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
18051         -masm=intel.
18052         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
18053         -masm=intel.
18054         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
18055         -masm=intel.
18056         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
18057         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18058         %g1.
18059         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18060         -masm=intel.
18061         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18062         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18063         %g1 and one with %0 and %1.
18064         (avx512er_vmrcp28<mode><round_saeonly_name>,
18065         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18066         %1 for -masm=intel.
18067         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18068         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18069         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18070         of %0 and %{%4%} for -masm=intel.
18071         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18072         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18073         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18074         order of %0 and %{%5%}%{z%} for -masm=intel.
18076 2018-04-17  Jan Hubicka  <jh@suse.cz>
18078         PR lto/85405
18079         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18081 2018-04-17  Martin Liska  <mliska@suse.cz>
18083         PR ipa/85329
18084         * multiple_target.c (create_dispatcher_calls): Set apostrophes
18085         for target_clone error message.  Make default implementation
18086         clone to be a local declaration.
18087         (separate_attrs): Add new argument and check for an empty
18088         string.
18089         (expand_target_clones): Handle it.
18090         (ipa_target_clone): Make redirection just for target_clones
18091         functions.
18093 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
18094             Tom de Vries  <tom@codesourcery.com>
18096         PR middle-end/84955
18097         * omp-expand.c (expand_oacc_for): Add dummy false branch for
18098         tiled basic blocks without omp continue statements.
18100 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
18102         PR target/83660
18103         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
18104         vec_extract expression as having side effects to make sure it gets
18105         a cleanup point.
18107 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
18109         PR target/85403
18110         * config/i386/i386.c (get_builtin_code_for_version): Check
18111         error_mark_node.
18113 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
18115         PR target/84331
18116         * config.gcc: Support "skylake".
18117         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18118         PROCESSOR_SKYLAKE.
18119         * config/i386/i386.c (m_SKYLAKE): Define.
18120         (processor_target_table): Add "skylake".
18121         (ix86_option_override_internal): Add "skylake".
18122         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
18123         PROCESSOR_CANNONLAKE.
18124         (get_builtin_code_for_version): Fix priority for
18125         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
18126         PROCESSOR_SKYLAKE-AVX512.
18127         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
18128         (processor_type): Add PROCESSOR_SKYLAKE.
18130 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
18131             Jason Merrill  <jason@redhat.com>
18133         PR c++/85112
18134         * convert.c (convert_to_integer_1): Use direct recursion for
18135         enumeral types and types with a precision less than the number
18136         of bits in their mode.
18138 2018-04-16  Julia Koval  <julia.koval@intel.com>
18140         PR target/84413
18141         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
18142         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
18144 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
18146         PR target/85293
18147         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
18148         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
18149         and -mno-direct-move.
18151 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
18153         PR target/83402
18154         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
18155         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
18156         Ensure negative shifts result in {0}.
18158 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
18160         PR rtl-optimization/79916
18161         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
18162         regs (if any) to define how to gnerate SD moves when LRA is in
18163         progress.
18165 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
18167         PR rtl-optimization/85393
18168         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
18169         * except.c (expand_dw2_landing_pad_for_region): Make static.
18170         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
18171         a label and unconditional jump to old_bb, rather than
18172         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
18173         basic block.
18175         PR rtl-optimization/85376
18176         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
18177         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
18178         instead of a specific value.
18180 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18181             Bin Cheng  <bin.cheng@arm.com>
18183         PR tree-optimization/82965
18184         PR tree-optimization/83991
18185         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
18186         by_profile_only parameter.
18187         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
18188         information if the loop was predicted to iterate too many times.
18189         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
18191 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18193         PR lto/71991
18194         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
18195         always inline.
18197 2018-04-13  Martin Liska  <mliska@suse.cz>
18198             Jakub Jelinek  <jakub@redhat.com>
18200         PR middle-end/81657
18201         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
18202         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
18203         * builtins.c (expand_builtin_memory_copy_args): Use
18204         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
18205         handle dest_addr == pc_rtx.
18207 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
18209         PR target/85291
18210         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
18211         asked to not generate direct moves.
18212         (fix_trunc<mode>si2_stfiwx): Similar.
18213         (fix_trunc<mode>si2_internal): Similar.
18215 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18217         PR debug/83157
18218         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
18219         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
18220         lookup if dest in some wider mode is known to be const0_rtx and
18221         if so, record permanent equivalence for it to be ZERO_EXTEND of
18222         the narrower mode destination.
18224 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18226         * lto-streamer-out.c (output_function): Revert 259346.
18227         * omp-expand.c (expand_oacc_for): Likewise.
18229 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
18231         PR rtl-optimization/85354
18232         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
18233         * sel-sched.c (sel_global_init): ... here.
18235 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
18237         PR target/85238
18238         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
18239         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
18240         mode for PE-COFF targets.
18241         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
18242         (i386_pe_asm_lto_end): Likewise.
18243         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
18244         (TARGET_ASM_LTO_END): Likewise.
18245         * config/i386/winnt.c (saved_debug_info_level): New static variable.
18246         (i386_pe_asm_lto_start): New function.
18247         (i386_pe_asm_lto_end): Likewise.
18249 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18250             Richard Biener  <rguenther@suse.de>
18252         PR middle-end/84955
18253         * lto-streamer-out.c (output_function): Fix CFG loop state before
18254         streaming out.
18255         * omp-expand.c (expand_oacc_for): Handle calls to internal
18256         functions like regular functions.
18258 2018-04-12  Richard Biener  <rguenther@suse.de>
18260         PR lto/85371
18261         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
18262         for the early LTO debug to properly generate references to it
18263         during DIE emission.  Do not re-use that for the skeleton for
18264         split-dwarf.
18265         (dwarf2out_early_finish): Likewise.
18267 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18269         PR target/85328
18270         * config/i386/sse.md
18271         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
18272         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
18273         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
18274         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
18275         and output is a reg, avoid creating invalid lowpart subreg, but
18276         instead split into a 512-bit move.  Don't split if not AVX512VL,
18277         input is xmm16+ reg and output is a mem.
18278         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
18279         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
18280         xmm16+ reg and output is a mem.
18282 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18284         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
18285         also for flag_dwarf2_cfi_asm.
18287 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18289         PR rtl-optimization/85342
18290         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
18291         a bool scalar var inside of the loop instead.  Don't try to update
18292         recog_data.operand after failed apply_change_group.
18294 2018-04-12  Tom de Vries  <tom@codesourcery.com>
18296         PR target/85296
18297         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
18298         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
18299         array with flexible array member as array without given dimension.
18300         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
18301         argument for undefined param to true.
18303 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
18305         PR target/85321
18306         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18307         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
18308         from PowerPC section.
18309         * config/rs6000/sysv4.opt (mcall-): Improve help text.
18310         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
18311         help text that is too long.
18312         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
18313         help text that is too long.
18314         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
18315         help text that is too long.
18317 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
18319         * config/alpha/alpha.md (stack_probe_internal): Rename
18320         from "probe_stack".  Update all callers.
18322 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18324         PR rtl-optimization/84566
18325         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
18326         sched_macro_fuse_insns.
18328 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18330         PR target/84301
18331         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
18332         (compute_block_dependences): ... from here.
18334 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18336         PR tree-optimization/85331
18337         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
18338         from int to HOST_WIDE_INT.
18340 2018-04-11  Martin Jambor  <mjambor@suse.cz>
18342         PR ipa/84149
18343         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
18344         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
18345         not the same as the source val.
18346         (cgraph_edge_brings_value_p): New parameter.
18347         (gather_edges_for_value): Pass destination value to
18348         cgraph_edge_brings_value_p.
18349         (perhaps_add_new_callers): Likewise.
18350         (get_info_about_necessary_edges): Likewise and exclude values brought
18351         only by self-recursive edges.
18352         (create_specialized_node): Redirect only clones of self-calling edges.
18353         (+self_recursive_pass_through_p): New function.
18354         (find_more_scalar_values_for_callers_subset): Use it.
18355         (find_aggregate_values_for_callers_subset): Likewise.
18356         (known_aggs_to_agg_replacement_list): Removed.
18357         (decide_whether_version_node): Re-calculate known constants for all
18358         remaining context clones.
18360 2018-04-11  Richard Biener  <rguenther@suse.de>
18362         PR lto/85339
18363         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
18364         from early DWARF output.
18365         (dwarf2out_early_finish): Output line info unconditionally into
18366         early DWARF and add reference to it.
18368 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18370         PR target/85281
18371         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
18372         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
18373         other than V2DFmode using iptr mode attribute.
18374         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
18376 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
18378         PR rtl-optimization/84659
18379         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
18381 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
18383         PR debug/85302
18384         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
18385         SIZEP is NULL.
18386         (output_loc_list): Pass address of a dummy size variable even in the
18387         locview handling loop.
18388         (index_location_lists): Add comment on why skip_loc_list_entry can't
18389         call size_of_locs.
18391 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18393         PR target/85261
18394         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
18395         into register.
18397 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
18399         PR target/85321
18400         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18401         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
18402         and -mstring-compare-inline-limit.
18404 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18406         PR target/85287
18407         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
18408         for stack clash protection in a register whenever we need it to be in
18409         a register.
18411 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18413         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
18414         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
18416 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
18418         PR target/85321
18419         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
18420         the help text.
18421         (mlong-double-): Ditto.
18422         * config/rs6000/sysv4.opt (msdata=): Ditto.
18423         (mtls-size=): Ditto.
18425 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18427         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18428         erroneous entries for
18429         "vector int vec_ldl (int, long int *)", and
18430         "vector unsigned int vec_ldl (int, unsigned long int *)".
18431         Add comments and entries for
18432         "vector bool char vec_ldl (int, bool char *)",
18433         "vector bool short vec_ldl (int, bool short *)",
18434         "vector bool int vec_ldl (int, bool int *)",
18435         "vector bool long long vec_ldl (int, bool long long *)",
18436         "vector pixel vec_ldl (int, pixel *)",
18437         "vector long long vec_ldl (int, long long *)",
18438         "vector unsigned long long vec_ldl (int, unsigned long long *)".
18439         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
18440         type tree bool_long_long_type_node and correct definition of
18441         bool_V2DI_type_node to make reference to this new type tree.
18442         (rs6000_mangle_type): Replace erroneous reference to
18443         bool_long_type_node with bool_long_long_type_node.
18444         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
18445         comments to emphasize sign distinctions for char and int types and
18446         replace RS6000_BTI_bool_long constant with
18447         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
18448         use of RS6000_BTI_pixel.
18449         (bool_long_type_node): Remove this macro definition.
18450         (bool_long_long_type_node): New macro definition
18452 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
18454         PR rtl-optimization/85300
18455         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
18456         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
18457         simplify_unary_operation fails.
18459 2018-04-10  Martin Liska  <mliska@suse.cz>
18461         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
18462         cgraph_edge and ipa_ref.
18464 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
18466         PR target/85177
18467         PR target/85255
18468         * config/i386/sse.md
18469         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
18470         computation of the VEC_MERGE selector from mask.
18471         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
18472         Fix decoding of the VEC_MERGE selector into mask.
18474 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
18476         PR tree-optimization/85286
18477         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
18479 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
18481         * final.c (final_1): Set insn_last_address as well as
18482         insn_current_address.
18484 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18486         PR target/85173
18487         * explow.c (emit_stack_probe): Call validize_mem on memory location
18488         before passing it to gen_probe_stack.  Create address operand and
18489         legitimize it for the probe_stack_address case.
18491 2018-04-09  Jan Hubicka  <jh@suse.cz>
18493         PR lto/85078
18494         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
18495         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
18496         * tree.c (free_lang_data_in_type): Fix handling of binfos;
18497         walk basetypes.
18498         (free_lang_data): Rebuild type inheritance graph.
18500 2018-04-09  Martin Sebor  <msebor@redhat.com>
18502         * invoke.texi (-finline-small-functions): Mention other optimization
18503         options.
18504         (-findirect-inlining, -fpartial-inlining): Same.
18505         (-finline-functions-called-once): Same.
18506         (-freorder-blocks-and-partition): Same.
18508 2018-04-09  Jan Hubicka  <jh@suse.cz>
18510         PR rtl/84058
18511         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
18512         jumps; choose last target that matches the criteria (i.e.
18513         no partition changes for non-crossing jumps).
18514         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
18515         support for redirecting crossing jumps to non-crossing.
18517 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
18519         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
18520         also for naked functions.
18522 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
18524         * config/arc/arc.md (add_shift): New pattern.
18525         (add_shift2): Likewise.
18526         (sub_shift): Likewise.
18527         (sub_shift_cmp0_noout): Likewise.
18528         (compare_si_ashiftsi): Likewise.
18529         (xbfu_cmp0_noout): New combine pattern.
18530         (xbfu_cmp0"): Likewise.
18531         (movsi_set_cc_insn): Place the predicable variant first.
18532         (commutative_binary_cmp0_noout): Remove clobber.
18533         (commutative_binary_cmp0): New pattern.
18534         (noncommutative_binary_cmp0): Likewise.
18535         (noncommutative_binary_cmp0_noout): Likewise.
18536         (noncommutative_binary_comparison_result_used): Removed.
18537         (rsub_cmp0): New pattern.
18538         (rsub_cmp0_noout): Likewise.
18539         (extzvsi): Changed, keep only meaningful variants.
18540         (SQH, SEZ): New iterators.
18541         (SQH_postfix): New mode attribute.
18542         (SEZ_prefix): New code attribute.
18543         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
18544         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
18545         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
18546         of numerical value.
18547         (noncommutative_operator): Check the availability of barrel
18548         shifter option.
18550 2018-04-09  Richard Biener  <rguenther@suse.de>
18552         PR tree-optimization/85284
18553         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
18554         Only use the niter constraining form of simple_iv when the exit
18555         is always executed.
18557 2018-04-09  Tom de Vries  <tom@codesourcery.com>
18559         PR target/84041
18560         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18561         (define_expand "*memory_barrier"): New define_expand.
18562         (define_insn "memory_barrier"): New insn.
18564 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18566         PR rtl-optimization/80463
18567         PR rtl-optimization/83972
18568         PR rtl-optimization/83480
18570         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
18571         correct producer for the insn.
18572         (tidy_control_flow): Fixup seqnos in case of debug insns.
18574 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18576         PR rtl-optimization/83913
18578         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
18579         different sched-times when merging exprs.
18581 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18583         PR rtl-optimization/83962
18585         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
18586         tidy_fallthru_edge and tidy_control_flow.
18588 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
18590         PR rtl-optimization/83530
18592         * sel-sched.c (force_next_insn): New global variable.
18593         (remove_insn_for_debug): When force_next_insn is true, also leave only
18594         next insn in the ready list.
18595         (sel_sched_region): When the region wasn't scheduled, make another pass
18596         over it with force_next_insn set to 1.
18598 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
18600         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
18601         into tm_file.
18602         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18603         for interrupt control.
18604         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18605         functions for interrupt control.
18606         * config/nds32/nds32-intrinsic.md: Likewise.
18607         * config/nds32/nds32_intrinsic.h: Likewise.
18608         * config/nds32/nds32.h (nds32_builtins): Likewise.
18610 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
18612         * config/nds32/nds32.c (nds32_init_machine_status,
18613         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
18614         strict_aligned_p field.
18615         (nds32_expand_to_rtl_hook): New function.
18616         (TARGET_EXPAND_TO_RTL_HOOK): Define.
18617         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
18619 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18620             Chung-Ju Wu  <jasonwucj@gmail.com>
18622         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
18623         * config/nds32/nds32-n7.md: New file.
18624         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
18625         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
18626         pipeline.
18627         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
18628         * config/nds32/nds32.md (pipeline_model): Add n7.
18629         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
18630         * config/nds32/pipelines.md: Include n7 settings.
18632 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18633             Chung-Ju Wu  <jasonwucj@gmail.com>
18635         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
18636         * config/nds32/nds32-e8.md: New file.
18637         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
18638         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
18639         pipeline.
18640         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
18641         * config/nds32/nds32.md (pipeline_model): Add e8.
18642         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
18643         * config/nds32/pipelines.md: Include e8 settings.
18645 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18646             Chung-Ju Wu  <jasonwucj@gmail.com>
18648         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
18649         * config/nds32/nds32-n8.md: New file.
18650         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
18651         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
18652         pipeline.
18653         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
18654         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
18655         * config/nds32/nds32.md (pipeline_model): Add n8.
18656         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
18657         * config/nds32/pipelines.md: Include n8 settings.
18659 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
18660             Chung-Ju Wu  <jasonwucj@gmail.com>
18662         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
18663         * config/nds32/nds32-n9-2r1w.md: New file.
18664         * config/nds32/nds32-n9-3r2w.md: New file.
18665         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
18666         nds32_register_ports): New or modify for cpu n9.
18667         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
18668         pipeline.
18669         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
18670         * config/nds32/nds32-utils.c: New file.
18671         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
18672         TARGET_MUL_SLOW): Define.
18673         * config/nds32/nds32.md (pipeline_model): New attribute.
18674         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
18675         New options that support cpu n9.
18676         * config/nds32/pipelines.md: Include n9 settings.
18677         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
18679 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
18681         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
18682         information if necessary.
18683         (output_cond_branch_compare_zero): Likewise.
18684         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
18685         (nds32_target_alignment): Refine for alignment.
18686         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
18687         (FUNCTION_BOUNDARY): Modify.
18688         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
18689         align case.
18690         * config/nds32/nds32.opt (malways-align, malign-functions): New.
18692 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
18694         * config/nds32/constants.md (unspec_volatile_element): Add values for
18695         TLB operation and data prefetch.
18696         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18697         functions for TLB operation and data prefetch.
18698         * config/nds32/nds32-intrinsic.md: Likewise.
18699         * config/nds32/nds32_intrinsic.h: Likewise.
18700         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
18701         (nds32_print_operand): Likewise.
18702         * config/nds32/nds32.h (nds32_builtins): Likewise.
18704 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
18705         Andrew Pinski <pinsika@gcc.gnu.org>
18707         PR middle-end/82976
18708         * match.pd: Use constant_boolean_node of correct type instead of
18709         boolean_true_node or boolean_false_node for simplifying
18710         pointer comparisons to zero.
18712 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
18714         PR tree-optimization/80021
18715         * tree.c (verify_type_variant): Make error call in verify_variant_match
18716         translatable and remove final full stop.
18718 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18720         * config/nds32/constants.md (unspec_volatile_element): Add
18721         UNSPEC_VOLATILE_EH_RETURN.
18722         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
18723         nds32_output_stack_pop): Support dwarf exception handling process.
18724         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
18725         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
18726         exception handling process.
18727         (nds32_compute_stack_frame): Likewise.
18728         (nds32_return_addr_rtx): Likewise.
18729         (nds32_initial_elimination_offset): Likewise.
18730         (nds32_expand_prologue): Likewise.
18731         (nds32_expand_epilogue): Likewise.
18732         (nds32_dynamic_chain_address): New function.
18733         * config/nds32/nds32.h (machine_function): Add fields for dwarf
18734         exception handling.
18735         (DYNAMIC_CHAIN_ADDRESS): Define.
18736         (EH_RETURN_DATA_REGNO): Define.
18737         (EH_RETURN_STACKADJ_RTX): Define.
18738         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
18739         patterns for dwarf exception handling.
18741 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18743         * config/nds32/nds32.h: Clean up obsolete macros.
18745 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18747         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18748         Add enum values for particular instructions.
18749         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18750         particular intrinsic functions.
18751         * config/nds32/nds32-intrinsic.md: Likewise.
18752         * config/nds32/nds32_intrinsic.h: Likewise.
18753         * config/nds32/nds32.h (nds32_builtins): Likewise.
18754         * config/nds32/nds32.md (type): Add pbsad and pbsada.
18755         (btst, ave): New patterns for particular instructions.
18757 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18759         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18760         Add enum values for atomic load/store and memory sync.
18761         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
18762         and memory sync.
18763         * config/nds32/nds32-intrinsic.md: Likewise.
18764         * config/nds32/nds32_intrinsic.h: Likewise.
18765         * config/nds32/nds32.h (nds32_builtins): Likewise.
18767 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
18769         PR tree-optimization/85257
18770         * fold-const.c (native_encode_vector): If not all elts could fit
18771         and off is -1, return 0 rather than offset.
18772         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
18773         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
18774         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
18775         adjust buffer in native_interpret_expr call.
18777 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18779         * config/nds32/constants.md (unspec_volatile_element): Add cache
18780         control enum values.
18781         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
18782         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
18783         * config/nds32/nds32.c (nds32_cctl_names): New.
18784         (nds32_print_operand): Handle cache control register names.
18785         * config/nds32/nds32.h (nds32_builtins): New enum values.
18786         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
18787         macros.
18788         * config/nds32/nds32.md (type): Add mmu.
18789         * config/nds32/pipelines.md (simple_insn): Add mmu.
18791 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18793         * config/nds32/nds32.md (type): Remove call.
18794         * config/nds32/pipelines.md (simple_insn): Likewise.
18796 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18798         * config/nds32/constants.md (unspec_volatile_element): Add
18799         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
18800         UNSPEC_VOLATILE_FMFCFG.
18801         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
18802         description for fmfcfg and fmfcsr.
18803         (bdesc_1arg): Add fmtcsr.
18804         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
18805         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18806         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18807         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18808         unspec_fmfcfg): New patterns.
18809         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18810         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18811         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18812         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18813         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18814         __nds32__fmfcfg): Define.
18816 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
18818         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18819         intrinsic register names.
18820         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18821         intrinsic register enum values and macros.
18823 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
18825         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18826         for load/store addressing form.
18827         (nds32_print_operand_address): Likewise.
18829 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
18831         PR target/85196
18832         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18833         based on LABEL_REF.  Remove useless assertion.
18834         (pic_address_needs_scratch): Fix formatting.
18835         (sparc_legitimize_pic_address): Minor tweaks.
18836         (sparc_delegitimize_address): Adjust assertion accordingly.
18837         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18838         into symbolic_operand.
18839         (movsi_high_pic_label_ref): Likewise.
18840         (movsi_lo_sum_pic_label_ref): Likewise.
18841         (movdi_pic_label_ref): Likewise.
18842         (movdi_high_pic_label_ref): Likewise.
18843         (movdi_lo_sum_pic_label_ref): Likewise.
18845 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
18847         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18848         custom LIB_SPEC setup.
18850 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
18851             Kito Cheng  <kito.cheng@gmail.com>
18853         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18854         * config/riscv/freebsd.h: New.
18856 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18858         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18859         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18860         file.
18862 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18863             Kito Cheng  <kito.cheng@gmail.com>
18865         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18866         nds32_output_call, nds32_symbol_binds_local_p): New functions.
18867         * config/nds32/nds32-protos.h (nds32_output_call,
18868         nds32_output_return): Declare.
18869         * config/nds32/nds32.md: Refine all the call and return patterns.
18871 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18873         PR debug/85252
18874         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18875         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18877         PR rtl-optimization/84872
18878         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18879         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18880         EDGE_CROSSING edge.
18882 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
18884         * expr.c (copy_blkmode_to_reg): Revert 254862.
18885         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18887 2018-04-06  Richard Biener  <rguenther@suse.de>
18889         PR middle-end/85244
18890         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18891         after seeing a component reference with an adjacent field.  Treat
18892         refs to arrays at struct end of external decls similar to
18893         refs to unconstrained commons.
18895 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
18897         PR sanitizer/85213
18898         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
18899         look through SAVE_EXPRs with non-side-effects argument.  Adjust
18900         recursive calls.
18901         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18902         save_p here.
18904 2018-04-06  Richard Biener  <rguenther@suse.de>
18906         PR middle-end/85180
18907         * alias.c (find_base_term): New wrapper around find_base_term
18908         unwinding CSELIB_VAL_PTR changes.
18909         (find_base_term): Do not restore CSELIB_VAL_PTR during the
18910         recursion.
18912 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18914         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18915         instructions.
18916         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
18917         constant definitions.
18918         ("nop"): lr 0,0 -> nopr r0
18919         ("nop_lr0", "nop_lr1"): New insn definitions.
18921 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
18923         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
18924         NDS32_V3PUSH_AVAILABLE_P macro.
18926 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
18927             Chung-Ju Wu  <jasonwucj@gmail.com>
18929         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18930         (nds32*-*-*): Add float and fpu_config into supported_defaults.
18931         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18932         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18933         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18934         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18935         * config/nds32/constraints.md: New constraints and checking for hard
18936         float configuration.
18937         * config/nds32/iterators.md: New mode iterator and attribute for hard
18938         float configuration.
18939         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18940         patterns.
18941         * config/nds32/nds32-fpu.md: New file.
18942         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18943         deal with hard float code generation.
18944         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18945         ARCH_V3S.
18946         (abi_type, float_reg_number): New enum type.
18947         * config/nds32/nds32-predicates.c: New predicates for hard float.
18948         * config/nds32/nds32-protos.h: Declare functions for hard float.
18949         * config/nds32/nds32.c: Implementation for hard float configuration.
18950         * config/nds32/nds32.h: Definitions for hard float configuration.
18951         * config/nds32/nds32.md: Include hard float machine description and
18952         modify patterns for hard float configuration.
18953         * config/nds32/nds32.opt: New options for hard float configuration.
18954         * config/nds32/predicates.md: New predicates for hard float
18955         configuration.
18957 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
18959         * common/config/nds32/nds32-common.c
18960         (nds32_option_optimization_table): Enable -mreleax-hint by default.
18962 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
18964         PR middle-end/85195
18965         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18966         CONSTRUCTOR_ELT (ctor, ...)->value.
18968 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
18970         PR target/85193
18971         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18973 2018-04-05  Tom de Vries  <tom@codesourcery.com>
18975         PR target/85204
18976         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18977         cond jump.
18979 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18980             Kito Cheng  <kito.cheng@gmail.com>
18982         * config/nds32/constraints.md (U33): Fine-tune checking condition.
18983         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18984         * config/nds32/nds32.h (nds32_16bit_address_type): Add
18985         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18987 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
18988             Kito Cheng  <kito.cheng@gmail.com>
18990         * config/nds32/constraints.md (Ufe): New memory constraint.
18991         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18992         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18993         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18994         operands.
18995         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18996         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18998 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19000         * config/nds32/nds32.md: Use optimize_size in the condition for
19001         alu-shift instructions.
19003 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19005         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19007 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19009         * config/nds32/nds32.md (negsi2): Refine pattern.
19011 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19012             Chung-Ju Wu  <jasonwucj@gmail.com>
19014         * config/nds32/iterators.md (shift_rotate): New code iterator.
19015         (shift): New code attribute.
19016         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19017         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19018         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19019         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19020         bit-wise operations.
19021         (andsi3, *andsi3): Ditto.
19022         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19023         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19024         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19025         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19026         nds32_ior_operand, nds32_xor_operand): New predicates.
19028 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19030         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19031         (addsi3, subsi3): ... this.
19033 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19035         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19037 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19039         * config/nds32/nds32.md: Adjust indention.
19041 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19043         * config/nds32/nds32.md (feature): New attribute.
19045 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19047         * config/nds32/nds32.md (subtype): New attribute.
19049 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19051         PR target/85203
19052         * config/arm/arm-builtins.c (arm_expand_builtin): Change
19053         expansion to perform a bitwise AND of the argument followed by a
19054         boolean negation of the result.
19056 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
19058         PR rtl-optimization/84878
19059         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19060         the basic block.  Assert the use reference is not artificial and that
19061         it has an associated insn.
19063 2018-04-04  Michael Matz  <matz@suse.de>
19065         * builtins.c (compute_objsize): Pass correct operand
19066         to array_at_struct_end_p.
19068 2018-04-04  Richard Biener  <rguenther@suse.de>
19070         PR lto/85176
19071         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19072         from contexts for DINFO_LEVEL_TERSE and below.
19074 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19076         * config/nds32/nds32-doubleword.md (move_<mode>): Require
19077         resiter_operand condition.
19078         * config/nds32/nds32.md (*move<mode>): Ditto.
19080 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19081             Monk Chiang  <sh.chiang04@gmail.com>
19083         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19085 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19087         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19089 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19090             Kito Cheng  <kito.cheng@gmail.com>
19092         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19093         nds32_cond_code_str, output_cond_branch,
19094         output_cond_branch_compare_zero, nds32_expand_cbranch,
19095         nds32_expand_cstore, nds32_expand_movcc,
19096         nds32_output_cbranchsi4_equality_zero,
19097         nds32_output_cbranchsi4_equality_reg,
19098         nds32_output_cbranchsi4_equality_reg_or_const_int,
19099         nds32_output_cbranchsi4_greater_less_zero: New functions.
19100         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19101         nds32_expand_cstore, nds32_expand_movcc,
19102         nds32_output_cbranchsi4_equality_zero,
19103         nds32_output_cbranchsi4_equality_reg,
19104         nds32_output_cbranchsi4_equality_reg_or_const_int,
19105         nds32_output_cbranchsi4_greater_less_zero): Declare.
19106         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19107         nds32_rimm11s_operand): New predicates.
19108         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
19109         * config/nds32/nds32.md: Rewrite all the branch and conditional move
19110         patterns.
19112 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19114         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
19115         * config/nds32/nds32.md: Ditto.
19116         * config/nds32/pipelines.md: Ditto.
19118 2018-04-04  Richard Biener  <rguenther@suse.de>
19120         PR tree-optimization/85168
19121         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
19122         propagating abnormals.
19124 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19126         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
19128 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19129             Kito Cheng  <kito.cheng@gmail.com>
19131         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
19132         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
19133         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
19134         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
19135         * config/nds32/nds32.md (sibcall_internal): New.
19136         (sibcall_register): Remove.
19137         (sibcall_immediate): Remove.
19138         (sibcall_value_internal): New.
19139         (sibcall_value_register): Remove.
19140         (sibcall_value_immediate): Remove.
19141         * config/nds32/predicates.md (nds32_general_register_operand): New.
19142         (nds32_call_address_operand): New.
19144 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19146         PR rtl-optimization/85167
19147         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
19148         bb_defs if *split_p, instead preinitialize it to NULL.
19150         PR tree-optimization/85156
19151         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
19152         evaluating the argument multiple times.
19154 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
19156         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
19157         than vector.
19158         (_mm_cvtpd_ps): Likewise.
19159         (_mm_cvttpd_epi32): Likewise.
19160         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
19161         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
19162         vector, pixel, and bool following altivec.h include.
19164 2018-04-03  Martin Sebor  <msebor@redhat.com>
19166         * doc/extend.texi (Common Function Attributes): Clarify.
19167         (const attribute): Likewise.
19168         (pure attribute): Likewise.
19170 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19172         PR target/85169
19173         * config/i386/i386.c (ix86_expand_vector_set): Use
19174         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
19176 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
19178         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
19179         instructions when changing rounding bits to preserve precision bits
19180         in the x87 control word.
19182 2018-04-03  Martin Liska  <mliska@suse.cz>
19184         PR tree-optimization/82491
19185         * rtl.h (strip_offset_and_add): Replace += suboffset with
19186         poly_uint64 () + suboffset.
19188 2018-03-29  Martin Liska  <mliska@suse.cz>
19189             Martin Jambor  <mjambor@suse.cz>
19191         PR ipa/84947
19192         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
19193         param_type is not an integral or pointer type.
19195 2018-04-03  Richard Biener  <rguenther@suse.de>
19197         * sese.h (recompute_all_dominators): Remove.
19199 2018-04-02  Martin Sebor  <msebor@redhat.com>
19201         * doc/invoke.texi (-Wrestrict): Fix typos.
19203 2018-04-02  Jim Wilson  <jimw@sifive.com>
19205         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
19206         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
19207         (<optab>di3, <optab>si3_extend): Likewise.
19208         (<optab>si3_mask, <optab>si3_mask_1): New.
19209         (<optab>di3_mask, <optab>di3_mask_1): New.
19210         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
19211         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
19212         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
19214 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
19216         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
19217         example.
19219 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19221         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
19222         (nds32_canonicalize_comparison): New function.
19224 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19225             Kito Cheng  <kito.cheng@gmail.com>
19226             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19228         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
19229         * config/nds32/constants.md (unspec_volatile_element): Add
19230         UNSPEC_VOLATILE_RELAX_GROUP.
19231         * config/nds32/nds32-relax-opt.c: New file.
19232         * config/nds32/nds32-predicates.c
19233         (nds32_symbol_load_store_p): New function.
19234         * config/nds32/nds32-protos.h
19235         (nds32_symbol_load_store_p): Declare function.
19236         (make_pass_nds32_relax_opt): Declare new rtl pass function.
19237         * config/nds32/nds32.c
19238         (nds32_register_pass): New function to register pass.
19239         (nds32_register_passes): New function to register passes.
19240         * config/nds32/nds32.md (relax_group): New pattern.
19241         * config/nds32/nds32.opt (mrelax-hint): New option.
19242         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
19244 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
19246         * config/nds32/t-nds32: Modify files dependency.
19248 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19250         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
19251         (PROFILE_HOOK): Define its implementation.
19253 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19255         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
19256         type and 32-bit size.
19258 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
19260         PR middle-end/85090
19261         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
19262         (V_128_256): New mode iterator.
19263         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
19264         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
19265         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
19266         of V.
19267         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
19268         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
19270 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
19272         PR target/83315
19273         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
19274         NaN inputs correctly.
19276 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
19278         PR target/80546
19279         * config/rs6000/vsx.md (??r): New mode attribute.
19280         (*vsx_mov<mode>_64bit): Use it.
19281         (*vsx_mov<mode>_32bit): Likewise.
19283 2018-03-30  Martin Sebor  <msebor@redhat.com>
19285         PR tree-optimization/84818
19286         * builtins.c (check_access): Use warning_n.
19288 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19290         PR target/83822
19291         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
19292         condition.
19293         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
19294         condition.
19296 2018-03-30  Julia Koval  <julia.koval@intel.com>
19298         PR target/84413
19299         * x86-tune.def (movx, partial_reg_dependency): Enable for
19300         m_SKYLAKE_AVX512.
19302 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
19304         PR inline-asm/84985
19305         * lra-constraints.c (process_alt_operands): Move setting
19306         this_alternative_matches below.
19308 2018-03-29  Martin Liska  <mliska@suse.cz>
19310         PR lto/84995.
19311         * doc/invoke.texi: Document how LTO works with debug info.
19312         Describe auto-load support of binutils.  Mention 'x86-64'
19313         as valid option value of -march option.
19315 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
19317         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
19319         PR c/85094
19320         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
19321         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
19322         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
19323         checking.
19325 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
19327         PR target/84912
19328         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
19329         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
19330         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
19331         for RS6000_BTM_POWERPC64.
19332         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
19333         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
19334         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
19335         definition.
19336         (DIVDE): Use it.
19337         (DIVDEU): Likewise.
19339 2018-03-28  Carl Love  <cel@us.ibm.com>
19341         Revert
19342         2017-09-27  Carl Love  <cel@us.ibm.com>
19344         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
19345         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
19346         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
19347         fctiw instruction.
19349 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19351         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
19352         instead of __vector bool.
19353         (_mm_max_pu8): Likewise.
19354         (_mm_min_pi16): Likewise.
19356 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
19358         PR target/84912
19359         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
19360         (DIVWEUO): Likewise.
19361         (DIVDEO): Likewise.
19362         (DIVDEUO): Likewise.
19363         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
19364         DIVWEUO and DIVDEUO.
19365         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
19366         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
19367         (div_extend): Likewise.
19368         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
19369         builtin function.
19370         (__builtin_divweuo): Likewise.
19371         (__builtin_divdeo): Likewise.
19372         (__builtin_divdeuo): Likewise.
19374 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
19376         PR target/85095
19377         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
19378         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
19380         PR tree-optimization/82004
19381         * gimple-match-head.c (optimize_pow_to_exp): New function.
19382         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
19383         Don't fold to exp if optimize_pow_to_exp is false.
19385 2018-03-28  Martin Liska  <mliska@suse.cz>
19387         PR other/84819
19388         * calls.c (initialize_argument_information): Fix trailing space.
19389         * common.opt: Fix typo and provide better explanation for
19390         -fsanitize-coverage option.
19391         * config/i386/i386.opt: Fix typo.
19393 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
19394             Martin Liska  <mliska@suse.cz>
19396         PR sanitizer/85081
19397         * gimplify.c (asan_poison_variable): Don't do the check for
19398         gimplify_omp_ctxp here.
19399         (gimplify_decl_expr): Do it here.
19400         (gimplify_target_expr): Likewise.
19402 2018-03-28  Martin Liska  <mliska@suse.cz>
19404         PR target/84988
19405         * config/i386/i386.c (ix86_function_arg_advance): Do not call
19406         chkp_type_bounds_count if MPX is not enabled.
19408 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
19410         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
19412 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
19414         PR target/84914
19415         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
19416         function to create the function decl for complex long double
19417         multiply and divide for -mabi=ieeelongdouble.
19418         (init_float128_ieee): Call it.
19420 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
19422         PR target/85044
19423         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
19424         -fcf-protection=branch -mibt.
19425         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
19427 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19429         PR target/81863
19430         * config/arm/arm.c (arm_valid_symbolic_address): Handle
19431         arm_word_relocations.
19433 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
19435         PR target/85056
19436         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
19437         extern array declarations.
19439 2018-03-27  Richard Biener  <rguenther@suse.de>
19441         PR middle-end/84067
19442         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
19443         explicit single_use checks.
19445 2018-03-27  Richard Biener  <rguenther@suse.de>
19447         PR tree-optimization/85082
19448         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
19449         Valueize the VUSE.
19451 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19453         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
19454         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
19455         Turn on fasynchronous-unwind-tables and funwind-tables.
19457 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
19459         PR target/85073
19460         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
19461         (*bmi_blsr_<mode>_ccz): Ditto.
19463 2018-03-26  Tom de Vries  <tom@codesourcery.com>
19465         PR tree-optimization/85063
19466         * omp-general.c (offloading_function_p): New function.  Factor out
19467         of ...
19468         * omp-offload.c (pass_omp_target_link::gate): ... here.
19469         * omp-general.h (offloading_function_p): Declare.
19470         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
19471         with attribute omp declare target for offloading functions.
19473 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
19475         PR tree-optimization/84005
19476         * tree-data-ref.h (get_base_for_alignment): Declare.
19477         * tree-data-ref.c (get_base_for_alignment_1): New function.
19478         (get_base_for_alignment): Likewise.
19479         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
19480         get_base_for_alignment to find a suitable base object, instead
19481         of always using drb->base_address.
19483 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
19485         PR inline-asm/85022
19486         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
19487         known size by default.
19489 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
19491         PR inline-asm/85030
19492         * lra-constraints.c (process_alt_operands): Don't match BLKmode
19493         and non BLKmode operands.
19495 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19497         PR target/85026
19498         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
19499         Clean up attributes.
19501 2018-03-23  Richard Biener  <rguenther@suse.de>
19503         PR debug/85020
19504         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
19505         we are going to emit early debug for LTO.
19507 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
19509         PR inline-asm/85034
19510         * function.c (match_asm_constraints_1): Don't optimize if input
19511         doesn't satisfy general_operand predicate for output's mode.
19513         PR inline-asm/85022
19514         * alias.c (write_dependence_p): Don't require for x_canonicalized
19515         non-VOIDmode if x has VOIDmode.
19517         PR sanitizer/85029
19518         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
19519         just don't try to optimize it rather than assert it never happens.
19521 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19523         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
19524         macro expansions for definition of ST_INTERNAL_<mode> and
19525         LD_INTERNAL_<mode> builtins.
19526         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
19527         Remove prototype.
19528         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
19529         function.
19530         (altivec_expand_st_builtin): Likewise.
19531         (altivec_expand_builtin): Remove calls to deleted functions.
19532         (rs6000_address_for_altivec): Delete this function.
19533         * config/rs6000/vector.md: Remove expands for
19534         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
19536 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
19538         PR target/84826
19539         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
19540         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
19541         re-computing once computed.
19542         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
19543         (arm_init_machine_status): Initialize
19544         machine->static_chain_stack_bytes.
19546 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19548         PR target/84760
19549         * doc/extend.texi: Add four new prototypes for vec_ld.
19550         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
19551         definitions for more logical presentation.
19552         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
19553         entries for V1TI variants of __builtin_altivec_ld builtin.
19554         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
19555         handling of V1TI variant of LVX icode pattern.
19556         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
19557         (rs6000_gimple_fold_builtin): Likewise.
19558         (altivec_init_builtins): Add code to define
19559         __builtin_altivec_lvx_v1ti function.
19561 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
19563         PR inline-asm/84941
19564         * function.c (match_asm_constraints_1): Don't do the optimization
19565         if input isn't a REG, SUBREG, MEM or constant.
19567 2018-03-22  Tom de Vries  <tom@codesourcery.com>
19569         PR tree-optimization/84956
19570         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
19571         bb_has_abnormal_pred.
19573 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
19575         PR sanitizer/85018
19576         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
19577         DECL_INITIAL (decl) to decl at the end.
19578         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
19579         adjust the comment.
19581 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
19583         * doc/extend.texi (__builtin_tgmath): Document when complex
19584         integer types are treated as _Complex _Float64.
19586 2018-03-21  Tom de Vries  <tom@codesourcery.com>
19588         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
19590 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19592         PR tree-optimization/84960
19593         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
19594         if it is ENTRY block, move them into single succ of ENTRY in that case.
19596 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
19598         PR tree-optimization/84811
19599         * poly-int.h (poly_span_traits): Remove the T3 parameter and
19600         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
19601         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
19602         (known_subrange_p): Update accordingly.  Cast each value involved
19603         in the size comparison, rather than casting the result of the
19604         subtraction.
19606 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19608         PR tree-optimization/84982
19609         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
19610         by flipping the least significant bit rather than all bits from
19611         bitpos to bitpos + bitsize - 1.
19613 2018-03-21  Nathan Sidwell  <nathan@acm.org>
19615         * doc/extend.texi (Deprecated Features): Remove mention of
19616         long-deleted deprecations.
19618 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
19620         PR jit/84288
19621         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
19622         * configure: Regenerate.
19624 2018-03-21  Tom de Vries  <tom@codesourcery.com>
19626         PR tree-optimization/83126
19627         * tree-parloops.c (num_phis): New function.
19628         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
19630 2018-03-21  Nathan Sidwell  <nathan@acm.org>
19632         * doc/extend.texi (Deprecated Features): Update deprecated flags,
19633         mention anon-struct/union members and trailing attributes.
19635 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
19637         PR tree-optimization/84969
19638         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
19639         builtin memset partitions if they set different rhs values.
19641 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
19643         PR rtl-optimization/84989
19644         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
19645         VEC_DUPLICATE with scalar result mode.
19647 2018-03-21  Martin Liska  <mliska@suse.cz>
19649         PR ipa/84963
19650         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
19651         not intended return statement.
19653 2018-03-21  Martin Liska  <mliska@suse.cz>
19655         PR target/84988
19656         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
19657         (chkp_find_bound_slots_1): Limit number of iterations.
19659 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
19661         PR target/84838
19662         * Minor grammar fixes for x86 options.
19664 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19666         PR debug/84875
19667         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
19668         holding REG_CFA_RESTORE notes, instead turn them into a USE.
19670 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
19672         PR target/83789
19673         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
19674         (altivec_lvx_<mode>_1op): Likewise.
19675         (altivec_stvx_<mode>_2op): Likewise.
19676         (altivec_stvx_<mode>_1op): Likewise.
19677         (altivec_lvx_<VM2:mode>): New define_expand.
19678         (altivec_stvx_<VM2:mode>): Likewise.
19679         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
19680         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19681         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
19682         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19683         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
19684         (rs6000_gen_lvx): Likewise.
19685         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
19686         (altivec_expand_stv_builtin): Likewise.
19687         (altivec_expand_builtin): Likewise.
19688         * config/rs6000/vector.md: Likewise.
19690 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19692         PR target/82518
19693         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
19694         BYTES_BIG_ENDIAN.
19696 2018-03-20  Richard Biener  <rguenther@suse.de>
19698         PR target/84986
19699         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
19700         sign-conversions as zero, fall back to standard scalar_stmt
19701         cost for the rest.
19703 2018-03-20  Martin Liska  <mliska@suse.cz>
19705         PR ipa/84825
19706         * predict.c (rebuild_frequencies): Handle case when we have
19707         PROFILE_ABSENT, but flag_guess_branch_prob is false.
19709 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19711         PR target/84990
19712         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
19713         flag_section_anchors.
19714         * varasm.c (use_blocks_for_decl_p): Remove hack for
19715         dw2_force_const_mem.
19717         PR target/84845
19718         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
19719         to ...
19720         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
19721         be created, use lowpart_subreg of operands[0] rather than operands[0]
19722         itself.
19723         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
19724         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
19725         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
19726         and n constraint instead of aarch64_shift_imm_di and Usd.
19727         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
19728         (*aarch64_<optab>_reg_minus<mode>3): ... this.
19730 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
19732         PR target/82989
19733         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
19734         to favor GPR over NEON registers.
19735         (<shift>di3_neon): Likewise.
19737 2018-03-20  Tom de Vries  <tom@codesourcery.com>
19739         PR target/84952
19740         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
19741         (nvptx_process_pars): Emit bar.sync asap and alap.
19743 2018-03-20  Tom de Vries  <tom@codesourcery.com>
19745         PR target/84954
19746         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
19747         seen_label if seen_label is already set.
19749 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
19751         PR target/84945
19752         * config/i386/i386.c (fold_builtin_cpu): For features above 31
19753         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
19754         Use 1U instead of 1.  Formatting fixes.
19756         PR c/84953
19757         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
19758         instead of TREE_TYPE (s1) for the return value.
19760 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
19762         PR tree-optimization/84946
19763         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
19764         bitsize + bitsize in poly_uint64 rather than poly_int64.
19766         PR sanitizer/78651
19767         * dwarf2asm.c: Include fold-const.c.
19768         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
19769         of decl rather than decl itself.
19771         PR rtl-optimization/84643
19772         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
19774 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
19776         PR sanitizer/78651
19777         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
19778         calling assemble_variable.
19780 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
19782         PR target/81647
19783         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
19784         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
19786 2018-03-19  Jim Wilson  <jimw@sifive.com>
19788         PR bootstrap/84856
19789         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
19790         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
19791         (riscv_first_stack_step): Likewise.
19792         (riscv_option_override): Use STACK_BOUNDARY instead of
19793         MIN_STACK_BOUNDARY.
19794         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
19795         MIN_STACK_BOUNDARY.
19796         (BIGGEST_ALIGNMENT): Set to 128.
19797         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19798         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
19799         STACK_BOUNDARY.
19801 2018-03-19  Richard Biener  <rguenther@suse.de>
19803         PR tree-optimization/84933
19804         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
19805         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19807 2018-03-19  Richard Biener  <rguenther@suse.de>
19809         PR tree-optimization/84859
19810         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19811         (cond_if_else_store_replacement): Perform sinking operation on
19812         single-store BBs regardless of MAX_STORES_TO_SINK setting.
19813         Generalize what a BB with a single eligible store is.
19815 2018-03-19  Richard Biener  <rguenther@suse.de>
19817         PR tree-optimization/84929
19818         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19819         chrec_is_positive against non-chrec arg.
19821 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
19823         PR target/84711
19824         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19826 2018-03-18  Martin Liska  <mliska@suse.cz>
19828         PR rtl-optimization/84635
19829         * regrename.c (build_def_use): Use matches_mode only when
19830         matches >= 0.
19832 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
19834         PR tree-optimization/84913
19835         * tree-vect-loop.c (vectorizable_reduction): Don't try to
19836         vectorize chains of COND_EXPRs.
19838 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19840         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19842 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19844         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19846 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
19848         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19850 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19851             Kito Cheng  <kito.cheng@gmail.com>
19853         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19854         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19855         (nds32_adjust_reg_alloc_order): New function.
19856         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19858 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
19860         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19861         nds32_print_operand, nds32_print_operand_address): Use
19862         HOST_WIDE_INT_PRINT_DEC instead.
19864 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
19866         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19868 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
19870         PR target/84902
19871         * config/i386/i386.c (initial_ix86_tune_features,
19872         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19873         unsigned long long.
19874         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19875         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19876         rather than 1u << ix86_tune.  Formatting fix.
19877         (ix86_option_override_internal): Change ix86_arch_mask from
19878         unsigned int to unsigned HOST_WIDE_INT, initialize to
19879         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19880         (ix86_function_specific_restore): Likewise.
19882 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19884         PR target/84899
19885         * postreload.c (reload_combine_recognize_pattern): Perform
19886         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19887         truncate_int_for_mode the result for the destination's mode.
19889         PR c/84909
19890         * hsa-gen.c (mem_type_for_type): Fix comment typo.
19891         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19892         Likewise.
19893         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19894         Likewise.
19896 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
19898         PR target/84876
19899         * lra-assigns.c (lra_split_hard_reg_for): Don't use
19900         regno_allocno_class_array and sorted_pseudos.
19901         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
19902         insns where regno is used.
19904 2018-03-16  Martin Liska  <mliska@suse.cz>
19906         PR ipa/84833
19907         * multiple_target.c (create_dispatcher_calls): Redirect
19908         reference in the symbol table.
19910 2018-03-16  Martin Liska  <mliska@suse.cz>
19912         PR ipa/84722
19913         * multiple_target.c (create_dispatcher_calls): Redirect also
19914         an alias.
19916 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19918         PR c++/79937
19919         PR c++/82410
19920         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
19921         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
19922         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
19924 2018-03-16  Julia Koval  <julia.koval@intel.com>
19926         * doc/invoke.texi (Skylake Server): Add CLWB.
19927         Cannonlake): Remove CLWB.
19929 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
19931         PR tree-optimization/84841
19932         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19933         1 << 3.
19934         (FLOAT_ONE_CONST_TYPE): Define.
19935         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19936         (sort_by_operand_rank): Put entries with higher constant_type last
19937         rather than first to match comments.
19939 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
19941         * config/nios2/nios2.md (movsi_internal): Fix thinko in
19942         split predicate.
19944 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
19946         PR c++/79085
19947         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19948         check and use address of target always.
19950 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
19952         PR target/84574
19953         * config/i386/i386.c (indirect_thunk_needed): Update comments.
19954         (indirect_thunk_bnd_needed): Likewise.
19955         (indirect_thunks_used): Likewise.
19956         (indirect_thunks_bnd_used): Likewise.
19957         (indirect_return_needed): New.
19958         (indirect_return_bnd_needed): Likewise.
19959         (output_indirect_thunk_function): Add a bool argument for
19960         function return.
19961         (output_indirect_thunk_function): Don't generate alias for
19962         function return thunk.
19963         (ix86_code_end): Call output_indirect_thunk_function to generate
19964         function return thunks.
19965         (ix86_output_function_return): Set indirect_return_bnd_needed
19966         and indirect_return_needed instead of indirect_thunk_bnd_needed
19967         and indirect_thunk_needed.
19969 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
19971         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19972         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19973         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19975 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
19976             Paul Hua <paul.hua.gm@gmail.com>
19978         PR c/84852
19979         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19981 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
19983         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19984         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19985         resp. SFmode cases.
19987 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
19989         PR target/84711
19990         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19991         instead of GET_MODE_SIZE when comparing Units.
19993 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
19995         PR target/68256
19996         * varasm.c (hash_section): Return an unchangeble hash value
19997         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19998         Return !aarch64_can_use_per_function_literal_pools_p ().
20000 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20002         PR target/84860
20003         * optabs.c (emit_conditional_move): Pass address of cmode's copy
20004         rather than address of cmode as last argument to prepare_cmp_insn.
20006 2018-03-15  Julia Koval  <julia.koval@intel.com>
20008         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20009         F_AVX512VNNI, F_AVX512BITALG): New.
20011 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
20013         PR target/83451
20014         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
20015         insn for floating-point loads and stores.
20017 2018-03-14  Carl Love  <cel@us.ibm.com>
20019         * config/rs6000/rs6000-c.c: Add macro definitions for
20020         ALTIVEC_BUILTIN_VEC_PERMXOR.
20021         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20022         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20023         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20024         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20025         UNSPEC_VPERMXOR.
20026         * config/doc/extend.texi: Add prototypes for vec_permxor.
20028 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
20030         PR c/84852
20031         * diagnostic-show-locus.c (class layout_point): Convert m_line
20032         from int to linenum_type.
20033         (line_span::comparator): Use linenum "compare" function when
20034         comparing line numbers.
20035         (test_line_span): New function.
20036         (layout_range::contains_point): Convert param "row" from int to
20037         linenum_type.
20038         (layout_range::intersects_line_p): Likewise.
20039         (layout::will_show_line_p): Likewise.
20040         (layout::print_source_line): Likewise.
20041         (layout::should_print_annotation_line_p): Likewise.
20042         (layout::print_annotation_line): Likewise.
20043         (layout::print_leading_fixits): Likewise.
20044         (layout::annotation_line_showed_range_p): Likewise.
20045         (struct line_corrections): Likewise for field m_row.
20046         (line_corrections::line_corrections): Likewise for param "row".
20047         (layout::print_trailing_fixits): Likewise.
20048         (layout::get_state_at_point): Likewise.
20049         (layout::get_x_bound_for_row): Likewise.
20050         (layout::print_line): Likewise.
20051         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
20052         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
20053         * input.c (selftest::test_linenum_comparisons): New function.
20054         (selftest::input_c_tests): Call it.
20055         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
20056         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
20057         * selftest.h (ASSERT_GT): New macro.
20058         (ASSERT_GT_AT): New macro.
20059         (ASSERT_LT): New macro.
20060         (ASSERT_LT_AT): New macro.
20062 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
20064         PR rtl-optimization/84780
20065         * combine.c (distribute_links): Don't make a link based on pc_rtx.
20067 2018-03-14  Martin Liska  <mliska@suse.cz>
20069         * tree.c (record_node_allocation_statistics): Use
20070         get_stats_node_kind.
20071         (get_stats_node_kind): New function extracted from
20072         record_node_allocation_statistics.
20073         (free_node): Use get_stats_node_kind.
20075 2018-03-14  Richard Biener  <rguenther@suse.de>
20077         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20078         that the value-set of ANTIC_IN doesn't grow.
20080         Revert
20081         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20082         member.
20083         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20084         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20086 2018-03-14  Julia Koval  <julia.koval@intel.com>
20088         * config.gcc (icelake-client, icelake-server): New.
20089         (icelake): Remove.
20090         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20091         (initial_ix86_arch_features): Ditto.
20092         (PTA_SKYLAKE): Add SGX.
20093         (PTA_ICELAKE): Remove.
20094         (PTA_ICELAKE_CLIENT): New.
20095         (PTA_ICELAKE_SERVER): New.
20096         (ix86_option_override_internal): Split up icelake on icelake client and
20097         icelake server.
20098         (get_builtin_code_for_version): Ditto.
20099         (fold_builtin_cpu): Ditto.
20100         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20101         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20102         * config/i386/i386.h (processor_type): Ditto.
20103         * doc/invoke.texi: Ditto.
20105 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
20107         PR sanitizer/83392
20108         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
20109         INTEGER_CST offset, add it together with bitpos / 8 and
20110         sign extend based on POINTER_SIZE.
20112         PR target/84844
20113         Revert
20114         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
20116         PR target/78090
20117         * config/i386/constraints.md (Yc): New register constraint.
20118         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
20119         Use Yc constraint for alternative 2 of operand 0.  Remove
20120         preferred_for_speed attribute.
20122 2018-03-14  Richard Biener  <rguenther@suse.de>
20124         PR tree-optimization/84830
20125         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
20126         with the old one to avoid oscillations.
20128 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
20130         PR target/83712
20131         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
20132         pseudos.
20133         (assign_by_spills): Return a flag of reload assignment failure.
20134         Do not process the reload assignment failures.  Do not spill other
20135         reload pseudos if they has the same reg class.  Update n if
20136         necessary.
20137         (lra_assign): Add a return arg.  Set up from the result of
20138         assign_by_spills call.
20139         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20140         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20141         usage_insns if it is not NULL.
20142         (spill_hard_reg_in_range): New function.
20143         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20144         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20145         function prototypes.
20146         (lra_assign): Change prototype.
20147         * lra.c (lra): Add code to deal with fails by splitting hard reg
20148         live ranges.
20150 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
20152         * config/riscv/riscv.opt (mrelax): New option.
20153         * config/riscv/riscv.c (riscv_file_start): Emit ".option
20154         "norelax" when riscv_mrelax is disabled.
20155         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
20157 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20159         PR target/84743
20160         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
20161         reassociation for int modes.
20163 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20165         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
20166         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
20167         for big-endian.
20168         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
20169         * config/aarch64/aarch64-sve.md
20170         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
20171         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
20172         (*extend<mode><Vwide>2): Rename to...
20173         (aarch64_sve_extend<mode><Vwide>2): ...this.
20174         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
20175         renaming the old pattern to...
20176         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
20177         unsigned packs.
20178         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
20179         define_expand, renaming the old pattern to...
20180         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
20181         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
20182         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
20183         account when deciding which SVE instruction the optab should use.
20184         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
20186 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20188         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
20189         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
20190         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
20191         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
20192         (tlsdesc_small_<mode>): Turn a define_expand and use
20193         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
20194         (tlsdesc_small_advsimd_<mode>): ...this.
20195         (tlsdesc_small_sve_<mode>): New pattern.
20197 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20199         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
20200         (UNSPEC_UMUL_HIGHPART): New constants.
20201         (MUL_HIGHPART): New int iteraor.
20202         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
20203         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20204         define_expand.
20205         (*<su>mul<mode>3_highpart): New define_insn.
20207 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20209         PR lto/84805
20210         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20211         incomplete types.
20213 2018-03-13  Martin Liska  <mliska@suse.cz>
20215         PR ipa/84658.
20216         * (sem_item_optimizer::sem_item_optimizer): Initialize new
20217         vector.
20218         (sem_item_optimizer::~sem_item_optimizer): Release it.
20219         (sem_item_optimizer::merge_classes): Register variable aliases.
20220         (sem_item_optimizer::fixup_pt_set): New function.
20221         (sem_item_optimizer::fixup_points_to_sets): Likewise.
20222         * ipa-icf.h: Declare new variables and functions.
20224 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
20226         PR middle-end/84834
20227         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
20228         integer_pow2p@2 and test integer_pow2p in condition.
20229         (A < 0 ? C : 0): Similarly for @1.
20231         PR middle-end/84831
20232         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
20233         characters starting at p contain '\0' character, don't look beyond
20234         that.
20236         PR target/84827
20237         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
20238         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
20240         PR target/84828
20241         * reg-stack.c (change_stack): Change update_end var from int to
20242         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
20243         also call set_block_for_insn on the newly added insns and rescan.
20245         PR target/84786
20246         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
20247         on the last operand.
20249         PR c++/84704
20250         * tree.c (stabilize_reference_1): Return save_expr (e) for
20251         STATEMENT_LIST even if it doesn't have side-effects.
20253 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
20255         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
20257 2018-03-12  Renlin Li  <renlin.li@arm.com>
20259         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
20260         aarch64_output_scalar_simd_mov_immediate.
20262 2018-03-12  Martin Sebor  <msebor@redhat.com>
20264         PR tree-optimization/83456
20265         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
20266         for perfectly overlapping calls to memcpy.
20267         (gimple_fold_builtin_memory_chk): Same.
20268         (gimple_fold_builtin_strcpy): Handle no-warning.
20269         (gimple_fold_builtin_stxcpy_chk): Same.
20270         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
20272 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20274         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
20275         parameter.  Use it for SFmode.
20276         (rs6000_function_arg_advance_1): Adjust.
20277         (rs6000_function_arg): Adjust.
20278         (rs6000_gimplify_va_arg): Pass false for that new parameter.
20280 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20282         PR rtl-optimization/84169
20283         PR rtl-optimization/84780
20284         * combine.c (can_combine_p): Check for a 2-insn combination whether
20285         the destination register is used between the two insns, too.
20287 2018-03-12  Richard Biener  <rguenther@suse.de>
20289         PR tree-optimization/84803
20290         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20291         for refs DR analysis didn't process.
20293 2018-03-12  Richard Biener  <rguenther@suse.de>
20295         PR tree-optimization/84777
20296         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
20297         force-vectorize loops ignore whether we are optimizing for size.
20299 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
20301         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
20302         (TARGET_MD_ASM_ADJUST): Define.
20304 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
20305             Kito Cheng  <kito.cheng@gmail.com>
20306             Chung-Ju Wu  <jasonwucj@gmail.com>
20308         * config/nds32/nds32.c (nds32_compute_stack_frame,
20309         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
20310         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
20311         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
20312         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
20313         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
20314         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
20315         * config/nds32/nds32.md (prologue, epilogue): Use macro
20316         NDS32_V3PUSH_AVAILABLE_P to do checking.
20318 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
20320         PR debug/58150
20321         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
20322         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
20323         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
20324         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
20325         addition of most attributes on !orig_type_die or the attribute not
20326         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
20328 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20329             Chung-Ju Wu  <jasonwucj@gmail.com>
20331         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
20332         __NDS32_VH__ macro.
20333         * config/nds32/nds32.opt (mvh): New option.
20335 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20336             Chung-Ju Wu  <jasonwucj@gmail.com>
20338         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
20339         function.
20340         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
20341         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
20342         definition.
20344 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
20345             Chung-Ju Wu  <jasonwucj@gmail.com>
20347         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
20348         function.
20349         * config/nds32/nds32-multiple.md (strlensi): New pattern.
20350         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
20352 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
20353             Kito Cheng  <kito.cheng@gmail.com>
20354             Chung-Ju Wu  <jasonwucj@gmail.com>
20356         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
20357         UNSPEC_FFMISM and UNSPEC_FLMISM.
20358         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
20359         for ffb, ffmism and flmism.
20360         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
20361         (unspec_ffmism): Ditto.
20362         (unspec_flmism): Ditto.
20363         (nds32_expand_builtin_impl): Check if string extension is available.
20364         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
20365         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
20367 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
20369         Reverting patch:
20370         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
20372         PR target/83712
20373         * lra-assigns.c (assign_by_spills): Return a flag of reload
20374         assignment failure.  Do not process the reload assignment
20375         failures.  Do not spill other reload pseudos if they has the same
20376         reg class.
20377         (lra_assign): Add a return arg.  Set up from the result of
20378         assign_by_spills call.
20379         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20380         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20381         usage_insns if it is not NULL.
20382         (spill_hard_reg_in_range): New function.
20383         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20384         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20385         function prototypes.
20386         (lra_assign): Change prototype.
20387         * lra.c (lra): Add code to deal with fails by splitting hard reg
20388         live ranges.
20390 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
20392         PR target/84807
20393         * config/i386/i386.opt: Replace Enforcment with Enforcement.
20395 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
20397         PR debug/84620
20398         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
20399         (dw_val_node): Add val_symbolic_view.
20400         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
20401         (symview_upper_bound): New.
20402         (new_line_info_table): Initialize symviews_since_reset.
20403         (dwarf2out_source_line): Count symviews_since_reset and set
20404         symview_upper_bound.
20405         (dw_val_equal_p): Handle symview.
20406         (add_AT_symview): New.
20407         (print_dw_val): Handle symview.
20408         (attr_checksum, attr_checksum_ordered): Likewise.
20409         (same_dw_val_p, size_of_die): Likewise.
20410         (value_format, output_die): Likewise.
20411         (add_high_low_attributes): Use add_AT_symview for entry_view.
20412         (dwarf2out_finish): Reset symview_upper_bound, clear
20413         zero_view_p.
20415 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
20417         PR target/83969
20418         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
20419         Add strict argument and use it.
20420         (rs6000_split_multireg_move): Update for new strict argument.
20421         (mem_operand_gpr): Disallow all non-offsettable addresses.
20422         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
20424 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
20426         PR target/84772
20427         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
20428         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
20429         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
20431         PR c++/84767
20432         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
20433         decl, use remap_type if we want to use the type.
20435 2018-03-09  Martin Sebor  <msebor@redhat.com>
20437         PR tree-optimization/84526
20438         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20439         Remove dead code.
20440         (builtin_access::generic_overlap): Be prepared to handle non-array
20441         base objects.
20443 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
20445         PR rtl-optimization/84682
20446         * lra-constraints.c (process_address_1): Check is_address flag
20447         for address constraints.
20448         (process_alt_operands): Likewise.
20449         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
20450         preprocess_constraints.
20451         * recog.h (preprocess_constraints): Add oploc parameter.
20452         Adjust callers.
20453         * recog.c (preprocess_constraints): Test address_operand for
20454         CT_ADDRESS constraints.
20456 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
20458         PR target/83712
20459         * lra-assigns.c (assign_by_spills): Return a flag of reload
20460         assignment failure.  Do not process the reload assignment
20461         failures.  Do not spill other reload pseudos if they has the same
20462         reg class.
20463         (lra_assign): Add a return arg.  Set up from the result of
20464         assign_by_spills call.
20465         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20466         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20467         usage_insns if it is not NULL.
20468         (spill_hard_reg_in_range): New function.
20469         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20470         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20471         function prototypes.
20472         (lra_assign): Change prototype.
20473         * lra.c (lra): Add code to deal with fails by splitting hard reg
20474         live ranges.
20476 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20478         PR target/83193
20479         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
20480         Accept complain bool parameter.  Only emit errors if it is true.
20481         (arm_parse_cpu_option_name): Likewise.
20482         (arm_target_thumb_only): Adjust callers of the above.
20483         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
20484         prototype to take a default true bool parameter.
20485         (arm_parse_arch_option_name): Likewise.
20487 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
20488             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
20490         PR jit/64089
20491         PR jit/84288
20492         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
20493         * configure: Regenerate.
20494         * configure.ac ("linker --version-script option"): New.
20495         ("linker soname option"): New.
20497 2018-03-09  Richard Biener  <rguenther@suse.de>
20499         PR tree-optimization/84775
20500         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
20501         immediate uses of predicate stmts and mark them modified.
20503         Revert
20504         PR tree-optimization/84178
20505         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20506         to caller.
20507         (version_loop_for_if_conversion): Delay update_ssa call.
20508         (tree_if_conversion): Delay update_ssa until after predicate
20509         insertion.
20511 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
20513         PR target/84763
20514         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
20515         when the function accesses prior frames.
20517 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
20519         PR debug/84456
20520         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
20521         gen_llsym, otherwise call maybe_gen_llsym.
20523         PR inline-asm/84742
20524         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
20525         has ',' character inside of it.
20527 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20529         PR target/84748
20530         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
20531         as clobbering CC_REGNUM.
20533 2018-03-08  Richard Biener  <rguenther@suse.de>
20535         PR middle-end/84552
20536         * tree-scalar-evolution.c: Include tree-into-ssa.h.
20537         (follow_copies_to_constant): Do not follow SSA names registered
20538         for update.
20540 2018-03-08  Richard Biener  <rguenther@suse.de>
20542         PR tree-optimization/84178
20543         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20544         to caller.
20545         (version_loop_for_if_conversion): Delay update_ssa call.
20546         (tree_if_conversion): Delay update_ssa until after predicate
20547         insertion.
20549 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
20551         PR tree-optimization/84178
20552         * tree-if-conv.c (release_bb_predicate): Remove the
20553         the assertion that the stmts have NULL use_ops.
20554         Discard the statements, asserting that they haven't
20555         yet been added to a BB.
20557 2018-03-08  Richard Biener  <rguenther@suse.de>
20559         PR tree-optimization/84746
20560         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
20561         (phi_translate): Pass in destination ANTIC_OUT set.
20562         (phi_translate_1): Likewise.  For a simplified result lookup
20563         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
20564         (phi_translate_set): Adjust.
20565         (do_pre_regular_insertion): Likewise.
20566         (do_pre_partial_partial_insertion): Likewise.
20568 2018-03-08  Martin Liska  <mliska@suse.cz>
20570         PR gcov-profile/84735
20571         * doc/gcov.texi: Document usage of profile files.
20572         * gcov-io.h: Document changes in the format.
20574 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
20576         PR debug/84404
20577         PR debug/84408
20578         * dwarf2out.c (struct dw_line_info_table): Update comments for
20579         view == -1.
20580         (FORCE_RESET_NEXT_VIEW): New.
20581         (FORCE_RESETTING_VIEW_P): New.
20582         (RESETTING_VIEW_P): Check for -1 too.
20583         (ZERO_VIEW_P): Likewise.
20584         (new_line_info_table): Force-reset next view.
20585         (dwarf2out_begin_function): Likewise.
20586         (dwarf2out_source_line): Simplify zero_view_p initialization.
20587         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
20588         view directly.  Omit view when omitting .loc at line 0.
20590 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
20592         PR tree-optimization/84740
20593         * tree-switch-conversion.c (process_switch): Call build_constructors
20594         only if info.phi_count is non-zero.
20596         PR tree-optimization/84739
20597         * tree-tailcall.c (find_tail_calls): Check call arguments against
20598         DECL_ARGUMENTS (current_function_decl) rather than
20599         DECL_ARGUMENTS (func) when checking for tail recursion.
20601 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
20603         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
20604         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
20605         Volker Reichelt's entry and add entries for people that perform
20606         GCC fuzzy testing and report numerous bugs.
20608 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
20610         PR target/82411
20611         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
20612         readonly data in sdata, if that is disabled.
20613         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
20614         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
20615         -mreadonly-in-sdata option.
20617 2018-03-07  Martin Sebor  <msebor@redhat.com>
20619         PR tree-optimization/84468
20620         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
20621         basic block when looking for nul assignment.
20623 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
20625         PR target/84277
20626         * except.h (output_function_exception_table): Adjust prototype.
20627         * except.c (output_function_exception_table): Remove FNNAME parameter
20628         and add SECTION parameter.  Ouput one part of the table at a time.
20629         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
20630         the first part of the exception table and emit unwind directives.
20631         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
20632         (i386_pe_seh_cold_init): Likewise.
20633         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
20634         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
20635         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
20636         (ix86_output_call_insn): Emit a nop in one more case for SEH.
20637         * config/i386/winnt.c: Include except.h.
20638         (struct seh_frame_state): Add reg_offset, after_prologue and
20639         in_cold_section fields.
20640         (i386_pe_seh_end_prologue): Set seh->after_prologue.
20641         (i386_pe_seh_cold_init): New function.
20642         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
20643         to seh->in_cold_section.
20644         (seh_emit_push): Record the offset of the push.
20645         (seh_emit_save): Record the offet of the save.
20646         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
20647         Test seh->after_prologue to disregard the epilogue.
20648         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
20649         (i386_pe_end_cold_function): New function.
20651 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
20653         PR fortran/84565
20654         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
20655         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
20657         PR c++/84704
20658         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
20659         on tmp_var.
20660         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
20661         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
20663         PR middle-end/84723
20664         * multiple_target.c: Include tree-inline.h and intl.h.
20665         (expand_target_clones): Diagnose and fail if node->definition and
20666         !tree_versionable_function_p (node->decl).
20668 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
20670         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
20671         sprint_ul.
20672         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
20673         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20674         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
20676 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
20678         PR target/84710
20679         * combine.c (try_combine): Use reg_or_subregno instead of handling
20680         just paradoxical SUBREGs and REGs.
20682 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
20684         * config/arc/arc.c (arc_finalize_pic): Remove function.
20685         (arc_must_save_register): We use single base PIC register, remove
20686         checks to save/restore the PIC register.
20687         (arc_expand_prologue): Likewise.
20688         * config/arc/arc-protos.h (arc_set_default_type_attributes):
20689         Remove.
20690         (arc_verify_short): Likewise.
20691         (arc_attr_type): Likewise.
20692         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
20693         (walk_stores): Likewise.
20694         (arc_address_cost): Make it static.
20695         (arc_verify_short): Likewise.
20696         (branch_dest): Likewise.
20697         (arc_attr_type): Likewise.
20698         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
20699         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
20700         (arc_final_prescan_insn): Remove inserting the nops due to
20701         hardware hazards.  It is done in reorg step.
20702         (insn_length_variant_t): Remove.
20703         (insn_length_parameters_t): Likewise.
20704         (arc_insn_length_parameters): Likewise.
20705         (arc_get_insn_variants): Likewise.
20706         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
20708 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
20710         PR inline-asm/84683
20711         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
20712         assertion failure.
20714         PR tree-optimization/84687
20715         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
20716         on new_node->decl.
20717         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
20719 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20721         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
20722         Rename to ppc_speculation_barrier.
20723         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
20724         __builtin_ppc_speculation_barrier.
20726 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
20728         PR target/84700
20729         * combine.c (combine_simplify_rtx): Don't try to simplify if
20730         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
20731         are equal to x.
20733 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
20735         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
20736         to 32 bytes when compiling for POWER9.
20738 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
20740         PR target/84564
20741         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
20742         regparm >= 3 with no arg reg available also for calls with
20743         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
20745         PR target/84524
20746         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
20747         orig,vex.
20748         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
20750 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
20752         PR target/84264
20753         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
20755 2018-03-05  Richard Biener  <rguenther@suse.de>
20757         PR tree-optimization/84486
20758         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
20759         When inserting a __builtin_assume_aligned call set the LHS
20760         SSA name alignment info accordingly.
20762 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
20764         PR tree-optimization/84114
20765         * config/aarch64/aarch64.c (aarch64_reassociation_width)
20766         Avoid reassociation of FLOAT_MODE addition.
20768 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
20770         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
20771         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
20772         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
20773         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
20774         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
20775         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
20776         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
20777         and -mwbnoinvd.
20778         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
20779         __builtin_ia32_wbinvd): New builtins.
20780         (SPECIAL_ARGS2): New.
20781         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
20782         (SPECIAL_ARGS2): New.
20783         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
20784         (ix86_valid_target_attribute_inner_p): Ditto.
20785         (ix86_init_mmx_sse_builtins): Add special_args2.
20786         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
20787         TARGET_WBNOINVD_P): New.
20788         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
20789         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
20790         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
20791         * config/i386/immintrin.h (_wbinvd): New intrinsic.
20792         * config/i386/pconfigintrin.h: New file.
20793         * config/i386/wbnoinvdintrin.h: Ditto.
20794         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
20795         wbnoinvdintrin.h.
20796         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
20798 2018-03-05  Richard Biener  <rguenther@suse.de>
20800         PR tree-optimization/84670
20801         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20802         member.
20803         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20804         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20805         (compute_antic_aux): Only assert the number of values in ANTIC_IN
20806         doesn't grow if all successors (recursively) were visited at least
20807         once.
20809 2018-03-05  Richard Biener  <rguenther@suse.de>
20811         PR tree-optimization/84650
20812         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20813         if executed in the loop pipeline.
20815 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
20817         * doc/configfiles.texi (Configuration Files): Move info about
20818         conditionalizing $target-protos.h to...
20819         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
20820         differs from $target-protos.h.
20822 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
20823             Chung-Ju Wu  <jasonwucj@gmail.com>
20825         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20826         * config/nds32/nds32-multiple.md (setmemsi): Define.
20827         * config/nds32/nds32-memory-manipulation.c
20828         (nds32_gen_dup_4_byte_to_word_value): New.
20829         (emit_setmem_word_loop): New.
20830         (emit_setmem_byte_loop): New.
20831         (nds32_expand_setmem_loop): New.
20832         (nds32_expand_setmem_loop_v3m): New.
20833         (nds32_expand_setmem_unroll): New.
20834         (nds32_expand_setmem): New.
20836 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20837             Chung-Ju Wu  <jasonwucj@gmail.com>
20839         * config/nds32/nds32-memory-manipulation.c
20840         (nds32_emit_load_store): New.
20841         (nds32_emit_post_inc_load_store): New.
20842         (nds32_emit_mem_move): New.
20843         (nds32_emit_mem_move_block): New.
20844         (nds32_expand_movmemsi_loop_unknown_size): New.
20845         (nds32_expand_movmemsi_loop_known_size): New.
20846         (nds32_expand_movmemsi_loop): New.
20847         (nds32_expand_movmemsi_unroll): New.
20848         (nds32_expand_movmemqi): Rename ...
20849         (nds32_expand_movmemsi): ... to this.
20850         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20851         (movmemsi): ... to this.
20852         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20853         (nds32_expand_movmemsi): ... to this.
20855 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20856             Monk Chiang  <sh.chiang04@gmail.com>
20857             Chung-Ju Wu  <jasonwucj@gmail.com>
20859         * config/nds32/nds32-protos.h
20860         (nds32_expand_load_multiple): New arguments.
20861         (nds32_expand_store_multiple): Ditto.
20862         (nds32_valid_multiple_load_store): Rename ...
20863         (nds32_valid_multiple_load_store_p): ... to this.
20864         * config/nds32/nds32-memory-manipulation.c
20865         (nds32_expand_load_multiple): Refine implementation.
20866         (nds32_expand_store_multiple): Ditto.
20867         * config/nds32/nds32-multiple.md
20868         (load_multiple): Update nds32_expand_load_multiple interface.
20869         (store_multiple): Update nds32_expand_store_multiple interface.
20870         * config/nds32/nds32-predicates.c
20871         (nds32_valid_multiple_load_store): Rename ...
20872         (nds32_valid_multiple_load_store_p): ... to this and refine
20873         implementation.
20874         * config/nds32/predicates.md
20875         (nds32_load_multiple_and_update_address_operation): New predicate.
20876         (nds32_store_multiple_and_update_address_operation): New predicate.
20878 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
20879             Chung-Ju Wu  <jasonwucj@gmail.com>
20881         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20882         (combo): New attribute.
20883         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20885 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
20887         * config/nds32/nds32.opt: Change -mcmodel= default value.
20889 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
20890             Monk Chiang  <sh.chiang04@gmail.com>
20891             Chung-Ju Wu  <jasonwucj@gmail.com>
20893         * config/nds32/constants.md (unspec_element): New enum.
20894         * config/nds32/constraints.md (Umw): New constraint.
20895         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
20896         * config/nds32/nds32-intrinsic.md: Likewise.
20897         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
20898         (nds32_valid_smw_lwm_base_p): New.
20899         (nds32_output_smw_single_word): New.
20900         (nds32_output_lmw_single_word): New.
20901         (nds32_expand_unaligned_load): New.
20902         (nds32_expand_unaligned_store): New.
20903         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20904         (nds32_output_smw_single_word): Declare.
20905         (nds32_output_lmw_single_word): Declare.
20906         (nds32_expand_unaligned_load): Declare.
20907         (nds32_expand_unaligned_store): Declare.
20908         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
20909         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
20910         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
20911         NDS32_BUILTIN_UASTORE_DW.
20912         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
20913         predicate.
20915 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
20916             Kito Cheng  <kito.cheng@gmail.com>
20917             Chung-Ju Wu  <jasonwucj@gmail.com>
20919         * config/nds32/nds32-intrinsic.c
20920         (nds32_expand_builtin_null_ftype_reg): Delete.
20921         (nds32_expand_builtin_reg_ftype_imm): Ditto.
20922         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
20923         (nds32_read_argument): New.
20924         (nds32_legitimize_target): Ditto.
20925         (nds32_legitimize_argument): Ditto.
20926         (nds32_check_constant_argument): Ditto.
20927         (nds32_expand_unop_builtin): Ditto.
20928         (nds32_expand_unopimm_builtin): Ditto.
20929         (nds32_expand_binop_builtin): Ditto.
20930         (nds32_builtin_decl_impl): Ditto.
20931         (builtin_description): Ditto.
20932         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20933         (nds32_init_builtins_impl): Ditto.
20934         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20935         (nds32_builtin_decl): New.
20936         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20937         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20939 2018-03-02  Jeff Law  <law@redhat.com>
20941         * reorg.c (stop_search_p): Handle DEBUG_INSN.
20942         (redundant_insn, fill_simple_delay_slots): Likewise.
20943         (fill_slots_from_thread): Likewise.
20944         * resource.c (mark_referenced_resources): Likewise.
20945         (mark_set_resources, find_dead_or_set_registers): Likewise.
20947 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20949         * substring-locations.h (format_warning_va): Formatting fix for
20950         ATTRIBUTE_GCC_DIAG.
20951         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20952         argument.
20953         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20954         * substring-locations.c: Include intl.h.
20955         (format_warning_va): Turned into small wrapper around
20956         format_warning_n_va, renamed to ...
20957         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20958         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20959         use ngettext.
20960         (format_warning_at_substring_n): New function.
20961         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20962         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
20963         format_warning_at_substring with just a shorter name instead of
20964         const function pointer.
20965         (fmtwarn_n): New function.
20966         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20967         appropriate, get rid of all the fmtstr temporaries, move conditionals
20968         with G_() wrapped string literals directly into fmtwarn arguments,
20969         cast dir.len to (int), formatting fixes.
20971 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
20973         * doc/invoke.texi: Remove "Cilk Plus" references.
20975 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20976             Richard Biener  <rguenther@suse.de>
20978         PR ipa/84628
20979         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20980         for error or warning attributes if CALL_FROM_THUNK_P is set.
20981         Formatting fixes.
20983 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
20985         PR target/56540
20986         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20987         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20989         PR target/56540
20990         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20991         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20993         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20994         instead of -1U in last predictors element's probability member.
20996 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20998         PR ipa/83983
20999         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21000         arguments if they are comparable.
21002 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21004         PR tree-optimization/84634
21005         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21006         masks and masked_loop_p with a single loop_masks, making sure it's
21007         null for bb vectorization.
21009 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21011         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21012         (vect_analyze_data_ref_access): Use loop->safe_len rather than
21013         loop->force_vectorize to check whether there is no alias.
21015 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21017         PR target/84614
21018         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21019         prototypes.
21020         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21021         comments.
21022         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21023         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21024         instead of a loop around prev_real_insn.
21025         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21026         prev_real_insn.
21028         PR inline-asm/84625
21029         * config/i386/i386.c (ix86_print_operand): Use conditional
21030         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21031         zero vector.
21033 2018-03-02  Richard Biener  <rguenther@suse.de>
21035         PR tree-optimization/84427
21036         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21037         (bitmap_set_subtract_values): Rewrite to handle multiple
21038         exprs per value.
21039         (clean): Likewise.
21040         (prune_clobbered_mems): Likewise.
21041         (phi_translate): Take edge instead of pred/phiblock.
21042         (phi_translate_1): Likewise.
21043         (phi_translate_set): Likewise.  Insert all translated
21044         exprs for a value into the set, keeping possibly multiple
21045         expressions per value.
21046         (compute_antic_aux): Adjust for phi_translate changes.
21047         When intersecting union the expressions and prune those
21048         not in the final value set, keeping possibly multiple
21049         expressions per value.  Do not use value-insertion
21050         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
21051         all expressions.  Add verification that the value-sets
21052         only shrink during iteration.
21053         (compute_partial_antic_aux): Adjust for the phi_translate changes.
21054         (do_pre_regular_insertion): Likewise.
21055         (do_pre_partial_partial_insertion): Likewise.
21057 2018-03-02  Richard Biener  <rguenther@suse.de>
21059         PR target/82005
21060         * config/darwin.c (saved_debug_info_level): New static global.
21061         (darwin_asm_lto_start): Disable debug info generation for LTO out.
21062         (darwin_asm_lto_end): Restore debug info generation settings.
21064 2018-03-01  Martin Liska  <mliska@suse.cz>
21066         PR sanitizer/82484
21067         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21068         volatile arguments.
21070 2018-03-01  Richard Biener  <rguenther@suse.de>
21072         PR debug/84645
21073         * dwarf2out.c (gen_variable_die): Properly handle late VLA
21074         type annotation with LTO when debug was disabled at compile-time.
21076 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
21078         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21079         XINT with INTVAL.
21080         (mips_final_postscan_insn): Likewise.
21082 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
21084         PR rtl-optimization/84528
21085         * alias.c (init_alias_target): Add commentary.
21086         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21087         a unique base value if the frame pointer is not eliminated
21088         to the stack pointer.
21090 2018-03-01  Tom de Vries  <tom@codesourcery.com>
21092         PR rtl-optimization/83327
21093         * lra-int.h (hard_regs_spilled_into): Declare.
21094         * lra.c (hard_regs_spilled_into): Define.
21095         (init_reg_info): Init hard_regs_spilled_into.
21096         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21097         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21098         (process_bb_lives): Handle hard_regs_spilled_into.
21099         (lra_create_live_ranges_1): Before doing liveness propagation, clear
21100         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21102 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
21104         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
21105         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
21106         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
21107         * config/rs6000/aix72.h: New file.
21109 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
21111         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
21112         instead of warning_at with conditional singular and plural messages
21113         where possible.
21115         PR target/52991
21116         * stor-layout.c (update_alignment_for_field): For
21117         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
21118         && !DECL_PACKED (field), do the alignment update, just use
21119         only desired_align instead of MAX (type_align, desired_align)
21120         as the alignment.
21121         (place_field): Don't do known_align < desired_align handling
21122         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
21123         is non-NULL, instead do it after rli->prev_field handling and
21124         only if not within a bitfield word.  For DECL_PACKED (field)
21125         use type_align of BITS_PER_UNIT.
21127 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
21129         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
21130         superfluous parentheses and trailing spaces.
21132 2018-02-28  Richard Biener  <rguenther@suse.de>
21134         PR tree-optimization/84584
21135         * graphite-scop-detection.c (scop_detection::add_scop): Discard
21136         SCoPs with fake exit edge.
21138 2018-02-28  Martin Liska  <mliska@suse.cz>
21140         PR testsuite/84597
21141         * timevar.c (timer::print): Fix format to properly print 100%
21142         values.
21144 2018-02-28  Richard Biener  <rguenther@suse.de>
21146         PR middle-end/84607
21147         * genmatch.c (capture_info::walk_match): Do not mark
21148         captured expressions without operands as expr_p given
21149         they act more like predicates and should be subject to
21150         "lost tail" side-effect preserving.
21152 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
21154         PR rtl-optimization/81611
21155         * auto-inc-dec.c (attempt_change): Move dead note from
21156         mem_insn if it's the next use of regno
21157         (find_address): Take address use of reg holding
21158         non-incremented value.  Add parm to limit search to the named
21159         reg only.
21160         (merge_in_block): Attempt to use a mem insn that is the next
21161         use of the original regno.
21163 2018-02-27  Martin Sebor  <msebor@redhat.com>
21165         PR c++/83871
21166         * doc/invoke.texi (-Wmissing-attributes): New option.
21167         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
21169 2018-02-27  Martin Sebor  <msebor@redhat.com>
21171         PR translation/84207
21172         * diagnostic-core.h (warning_n, error_n, inform_n): Change
21173         n argument to unsigned HOST_WIDE_INT.
21174         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
21175         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
21176         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
21177         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
21179 2018-02-27  Richard Biener  <rguenther@suse.de>
21181         PR tree-optimization/84512
21182         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21183         Do not use the estimate returned from record_stmt_cost for
21184         the scalar iteration cost but sum properly using add_stmt_cost.
21186 2018-02-27  Richard Biener  <rguenther@suse.de>
21188         PR tree-optimization/84466
21189         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21190         Adjust last change to less strictly validate use operands.
21192 2018-02-27  Martin Liska  <mliska@suse.cz>
21194         PR gcov-profile/84548
21195         * gcov.c (process_file): Allow partial overlap and consider it
21196         also as group functions.
21197         (output_lines): Properly calculate range of lines for a group.
21199 2018-02-27  Martin Liska  <mliska@suse.cz>
21201         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
21202         'ggc' suffixes.  Change first column width.
21203         (timer::print): Fix formatting of the column.
21205 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
21207         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
21208         preserve inline entry blocks for the sake of debug inline
21209         entry point markers alone.
21210         (remove_unused_locals): Suggest in comments a better place to
21211         force the preservation of inline entry blocks that are
21212         otherwise unused, but do not preserve them.
21214 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21216         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
21218 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21220         PR target/84039
21221         * config/i386/constraints.md (Bs): Replace
21222         ix86_indirect_branch_register with
21223         TARGET_INDIRECT_BRANCH_REGISTER.
21224         (Bw): Likewise.
21225         * config/i386/i386.md (indirect_jump): Likewise.
21226         (tablejump): Likewise.
21227         (*sibcall_memory): Likewise.
21228         (*sibcall_value_memory): Likewise.
21229         Peepholes of indirect call and jump via memory: Likewise.
21230         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
21231         (*sibcall_value_GOT_32): Likewise.
21232         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21233         (GOT_memory_operand): Likewise.
21234         (call_insn_operand): Likewise.
21235         (sibcall_insn_operand): Likewise.
21236         (GOT32_symbol_operand): Likewise.
21237         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
21239 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21241         PR rtl-optimization/83496
21242         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
21243         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
21244         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
21245         redundant insn, if any.
21246         (relax_delay_slots): Likewise.
21247         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
21249 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
21251         PR tree-optimization/83965
21252         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
21253         that grouped statements are part of a reduction chain.  Return
21254         true if the statement is not marked as a reduction itself but
21255         is part of a group.
21256         (vect_recog_dot_prod_pattern): Don't check whether the statement
21257         is part of a group here.
21258         (vect_recog_sad_pattern): Likewise.
21259         (vect_recog_widen_sum_pattern): Likewise.
21261 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21263         PR debug/84545
21264         * final.c (rest_of_clean_state): Also look for calls inside sequences.
21266 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21268         PR target/84530
21269         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
21270         the bool argument.
21271         (ix86_output_indirect_function_return): New prototype.
21272         (ix86_split_simple_return_pop_internal): Likewise.
21273         * config/i386/i386.c (indirect_return_via_cx): New.
21274         (indirect_return_via_cx_bnd): Likewise.
21275         (indirect_thunk_name): Handle return va CX_REG.
21276         (output_indirect_thunk_function): Create alias for
21277         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
21278         (ix86_output_indirect_jmp): Remove the bool argument.
21279         (ix86_output_indirect_function_return): New function.
21280         (ix86_split_simple_return_pop_internal): Likewise.
21281         * config/i386/i386.md (*indirect_jump): Don't pass false
21282         to ix86_output_indirect_jmp.
21283         (*tablejump_1): Likewise.
21284         (simple_return_pop_internal): Change it to define_insn_and_split.
21285         Call ix86_split_simple_return_pop_internal to split it for
21286         -mfunction-return=.
21287         (simple_return_indirect_internal): Call
21288         ix86_output_indirect_function_return instead of
21289         ix86_output_indirect_jmp.
21291 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
21293         PR bootstrap/84405
21294         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
21295         memset and value initialization afterwards.
21297 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
21299         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
21301 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
21303         PR target/84521
21304         * common/config/aarch64/aarch64-common.c
21305         (aarch_option_optimization_table[]): Switch
21306         off fomit-frame-pointer
21308 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
21309             Chung-Ju Wu  <jasonwucj@gmail.com>
21311         * config/nds32/nds32-multiple.md (load_multiple): Disallow
21312         volatile memory.
21313         (store_multiple): Ditto.
21315 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
21317         * config.gcc: Add --with-cpu support for nds32 target.
21318         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
21319         * config/nds32/nds32.opt: Add -mcpu= option.
21321 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
21323         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
21324         isel=yes): Warn for these deprecated options.
21326 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
21328         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
21329         ISA_2_5_MASKS_EMBEDDED.
21331 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
21333         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
21334         p->max as pointers rather than using iterative_hash_expr.
21336 2018-02-23  Carl Love  <cel@us.ibm.com>
21338         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
21339         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
21340         BU_P8V_OVERLOAD_2.
21341         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
21342         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
21343         P8V_BUILTIN_VEC_VUNSIGNED2.
21345 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
21347         PR target/81572
21348         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
21349         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
21350         LRA_UNKNOWN_ALT.
21351         * lra-constraints.c (curr_insn_transform): Set up
21352         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
21353         LRA_UNKNOWN_ALT.
21354         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
21355         * lra-eliminations.c (spill_pseudos): Ditto.
21356         (process_insn_for_elimination): Ditto.
21357         * lra-lives.c (reg_early_clobber_p): Use the new macros.
21358         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
21359         LRA_NON_CLOBBERED_ALT.
21361 2018-02-22  Martin Sebor  <msebor@redhat.com>
21363         PR tree-optimization/84480
21364         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
21365         to maybe_diag_stxncpy_trunc.  Call it.
21366         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
21367         from gimple_fold_builtin_strcpy.  Print inlining stack.
21368         (handle_builtin_stxncpy): Print inlining stack.
21369         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
21371 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
21373         PR target/84176
21374         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
21375         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
21376         and -fcheck-pointer-bounds are used together.
21377         (indirect_thunk_prefix): New enum.
21378         (indirect_thunk_need_prefix): New function.
21379         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
21380         "_nt" instead of "_bnd" for NOTRACK prefix.
21381         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
21382         (output_indirect_thunk_function): Likewise.
21383         (): Likewise.
21384         (ix86_code_end): Update output_indirect_thunk_function calls.
21385         (ix86_output_indirect_branch_via_reg): Replace
21386         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
21387         (ix86_output_indirect_branch_via_push): Likewise.
21388         (ix86_output_function_return): Likewise.
21389         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
21390         incompatible with -fcf-protection=branch and
21391         -fcheck-pointer-bounds.
21393 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
21395         PR target/83335
21396         * config/aarch64/aarch64.c (aarch64_print_address_internal):
21397         Change gcc_assert call to output_operand_lossage.
21399 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
21401         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
21403 2018-02-22  DJ Delorie  <dj@redhat.com>
21404             Sebastian Perta  <sebastian.perta@renesas.com>
21405             Oleg Endo  <olegendo@gcc.gnu.org>
21407         * config/rx/rx.c (rx_rtx_costs): New function.
21408         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
21410 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21412         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
21414 2018-02-22  Martin Liska  <mliska@suse.cz>
21416         PR driver/83193
21417         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
21418         Add "native" as a possible value.
21420 2018-02-22  Martin Liska  <mliska@suse.cz>
21422         PR driver/83193
21423         * config/i386/i386.c (ix86_option_override_internal):
21424         Add "native" as a possible value for -march and -mtune.
21426 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
21428         PR target/84502
21429         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
21430         to all type variants.
21432         PR tree-optimization/84503
21433         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
21434         width as info->bitpos + info->bitsize - start.
21435         (merged_store_group::merge_overlapping): Simplify width computation.
21436         (check_no_overlap): New function.
21437         (imm_store_chain_info::try_coalesce_bswap): Compute expected
21438         start + width and last_order of the group, fail if check_no_overlap
21439         fails.
21440         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
21441         to group if check_no_overlap fails.
21443 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
21445         * config/rs6000/altivec.md: Delete contraint arguments to
21446         define_expand, define_split, and define_peephole2, and in
21447         define_insn_and_split if always unused.
21448         * config/rs6000/darwin.md: Ditto.
21449         * config/rs6000/dfp.md: Ditto.
21450         * config/rs6000/rs6000.md: Ditto.
21451         * config/rs6000/sync.md: Ditto.
21452         * config/rs6000/vector.md: Ditto.
21453         * config/rs6000/vsx.md: Ditto.
21455 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
21457         * config/rs6000/altivec.md: Write output control strings as braced
21458         blocks instead of double-quoted strings.
21459         * config/rs6000/darwin.md: Ditto.
21460         * config/rs6000/rs6000.md: Ditto.
21461         * config/rs6000/vector.md: Ditto.
21462         * config/rs6000/vsx.md: Ditto.
21464 2018-02-21  Jason Merrill  <jason@redhat.com>
21466         PR c++/84314 - ICE with templates and fastcall attribute.
21467         * attribs.c (build_type_attribute_qual_variant): Remove assert.
21469 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
21471         * ipa-cp.c (determine_versionability): Fix comment typos.
21473 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
21475         PR c/84229
21476         * ipa-cp.c (determine_versionability): Do not version functions caling
21477         va_arg_pack.
21479 2018-02-21  Martin Liska  <mliska@suse.cz>
21481         PR driver/83193
21482         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
21483         Add "native" as a possible value.
21484         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
21485         the macro when native cpu detection is available.
21487 2018-02-21  Martin Liska  <mliska@suse.cz>
21489         PR driver/83193
21490         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
21491         Add "native" as a possible value.
21492         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
21493         when native cpu detection is available.
21495 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
21496             Martin Sebor  <msebor@redhat.com>
21498         PR tree-optimization/84478
21499         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
21500         false.
21501         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
21502         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
21503         support which is conservatively correct, for 2 only stay conservative
21504         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
21505         argument to the 2 argument get_range_strlen, adjust 6 arg
21506         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
21507         false.
21508         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
21509         (gimple_fold_builtin_strlen): Pass true as last argument to
21510         get_range_strlen.
21512 2018-02-20  Martin Sebor  <msebor@redhat.com>
21514         PR middle-end/84095
21515         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
21516         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
21517         (builtin_memref::builtin_memref): Factor out parts into
21518         set_base_and_offset and call it.
21520 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
21522         PR middle-end/84406
21523         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
21524         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
21525         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
21526         search at the associated MODE_INT.
21528 2018-02-20  Jeff Law  <law@redhat.com>
21530         PR middle-end/82123
21531         PR tree-optimization/81592
21532         PR middle-end/79257
21533         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
21534         for range data rather than using global data.
21535         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
21536         range data rather than using global data.
21537         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
21538         pass it to children as needed.
21539         (struct directive::fmtresult): Similarly.
21540         (struct directive::set_width): Similarly.
21541         (struct directive::set_precision): Similarly.
21542         (format_integer, format_directive, parse_directive): Similarly.
21543         (format_none): Accept unnamed vr_values parameter.
21544         (format_percent, format_floating, format_character): Similarly.
21545         (format_string, format_plain): Similarly.
21546         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
21547         the EVRP range analyzer for range data rather than using global data.
21548         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
21549         gimple-ssa-evrp-analyze.h
21550         (class sprintf_dom_walker): Add after_dom_children member function.
21551         Add evrp_range_analyzer member.
21552         (sprintf_dom_walker::before_dom_children): Call into the EVRP
21553         range analyzer as needed.
21554         (sprintf_dom_walker::after_dom_children): New member function.
21555         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
21556         if not optimizing.
21557         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
21558         (evrp_range_analyzer::pop_to_marker): Likewise.
21560 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
21562         PR tree-optimization/84419
21563         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
21564         with the required type if its current type is compatible but
21565         different.
21567 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
21569         PR middle-end/82004
21570         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
21571         after vectorization.
21573 2018-02-20  Martin Liska  <mliska@suse.cz>
21575         PR driver/83193
21576         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
21577         possible values if we don't have a hint.
21579 2018-02-20  Martin Liska  <mliska@suse.cz>
21581         PR c/84310
21582         PR target/79747
21583         * final.c (shorten_branches): Build align_tab array with one
21584         more element.
21585         * opts.c (finish_options): Add alignment option limit check.
21586         (MAX_CODE_ALIGN): Likewise.
21587         (MAX_CODE_ALIGN_VALUE): Likewise.
21588         * doc/invoke.texi: Document maximum allowed option value for
21589         all -falign-* options.
21591 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
21593         PR target/84146
21594         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
21595         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
21596         * var-tracking.c (emit_note_insn_var_location): Remove all references
21597         to NOTE_INSN_CALL_ARG_LOCATION.
21598         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
21599         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
21600         Use copy_rtx_if_shared.
21601         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
21602         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
21603         (dwarf2out_var_location): Remove handling of
21604         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
21605         on call_insn.
21606         * final.c (final_scan_insn): Remove all references to
21607         NOTE_INSN_CALL_ARG_LOCATION.
21608         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
21609         before dumping final insns.
21610         * except.c (emit_note_eh_region_end): Remove all references to
21611         NOTE_INSN_CALL_ARG_LOCATION.
21612         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
21613         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
21614         * config/arc/arc.c (hwloop_optimize): Likewise.
21615         * config/arm/arm.c (create_fix_barrier): Likewise.
21616         * config/s390/s390.c (s390_chunkify_start): Likewise.
21617         * config/sh/sh.c (find_barrier): Likewise.
21618         * config/i386/i386.c (rest_of_insert_endbranch,
21619         ix86_seh_fixup_eh_fallthru): Likewise.
21620         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
21621         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
21622         * config/frv/frv.c (frv_function_prologue): Likewise.
21623         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
21624         reg note.
21625         (note_outside_basic_block_p): Remove all references to
21626         NOTE_INSN_CALL_ARG_LOCATION.
21627         * gengtype.c (adjust_field_rtx_def): Likewise.
21628         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
21629         Likewise.
21630         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
21631         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
21633         PR c++/84444
21634         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
21635         is ADDR_EXPR.
21637         PR tree-optimization/84452
21638         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
21639         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
21640         is NULL.
21642 2018-02-19  Martin Liska  <mliska@suse.cz>
21644         PR sanitizer/82183
21645         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
21647 2018-02-19  Martin Liska  <mliska@suse.cz>
21648             Richard Sandiford  <richard.sandiford@linaro.org>
21650         PR tree-optimization/82491
21651         * gimple-fold.c (get_base_constructor): Make earlier bail out
21652         to prevent ubsan.
21654 2018-02-19  Carl Love  <cel@us.ibm.com>
21656         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
21657         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
21658         BU_P8V_OVERLOAD_1.
21659         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
21660         P8V_BUILTIN_VEC_NEG.
21662 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
21664         * config/rl78/rl78.md (movdf): New define expand.
21666 2018-02-19  Martin Liska  <mliska@suse.cz>
21668         PR other/80589
21669         * doc/invoke.texi: Fix typo.
21670         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
21672 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
21674         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
21675         handle rs6000_single_float and rs6000_double_float specially for
21676         e500 family CPUs.
21678 2018-02-16  Jeff Law  <law@redhat.com>
21680         * config/rx/rx.c (add_pop_cfi_notes): New function.;
21681         (pop_regs): Use it.
21683 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
21685         PR ipa/84425
21686         * ipa-inline.c (inline_small_functions): Fix a typo.
21688 2018-02-16  Nathan Sidwell  <nathan@acm.org>
21690         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
21692 2018-02-16  Carl Love  <cel@us.ibm.com>
21694         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
21695         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
21696         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
21697         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
21698         expansion to P8V_BUILTIN_VEC_FLOAT2.
21700 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
21702         PR rtl-optimization/70023
21703         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
21704         src_regno into account.
21706 2018-02-16  Carl Love  <cel@us.ibm.com>
21708         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
21709         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
21710         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
21711         * config/rs6000/rs6000.c: Remove case statements for
21712         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
21713         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
21714         and P9V_BUILTIN_VEC_VINSERT4B.
21715         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
21716         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
21717         * config/rs6000/vsx.md:
21718         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
21719         vec_insert4b.
21721 2018-02-16  Carl Love  <cel@us.ibm.com>
21723         * config/rs6000/altivec.h: Add builtin names vec_extract4b
21724         vec_insert4b.
21725         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
21726         definitions.
21727         * config/rs6000/rs6000-c.c: Add the definitions for
21728         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
21729         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
21730         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
21731         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
21732         definition for insert4b and define insn *insert3b_internal.
21733         * doc/extend.texi: Add documentation for vec_extract4b.
21735 2018-02-16  Nathan Sidwell  <nathan@acm.org>
21737         * doc/extend.texi (Backwards Compatibility): Mention friend
21738         injection.  Note for-scope is deprecated.
21739         * doc/invoke.texi (-ffriend-injection): Deprecate.
21741 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
21743         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
21744         that moved to I2, also allow destinations that are a paradoxical
21745         subreg (instead of a normal reg).
21747 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
21749         PR target/83831
21750         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
21751         to QImode.
21753 2018-02-16  Richard Biener  <rguenther@suse.de>
21755         PR tree-optimization/84037
21756         PR tree-optimization/84016
21757         PR target/82862
21758         * config/i386/i386.c (ix86_builtin_vectorization_cost):
21759         Adjust vec_construct for the fact we need additional higher latency
21760         128bit inserts for AVX256 and AVX512 vector builds.
21761         (ix86_add_stmt_cost): Scale vector construction cost for
21762         elementwise loads.
21764 2018-02-16  Richard Biener  <rguenther@suse.de>
21766         PR tree-optimization/84417
21767         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
21768         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
21769         (non_rewritable_lvalue_p): Likewise, use poly-ints.
21771 2018-02-16  Martin Liska  <mliska@suse.cz>
21773         PR sanitizer/84307
21774         * internal-fn.def (ASAN_CHECK): Set proper flags.
21775         (ASAN_MARK): Likewise.
21777 2018-02-16  Julia Koval  <julia.koval@intel.com>
21779         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
21780         from PTA_CANNONLAKE.
21782 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
21784         PR target/84272
21785         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
21786         Use ++iter rather than iter++ for std::list iterators.
21787         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
21788         defer deleting them until all nodes in the forest are processed.  Do
21789         free even leaf nodes.  Change to_process into auto_vec.
21791         PR bootstrap/84405
21792         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
21793         * vec.h (vec_default_construct): Use memset instead of placement new
21794         if BROKEN_VALUE_INITIALIZATION is defined.
21795         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
21796         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
21797         is defined.
21799         PR rtl-optimization/83723
21800         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
21801         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
21802         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
21803         recursive calls.
21804         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21805         callers.
21806         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21808 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
21810         PR rtl-optimization/81443
21811         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21812         from inner REGs to paradoxical SUBREGs.
21814 2018-02-16  Richard Biener  <rguenther@suse.de>
21816         PR tree-optimization/84399
21817         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21818         For operands we can analyze at their definition make sure we can
21819         analyze them at each use as well.
21821 2018-02-16  Richard Biener  <rguenther@suse.de>
21823         PR tree-optimization/84190
21824         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21825         volatile accesses if the decl isn't volatile.
21827 2018-02-15  Jason Merrill  <jason@redhat.com>
21829         PR c++/84314 - ICE with templates and fastcall attribute.
21830         * attribs.c (build_type_attribute_qual_variant): Don't clobber
21831         TYPE_CANONICAL on an existing type.
21833 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
21835         PR tree-optimization/84383
21836         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21837         dstoff nor call operand_equal_p if dstbase is NULL.
21839         PR tree-optimization/84334
21840         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21841         also a CONSTANT_CLASS_P, punt.
21843 2018-02-14  Jim Wilson  <jimw@sifive.com>
21845         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21846         first SMALL_OPERAND check.  New local min_second_step.  Move assert
21847         to where locals are set.  Add TARGET_RVC support.
21848         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21850 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
21852         * doc/invoke.texi: Correct -Wformat-overflow code sample.
21854 2018-02-14  Martin Sebor  <msebor@redhat.com>
21856         PR tree-optimization/83698
21857         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21858         arrays constrain the offset range to their bounds.
21859         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21860         (builtin_access::overlap): Avoid setting the size of overlap if it's
21861         already been set.
21862         (maybe_diag_overlap): Also consider arrays when deciding what values
21863         of offsets to include in diagnostics.
21865 2018-02-14  Martin Sebor  <msebor@redhat.com>
21867         PR c/84108
21868         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21869         that correspond to the kind of a declaration.
21871 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
21873         PR target/83984
21874         * config/pa/pa.md: Load address of PIC label using the linkage table
21875         if the label is nonlocal.
21877 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21879         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21880         warning message if user requests -maltivec=be.
21881         * doc/invoke.texi: Document deprecation of -maltivec=be.
21883 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
21885         PR target/84220
21886         * config/rs6000/rs6000-c.c: Update definitions for
21887         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21888         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21890 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21892         PR target/84239
21893         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
21894         add _get_ssp intrinsics. Remove argument from
21895         __builtin_ia32_rdssp[d|q].
21896         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
21897         * config/i386/i386-builtin.def: Remove argument from
21898         __builtin_ia32_rdssp[d|q].
21899         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
21900         ix86_expand_special_args_builtin for _rdssp[d|q].
21901         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
21902         Clear register before usage.
21903         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
21904         Add documentation for new _get_ssp and _inc_ssp intrinsics.
21906 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
21908         PR tree-optimization/84357
21909         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
21910         operand 1 of an ARRAY_REF too.
21912 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
21914         PR target/83831
21915         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
21916         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21917         declarations.
21918         (set_of_reg): New struct.
21919         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
21920         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
21921         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21922         functions.
21923         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
21924         Split into bitclr, bitset, bitinvert patterns if appropriate.
21925         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21926         use rx_fuse_in_memory_bitop.
21927         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21928         to named insn, correct maximum insn length.
21930 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
21932         PR target/79242
21933         * machmode.def: Define a complex mode for PARTIAL_INT.
21934         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21935         MODE_PARTIAL_INT.
21936         * doc/rtl.texi: Document CSPImode.
21937         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21938         handling.
21939         (msp430_hard_regno_nregs_with_padding): Likewise.
21941 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
21943         PR target/84279
21944         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21946 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
21948         PR rtl-optimization/84169
21949         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
21950         we generated a parallel as new i3 and we split that to new i2 and i3
21951         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
21952         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
21953         those to i2, not i1.  Partially rewrite this scan code.
21955 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
21957         PR c/82210
21958         * stor-layout.c (place_field): For variable length fields, adjust
21959         offset_align afterwards not just based on the field's alignment,
21960         but also on the size.
21962         PR middle-end/84309
21963         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21964         of exps and logs in the use_exp2 case.
21966 2018-02-13  Jeff Law  <law@redhat.com>
21968         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21969         entry for "vector".
21971         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21972         ARGS as unused.
21974 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
21976         PR debug/84342
21977         PR debug/84319
21978         * common.opt (gas-loc-support, gas-locview-support): New.
21979         (ginline-points, ginternal-reset-location-views): New.
21980         * doc/invoke.texi: Document them.  Use @itemx where intended.
21981         (gvariable-location-views): Adjust.
21982         * target.def (reset_location_view): New.
21983         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21984         (TARGET_RESET_LOCATION_VIEW): New.
21985         * doc/tm.texi: Rebuilt.
21986         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21987         (dwarf2out_default_as_locview_support): New.
21988         (output_asm_line_debug_info): Use option variables.
21989         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21990         (output_loc_list): Likewise.
21991         (add_high_low_attributes): Check option variables.
21992         Don't output entry view attribute in strict mode.
21993         (gen_inlined_subroutine_die): Check option variables.
21994         (dwarf2out_inline_entry): Likewise.
21995         (init_sections_and_labels): Likewise.
21996         (dwarf2out_early_finish): Likewise.
21997         (maybe_reset_location_view): New, from...
21998         (dwarf2out_var_location): ... here.  Call it.
21999         * debug.h (dwarf2out_default_as_loc_support): Declare.
22000         (dwarf2out_default_as_locview_support): Declare.
22001         * hooks.c (hook_int_rtx_insn_0): New.
22002         * hooks.h (hook_int_rtx_insn_0): Declare.
22003         * toplev.c (process_options): Take -gas-loc-support and
22004         -gas-locview-support from dwarf2out.  Enable
22005         -gvariable-location-views by default only with locview
22006         assembler support.  Enable -ginternal-reset-location-views by
22007         default only if the target defines the corresponding hook.
22008         Enable -ginline-points by default if location views are
22009         enabled; force it disabled if statement frontiers are
22010         disabled.
22011         * tree-inline.c (expand_call_inline): Check option variables.
22012         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22014 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
22016         PR tree-optimization/84321
22017         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22018         handling.  Also check whether the anti-range contains any values
22019         that satisfy the mask; switch to a VR_RANGE if not.
22021 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
22023         PR sanitizer/84340
22024         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22026 2018-02-13  Martin Jambor  <mjambor@suse.cz>
22028         PR c++/83990
22029         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22030         of call statements, also set location of a load to a temporary.
22032 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
22034         * config/rl78/rl78.c (add_vector_labels): New function.
22035         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22036         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22037         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22038         which checks that no arguments are passed.
22039         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22040         * doc/extend.texi: Documentation for the new attribute.
22042 2018-02-13  Andreas Schwab  <schwab@suse.de>
22044         * config/riscv/linux.h (CPP_SPEC): Define.
22046 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22048         PR target/84335
22049         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
22050         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
22051         OPTION_MASK_ISA_AES as first argument to def_builtin_const
22052         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
22053         instead of OPTION_MASK_ISA_PCLMUL as first argument to
22054         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
22055         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
22056         temporarily for AES and PCLMUL builtins.
22058         PR tree-optimization/84339
22059         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22060         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22061         Formatting fixes.
22063         PR middle-end/84309
22064         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22065         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22066         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22067         inline function.
22068         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22069         inline function.
22070         * omp-simd-clone.h: New file.
22071         * omp-simd-clone.c: Include omp-simd-clone.h.
22072         (expand_simd_clones): No longer static.
22073         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22074         cgraph.h and omp-simd-clone.h.
22075         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22076         (vect_recog_widen_shift_pattern): Formatting fix.
22077         (vect_pattern_recog_1): Don't check optab for calls.
22079         PR target/84336
22080         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22081         operands[2] into a REG before using gen_lowpart on it.
22083 2018-02-12  Jeff Law  <law@redhat.com>
22085         PR target/83760
22086         * config/sh/sh.c (find_barrier): Consider a sibling call
22087         a barrier as well.
22089         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22090         successfully back substituting a reg.
22092 2018-02-12  Richard Biener  <rguenther@suse.de>
22094         PR tree-optimization/84037
22095         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22096         parameter, move visited init to caller.
22097         (vect_slp_analyze_operations): Separate cost from validity
22098         check, initialize visited once for all instances.
22099         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22100         for all instances.
22101         * tree-vect-stmts.c (vect_model_simple_cost): Make early
22102         out an assert.
22103         (vect_model_promotion_demotion_cost): Likewise.
22104         (vectorizable_bswap): Guard cost modeling with !slp_node
22105         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
22106         SLP stmts.
22107         (vectorizable_call): Likewise.
22108         (vectorizable_conversion): Likewise.
22109         (vectorizable_assignment): Likewise.
22110         (vectorizable_shift): Likewise.
22111         (vectorizable_operation): Likewise.
22112         (vectorizable_store): Likewise.
22113         (vectorizable_load): Likewise.
22114         (vectorizable_condition): Likewise.
22115         (vectorizable_comparison): Likewise.
22117 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
22119         PR sanitizer/84307
22120         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
22121         (ASAN_MARK): Fix fnspec to account for return value, change pointer
22122         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
22124 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22126         PR middle-end/83665
22127         * params.def (inline-min-speedup): Increase from 8 to 15.
22128         (max-inline-insns-auto): Decrease from 40 to 30.
22129         * ipa-split.c (consider_split): Add some buffer for function to
22130         be considered inlining candidate.
22131         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
22132         default values.
22134 2018-02-12  Richard Biener  <rguenther@suse.de>
22136         PR tree-optimization/84037
22137         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
22138         matched stmts if we cannot swap the non-matched ones.
22140 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22142         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
22143         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
22144         _mm_maskz_scalef_round_ss): New intrinsics.
22145         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
22146         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
22147         __builtin_ia32_scalefss_round): Remove.
22148         (__builtin_ia32_scalefsd_mask_round,
22149         __builtin_ia32_scalefss_mask_round): New intrinsics.
22150         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
22151         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
22152         ((match_operand:VF_128 2 "<round_nimm_predicate>"
22153         "<round_constraint>")): Changed to ...
22154         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
22155         "<round_scalar_constraint>")): ... this.
22156         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
22157         %0, %1, %2<round_op3>}"): Changed to ...
22158         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
22159         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
22160         %2<round_scalar_mask_op3>}"): ... this.
22161         * config/i386/subst.md (round_scalar_nimm_predicate): New.
22163 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22165         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
22166         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
22167         (_mm_maskz_sqrt_round_ss): New intrinsics.
22168         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
22169         (__builtin_ia32_sqrtsd_mask_round)
22170         (__builtin_ia32_sqrtss_mask_round): New builtins.
22171         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
22172         (__builtin_ia32_sqrtss_round): Remove.
22173         (__builtin_ia32_sqrtsd_mask_round)
22174         (__builtin_ia32_sqrtss_mask_round): New builtins.
22175         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
22176         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
22177         ((match_operand:VF_128 1 "vector_operand"
22178         "xBm,<round_constraint>")): Changed to ...
22179         ((match_operand:VF_128 1 "vector_operand"
22180         "xBm,<round_scalar_constraint>")): ... this.
22181         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
22182         %0, %2, %<iptr>1<round_op3>}): Changed to ...
22183         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
22184         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
22185         %<iptr>1<round_scalar_mask_op3>}): ... this.
22186         ((set_attr "prefix" "<round_prefix>")): Changed to ...
22187         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
22189 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
22191         PR target/84266
22192         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
22193         Cast vec_cmpeq result to correct type.
22194         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
22195         Cast vec_cmpgt result to correct type.
22197 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
22199         * final.c (final_scan_insn_1): Renamed from...
22200         (final_scan_insn): ... this.  New wrapper, to recover
22201         seen from the outermost call in recursive ones.
22202         * config/sparc/sparc.c (output_return): Drop seen from call.
22203         (output_sibcall): Likewise.
22204         * config/visium/visium.c (output_branch): Likewise.
22206 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
22208         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
22209         function label.
22211 2018-02-10  Alan Modra  <amodra@gmail.com>
22213         PR target/84300
22214         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
22215         Specify LR as an input.
22217 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
22219         PR sanitizer/83987
22220         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
22221         remove_member_access_dummy_vars): New functions.
22222         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
22223         lower_omp_1, execute_lower_omp): Use them.
22225         PR rtl-optimization/84308
22226         * shrink-wrap.c (spread_components): Release todo vector.
22228 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
22230         PR rtl-optimization/57193
22231         * ira-color.c (struct allocno_color_data): Add member
22232         conflict_allocno_hard_prefs.
22233         (update_conflict_allocno_hard_prefs): New.
22234         (bucket_allocno_compare_func): Add a preference based on
22235         conflict_allocno_hard_prefs.
22236         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
22237         (color_allocnos): Remove a dead code.  Initiate
22238         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
22240 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22242         PR target/84226
22243         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
22244         constraint from =wa to wa.  Avoid a subreg on the output operand,
22245         instead use a pseudo and subreg it in a move.
22246         (p9_xxbrd_<mode>): Changed to ...
22247         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
22248         (p9_xxbrd_v2df): New expander.
22249         (p9_xxbrw_<mode>): Changed to ...
22250         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
22251         (p9_xxbrw_v4sf): New expander.
22253 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22255         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
22257 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
22259         PR target/83926
22260         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
22261         multiply in 32-bit mode.
22262         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
22263         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
22264         mode.
22266 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22268         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
22269         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
22270         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
22271         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
22273 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
22275         PR lto/84213
22276         * dwarf2out.c (is_trivial_indirect_ref): New function.
22277         (dwarf2out_late_global_decl): Do not generate a location
22278         attribute for variables that have a non-trivial DECL_VALUE_EXPR
22279         and that are not defined in the current unit.
22281 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22283         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
22284         instead of a libcall for UNORDERED.
22286 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
22288         PR target/82641
22289         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
22290         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
22292 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22294         PR target/PR84295
22295         * config/s390/s390.c (s390_set_current_function): Invoke
22296         s390_indirect_branch_settings also if fndecl didn't change.
22298 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22300         * config/rs6000/rs6000.md (blockage): Set length to zero.
22302 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22304         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
22306 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22308         PR sanitizer/84285
22309         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
22310         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
22311         -static-lib*san.
22313         PR debug/84252
22314         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
22315         PARALLEL incoming that failed vt_get_decl_and_offset check.
22317         PR middle-end/84237
22318         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
22319         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
22320         TREE_READONLY bit.
22321         (get_variable_section): For decls in named .bss* sections pass true as
22322         second argument to bss_initializer_p.
22324 2018-02-09  Marek Polacek  <polacek@redhat.com>
22325             Jakub Jelinek  <jakub@redhat.com>
22327         PR c++/83659
22328         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
22329         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
22330         Sync some changes from cxx_fold_indirect_ref.
22332 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22334         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
22335         markers.
22336         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
22337         (BLOCK_INLINE_ENTRY_LABEL): New.
22338         (dwarf2out_var_location): Disregard inline entry markers.
22339         (inline_entry_data): New struct.
22340         (inline_entry_data_hasher): New hashtable type.
22341         (inline_entry_data_hasher::hash): New.
22342         (inline_entry_data_hasher::equal): New.
22343         (inline_entry_data_table): New variable.
22344         (add_high_low_attributes): Add DW_AT_entry_pc and
22345         DW_AT_GNU_entry_view attributes if a pending entry is found
22346         in inline_entry_data_table.  Add old entry_pc attribute only
22347         if debug nonbinding markers are disabled.
22348         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
22349         markers are enabled.
22350         (block_within_block_p, dwarf2out_inline_entry): New.
22351         (dwarf2out_finish): Check that no entries remained in
22352         inline_entry_data_table.
22353         * final.c (reemit_insn_block_notes): Handle inline entry notes.
22354         (final_scan_insn, notice_source_line): Likewise.
22355         (rest_of_clean_state): Skip inline entry markers.
22356         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
22357         markers.
22358         * gimple.c (gimple_build_debug_inline_entry): New.
22359         * gimple.h (enum gimple_debug_subcode): Add
22360         GIMPLE_DEBUG_INLINE_ENTRY.
22361         (gimple_build_debug_inline_entry): Declare.
22362         (gimple_debug_inline_entry_p): New.
22363         (gimple_debug_nonbind_marker_p): Adjust.
22364         * insn-notes.def (INLINE_ENTRY): New.
22365         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
22366         inline entry marker notes.
22367         (print_insn): Likewise.
22368         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
22369         (INSN_DEBUG_MARKER_KIND): Likewise.
22370         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
22371         * tree-inline.c (expand_call_inline): Build and insert
22372         debug_inline_entry stmt.
22373         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
22374         inline entry blocks early, if nonbind markers are enabled.
22375         (dump_scope_block): Dump fragment info.
22376         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
22377         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
22378         (gimple_build_debug_inline_entry): New.
22379         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
22380         Enable/disable inline entry points too.
22381         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
22382         (DEBUG_INSN): Describe inline entry markers.
22384         * common.opt (gvariable-location-views): New.
22385         (gvariable-location-views=incompat5): New.
22386         * config.in: Rebuilt.
22387         * configure: Rebuilt.
22388         * configure.ac: Test assembler for view support.
22389         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
22390         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
22391         * dwarf2out.c (var_loc_view): New typedef.
22392         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
22393         (dwarf2out_locviews_in_attribute): New.
22394         (dwarf2out_locviews_in_loclist): New.
22395         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
22396         (enum dw_line_info_opcode): Add LI_adv_address.
22397         (struct dw_line_info_table): Add view.
22398         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
22399         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
22400         (zero_view_p): New variable.
22401         (ZERO_VIEW_P): New macro.
22402         (output_asm_line_debug_info): New.
22403         (struct var_loc_node): Add view.
22404         (add_AT_view_list, AT_loc_list): New.
22405         (add_var_loc_to_decl): Add view param.  Test it against last.
22406         (new_loc_list): Add view params.  Record them.
22407         (AT_loc_list_ptr): Handle loc and view lists.
22408         (view_list_to_loc_list_val_node): New.
22409         (print_dw_val): Handle dw_val_class_view_list.
22410         (size_of_die): Likewise.
22411         (value_format): Likewise.
22412         (loc_list_has_views): New.
22413         (gen_llsym): Set vl_symbol too.
22414         (maybe_gen_llsym, skip_loc_list_entry): New.
22415         (dwarf2out_maybe_output_loclist_view_pair): New.
22416         (output_loc_list): Output view list or entries too.
22417         (output_view_list_offset): New.
22418         (output_die): Handle dw_val_class_view_list.
22419         (output_dwarf_version): New.
22420         (output_compilation_unit_header): Use it.
22421         (output_skeleton_debug_sections): Likewise.
22422         (output_rnglists, output_line_info): Likewise.
22423         (output_pubnames, output_aranges): Update version comments.
22424         (output_one_line_info_table): Output view numbers in asm comments.
22425         (dw_loc_list): Determine current endview, pass it to new_loc_list.
22426         Call maybe_gen_llsym.
22427         (loc_list_from_tree_1): Adjust.
22428         (add_AT_location_description): Create view list attribute if
22429         needed, check it's absent otherwise.
22430         (convert_cfa_to_fb_loc_list): Adjust.
22431         (maybe_emit_file): Call output_asm_line_debug_info for test.
22432         (dwarf2out_var_location): Reset views as needed.  Precompute
22433         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
22434         attribute.  Set view.
22435         (new_line_info_table): Reset next view.
22436         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
22437         (dwarf2out_source_line): Likewise.  Output view resets and labels to
22438         the assembler, or select appropriate line info opcodes.
22439         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
22440         (optimize_string_length): Catch it.  Adjust.
22441         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
22442         dw_val_class_view_list, and remove it if no longer needed.
22443         (hash_loc_list): Hash view numbers.
22444         (loc_list_hasher::equal): Compare them.
22445         (optimize_location_lists): Check whether a view list symbol is
22446         needed, and whether the locview attribute is present, and
22447         whether they match.  Remove the locview attribute if no longer
22448         needed.
22449         (index_location_lists): Call skip_loc_list_entry for test.
22450         (dwarf2out_finish): Call output_asm_line_debug_info for test.
22451         Use output_dwarf_version.
22452         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
22453         (struct dw_val_node): Add val_view_list.
22454         * final.c (SEEN_NEXT_VIEW): New.
22455         (set_next_view_needed): New.
22456         (clear_next_view_needed): New.
22457         (maybe_output_next_view): New.
22458         (final_start_function): Rename to...
22459         (final_start_function_1): ... this.  Take pointer to FIRST,
22460         add SEEN parameter.  Emit param bindings in the initial view.
22461         (final_start_function): Reintroduce SEEN-less interface.
22462         (final): Rename to...
22463         (final_1): ... this.  Take SEEN parameter.  Output final pending
22464         next view at the end.
22465         (final): Reintroduce seen-less interface.
22466         (final_scan_insn): Output pending next view before switching
22467         sections or ending a block.  Mark the next view as needed when
22468         outputting variable locations.  Notify debug backend of section
22469         changes, and of location view changes.
22470         (rest_of_handle_final): Adjust.
22471         * toplev.c (process_options): Autodetect value for debug variable
22472         location views option.  Warn on incompat5 without -gdwarf-5.
22473         * doc/invoke.texi (gvariable-location-views): New.
22474         (gvariable-location-views=incompat5): New.
22475         (gno-variable-location-views): New.
22477 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
22479         PR tree-optimization/84136
22480         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
22481         that the result of find_edge is non-NULL.
22483 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
22485         PR target/83008
22486         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
22487         storing integer register in SImode.  Fix cost of 256 and 512
22488         byte aligned SSE register store.
22490 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
22492         * config/i386/i386.c (ix86_multiplication_cost): Fix
22493         multiplication cost for TARGET_AVX512DQ.
22495 2018-02-08  Marek Polacek  <polacek@redhat.com>
22497         PR tree-optimization/84238
22498         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
22499         get_range_strlen.
22501 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22503         PR tree-optimization/84265
22504         * tree-vect-stmts.c (vectorizable_store): Don't treat
22505         VMAT_CONTIGUOUS accesses as grouped.
22506         (vectorizable_load): Likewise.
22508 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22510         PR tree-optimization/81635
22511         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
22512         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
22513         (test_round_for_mask): New functions.
22514         (wide_int_cc_tests): Call test_round_for_mask.
22515         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
22516         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
22517         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
22518         range returned by get_range_info.
22520 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22522         PR ipa/81360
22523         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
22524         * symtab.c: Include builtins.h
22525         (symtab_node::output_to_lto_symbol_table_p): Move here
22526         from lto-streamer-out.c:output_symbol_p.
22527         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
22528         (output_symbol_p): Move all logic to symtab.c
22529         (produce_symtab): Update.
22531 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22533         * config/s390/s390-opts.h (enum indirect_branch): Define.
22534         * config/s390/s390-protos.h (s390_return_addr_from_memory)
22535         (s390_indirect_branch_via_thunk)
22536         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
22537         (enum s390_indirect_branch_type): Define.
22538         * config/s390/s390.c (struct s390_frame_layout, struct
22539         machine_function): Remove.
22540         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
22541         (indirect_branch_table_label_no, indirect_branch_table_name):
22542         Define variables.
22543         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
22544         (enum s390_indirect_branch_option): Define.
22545         (s390_return_addr_from_memory): New function.
22546         (s390_handle_string_attribute): New function.
22547         (s390_attribute_table): Add new attribute handler.
22548         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
22549         (s390_indirect_branch_via_thunk): New function.
22550         (s390_indirect_branch_via_inline_thunk): New function.
22551         (s390_function_ok_for_sibcall): When jumping via thunk disallow
22552         sibling call optimization for non z10 compiles.
22553         (s390_emit_call): Force indirect branch target to be a single
22554         register.  Add r1 clobber for non-z10 compiles.
22555         (s390_emit_epilogue): Emit return jump via return_use expander.
22556         (s390_reorg): Handle JUMP_INSNs as execute targets.
22557         (s390_option_override_internal): Perform validity checks for the
22558         new command line options.
22559         (s390_indirect_branch_attrvalue): New function.
22560         (s390_indirect_branch_settings): New function.
22561         (s390_set_current_function): Invoke s390_indirect_branch_settings.
22562         (s390_output_indirect_thunk_function):  New function.
22563         (s390_code_end): Implement target hook.
22564         (s390_case_values_threshold): Implement target hook.
22565         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
22566         macros.
22567         * config/s390/s390.h (struct s390_frame_layout)
22568         (struct machine_function): Move here from s390.c.
22569         (TARGET_INDIRECT_BRANCH_NOBP_RET)
22570         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
22571         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
22572         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
22573         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
22574         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
22575         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
22576         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
22577         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
22578         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
22579         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
22580         (mnemonic attribute): Add values which aren't recognized
22581         automatically.
22582         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
22583         pattern for branch conversion.  Fix mnemonic attribute.
22584         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
22585         indirect branch via thunk if requested.
22586         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
22587         ("*indirect_jump"): Disable for branch conversion using out of
22588         line thunks.
22589         ("indirect_jump_via_thunk<mode>_z10")
22590         ("indirect_jump_via_thunk<mode>")
22591         ("indirect_jump_via_inlinethunk<mode>_z10")
22592         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
22593         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
22594         ("casesi_jump_via_inlinethunk<mode>_z10")
22595         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
22596         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
22597         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
22598         ("*indirect2_jump"): Disable for branch conversion.
22599         ("casesi_jump"): Turn into expander and expand patterns for branch
22600         conversion.
22601         ("return_use"): New expander.
22602         ("*return"): Emit return via thunk and rename it to ...
22603         ("*return<mode>"): ... this one.
22604         * config/s390/s390.opt: Add new options and and enum for the
22605         option values.
22607 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
22609         * lra-constraints.c (match_reload): Unconditionally use
22610         gen_lowpart_SUBREG, rather than selecting between that
22611         and equivalent gen_rtx_SUBREG code.
22613 2018-02-08  Richard Biener  <rguenther@suse.de>
22615         PR tree-optimization/84233
22616         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
22617         changed flag instead of boguously re-using phi_inserted.
22619 2018-02-08  Martin Jambor  <mjambor@suse.cz>
22621         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
22622         static local variables.
22624 2018-02-08  Richard Biener  <rguenther@suse.de>
22626         PR tree-optimization/84278
22627         * tree-vect-stmts.c (vectorizable_store): When looking for
22628         smaller vector types to perform grouped strided loads/stores
22629         make sure the mode is supported by the target.
22630         (vectorizable_load): Likewise.
22632 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
22634         * config/aarch64/aarch64.c (aarch64_components_for_bb):
22635         Increase LDP/STP opportunities by adding adjacent callee-saves.
22637 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
22639         PR rtl-optimization/84068
22640         PR rtl-optimization/83459
22641         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
22643 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
22645         PR tree-optimization/84224
22646         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
22647         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
22648         non-zero arguments.
22650 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
22652         PR target/84113
22653         * config/rs6000/altivec.md (*restore_world): Remove LR use.
22654         * config/rs6000/predicates.md (restore_world_operation): Adjust op
22655         count, remove one USE.
22657 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22659         * doc/install.texi (Configuration): Document the
22660         --with-long-double-format={ibm,ieee} PowerPC configuration
22661         options.
22663         PR target/84154
22664         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
22665         Convert from define_expand to be define_insn_and_split.  Rework
22666         float/double/_Float128 conversions to QI/HI/SImode to work with
22667         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
22668         conversions to QI/HImode types did a store and then a load to
22669         truncate the value.  For conversions to VSX registers, don't split
22670         the insn, instead emit the code directly.  Use the code iterator
22671         any_fix to combine signed and unsigned conversions.
22672         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
22673         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
22674         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22675         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
22676         (fix_<mode>di2_hw): Likewise.
22677         (fixuns_<mode>di2_hw): Likewise.
22678         (fix_<mode>si2_hw): Likewise.
22679         (fixuns_<mode>si2_hw): Likewise.
22680         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
22681         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22682         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
22683         fix<uns>_trunc<SFDF:mode>si2_p8.
22684         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
22685         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
22686         (fix<uns>_<mode>_mem): Likewise.
22687         (fctiw<u>z_<mode>_mem): Likewise.
22688         (fix<uns>_<mode>_mem): Likewise.
22689         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
22690         the register allocator from doing a direct move to the GPRs to do
22691         a store, and instead use the ISA 3.0 store byte/half-word from
22692         vector register instruction.  For IEEE 128-bit floating point,
22693         also optimize stores of 32-bit ints.
22694         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
22696 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
22698         * genextract.c (push_pathstr_operand): New function to support
22699         [a-zA-Z].
22700         (walk_rtx): Call push_pathstr_operand.
22701         (print_path): Support [a-zA-Z].
22703 2018-02-07  Richard Biener  <rguenther@suse.de>
22705         PR tree-optimization/84037
22706         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
22707         (cse_and_gimplify_to_preheader): Declare.
22708         (vect_get_place_in_interleaving_chain): Likewise.
22709         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22710         ivexpr_map.
22711         (_loop_vec_info::~_loop_vec_info): Delete it.
22712         (cse_and_gimplify_to_preheader): New function.
22713         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
22714         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
22715         (vectorizable_load): Likewise.  For grouped stores always base
22716         the IV on the first element.
22717         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
22718         condition before gimplifying.
22720 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
22722         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
22723         *DIV_EXPR and *MOD_EXPR.
22725 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
22727         PR target/84248
22728         * config/i386/i386.c (ix86_option_override_internal): Mask out
22729         the CF_SET bit when checking -fcf-protection.
22731 2018-02-07  Tom de Vries  <tom@codesourcery.com>
22733         PR libgomp/84217
22734         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
22735         enough.
22737 2018-02-07  Richard Biener  <rguenther@suse.de>
22739         PR tree-optimization/84204
22740         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
22741         this place.
22743         PR tree-optimization/84205
22744         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
22745         special-case isl_ast_op_zdiv_r.
22747         PR tree-optimization/84223
22748         * graphite-scop-detection.c (gather_bbs::before_dom_children):
22749         Only add conditions from within the region.
22750         (gather_bbs::after_dom_children): Adjust.
22752 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
22754         PR target/84209
22755         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
22756         * config/avr/avr.md: Only post-reload split REG-REG moves if
22757         either register is GENERAL_REG_P.
22759 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
22761         PR tree-optimization/84235
22762         * tree-ssa-scopedtables.c
22763         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
22764         if the subtraction is performed in floating point type where NaNs are
22765         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
22766         build 1.  Formatting fix.
22768 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
22770         PR target/84146
22771         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
22772         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
22773         and skip it regardless of bb boundaries.  Use CALL_P macro,
22774         don't test INSN_P (insn) together with CALL_P or JUMP_P check
22775         unnecessarily, formatting fix.
22777 2018-02-06  Michael Collison  <michael.collison@arm.com>
22779         * config/arm/thumb2.md:
22780         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
22781         (*thumb_mov_notscc): Ditto.
22783 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
22785         PR target/84154
22786         * config/rs6000/rs6000.md (su code attribute): Use "u" for
22787         unsigned_fix, not "s".
22789 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22791         * configure.ac (gcc_fn_eh_frame_ro): New function.
22792         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
22793         correct .eh_frame permissions.
22794         * configure: Regenerate.
22796 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
22798         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
22799         irrelevant options.
22801 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22803         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22804         Display warning message for -mno-speculate-indirect-jumps.
22806 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
22808         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22809         Undocumented.
22810         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22812 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
22814         PR tree-optimization/84225
22815         * tree-eh.c (find_trapping_overflow): Only call
22816         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22818 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22820         PR target/84145
22821         * config/i386/i386.c: Reimplement the check of possible options
22822         -mibt/-mshstk conbination. Change error messages.
22823         * doc/invoke.texi: Fix a typo: remove extra '='.
22825 2018-02-06  Marek Polacek  <polacek@redhat.com>
22827         PR tree-optimization/84228
22828         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22830 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
22832         PR target/82641
22833         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22834         emitted arch directives.
22835         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22836         __ARM_FEATURE_COPROC before changing architectures.
22838 2018-02-06  Richard Biener  <rguenther@suse.de>
22840         * config/i386/i386.c (print_reg): Fix typo.
22841         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22843 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
22845         * configure: Regenerate.
22847 2018-02-05  Martin Sebor  <msebor@redhat.com>
22849         PR tree-optimization/83369
22850         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22851         inlining context.
22853 2018-02-05  Martin Liska  <mliska@suse.cz>
22855         * doc/invoke.texi: Cherry-pick upstream r323995.
22857 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
22859         * ira.c (ira_init_register_move_cost): Adjust comment.
22861 2018-02-05  Martin Liska  <mliska@suse.cz>
22863         PR gcov-profile/84137
22864         * doc/gcov.texi: Fix typo in documentation.
22866 2018-02-05  Martin Liska  <mliska@suse.cz>
22868         PR gcov-profile/83879
22869         * doc/gcov.texi: Document necessity of --dynamic-list-data when
22870         using dlopen functionality.
22872 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
22874         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22875         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22876         _mm_maskz_range_ss, _mm_mask_range_round_ss,
22877         _mm_maskz_range_round_ss): New intrinsics.
22878         (__builtin_ia32_rangesd128_round)
22879         (__builtin_ia32_rangess128_round): Remove.
22880         (__builtin_ia32_rangesd128_mask_round,
22881         __builtin_ia32_rangess128_mask_round): New builtins.
22882         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22883         __builtin_ia32_rangess128_round): Remove.
22884         (__builtin_ia32_rangesd128_mask_round,
22885         __builtin_ia32_rangess128_mask_round): New builtins.
22886         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22887         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22888         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22889         "<round_saeonly_constraint>")): Changed to ...
22890         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22891         "<round_saeonly_scalar_constraint>")): ... this.
22892         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
22893         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
22894         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
22895         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
22896         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
22898 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
22900         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
22901         options.
22902         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
22903         Remove all values except native, 8540 and 8548.
22905 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
22907         * config/i386/i386.c (ix86_output_function_return): Pass
22908         INVALID_REGNUM, instead of -1, as invalid register number to
22909         indirect_thunk_name and output_indirect_thunk.
22911 2018-02-02  Julia Koval  <julia.koval@intel.com>
22913         * config.gcc: Add -march=icelake.
22914         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
22915         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
22916         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
22917         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
22918         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
22919         (processor_target_table): Add icelake.
22920         (ix86_option_override_internal): Handle new PTAs.
22921         (get_builtin_code_for_version): Handle icelake.
22922         (M_INTEL_COREI7_ICELAKE): New.
22923         (fold_builtin_cpu): Handle icelake.
22924         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
22925         * doc/invoke.texi: Add -march=icelake.
22927 2018-02-02  Julia Koval  <julia.koval@intel.com>
22929         * config/i386/i386.c (ix86_option_override_internal): Change flags type
22930         to wide_int_bitmask.
22931         * wide-int-bitmask.h: New.
22933 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22935         PR target/84066
22936         * config/i386/i386.md: Replace Pmode with word_mode in
22937         builtin_setjmp_setup and builtin_longjmp to support x32.
22939 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
22941         PR target/56010
22942         PR target/83743
22943         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22944         #include "opts.h".
22945         (rs6000_supported_cpu_names): New static variable.
22946         (linux_cpu_translation_table): Likewise.
22947         (elf_platform) <cpu>: Define new static variable and use it.
22948         Translate kernel AT_PLATFORM name to canonical name if needed.
22949         Error if platform name is unknown.
22951 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
22953         PR target/84089
22954         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22956 2018-02-01  Jeff Law  <law@redhat.com>
22958         PR target/84128
22959         * config/i386/i386.c (release_scratch_register_on_entry): Add new
22960         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
22961         the scratch if RELEASE_VIA_POP is false.
22962         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22963         If we have to save a temporary register, decrement SIZE appropriately.
22964         Pass new arguments to release_scratch_register_on_entry.
22965         (ix86_adjust_stack_and_probe): Likewise.
22966         (ix86_emit_probe_stack_range): Pass new arguments to
22967         release_scratch_register_on_entry.
22969 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
22971         PR rtl-optimization/84157
22972         * combine.c (change_zero_ext): Use REG_P predicate in
22973         front of HARD_REGISTER_P predicate.
22975 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
22977         * config/avr/avr.c (avr_option_override): Move disabling of
22978         -fdelete-null-pointer-checks to...
22979         * common/config/avr/avr-common.c (avr_option_optimization_table):
22980         ...here.
22982 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
22984         PR tree-optimization/81635
22985         * tree-data-ref.c (split_constant_offset_1): For types that
22986         wrap on overflow, try to use range info to prove that wrapping
22987         cannot occur.
22989 2018-02-01  Renlin Li  <renlin.li@arm.com>
22991         PR target/83370
22992         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22993         TAILCALL_ADDR_REGS.
22994         (aarch64_register_move_cost): Likewise.
22995         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22996         TAILCALL_ADDR_REGS.
22997         (REG_CLASS_NAMES): Likewise.
22998         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22999         TAILCALL_ADDR_REGS. Remove IP registers.
23000         * config/aarch64/aarch64.md (Ucs): Update register constraint.
23002 2018-02-01  Richard Biener  <rguenther@suse.de>
23004         * domwalk.h (dom_walker::dom_walker): Add additional constructor
23005         for specifying RPO order and allow NULL for that.
23006         * domwalk.c (dom_walker::dom_walker): Likewise.
23007         (dom_walker::walk): Handle NULL RPO order.
23008         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23009         in RPO order.
23010         (rewrite_update_dom_walker): Likewise.
23011         (mark_def_dom_walker): Likewise.
23013 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23015         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23016         (aarch64_maybe_expand_sve_subreg_move): Declare.
23017         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23018         * config/aarch64/predicates.md (aarch64_any_register_operand): New
23019         predicate.
23020         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23021         that are semantically a reverse operation.
23022         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23023         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23024         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23025         functions.
23026         (aarch64_can_change_mode_class): For big-endian, forbid changes
23027         between two SVE modes if they have different element sizes.
23029 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23031         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23032         the TImode handling for big-endian targets.
23034 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23036         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23037         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
23038         not just bytes.
23039         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23040         Remove BSWAP handing for big-endian targets and use the form of
23041         LD1RQ appropariate for the mode.
23043 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23045         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23046         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
23047         duplicated element.
23049 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23051         PR tearget/83845
23052         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
23053         check for operands that need to go through aarch64_sve_reload_be.
23055 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
23057         PR tree-optimization/81661
23058         PR tree-optimization/84117
23059         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23060         * tree-eh.c: Include gimplify.h.
23061         (find_trapping_overflow, replace_trapping_overflow,
23062         rewrite_to_non_trapping_overflow): New functions.
23063         * tree-vect-loop.c: Include tree-eh.h.
23064         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23065         * tree-data-ref.c: Include tree-eh.h.
23066         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23068 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
23070         PR rtl-optimization/84123
23071         * combine.c (change_zero_ext): Check if hard register satisfies
23072         can_change_dest_mode before calling gen_lowpart_SUBREG.
23074 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
23076         PR target/82444
23077         * ira.c (ira_init_register_move_cost): Remove assert.
23079 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23081         PR rtl-optimization/84071
23082         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23083         * doc/tm.texi: Regenerate.
23085 2018-01-31  Richard Biener  <rguenther@suse.de>
23087         PR tree-optimization/84132
23088         * tree-data-ref.c (analyze_miv_subscript): Properly
23089         check whether evolution_function_is_affine_multivariate_p
23090         before calling gcd_of_steps_may_divide_p.
23092 2018-01-31  Julia Koval  <julia.koval@intel.com>
23094         PR target/83618
23095         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23096         * config/i386/i386.md (rdpid_rex64) New.
23097         (rdpid): Make 32bit only.
23099 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
23101         PR lto/84105
23102         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23103         an IDENTIFIER_NODE for FUNCTION_TYPE's.
23105 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23107         Revert
23108         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23110         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23112 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23114         PR rtl-optimization/84071
23115         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
23116         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
23118 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23120         * config/arc/arc.c (arc_handle_aux_attribute): New function.
23121         (arc_attribute_table): Add 'aux' attribute.
23122         (arc_in_small_data_p): Consider aux like variables.
23123         (arc_is_aux_reg_p): New function.
23124         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
23125         (arc_get_aux_arg): New function.
23126         (prepare_move_operands): Handle aux-register access.
23127         (arc_handle_aux_attribute): New function.
23128         * doc/extend.texi (ARC Variable attributes): Add subsection.
23130 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23132         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
23133         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
23134         (arc_attribute_table): Add 'uncached' attribute.
23135         (arc_print_operand): Print '.di' flag for uncached memory
23136         accesses.
23137         (arc_in_small_data_p): Do not consider for small data the uncached
23138         types.
23139         (arc_is_uncached_mem_p): New function.
23140         * config/arc/predicates.md (compact_store_memory_operand): Check
23141         for uncached memory accesses.
23142         (nonvol_nonimm_operand): Likewise.
23143         * doc/extend.texi (ARC Type Attribute): New subsection.
23145 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
23147         PR c/84100
23148         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
23149         falign-loops=): Add Optimization flag.
23151 2018-01-30  Jeff Law  <law@redhat.com>
23153         PR target/84064
23154         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
23155         INT_REGISTERS_SAVED.  Check it prior to calling
23156         get_scratch_register_on_entry.
23157         (ix86_adjust_stack_and_probe): Similarly.
23158         (ix86_emit_probe_stack_range): Similarly.
23159         (ix86_expand_prologue): Corresponding changes.
23161 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23163         PR target/40411
23164         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
23165         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
23167 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
23169         PR target/84112
23170         * lra-constraints.c (curr_insn_transform): Process AND in the
23171         address.
23173 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
23175         PR rtl-optimization/83986
23176         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
23177         dependence against last_pending_memory_flush in addition to
23178         pending_jump_insns.
23180 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
23182         PR tree-optimization/81611
23183         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
23184         copies.
23186 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23188         PR target/83758
23189         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
23190         a reg rtx.
23192 2018-01-30  Richard Biener  <rguenther@suse.de>
23193             Jakub Jelinek  <jakub@redhat.com>
23195         PR tree-optimization/84111
23196         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
23197         inner loops added during recursion, as they don't have up-to-date
23198         SSA form.
23200 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23202         PR ipa/81360
23203         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
23204         (can_inline_edge_by_limits_p): ... here.
23205         (can_early_inline_edge_p, check_callers,
23206         update_caller_keys, update_callee_keys, recursive_inlining,
23207         add_new_edges_to_heap, speculation_useful_p,
23208         inline_small_functions,
23209         inline_small_functions, flatten_function,
23210         inline_to_all_callers_1): Update.
23212 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23214         * profile-count.c (profile_count::combine_with_ipa_count): Handle
23215         zeros correctly.
23217 2018-01-30  Richard Biener  <rguenther@suse.de>
23219         PR tree-optimization/83008
23220         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
23221         invariant and constant vector uses in stmts when they need
23222         more than one stmt.
23224 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23226         PR bootstrap/84017
23227         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
23228         * configure: Regenerate.
23230 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23232         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
23233         pattern.
23234         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
23235         Use gen_rtx_REG rather than gen_lowpart.
23237 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23239         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
23240         rather than 0 when creating partial subregs.
23242 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23244         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
23245         of usage.
23247 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
23249         PR target/81550
23250         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
23251         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
23252         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
23253         flags.  This restores the settings used before the 2017-07-24.
23254         Turning off pre increment/decrement/modify allows IVOPTS to
23255         optimize DF/SF loops where the index is an int.
23257 2018-01-29  Richard Biener  <rguenther@suse.de>
23258             Kelvin Nilsen  <kelvin@gcc.gnu.org>
23260         PR bootstrap/80867
23261         * tree-vect-stmts.c (vectorizable_call): Don't call
23262         targetm.vectorize_builtin_md_vectorized_function if callee is
23263         NULL.
23265 2018-01-22  Carl Love  <cel@us.ibm.com>
23267         * doc/extend.tex: Fix typo in second arg in
23268         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
23270 2018-01-29  Richard Biener  <rguenther@suse.de>
23272         PR tree-optimization/84086
23273         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
23274         (flush_ssaname_freelist): When SSA names were released reset
23275         the SCEV hash table.
23277 2018-01-29  Richard Biener  <rguenther@suse.de>
23279         PR tree-optimization/84057
23280         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
23281         removed paths when removing edges.
23283 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
23285         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
23286         -mfunction-return=@var{choice}.
23288 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
23290         PR diagnostic/84034
23291         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
23292         Handle CR like TAB.
23293         (layout::print_source_line): Likewise.
23294         (test_get_line_width_without_trailing_whitespace): Add test cases.
23296 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
23298         PR middle-end/84040
23299         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
23300         debug insns.
23302 2018-01-26  Jim Wilson  <jimw@sifive.com>
23304         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
23306         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
23307         specified.
23309 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23311         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
23312         and CMP + SUB-immediate -> SUBS.
23314 2018-01-26  Martin Sebor  <msebor@redhat.com>
23316         PR tree-optimization/83896
23317         * tree-ssa-strlen.c (get_string_len): Rename...
23318         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
23319         Avoid assuming length is constant.
23320         (handle_char_store): Use HOST_WIDE_INT for string length.
23322 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
23324         PR target/81763
23325         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
23326         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
23328 2018-01-26  Richard Biener  <rguenther@suse.de>
23330         PR rtl-optimization/84003
23331         * dse.c (record_store): Only record redundant stores when
23332         the earlier store aliases at least all accesses the later one does.
23334 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
23336         PR rtl-optimization/83985
23337         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
23338         REG_CFA_RESTORE insns.
23339         (delete_unmarked_insns): Don't ignore separate shrink wrapping
23340         REG_CFA_RESTORE insns here.
23342         PR c/83989
23343         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
23344         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
23346 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23348         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
23349         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
23350         (arc_init): Likewise.
23351         (arc_override_options): Likewise.
23352         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
23353         value.
23354         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
23355         support.
23356         * config/arc/arc.h (TARGET_DBNZ): Define.
23357         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
23358         properly set the tune attribute.
23359         (dbnz): Use TARGET_DBNZ guard.
23360         * config/arc/arc.opt (mtune): Add core3 option.
23362 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23364         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
23365         recognize new pic like addresses.
23366         (arc_delegitimize_address): Clean up.
23368 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23370         * config/arc/arc-arches.def: Option mrf16 valid for all
23371         architectures.
23372         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
23373         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
23374         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
23375         * config/arc/arc-tables.opt: Regenerate.
23376         * config/arc/arc.c (arc_conditional_register_usage): Handle
23377         reduced register file case.
23378         (arc_file_start): Set must have build attributes.
23379         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
23380         mrf16 option value.
23381         * config/arc/arc.opt (mrf16): Add new option.
23382         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
23383         * config/arc/genmultilib.awk: Handle new mrf16 option.
23384         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
23385         * config/arc/t-multilib: Regenerate.
23386         * doc/invoke.texi (ARC Options): Document mrf16 option.
23388 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23390         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
23391         * config/arc/arc.c (arc_handle_secure_attribute): New function.
23392         (arc_attribute_table): Add 'secure_call' attribute.
23393         (arc_print_operand): Print secure call operand.
23394         (arc_function_ok_for_sibcall): Don't optimize tail calls when
23395         secure.
23396         (arc_is_secure_call_p): New function.  * config/arc/arc.md
23397         (call_i): Add support for sjli instruction.
23398         (call_value_i): Likewise.
23399         * config/arc/constraints.md (Csc): New constraint.
23401 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
23402             John Eric Martin  <John.Martin@emmicro-us.com>
23404         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
23405         * config/arc/arc.c (_arc_jli_section): New struct.
23406         (arc_jli_section): New type.
23407         (rc_jli_sections): New static variable.
23408         (arc_handle_jli_attribute): New function.
23409         (arc_attribute_table): Add jli_always and jli_fixed attribute.
23410         (arc_file_end): New function.
23411         (TARGET_ASM_FILE_END): Define.
23412         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
23413         (arc_add_jli_section): New function.
23414         (jli_call_scan): Likewise.
23415         (arc_reorg): Call jli_call_scan.
23416         (arc_output_addsi): Remove 'S' from printing asm operand.
23417         (arc_is_jli_call_p): New function.
23418         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
23419         operand.
23420         (movhi_insn): Likewise.
23421         (movsi_insn): Likewise.
23422         (movsi_set_cc_insn): Likewise.
23423         (loadqi_update): Likewise.
23424         (load_zeroextendqisi_update): Likewise.
23425         (load_signextendqisi_update): Likewise.
23426         (loadhi_update): Likewise.
23427         (load_zeroextendhisi_update): Likewise.
23428         (load_signextendhisi_update): Likewise.
23429         (loadsi_update): Likewise.
23430         (loadsf_update): Likewise.
23431         (movsicc_insn): Likewise.
23432         (bset_insn): Likewise.
23433         (bxor_insn): Likewise.
23434         (bclr_insn): Likewise.
23435         (bmsk_insn): Likewise.
23436         (bicsi3_insn): Likewise.
23437         (cmpsi_cc_c_insn): Likewise.
23438         (movsi_ne): Likewise.
23439         (movsi_cond_exec): Likewise.
23440         (clrsbsi2): Likewise.
23441         (norm_f): Likewise.
23442         (normw): Likewise.
23443         (swap): Likewise.
23444         (divaw): Likewise.
23445         (flag): Likewise.
23446         (sr): Likewise.
23447         (kflag): Likewise.
23448         (ffs): Likewise.
23449         (ffs_f): Likewise.
23450         (fls): Likewise.
23451         (call_i): Remove 'S' asm letter, add jli instruction.
23452         (call_value_i): Likewise.
23453         * config/arc/arc.op (mjli-always): New option.
23454         * config/arc/constraints.md (Cji): New constraint.
23455         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
23456         operand.
23457         (subsf3_fpx): Likewise.
23458         (mulsf3_fpx): Likewise.
23459         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
23460         asm operand.
23461         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
23462         function attrbutes.
23463         * doc/invoke.texi (ARC): Document mjli-always option.
23465 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
23467         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
23468         avoid addition with 0 and use incw and decw where possible.
23470 2018-01-26  Richard Biener  <rguenther@suse.de>
23472         PR tree-optimization/81082
23473         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
23474         association if it requires casting to unsigned.
23475         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
23476         from fold_plusminus_mult_expr to catch important cases late when
23477         range info is available.
23479 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23481         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
23482         * configure.ac (hidden_linkonce): New test.
23483         * configure: Regenerate.
23484         * config.in: Regenerate.
23486 2018-01-26  Julia Koval  <julia.koval@intel.com>
23488         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
23489         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
23490         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
23491         _mm_mask_bitshuffle_epi64_mask): Fix type.
23492         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
23493         USI_FTYPE_V4DI_V4DI_USI): Remove.
23494         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
23495         __builtin_ia32_vpshufbitqmb256_mask,
23496         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
23497         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
23498         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
23500 2018-01-26  Alan Modra  <amodra@gmail.com>
23502         PR target/84033
23503         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
23504         UNSPEC_VBPERMQ.  Sort other unspecs.
23506 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
23508         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
23510 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
23512         PR middle-end/83055
23513         * predict.c (drop_profile): Do not push/pop cfun; update also
23514         node->count.
23515         (handle_missing_profiles): Fix logic looking for zero profiles.
23517 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
23519         PR middle-end/83977
23520         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
23521         on functions with #pragma omp declare simd or functions with simd
23522         attribute.
23523         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
23524         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
23525         Remove trailing \n from warning_at calls.
23527 2018-01-25  Tom de Vries  <tom@codesourcery.com>
23529         PR target/84028
23530         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23531         for neutered workers.
23533 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
23535         PR target/68467
23536         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
23537         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
23539 2018-01-24  Jeff Law  <law@redhat.com>
23541         PR target/83994
23542         * i386.c (get_probe_interval): Move to earlier point.
23543         (ix86_compute_frame_layout): If -fstack-clash-protection and
23544         the frame is larger than the probe interval, then use pushes
23545         to save registers rather than reg->mem moves.
23546         (ix86_expand_prologue): Remove conditional for int_registers_saved
23547         assertion.
23549 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
23551         PR target/84014
23552         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
23553         min/max for never referenced object.
23555 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
23557         PR middle-end/83977
23558         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
23559         here.
23560         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
23561         attributes from DECL_ATTRIBUTES (decl) without affecting
23562         DECL_ATTRIBUTES (current_function_decl).
23563         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
23564         functions with non-NULL DECL_ABSTRACT_ORIGIN.
23566 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
23568         PR tree-optimization/83979
23569         * fold-const.c (fold_comparison): Use constant_boolean_node
23570         instead of boolean_{true,false}_node.
23572 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
23574         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
23575         with zero counts.
23577 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23579         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23580         Simplify the clause that sets the length attribute.
23581         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23582         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
23583         clause that sets the length attribute.
23584         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23586 2018-01-24  Tom de Vries  <tom@codesourcery.com>
23588         PR target/83589
23589         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
23590         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
23591         Add strict parameter.
23592         (prevent_branch_around_nothing): Insert dummy insn between branch to
23593         label and label with no ptx insn inbetween.
23594         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
23596 2018-01-24  Tom de Vries  <tom@codesourcery.com>
23598         PR target/81352
23599         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23600         for neutered threads in warp.
23601         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
23603 2018-01-24  Richard Biener  <rguenther@suse.de>
23605         PR tree-optimization/83176
23606         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
23607         operands.
23609 2018-01-24  Richard Biener  <rguenther@suse.de>
23611         PR tree-optimization/82819
23612         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
23613         code generating pluses that are no-ops in the target precision.
23615 2018-01-24  Richard Biener  <rguenther@suse.de>
23617         PR middle-end/84000
23618         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
23620 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23622         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
23623         to merge probabilities.
23624         * predict.c (probably_never_executed): Also mark as cold functions
23625         with global 0 profile and guessed local profile.
23626         * profile-count.c (profile_probability::combine_with_count): New
23627         member function.
23628         * profile-count.h (profile_probability::operator*,
23629         profile_probability::operator*=, profile_probability::operator/,
23630         profile_probability::operator/=): Reduce precision to adjusted
23631         and set value to guessed on contradictory divisions.
23632         (profile_probability::combine_with_freq): Remove.
23633         (profile_probability::combine_wiht_count): Declare.
23634         (profile_count::force_nonzero):: Set to adjusted.
23635         (profile_count::probability_in):: Set quality to adjusted.
23636         * tree-ssa-tail-merge.c (replace_block_by): Use
23637         combine_with_count.
23639 2018-01-23  Andrew Waterman  <andrew@sifive.com>
23640             Jim Wilson  <jimw@sifive.com>
23642         * config/riscv/riscv.c (riscv_stack_boundary): New.
23643         (riscv_option_override): Set riscv_stack_boundary.  Handle
23644         riscv_preferred_stack_boundary_arg.
23645         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
23646         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
23647         (STACK_BOUNDARY): Set to riscv_stack_boundary.
23648         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
23649         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
23650         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
23652 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
23654         PR target/83905
23655         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
23656         of struct ix86_frame.
23657         (ix86_expand_epilogue): Likewise.  Add a local variable for
23658         the reg_save_offset field in struct ix86_frame.
23660 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
23662         PR tree-optimization/82604
23663         * tree-loop-distribution.c (enum partition_kind): New enum item
23664         PKIND_PARTIAL_MEMSET.
23665         (partition_builtin_p): Support above new enum item.
23666         (generate_code_for_partition): Ditto.
23667         (compute_access_range): Differentiate cases that equality can be
23668         proven at all loops, the innermost loops or no loops.
23669         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
23670         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
23671         (finalize_partitions, distribute_loop): Don't fuse partition of
23672         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
23673         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
23674         parloop is enabled.
23676 2018-01-23  Martin Liska  <mliska@suse.cz>
23678         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
23679         order to ignore the predictor.
23680         (PRED_POLYMORPHIC_CALL): Likewise.
23681         (PRED_RECURSIVE_CALL): Likewise.
23683 2018-01-23  Martin Liska  <mliska@suse.cz>
23685         * tree-profile.c (tree_profiling): Print function header to
23686         aware reader which function we are working on.
23687         * value-prof.c (gimple_find_values_to_profile): Do not print
23688         not interesting value histograms.
23690 2018-01-23  Martin Liska  <mliska@suse.cz>
23692         * profile-count.h (enum profile_quality): Add
23693         profile_uninitialized as the first value. Do not number values
23694         as they are zero based.
23695         (profile_count::verify): Update sanity check.
23696         (profile_probability::verify): Likewise.
23698 2018-01-23  Nathan Sidwell  <nathan@acm.org>
23700         * doc/invoke.texi (ffor-scope): Deprecate.
23702 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23704         PR tree-optimization/83510
23705         * domwalk.c (set_all_edges_as_executable): New function.
23706         (dom_walker::dom_walker): Convert bool param
23707         "skip_unreachable_blocks" to enum reachability.  Move setup of
23708         edge flags to set_all_edges_as_executable and only do it when
23709         reachability is REACHABLE_BLOCKS.
23710         * domwalk.h (enum dom_walker::reachability): New enum.
23711         (dom_walker::dom_walker): Convert bool param
23712         "skip_unreachable_blocks" to enum reachability.
23713         (set_all_edges_as_executable): New decl.
23714         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
23715         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
23716         "reachability".
23717         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
23718         but converting true to REACHABLE_BLOCKS.
23719         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
23720         * tree-vrp.c
23721         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
23722         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
23723         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
23724         REACHABLE_BLOCKS.
23725         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
23726         if check_all_array_refs will be called.
23728 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23730         * tree.c (selftest::test_location_wrappers): Add more test
23731         coverage.
23733 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
23735         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
23736         (selftest::test_bit_in_range): Likewise.
23738 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
23740         PR testsuite/83888
23741         * doc/sourcebuild.texi (vect_float): Say that the selector
23742         only describes the situation when -funsafe-math-optimizations is on.
23743         (vect_float_strict): Document.
23745 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
23747         PR tree-optimization/83965
23748         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
23749         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
23750         instead of checking only for a reduction.
23751         (vect_recog_widen_sum_pattern): Likewise.
23753 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23755         * predict.c (probably_never_executed): Only use precise profile info.
23756         (compute_function_frequency): Skip after inlining hack since we now
23757         have quality checking.
23759 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
23761         * profile-count.h (profile_probability::very_unlikely,
23762         profile_probability::unlikely, profile_probability::even): Set
23763         precision to guessed.
23765 2018-01-23  Richard Biener  <rguenther@suse.de>
23767         PR tree-optimization/83963
23768         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
23769         Properly terminate dominator walk when crossing the exit edge not
23770         when visiting its source block.
23772 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
23774         PR c++/83918
23775         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
23776         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
23778 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
23780         PR tree-optimization/83957
23781         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
23782         semicolon after for body surrounded by braces.
23784         PR tree-optimization/83081
23785         * profile-count.h (profile_probability::split): New method.
23786         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
23787         Use profile_probability::split.
23788         (do_compare_rtx_and_jump): Fix adjustment of probabilities
23789         when splitting a single conditional jump into 2.
23791 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
23793         PR tree-optimization/69452
23794         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
23795         decl.
23797 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23799         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
23800         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
23801         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
23803 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23805         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23806         declaration.
23807         * config/rl78/rl78.md (movdi): New define_expand.
23808         * config/rl78/rl78.c (rl78_split_movdi): New function.
23810 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
23812         PR target/83862
23813         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23814         no longer used.
23815         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23816         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23817         128-bit to produce an UNSPEC move to get the double word with the
23818         signbit and then a shift directly to do signbit.
23819         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23820         implementation with a new version that just does either a direct
23821         move or a regular move.  Move memory interface to separate insns.
23822         Move insns so they are next to the expander.
23823         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23824         with signbit move.  Split big and little endian case.
23825         (signbit<mode>2_dm_mem_le): Likewise.
23826         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23827         (signbit<mode>2_dm2): Likewise.
23829 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23831         * config/rl78/rl78.md (anddi3): New define_expand.
23833 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23835         * config/rl78/rl78.md (umindi3): New define_expand.
23837 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23839         * config/rl78/rl78.md (smindi3): New define_expand.
23841 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23843         * config/rl78/rl78.md (smaxdi3): New define_expand.
23845 2018-01-22  Carl Love  <cel@us.ibm.com>
23847         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23848         LVX_V1TI): Add macro expansion.
23849         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23850         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23851         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23852         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23853         Change check to determine if the instruction is a byte reversing
23854         entry.  Fix typo in comment.
23855         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23856         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23857         Add def_builtin calls for new builtins.
23858         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23859         Add define_insn expansion.
23861 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23863         * config/rl78/rl78.md (umaxdi3): New define_expand.
23865 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
23867         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23868         for non-QImode registers.
23870 2018-01-22  Richard Biener  <rguenther@suse.de>
23872         PR tree-optimization/83963
23873         * graphite-scop-detection.c (scop_detection::get_sese): Delay
23874         including the loop exit block.
23875         (scop_detection::merge_sese): Likewise.
23876         (scop_detection::add_scop): Do it here instead.
23878 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23880         * doc/sourcebuild.texi (arm_softfloat): Document.
23882 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
23884         PR gcc/77734
23885         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23886         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23887         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23889 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23890             David Edelsohn  <dje.gcc@gmail.com>
23892         PR target/83946
23893         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23894         Change "crset eq" to "crset 2".
23895         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23896         (*call_indirect_aix<mode>_nospec): Likewise.
23897         (*call_value_indirect_aix<mode>_nospec): Likewise.
23898         (*call_indirect_elfv2<mode>_nospec): Likewise.
23899         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
23900         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
23901         change assembly output from . to $.
23902         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23903         (indirect_jump<mode>_nospec): Change assembly output from . to $.
23904         (*tablejump<mode>_internal1_nospec): Likewise.
23906 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
23908         PR target/80870
23909         * config/sh/sh_optimize_sett_clrt.cc:
23910         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
23912 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
23914         PR tree-optimization/83940
23915         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
23916         offset_dt to vect_constant_def rather than vect_unknown_def_type.
23917         (vect_check_load_store_mask): Add a mask_dt_out parameter and
23918         use it to pass back the definition type.
23919         (vect_check_store_rhs): Likewise rhs_dt_out.
23920         (vect_build_gather_load_calls): Add a mask_dt argument and use
23921         it instead of a call to vect_is_simple_use.
23922         (vectorizable_store): Update calls to vect_check_load_store_mask
23923         and vect_check_store_rhs.  Use the dt returned by the latter instead
23924         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
23925         instead of calls to vect_is_simple_use.  Pass the scalar rather
23926         than the vector operand to vect_is_simple_use when handling
23927         second and subsequent copies of an rhs value.
23928         (vectorizable_load): Update calls to vect_check_load_store_mask
23929         and vect_build_gather_load_calls.  Use the cached mask_dt and
23930         gs_info.offset_dt instead of calls to vect_is_simple_use.
23932 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
23934         PR middle-end/83945
23935         * tree-emutls.c: Include gimplify.h.
23936         (lower_emutls_2): New function.
23937         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23938         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23939         it before further processing.
23941         PR target/83930
23942         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23943         UINTVAL (trueop1) instead of INTVAL (op1).
23945 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
23947         PR debug/81570
23948         PR debug/83728
23949         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23950         INCOMING_FRAME_SP_OFFSET if not defined.
23951         (scan_trace): Add ENTRY argument.  If true and
23952         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23953         emit a note to adjust the CFA offset.
23954         (create_cfi_notes): Adjust scan_trace callers.
23955         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23956         INCOMING_FRAME_SP_OFFSET in the CIE.
23957         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23958         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23959         Likewise.
23960         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23961         * doc/tm.texi: Regenerated.
23963 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23965         PR rtl-optimization/83147
23966         * lra-constraints.c (remove_inheritance_pseudos): Use
23967         lra_substitute_pseudo_within_insn.
23969 2018-01-19  Tom de Vries  <tom@codesourcery.com>
23970             Cesar Philippidis  <cesar@codesourcery.com>
23972         PR target/83920
23973         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23975 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
23977         PR target/83790
23978         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23979         spaces for function labels.
23981 2018-01-19  Martin Liska  <mliska@suse.cz>
23983         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23984         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23985         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23986         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23987         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23988         (PRED_CONST_RETURN): Change from 69 to 65.
23989         (PRED_NULL_RETURN): Change from 91 to 71.
23990         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23991         (PRED_LOOP_GUARD): Change from 66 to 73.
23993 2018-01-19  Martin Liska  <mliska@suse.cz>
23995         * predict.c (predict_insn_def): Add new assert.
23996         (struct branch_predictor): Change type to signed integer.
23997         (test_prediction_value_range): Amend test to cover
23998         PROB_UNINITIALIZED.
23999         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24000         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24001         (PRED_LOOP_ITERATIONS_MAX): Likewise.
24002         (PRED_LOOP_IV_COMPARE): Likewise.
24003         * predict.h (PROB_UNINITIALIZED): Define new constant.
24005 2018-01-19  Martin Liska  <mliska@suse.cz>
24007         * predict.c (dump_prediction): Add new format for
24008         analyze_brprob.py script which is enabled with -details
24009         suboption.
24010         * profile-count.h (precise_p): New function.
24012 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24014         PR tree-optimization/83922
24015         * tree-vect-loop.c (vect_verify_full_masking): Return false if
24016         there are no statements that need masking.
24017         (vect_active_double_reduction_p): New function.
24018         (vect_analyze_loop_operations): Use it when handling phis that
24019         are not in the loop header.
24021 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24023         PR tree-optimization/83914
24024         * tree-vect-loop.c (vectorizable_induction): Don't convert
24025         init_expr or apply the peeling adjustment for inductions
24026         that are nested within the vectorized loop.
24028 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24030         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24031         instead of NEG.
24033 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24035         PR sanitizer/81715
24036         PR testsuite/83882
24037         * function.h (gimplify_parameters): Add gimple_seq * argument.
24038         * function.c: Include gimple.h and options.h.
24039         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24040         for the added local temporaries if needed.
24041         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24042         if there are any parameter cleanups, wrap whole body into a
24043         try/finally with the cleanups.
24045 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
24047         PR target/82964
24048         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
24049         Use GET_MODE_CLASS for scalar floating point.
24051 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24053         PR ipa/82256
24054         patch by PaX Team
24055         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
24056         Fix call of call_cgraph_insertion_hooks.
24058 2018-01-18  Martin Sebor  <msebor@redhat.com>
24060         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24062 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24064         PR ipa/83619
24065         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24066         frequencies.
24068 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
24070         PR other/70268
24071         * common.opt: (-ffile-prefix-map): New option.
24072         * opts.c (common_handle_option): Defer it.
24073         * opts-global.c (handle_common_deferred_options): Handle it.
24074         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24075         * file-prefix-map.h: New file.
24076         (remap_debug_filename, add_debug_prefix_map): ...here.
24077         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24078         * final.c (debug_prefix_map, add_debug_prefix_map
24079         remap_debug_filename): Move to...
24080         * file-prefix-map.c: New file.
24081         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24082         generalize, get rid of alloca(), use strrchr() instead of strchr().
24083         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24084         Implement in terms of add_prefix_map().
24085         (remap_macro_filename, remap_debug_filename): Implement in term of
24086         remap_filename().
24087         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24088         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24089         * dbxout.c: Include file-prefix-map.h.
24090         * varasm.c: Likewise.
24091         * vmsdbgout.c: Likewise.
24092         * xcoffout.c: Likewise.
24093         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24094         * doc/cppopts.texi (-fmacro-prefix-map): Document.
24095         * doc/invoke.texi (-ffile-prefix-map): Document.
24096         (-fdebug-prefix-map): Update description.
24098 2018-01-18  Martin Liska  <mliska@suse.cz>
24100         * config/i386/i386.c (indirect_thunk_name): Document that also
24101         lfence is emitted.
24102         (output_indirect_thunk): Document why both instructions
24103         (pause and lfence) are generated.
24105 2018-01-18  Richard Biener  <rguenther@suse.de>
24107         PR tree-optimization/83887
24108         * graphite-scop-detection.c
24109         (scop_detection::get_nearest_dom_with_single_entry): Remove.
24110         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
24111         (scop_detection::merge_sese): Re-implement with a flood-fill
24112         algorithm that properly finds a SESE region if it exists.
24114 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24116         PR c/61240
24117         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
24118         pointer_diff optimizations use view_convert instead of convert.
24120 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24122         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24123         Generate different code for -mno-speculate-indirect-jumps.
24124         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24125         (*call_indirect_aix<mode>): Disable for
24126         -mno-speculate-indirect-jumps.
24127         (*call_indirect_aix<mode>_nospec): New define_insn.
24128         (*call_value_indirect_aix<mode>): Disable for
24129         -mno-speculate-indirect-jumps.
24130         (*call_value_indirect_aix<mode>_nospec): New define_insn.
24131         (*sibcall_nonlocal_sysv<mode>): Generate different code for
24132         -mno-speculate-indirect-jumps.
24133         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24135 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
24137         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
24138         long double type, set the flags for noting the default long double
24139         type, even if we don't pass or return a long double type.
24141 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
24143         PR ipa/83051
24144         * ipa-inline.c (flatten_function): Do not overwrite final inlining
24145         failure.
24147 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
24149         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
24150         support for merge[hl].
24151         (fold_mergehl_helper): New helper function.
24152         (tree-vector-builder.h): New #include for tree_vector_builder usage.
24153         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
24154         (altivec_vmrglw_direct): Add xxmrglw insn.
24156 2018-01-17  Andrew Waterman  <andrew@sifive.com>
24158         * config/riscv/riscv.c (riscv_conditional_register_usage): If
24159         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
24161 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
24163         PR lto/83121
24164         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
24165         call the lto_location_cache before reading the
24166         DECL_SOURCE_LOCATION of the types.
24168 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
24169             Richard Sandiford  <richard.sandiford@linaro.org>
24171         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
24172         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
24173         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
24174         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
24175         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
24176         Add declaration.
24177         * config/aarch64/constraints.md (aarch64_movti_operand):
24178         Limit immediates.
24179         * config/aarch64/predicates.md (Uti): Add new constraint.
24181 2018-01-17  Carl Love  <cel@us.ibm.com>
24183         * config/rs6000/vsx.md (define_expand xl_len_r,
24184         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
24185         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
24186         lxvll.
24187         (define_expand, define_insn): Move the shift left from  the
24188         define_insn to the define_expand for lxvl and stxvl instructions.
24189         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
24190         and XL_LEN_R definitions to PURE.
24192 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
24194         * config/i386/i386.c (indirect_thunk_name): Declare regno
24195         as unsigned int.  Compare regno with INVALID_REGNUM.
24196         (output_indirect_thunk): Ditto.
24197         (output_indirect_thunk_function): Ditto.
24198         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
24199         in the call to output_indirect_thunk_function.
24201 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
24203         PR middle-end/83884
24204         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
24205         rather than the size of inner_type to determine the stack slot size
24206         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
24208 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
24210         PR target/83546
24211         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
24212         to PTA_SILVERMONT.
24214 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
24216         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
24217         endian Linux systems to optionally enable multilibs for selecting
24218         the long double type if the user configured an explicit type.
24219         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
24220         have no long double multilibs if not defined.
24221         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
24222         warn if the user used -mabi={ieee,ibm}longdouble and we built
24223         multilibs for long double.
24224         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
24225         appropriate multilib option.
24226         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
24227         multilib options.
24228         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
24229         for building long double multilibs.
24230         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
24232 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
24234         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
24235         copies.
24237         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
24238         64 bits.
24239         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
24240         128 bits.
24242         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
24243         variables.
24245         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
24246         return value.
24248 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24250         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
24251         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
24253 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24255         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
24256         different rtl trees depending on TARGET_64BIT.
24257         (rs6000_gen_lvx): Likewise.
24259 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24261         * config/visium/visium.md (nop): Tweak comment.
24262         (hazard_nop): Likewise.
24264 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24266         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
24267         -mspeculate-indirect-jumps.
24268         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
24269         for -mno-speculate-indirect-jumps.
24270         (*call_indirect_elfv2<mode>_nospec): New define_insn.
24271         (*call_value_indirect_elfv2<mode>): Disable for
24272         -mno-speculate-indirect-jumps.
24273         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
24274         (indirect_jump): Emit different RTL for
24275         -mno-speculate-indirect-jumps.
24276         (*indirect_jump<mode>): Disable for
24277         -mno-speculate-indirect-jumps.
24278         (*indirect_jump<mode>_nospec): New define_insn.
24279         (tablejump): Emit different RTL for
24280         -mno-speculate-indirect-jumps.
24281         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
24282         (tablejumpsi_nospec): New define_expand.
24283         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
24284         (tablejumpdi_nospec): New define_expand.
24285         (*tablejump<mode>_internal1): Disable for
24286         -mno-speculate-indirect-jumps.
24287         (*tablejump<mode>_internal1_nospec): New define_insn.
24288         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
24289         option.
24291 2018-01-16  Artyom Skrobov tyomitch@gmail.com
24293         * caller-save.c (insert_save): Drop unnecessary parameter.  All
24294         callers updated.
24296 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24297             Richard Biener  <rguenth@suse.de>
24299         PR libgomp/83590
24300         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
24301         return early, inline manually is_gimple_sizepos.  Make sure if we
24302         call gimplify_expr we don't end up with a gimple constant.
24303         * tree.c (variably_modified_type_p): Don't return true for
24304         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
24305         * gimplify.h (is_gimple_sizepos): Remove.
24307 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24309         PR tree-optimization/83857
24310         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
24311         vectorizable_live_operation for pure SLP statements.
24312         (vectorizable_live_operation): Handle PHIs.
24314 2018-01-16  Richard Biener  <rguenther@suse.de>
24316         PR tree-optimization/83867
24317         * tree-vect-stmts.c (vect_transform_stmt): Precompute
24318         nested_in_vect_loop_p since the scalar stmt may get invalidated.
24320 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24322         PR c/83844
24323         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
24324         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
24325         If off is not INTEGER_CST, issue a may not be aligned warning
24326         rather than isn't aligned.  Use isn%'t rather than isn't.
24327         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
24328         into MULT_EXPR.
24329         <case MULT_EXPR>: Improve the case when bottom and one of the
24330         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
24331         operand, in that case check if the other operand is multiple of
24332         bottom divided by the INTEGER_CST operand.
24334 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24336         PR target/83858
24337         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
24338         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
24339         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
24340         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
24341         * config/pa/pa.c (pa_function_arg_advance): Likewise.
24342         (pa_function_arg, pa_arg_partial_bytes): Likewise.
24343         (pa_function_arg_size): New function.
24345 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24347         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
24348         in a separate statement.
24350 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
24352         PR tree-optimization/83847
24353         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
24354         group gathers and scatters.
24356 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24358         PR rtl-optimization/86620
24359         * params.def (max-sched-ready-insns): Bump minimum value to 1.
24361         PR rtl-optimization/83213
24362         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
24363         to last if both are JUMP_INSNs.
24365         PR tree-optimization/83843
24366         * gimple-ssa-store-merging.c
24367         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
24368         store_immediate_info for bswap/nop orig_stores.
24370 2018-01-15  Andrew Waterman  <andrew@sifive.com>
24372         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
24373         !TARGET_MUL.
24374         <UDIV>: Increase cost if !TARGET_DIV.
24376 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
24378         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
24379         (define_attr "cr_logical_3op"): New.
24380         (cceq_ior_compare): Adjust.
24381         (cceq_ior_compare_complement): Adjust.
24382         (*cceq_rev_compare): Adjust.
24383         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
24384         (is_cracked_insn): Adjust.
24385         (insn_must_be_first_in_group): Adjust.
24386         * config/rs6000/40x.md: Adjust.
24387         * config/rs6000/440.md: Adjust.
24388         * config/rs6000/476.md: Adjust.
24389         * config/rs6000/601.md: Adjust.
24390         * config/rs6000/603.md: Adjust.
24391         * config/rs6000/6xx.md: Adjust.
24392         * config/rs6000/7450.md: Adjust.
24393         * config/rs6000/7xx.md: Adjust.
24394         * config/rs6000/8540.md: Adjust.
24395         * config/rs6000/cell.md: Adjust.
24396         * config/rs6000/e300c2c3.md: Adjust.
24397         * config/rs6000/e500mc.md: Adjust.
24398         * config/rs6000/e500mc64.md: Adjust.
24399         * config/rs6000/e5500.md: Adjust.
24400         * config/rs6000/e6500.md: Adjust.
24401         * config/rs6000/mpc.md: Adjust.
24402         * config/rs6000/power4.md: Adjust.
24403         * config/rs6000/power5.md: Adjust.
24404         * config/rs6000/power6.md: Adjust.
24405         * config/rs6000/power7.md: Adjust.
24406         * config/rs6000/power8.md: Adjust.
24407         * config/rs6000/power9.md: Adjust.
24408         * config/rs6000/rs64.md: Adjust.
24409         * config/rs6000/titan.md: Adjust.
24411 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24413         * config/i386/predicates.md (indirect_branch_operand): Rewrite
24414         ix86_indirect_branch_register logic.
24416 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24418         * config/i386/constraints.md (Bs): Update
24419         ix86_indirect_branch_register check.  Don't check
24420         ix86_indirect_branch_register with GOT_memory_operand.
24421         (Bw): Likewise.
24422         * config/i386/predicates.md (GOT_memory_operand): Don't check
24423         ix86_indirect_branch_register here.
24424         (GOT32_symbol_operand): Likewise.
24426 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24428         * config/i386/predicates.md (constant_call_address_operand):
24429         Rewrite ix86_indirect_branch_register logic.
24430         (sibcall_insn_operand): Likewise.
24432 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24434         * config/i386/constraints.md (Bs): Replace
24435         ix86_indirect_branch_thunk_register with
24436         ix86_indirect_branch_register.
24437         (Bw): Likewise.
24438         * config/i386/i386.md (indirect_jump): Likewise.
24439         (tablejump): Likewise.
24440         (*sibcall_memory): Likewise.
24441         (*sibcall_value_memory): Likewise.
24442         Peepholes of indirect call and jump via memory: Likewise.
24443         * config/i386/i386.opt: Likewise.
24444         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24445         (GOT_memory_operand): Likewise.
24446         (call_insn_operand): Likewise.
24447         (sibcall_insn_operand): Likewise.
24448         (GOT32_symbol_operand): Likewise.
24450 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
24452         PR middle-end/83837
24453         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
24454         type rather than type addr's type points to.
24455         (expand_omp_atomic_mutex): Likewise.
24456         (expand_omp_atomic): Likewise.
24458 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
24460         PR target/83839
24461         * config/i386/i386.c (output_indirect_thunk_function): Use
24462         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
24463         for  __x86_return_thunk.
24465 2018-01-15  Richard Biener  <rguenther@suse.de>
24467         PR middle-end/83850
24468         * expmed.c (extract_bit_field_1): Fix typo.
24470 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24472         PR target/83687
24473         * config/arm/iterators.md (VF): New mode iterator.
24474         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
24475         Remove integer-related logic from pattern.
24476         (neon_vabd<mode>_3): Likewise.
24478 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
24480         PR middle-end/82694
24481         * common.opt (fstrict-overflow): No longer an alias.
24482         (fwrapv-pointer): New option.
24483         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
24484         also for pointer types based on flag_wrapv_pointer.
24485         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
24486         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
24487         opts->x_flag_wrapv got set.
24488         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
24489         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
24490         POINTER_TYPE_OVERFLOW_UNDEFINED.
24491         * match.pd: Likewise in address comparison pattern.
24492         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
24494 2018-01-15  Richard Biener  <rguenther@suse.de>
24496         PR lto/83804
24497         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
24498         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
24499         Reset type names to their identifier if their TYPE_DECL doesn't
24500         have linkage (and thus is used for ODR and devirt).
24501         (save_debug_info_for_decl): Remove.
24502         (save_debug_info_for_type): Likewise.
24503         (add_tree_to_fld_list): Adjust.
24504         * tree-pretty-print.c (dump_generic_node): Make dumping of
24505         type names more robust.
24507 2018-01-15  Richard Biener  <rguenther@suse.de>
24509         * BASE-VER: Bump to 8.0.1.
24511 2018-01-14  Martin Sebor  <msebor@redhat.com>
24513         PR other/83508
24514         * builtins.c (check_access): Avoid warning when the no-warning bit
24515         is set.
24517 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
24519         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
24520         * ira-color (allocno_hard_regs_compare): Likewise.
24522 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
24524         PR target/83013
24525         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
24526         Use .pushsection/.popsection.
24528 2018-01-14  Martin Sebor  <msebor@redhat.com>
24530         PR c++/81327
24531         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
24533 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
24535         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
24536         entry from extra_headers.
24537         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
24538         extra_headers, make the list bitwise identical to the i?86-*-* one.
24540 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24542         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
24543         -mcmodel=large with -mindirect-branch=thunk,
24544         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
24545         -mfunction-return=thunk-extern.
24546         * doc/invoke.texi: Document -mcmodel=large is incompatible with
24547         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
24548         -mfunction-return=thunk and -mfunction-return=thunk-extern.
24550 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24552         * config/i386/i386.c (print_reg): Print the name of the full
24553         integer register without '%'.
24554         (ix86_print_operand): Handle 'V'.
24555         * doc/extend.texi: Document 'V' modifier.
24557 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24559         * config/i386/constraints.md (Bs): Disallow memory operand for
24560         -mindirect-branch-register.
24561         (Bw): 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.
24567         * config/i386/i386.md (indirect_jump): Call convert_memory_address
24568         for -mindirect-branch-register.
24569         (tablejump): Likewise.
24570         (*sibcall_memory): Likewise.
24571         (*sibcall_value_memory): Likewise.
24572         Disallow peepholes of indirect call and jump via memory for
24573         -mindirect-branch-register.
24574         (*call_pop): Replace m with Bw.
24575         (*call_value_pop): Likewise.
24576         (*sibcall_pop_memory): Replace m with Bs.
24577         * config/i386/i386.opt (mindirect-branch-register): New option.
24578         * doc/invoke.texi: Document -mindirect-branch-register option.
24580 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24582         * config/i386/i386-protos.h (ix86_output_function_return): New.
24583         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
24584         set function_return_type.
24585         (indirect_thunk_name): Add ret_p to indicate thunk for function
24586         return.
24587         (output_indirect_thunk_function): Pass false to
24588         indirect_thunk_name.
24589         (ix86_output_indirect_branch_via_reg): Likewise.
24590         (ix86_output_indirect_branch_via_push): Likewise.
24591         (output_indirect_thunk_function): Create alias for function
24592         return thunk if regno < 0.
24593         (ix86_output_function_return): New function.
24594         (ix86_handle_fndecl_attribute): Handle function_return.
24595         (ix86_attribute_table): Add function_return.
24596         * config/i386/i386.h (machine_function): Add
24597         function_return_type.
24598         * config/i386/i386.md (simple_return_internal): Use
24599         ix86_output_function_return.
24600         (simple_return_internal_long): Likewise.
24601         * config/i386/i386.opt (mfunction-return=): New option.
24602         (indirect_branch): Mention -mfunction-return=.
24603         * doc/extend.texi: Document function_return function attribute.
24604         * doc/invoke.texi: Document -mfunction-return= option.
24606 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
24608         * config/i386/i386-opts.h (indirect_branch): New.
24609         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
24610         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
24611         with local indirect jump when converting indirect call and jump.
24612         (ix86_set_indirect_branch_type): New.
24613         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
24614         (indirectlabelno): New.
24615         (indirect_thunk_needed): Likewise.
24616         (indirect_thunk_bnd_needed): Likewise.
24617         (indirect_thunks_used): Likewise.
24618         (indirect_thunks_bnd_used): Likewise.
24619         (INDIRECT_LABEL): Likewise.
24620         (indirect_thunk_name): Likewise.
24621         (output_indirect_thunk): Likewise.
24622         (output_indirect_thunk_function): Likewise.
24623         (ix86_output_indirect_branch_via_reg): Likewise.
24624         (ix86_output_indirect_branch_via_push): Likewise.
24625         (ix86_output_indirect_branch): Likewise.
24626         (ix86_output_indirect_jmp): Likewise.
24627         (ix86_code_end): Call output_indirect_thunk_function if needed.
24628         (ix86_output_call_insn): Call ix86_output_indirect_branch if
24629         needed.
24630         (ix86_handle_fndecl_attribute): Handle indirect_branch.
24631         (ix86_attribute_table): Add indirect_branch.
24632         * config/i386/i386.h (machine_function): Add indirect_branch_type
24633         and has_local_indirect_jump.
24634         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
24635         to true.
24636         (tablejump): Likewise.
24637         (*indirect_jump): Use ix86_output_indirect_jmp.
24638         (*tablejump_1): Likewise.
24639         (simple_return_indirect_internal): Likewise.
24640         * config/i386/i386.opt (mindirect-branch=): New option.
24641         (indirect_branch): New.
24642         (keep): Likewise.
24643         (thunk): Likewise.
24644         (thunk-inline): Likewise.
24645         (thunk-extern): Likewise.
24646         * doc/extend.texi: Document indirect_branch function attribute.
24647         * doc/invoke.texi: Document -mindirect-branch= option.
24649 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
24651         PR ipa/83051
24652         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
24654 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
24656         * ipa-inline.c (want_inline_small_function_p): Return false if
24657         inlining has already failed with CIF_FINAL_ERROR.
24658         (update_caller_keys): Call want_inline_small_function_p before
24659         can_inline_edge_p.
24660         (update_callee_keys): Likewise.
24662 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24664         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
24665         New function.
24666         (rs6000_quadword_masked_address_p): Likewise.
24667         (quad_aligned_load_p): Likewise.
24668         (quad_aligned_store_p): Likewise.
24669         (const_load_sequence_p): Add comment to describe the outer-most loop.
24670         (mimic_memory_attributes_and_flags): New function.
24671         (rs6000_gen_stvx): Likewise.
24672         (replace_swapped_aligned_store): Likewise.
24673         (rs6000_gen_lvx): Likewise.
24674         (replace_swapped_aligned_load): Likewise.
24675         (replace_swapped_load_constant): Capitalize argument name in
24676         comment describing this function.
24677         (rs6000_analyze_swaps): Add a third pass to search for vector loads
24678         and stores that access quad-word aligned addresses and replace
24679         with stvx or lvx instructions when appropriate.
24680         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
24681         New function prototype.
24682         (rs6000_quadword_masked_address_p): Likewise.
24683         (rs6000_gen_lvx): Likewise.
24684         (rs6000_gen_stvx): Likewise.
24685         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
24686         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
24687         when memory address is aligned.
24688         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
24689         this split to select lvx instruction when memory address is aligned.
24690         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
24691         instruction when memory address is aligned.
24692         (*vsx_le_perm_load_v16qi): Likewise.
24693         (four unnamed splitters): Modify to select the stvx instruction
24694         when memory is aligned.
24696 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
24698         * predict.c (determine_unlikely_bbs): Handle correctly BBs
24699         which appears in the queue multiple times.
24701 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24702             Alan Hayward  <alan.hayward@arm.com>
24703             David Sherwood  <david.sherwood@arm.com>
24705         * tree-vectorizer.h (vec_lower_bound): New structure.
24706         (_loop_vec_info): Add check_nonzero and lower_bounds.
24707         (LOOP_VINFO_CHECK_NONZERO): New macro.
24708         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
24709         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
24710         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
24711         fields.  Make seg_len the distance travelled, not including the
24712         access size.
24713         (dr_direction_indicator): Declare.
24714         (dr_zero_step_indicator): Likewise.
24715         (dr_known_forward_stride_p): Likewise.
24716         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
24717         tree-ssanames.h.
24718         (runtime_alias_check_p): Allow runtime alias checks with
24719         variable strides.
24720         (operator ==): Compare access_size and align.
24721         (prune_runtime_alias_test_list): Rework for new distinction between
24722         the access_size and seg_len.
24723         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
24724         segment lengths.
24725         (get_segment_min_max): New function.
24726         (create_intersect_range_checks): Use it.
24727         (dr_step_indicator): New function.
24728         (dr_direction_indicator): Likewise.
24729         (dr_zero_step_indicator): Likewise.
24730         (dr_known_forward_stride_p): Likewise.
24731         * tree-loop-distribution.c (data_ref_segment_size): Return
24732         DR_STEP * (niters - 1).
24733         (compute_alias_check_pairs): Update call to the dr_with_seg_len
24734         constructor.
24735         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
24736         (vect_preserves_scalar_order_p): New function, split out from...
24737         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
24738         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
24739         (vect_vfa_access_size): New function.
24740         (vect_vfa_align): Likewise.
24741         (vect_compile_time_alias): Take access_size_a and access_b arguments.
24742         (dump_lower_bound): New function.
24743         (vect_check_lower_bound): Likewise.
24744         (vect_small_gap_p): Likewise.
24745         (vectorizable_with_step_bound_p): Likewise.
24746         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
24747         depencies if the vectorization factor is 1.  Convert the checks
24748         for nonzero steps into checks on the bounds of DR_STEP.  Try using
24749         a bunds check for variable steps if the minimum required step is
24750         relatively small. Update calls to the dr_with_seg_len
24751         constructor and to vect_compile_time_alias.
24752         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
24753         function.
24754         (vect_loop_versioning): Call it.
24755         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
24756         when retrying.
24757         (vect_estimate_min_profitable_iters): Account for any bounds checks.
24759 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24760             Alan Hayward  <alan.hayward@arm.com>
24761             David Sherwood  <david.sherwood@arm.com>
24763         * doc/sourcebuild.texi (vect_scatter_store): Document.
24764         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
24765         optabs.
24766         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
24767         Document.
24768         * genopinit.c (main): Add supports_vec_scatter_store and
24769         supports_vec_scatter_store_cached to target_optabs.
24770         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
24771         IFN_MASK_SCATTER_STORE.
24772         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
24773         functions.
24774         * internal-fn.h (internal_store_fn_p): Declare.
24775         (internal_fn_stored_value_index): Likewise.
24776         * internal-fn.c (scatter_store_direct): New macro.
24777         (expand_scatter_store_optab_fn): New function.
24778         (direct_scatter_store_optab_supported_p): New macro.
24779         (internal_store_fn_p): New function.
24780         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
24781         IFN_MASK_SCATTER_STORE.
24782         (internal_fn_mask_index): Likewise.
24783         (internal_fn_stored_value_index): New function.
24784         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
24785         for scatter stores.
24786         * optabs-query.h (supports_vec_scatter_store_p): Declare.
24787         * optabs-query.c (supports_vec_scatter_store_p): New function.
24788         * tree-vectorizer.h (vect_get_store_rhs): Declare.
24789         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
24790         true for scatter stores.
24791         (vect_gather_scatter_fn_p): Handle scatter stores too.
24792         (vect_check_gather_scatter): Consider using scatter stores if
24793         supports_vec_scatter_store_p.
24794         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
24795         scatter stores too.
24796         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24797         internal_fn_stored_value_index.
24798         (check_load_store_masking): Handle scatter stores too.
24799         (vect_get_store_rhs): Make public.
24800         (vectorizable_call): Use internal_store_fn_p.
24801         (vectorizable_store): Handle scatter store internal functions.
24802         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
24803         when deciding whether the end of the group has been reached.
24804         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
24805         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24806         (mask_scatter_store<mode>): New insns.
24808 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24809             Alan Hayward  <alan.hayward@arm.com>
24810             David Sherwood  <david.sherwood@arm.com>
24812         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24813         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24814         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24815         function.
24816         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24817         Use vect_truncate_gather_scatter_offset if we can't treat the
24818         operation as a normal gather load or scatter store.
24819         (get_group_load_store_type): Take the gather_scatter_info
24820         as argument.  Try using a gather load or scatter store for
24821         single-element groups.
24822         (get_load_store_type): Update calls to get_group_load_store_type
24823         and vect_use_strided_gather_scatters_p.
24825 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24826             Alan Hayward  <alan.hayward@arm.com>
24827             David Sherwood  <david.sherwood@arm.com>
24829         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24830         optional tree argument.
24831         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24832         null target hooks.
24833         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24834         but continue to use the current value as a fallback.
24835         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24836         to compare the updates.
24837         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24838         (get_load_store_type): Use it when handling a strided access.
24839         (vect_get_strided_load_store_ops): New function.
24840         (vect_get_data_ptr_increment): Likewise.
24841         (vectorizable_load): Handle strided gather loads.  Always pass
24842         a step to vect_create_data_ref_ptr and bump_vector_ptr.
24844 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24845             Alan Hayward  <alan.hayward@arm.com>
24846             David Sherwood  <david.sherwood@arm.com>
24848         * doc/md.texi (gather_load@var{m}): Document.
24849         (mask_gather_load@var{m}): Likewise.
24850         * genopinit.c (main): Add supports_vec_gather_load and
24851         supports_vec_gather_load_cached to target_optabs.
24852         * optabs-tree.c (init_tree_optimization_optabs): Use
24853         ggc_cleared_alloc to allocate target_optabs.
24854         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24855         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24856         functions.
24857         * internal-fn.h (internal_load_fn_p): Declare.
24858         (internal_gather_scatter_fn_p): Likewise.
24859         (internal_fn_mask_index): Likewise.
24860         (internal_gather_scatter_fn_supported_p): Likewise.
24861         * internal-fn.c (gather_load_direct): New macro.
24862         (expand_gather_load_optab_fn): New function.
24863         (direct_gather_load_optab_supported_p): New macro.
24864         (direct_internal_fn_optab): New function.
24865         (internal_load_fn_p): Likewise.
24866         (internal_gather_scatter_fn_p): Likewise.
24867         (internal_fn_mask_index): Likewise.
24868         (internal_gather_scatter_fn_supported_p): Likewise.
24869         * optabs-query.c (supports_at_least_one_mode_p): New function.
24870         (supports_vec_gather_load_p): Likewise.
24871         * optabs-query.h (supports_vec_gather_load_p): Declare.
24872         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24873         and memory_type field.
24874         (NUM_PATTERNS): Bump to 15.
24875         * tree-vect-data-refs.c: Include internal-fn.h.
24876         (vect_gather_scatter_fn_p): New function.
24877         (vect_describe_gather_scatter_call): Likewise.
24878         (vect_check_gather_scatter): Try using internal functions for
24879         gather loads.  Recognize existing calls to a gather load function.
24880         (vect_analyze_data_refs): Consider using gather loads if
24881         supports_vec_gather_load_p.
24882         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24883         (vect_get_gather_scatter_offset_type): Likewise.
24884         (vect_convert_mask_for_vectype): Likewise.
24885         (vect_add_conversion_to_patterm): Likewise.
24886         (vect_try_gather_scatter_pattern): Likewise.
24887         (vect_recog_gather_scatter_pattern): New pattern recognizer.
24888         (vect_vect_recog_func_ptrs): Add it.
24889         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24890         internal_fn_mask_index and internal_gather_scatter_fn_p.
24891         (check_load_store_masking): Take the gather_scatter_info as an
24892         argument and handle gather loads.
24893         (vect_get_gather_scatter_ops): New function.
24894         (vectorizable_call): Check internal_load_fn_p.
24895         (vectorizable_load): Likewise.  Handle gather load internal
24896         functions.
24897         (vectorizable_store): Update call to check_load_store_masking.
24898         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
24899         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
24900         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
24901         (aarch64_gather_scale_operand_d): New predicates.
24902         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
24903         (mask_gather_load<mode>): New insns.
24905 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24906             Alan Hayward  <alan.hayward@arm.com>
24907             David Sherwood  <david.sherwood@arm.com>
24909         * optabs.def (fold_left_plus_optab): New optab.
24910         * doc/md.texi (fold_left_plus_@var{m}): Document.
24911         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
24912         * internal-fn.c (fold_left_direct): Define.
24913         (expand_fold_left_optab_fn): Likewise.
24914         (direct_fold_left_optab_supported_p): Likewise.
24915         * fold-const-call.c (fold_const_fold_left): New function.
24916         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
24917         * tree-parloops.c (valid_reduction_p): New function.
24918         (gather_scalar_reductions): Use it.
24919         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
24920         (vect_finish_replace_stmt): Declare.
24921         * tree-vect-loop.c (fold_left_reduction_fn): New function.
24922         (needs_fold_left_reduction_p): New function, split out from...
24923         (vect_is_simple_reduction): ...here.  Accept reductions that
24924         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
24925         (vect_force_simple_reduction): Also store the reduction type in
24926         the assignment's STMT_VINFO_REDUC_TYPE.
24927         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24928         (merge_with_identity): New function.
24929         (vect_expand_fold_left): Likewise.
24930         (vectorize_fold_left_reduction): Likewise.
24931         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
24932         scalar phi in place for it.  Check for target support and reject
24933         cases that would reassociate the operation.  Defer the transform
24934         phase to vectorize_fold_left_reduction.
24935         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24936         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24937         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24939 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24941         * tree-if-conv.c (predicate_mem_writes): Remove redundant
24942         call to ifc_temp_var.
24944 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24945             Alan Hayward  <alan.hayward@arm.com>
24946             David Sherwood  <david.sherwood@arm.com>
24948         * target.def (legitimize_address_displacement): Take the original
24949         offset as a poly_int.
24950         * targhooks.h (default_legitimize_address_displacement): Update
24951         accordingly.
24952         * targhooks.c (default_legitimize_address_displacement): Likewise.
24953         * doc/tm.texi: Regenerate.
24954         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24955         as an argument, moving assert of ad->disp == ad->disp_term to...
24956         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
24957         Try calling targetm.legitimize_address_displacement before expanding
24958         the address rather than afterwards, and adjust for the new interface.
24959         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24960         Match the new hook interface.  Handle SVE addresses.
24961         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24962         new hook interface.
24964 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24966         * Makefile.in (OBJS): Add early-remat.o.
24967         * target.def (select_early_remat_modes): New hook.
24968         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24969         * doc/tm.texi: Regenerate.
24970         * targhooks.h (default_select_early_remat_modes): Declare.
24971         * targhooks.c (default_select_early_remat_modes): New function.
24972         * timevar.def (TV_EARLY_REMAT): New timevar.
24973         * passes.def (pass_early_remat): New pass.
24974         * tree-pass.h (make_pass_early_remat): Declare.
24975         * early-remat.c: New file.
24976         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24977         function.
24978         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24980 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24981             Alan Hayward  <alan.hayward@arm.com>
24982             David Sherwood  <david.sherwood@arm.com>
24984         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24985         vfm1 with a bound_epilog parameter.
24986         (vect_do_peeling): Update calls accordingly, and move the prologue
24987         call earlier in the function.  Treat the base bound_epilog as 0 for
24988         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
24989         this base when peeling for gaps.
24990         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24991         with fully-masked loops.
24992         (vect_estimate_min_profitable_iters): Handle the single peeled
24993         iteration in that case.
24995 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
24996             Alan Hayward  <alan.hayward@arm.com>
24997             David Sherwood  <david.sherwood@arm.com>
24999         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25000         single-element interleaving even if the size is not a power of 2.
25001         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25002         accesses for single-element interleaving if the group size is
25003         not a power of 2.
25005 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25006             Alan Hayward  <alan.hayward@arm.com>
25007             David Sherwood  <david.sherwood@arm.com>
25009         * doc/md.texi (fold_extract_last_@var{m}): Document.
25010         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25011         * optabs.def (fold_extract_last_optab): New optab.
25012         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25013         * internal-fn.c (fold_extract_direct): New macro.
25014         (expand_fold_extract_optab_fn): Likewise.
25015         (direct_fold_extract_optab_supported_p): Likewise.
25016         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25017         * tree-vect-loop.c (vect_model_reduction_cost): Handle
25018         EXTRACT_LAST_REDUCTION.
25019         (get_initial_def_for_reduction): Do not create an initial vector
25020         for EXTRACT_LAST_REDUCTION reductions.
25021         (vectorizable_reduction): Leave the scalar phi in place for
25022         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
25023         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
25024         epilogue code for EXTRACT_LAST_REDUCTION and defer the
25025         transform phase to vectorizable_condition.
25026         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25027         split out from...
25028         (vect_finish_stmt_generation): ...here.
25029         (vect_finish_replace_stmt): New function.
25030         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25031         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25032         pattern.
25033         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25035 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25036             Alan Hayward  <alan.hayward@arm.com>
25037             David Sherwood  <david.sherwood@arm.com>
25039         * doc/md.texi (extract_last_@var{m}): Document.
25040         * optabs.def (extract_last_optab): New optab.
25041         * internal-fn.def (EXTRACT_LAST): New internal function.
25042         * internal-fn.c (cond_unary_direct): New macro.
25043         (expand_cond_unary_optab_fn): Likewise.
25044         (direct_cond_unary_optab_supported_p): Likewise.
25045         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25046         loops using EXTRACT_LAST.
25047         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
25048         (extract_last_<mode>): ...this optab.
25049         (vec_extract<mode><Vel>): Update accordingly.
25051 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25052             Alan Hayward  <alan.hayward@arm.com>
25053             David Sherwood  <david.sherwood@arm.com>
25055         * target.def (empty_mask_is_expensive): New hook.
25056         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
25057         * doc/tm.texi: Regenerate.
25058         * targhooks.h (default_empty_mask_is_expensive): Declare.
25059         * targhooks.c (default_empty_mask_is_expensive): New function.
25060         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25061         if the target says that empty masks are expensive.
25062         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25063         New function.
25064         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25066 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25067             Alan Hayward  <alan.hayward@arm.com>
25068             David Sherwood  <david.sherwood@arm.com>
25070         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25071         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25072         (vect_use_loop_mask_for_alignment_p): New function.
25073         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25074         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25075         niters_skip argument.  Make sure that the first niters_skip elements
25076         of the first iteration are inactive.
25077         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25078         Update call to vect_set_loop_masks_directly.
25079         (get_misalign_in_elems): New function, split out from...
25080         (vect_gen_prolog_loop_niters): ...here.
25081         (vect_update_init_of_dr): Take a code argument that specifies whether
25082         the adjustment should be added or subtracted.
25083         (vect_update_init_of_drs): Likewise.
25084         (vect_prepare_for_masked_peels): New function.
25085         (vect_do_peeling): Skip prologue peeling if we're using a mask
25086         instead.  Update call to vect_update_inits_of_drs.
25087         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25088         mask_skip_niters.
25089         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25090         alignment.  Do not include the number of peeled iterations in
25091         the minimum threshold in that case.
25092         (vectorizable_induction): Adjust the start value down by
25093         LOOP_VINFO_MASK_SKIP_NITERS iterations.
25094         (vect_transform_loop): Call vect_prepare_for_masked_peels.
25095         Take the number of skipped iterations into account when calculating
25096         the loop bounds.
25097         * tree-vect-stmts.c (vect_gen_while_not): New function.
25099 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25100             Alan Hayward  <alan.hayward@arm.com>
25101             David Sherwood  <david.sherwood@arm.com>
25103         * doc/sourcebuild.texi (vect_fully_masked): Document.
25104         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
25105         default value to 0.
25106         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
25107         split out from...
25108         (vect_analyze_loop_2): ...here. Don't check the vectorization
25109         factor against the number of loop iterations if the loop is
25110         fully-masked.
25112 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25113             Alan Hayward  <alan.hayward@arm.com>
25114             David Sherwood  <david.sherwood@arm.com>
25116         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
25117         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
25118         (dump_groups): Update accordingly.
25119         (iv_use::mem_type): New member variable.
25120         (address_p): New function.
25121         (record_use): Add a mem_type argument and initialize the new
25122         mem_type field.
25123         (record_group_use): Add a mem_type argument.  Use address_p.
25124         Remove obsolete null checks of base_object.  Update call to record_use.
25125         (find_interesting_uses_op): Update call to record_group_use.
25126         (find_interesting_uses_cond): Likewise.
25127         (find_interesting_uses_address): Likewise.
25128         (get_mem_type_for_internal_fn): New function.
25129         (find_address_like_use): Likewise.
25130         (find_interesting_uses_stmt): Try find_address_like_use before
25131         calling find_interesting_uses_op.
25132         (addr_offset_valid_p): Use the iv mem_type field as the type
25133         of the addressed memory.
25134         (add_autoinc_candidates): Likewise.
25135         (get_address_cost): Likewise.
25136         (split_small_address_groups_p): Use address_p.
25137         (split_address_groups): Likewise.
25138         (add_iv_candidate_for_use): Likewise.
25139         (autoinc_possible_for_pair): Likewise.
25140         (rewrite_groups): Likewise.
25141         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
25142         (determine_group_iv_cost): Update after split of USE_ADDRESS.
25143         (get_alias_ptr_type_for_ptr_address): New function.
25144         (rewrite_use_address): Rewrite address uses in calls that were
25145         identified by find_address_like_use.
25147 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25148             Alan Hayward  <alan.hayward@arm.com>
25149             David Sherwood  <david.sherwood@arm.com>
25151         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
25152         TARGET_MEM_REFs.
25153         * gimple-expr.h (is_gimple_addressable: Likewise.
25154         * gimple-expr.c (is_gimple_address): Likewise.
25155         * internal-fn.c (expand_call_mem_ref): New function.
25156         (expand_mask_load_optab_fn): Use it.
25157         (expand_mask_store_optab_fn): Likewise.
25159 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25160             Alan Hayward  <alan.hayward@arm.com>
25161             David Sherwood  <david.sherwood@arm.com>
25163         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
25164         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
25165         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
25166         (cond_umax@var{mode}): Document.
25167         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
25168         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
25169         (cond_umin_optab, cond_umax_optab): New optabs.
25170         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
25171         (COND_IOR, COND_XOR): New internal functions.
25172         * internal-fn.h (get_conditional_internal_fn): Declare.
25173         * internal-fn.c (cond_binary_direct): New macro.
25174         (expand_cond_binary_optab_fn): Likewise.
25175         (direct_cond_binary_optab_supported_p): Likewise.
25176         (get_conditional_internal_fn): New function.
25177         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
25178         Cope with reduction statements that are vectorized as calls rather
25179         than assignments.
25180         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
25181         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
25182         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
25183         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
25184         (UNSPEC_COND_EOR): New unspecs.
25185         (optab): Add mappings for them.
25186         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
25187         (sve_int_op, sve_fp_op): New int attributes.
25189 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25190             Alan Hayward  <alan.hayward@arm.com>
25191             David Sherwood  <david.sherwood@arm.com>
25193         * optabs.def (while_ult_optab): New optab.
25194         * doc/md.texi (while_ult@var{m}@var{n}): Document.
25195         * internal-fn.def (WHILE_ULT): New internal function.
25196         * internal-fn.h (direct_internal_fn_supported_p): New override
25197         that takes two types as argument.
25198         * internal-fn.c (while_direct): New macro.
25199         (expand_while_optab_fn): New function.
25200         (convert_optab_supported_p): Likewise.
25201         (direct_while_optab_supported_p): New macro.
25202         * wide-int.h (wi::udiv_ceil): New function.
25203         * tree-vectorizer.h (rgroup_masks): New structure.
25204         (vec_loop_masks): New typedef.
25205         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
25206         and fully_masked_p.
25207         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
25208         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
25209         (vect_max_vf): New function.
25210         (slpeel_make_loop_iterate_ntimes): Delete.
25211         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
25212         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
25213         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
25214         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
25215         internal-fn.h, stor-layout.h and optabs-query.h.
25216         (vect_set_loop_mask): New function.
25217         (add_preheader_seq): Likewise.
25218         (add_header_seq): Likewise.
25219         (interleave_supported_p): Likewise.
25220         (vect_maybe_permute_loop_masks): Likewise.
25221         (vect_set_loop_masks_directly): Likewise.
25222         (vect_set_loop_condition_masked): Likewise.
25223         (vect_set_loop_condition_unmasked): New function, split out from
25224         slpeel_make_loop_iterate_ntimes.
25225         (slpeel_make_loop_iterate_ntimes): Rename to..
25226         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
25227         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
25228         (vect_do_peeling): Update call accordingly.
25229         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
25230         loops.
25231         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25232         mask_compare_type, can_fully_mask_p and fully_masked_p.
25233         (release_vec_loop_masks): New function.
25234         (_loop_vec_info): Use it to free the loop masks.
25235         (can_produce_all_loop_masks_p): New function.
25236         (vect_get_max_nscalars_per_iter): Likewise.
25237         (vect_verify_full_masking): Likewise.
25238         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
25239         retries, and free the mask rgroups before retrying.  Check loop-wide
25240         reasons for disallowing fully-masked loops.  Make the final decision
25241         about whether use a fully-masked loop or not.
25242         (vect_estimate_min_profitable_iters): Do not assume that peeling
25243         for the number of iterations will be needed for fully-masked loops.
25244         (vectorizable_reduction): Disable fully-masked loops.
25245         (vectorizable_live_operation): Likewise.
25246         (vect_halve_mask_nunits): New function.
25247         (vect_double_mask_nunits): Likewise.
25248         (vect_record_loop_mask): Likewise.
25249         (vect_get_loop_mask): Likewise.
25250         (vect_transform_loop): Handle the case in which the final loop
25251         iteration might handle a partial vector.  Call vect_set_loop_condition
25252         instead of slpeel_make_loop_iterate_ntimes.
25253         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
25254         (check_load_store_masking): New function.
25255         (prepare_load_store_mask): Likewise.
25256         (vectorizable_store): Handle fully-masked loops.
25257         (vectorizable_load): Likewise.
25258         (supportable_widening_operation): Use vect_halve_mask_nunits for
25259         booleans.
25260         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
25261         (vect_gen_while): New function.
25262         * config/aarch64/aarch64.md (umax<mode>3): New expander.
25263         (aarch64_uqdec<mode>): New insn.
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         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
25270         (reduc_xor_scal_optab): New optabs.
25271         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
25272         (reduc_xor_scal_@var{m}): Document.
25273         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
25274         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
25275         internal functions.
25276         * fold-const-call.c (fold_const_call): Handle them.
25277         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
25278         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
25279         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
25280         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
25281         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
25282         (UNSPEC_XORV): New unspecs.
25283         (optab): Add entries for them.
25284         (BITWISEV): New int iterator.
25285         (bit_reduc_op): New int attributes.
25287 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25288             Alan Hayward  <alan.hayward@arm.com>
25289             David Sherwood  <david.sherwood@arm.com>
25291         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
25292         * internal-fn.def (VEC_SHL_INSERT): New internal function.
25293         * optabs.def (vec_shl_insert_optab): New optab.
25294         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
25295         (duplicate_and_interleave): Likewise.
25296         * tree-vect-loop.c: Include internal-fn.h.
25297         (neutral_op_for_slp_reduction): New function, split out from
25298         get_initial_defs_for_reduction.
25299         (get_initial_def_for_reduction): Handle option 2 for variable-length
25300         vectors by loading the neutral value into a vector and then shifting
25301         the initial value into element 0.
25302         (get_initial_defs_for_reduction): Replace the code argument with
25303         the neutral value calculated by neutral_op_for_slp_reduction.
25304         Use gimple_build_vector for constant-length vectors.
25305         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
25306         but the first group_size elements have a neutral value.
25307         Use duplicate_and_interleave otherwise.
25308         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
25309         Update call to get_initial_defs_for_reduction.  Handle SLP
25310         reductions for variable-length vectors by creating one vector
25311         result for each scalar result, with the elements associated
25312         with other scalar results stubbed out with the neutral value.
25313         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
25314         Require IFN_VEC_SHL_INSERT for double reductions on
25315         variable-length vectors, or SLP reductions that have
25316         a neutral value.  Require can_duplicate_and_interleave_p
25317         support for variable-length unchained SLP reductions if there
25318         is no neutral value, such as for MIN/MAX reductions.  Also require
25319         the number of vector elements to be a multiple of the number of
25320         SLP statements when doing variable-length unchained SLP reductions.
25321         Update call to vect_create_epilog_for_reduction.
25322         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
25323         and remove initial values.
25324         (duplicate_and_interleave): Make public.
25325         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
25326         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
25328 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25329             Alan Hayward  <alan.hayward@arm.com>
25330             David Sherwood  <david.sherwood@arm.com>
25332         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
25333         (can_duplicate_and_interleave_p): New function.
25334         (vect_get_and_check_slp_defs): Take the vector of statements
25335         rather than just the current one.  Remove excess parentheses.
25336         Restriction rejectinon of vect_constant_def and vect_external_def
25337         for variable-length vectors to boolean types, or types for which
25338         can_duplicate_and_interleave_p is false.
25339         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
25340         (duplicate_and_interleave): New function.
25341         (vect_get_constant_vectors): Use gimple_build_vector for
25342         constant-length vectors and suitable variable-length constant
25343         vectors.  Use duplicate_and_interleave for other variable-length
25344         vectors.  Don't defer the update when inserting new statements.
25346 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25347             Alan Hayward  <alan.hayward@arm.com>
25348             David Sherwood  <david.sherwood@arm.com>
25350         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
25351         min_profitable_iters doesn't go negative.
25353 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25354             Alan Hayward  <alan.hayward@arm.com>
25355             David Sherwood  <david.sherwood@arm.com>
25357         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
25358         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
25359         * optabs.def (vec_mask_load_lanes_optab): New optab.
25360         (vec_mask_store_lanes_optab): Likewise.
25361         * internal-fn.def (MASK_LOAD_LANES): New internal function.
25362         (MASK_STORE_LANES): Likewise.
25363         * internal-fn.c (mask_load_lanes_direct): New macro.
25364         (mask_store_lanes_direct): Likewise.
25365         (expand_mask_load_optab_fn): Handle masked operations.
25366         (expand_mask_load_lanes_optab_fn): New macro.
25367         (expand_mask_store_optab_fn): Handle masked operations.
25368         (expand_mask_store_lanes_optab_fn): New macro.
25369         (direct_mask_load_lanes_optab_supported_p): Likewise.
25370         (direct_mask_store_lanes_optab_supported_p): Likewise.
25371         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
25372         parameter.
25373         (vect_load_lanes_supported): Likewise.
25374         * tree-vect-data-refs.c (strip_conversion): New function.
25375         (can_group_stmts_p): Likewise.
25376         (vect_analyze_data_ref_accesses): Use it instead of checking
25377         for a pair of assignments.
25378         (vect_store_lanes_supported): Take a masked_p parameter.
25379         (vect_load_lanes_supported): Likewise.
25380         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
25381         vect_store_lanes_supported and vect_load_lanes_supported.
25382         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
25383         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
25384         parameter.  Don't allow gaps for masked accesses.
25385         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
25386         and vect_load_lanes_supported.
25387         (get_load_store_type): Take a masked_p parameter and update
25388         call to get_group_load_store_type.
25389         (vectorizable_store): Update call to get_load_store_type.
25390         Handle IFN_MASK_STORE_LANES.
25391         (vectorizable_load): Update call to get_load_store_type.
25392         Handle IFN_MASK_LOAD_LANES.
25394 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25395             Alan Hayward  <alan.hayward@arm.com>
25396             David Sherwood  <david.sherwood@arm.com>
25398         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
25399         modes for SVE.
25400         * config/aarch64/aarch64-protos.h
25401         (aarch64_sve_struct_memory_operand_p): Declare.
25402         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
25403         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
25404         (VPRED, vpred): Handle SVE structure modes.
25405         * config/aarch64/constraints.md (Utx): New constraint.
25406         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
25407         (aarch64_sve_struct_nonimmediate_operand): New predicates.
25408         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
25409         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
25410         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
25411         structure modes.  Split into pieces after RA.
25412         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
25413         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
25414         New patterns.
25415         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
25416         SVE structure modes.
25417         (aarch64_classify_address): Likewise.
25418         (sizetochar): Move earlier in file.
25419         (aarch64_print_operand): Handle SVE register lists.
25420         (aarch64_array_mode): New function.
25421         (aarch64_sve_struct_memory_operand_p): Likewise.
25422         (TARGET_ARRAY_MODE): Redefine.
25424 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25425             Alan Hayward  <alan.hayward@arm.com>
25426             David Sherwood  <david.sherwood@arm.com>
25428         * target.def (array_mode): New target hook.
25429         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
25430         * doc/tm.texi: Regenerate.
25431         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
25432         * hooks.c (hook_optmode_mode_uhwi_none): New function.
25433         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
25434         targetm.array_mode.
25435         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
25436         type sizes.
25438 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25439             Alan Hayward  <alan.hayward@arm.com>
25440             David Sherwood  <david.sherwood@arm.com>
25442         * fold-const.c (fold_binary_loc): Check the argument types
25443         rather than the result type when testing for a vector operation.
25445 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25447         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
25448         * doc/tm.texi: Regenerate.
25450 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25451             Alan Hayward  <alan.hayward@arm.com>
25452             David Sherwood  <david.sherwood@arm.com>
25454         * doc/invoke.texi (-msve-vector-bits=): Document new option.
25455         (sve): Document new AArch64 extension.
25456         * doc/md.texi (w): Extend the description of the AArch64
25457         constraint to include SVE vectors.
25458         (Upl, Upa): Document new AArch64 predicate constraints.
25459         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
25460         enum.
25461         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
25462         (msve-vector-bits=): New option.
25463         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
25464         SVE when these are disabled.
25465         (sve): New extension.
25466         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
25467         modes.  Adjust their number of units based on aarch64_sve_vg.
25468         (MAX_BITSIZE_MODE_ANY_MODE): Define.
25469         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
25470         aarch64_addr_query_type.
25471         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
25472         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
25473         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
25474         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
25475         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
25476         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
25477         (aarch64_simd_imm_zero_p): Delete.
25478         (aarch64_check_zero_based_sve_index_immediate): Declare.
25479         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25480         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25481         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25482         (aarch64_sve_float_mul_immediate_p): Likewise.
25483         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25484         rather than an rtx.
25485         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
25486         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
25487         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
25488         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
25489         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
25490         (aarch64_regmode_natural_size): Likewise.
25491         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
25492         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
25493         left one place.
25494         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
25495         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
25496         for VG and the SVE predicate registers.
25497         (V_ALIASES): Add a "z"-prefixed alias.
25498         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
25499         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
25500         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
25501         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
25502         (REG_CLASS_NAMES): Add entries for them.
25503         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
25504         and the predicate registers.
25505         (aarch64_sve_vg): Declare.
25506         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
25507         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
25508         (REGMODE_NATURAL_SIZE): Define.
25509         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
25510         SVE macros.
25511         * config/aarch64/aarch64.c: Include cfgrtl.h.
25512         (simd_immediate_info): Add a constructor for series vectors,
25513         and an associated step field.
25514         (aarch64_sve_vg): New variable.
25515         (aarch64_dbx_register_number): Handle VG and the predicate registers.
25516         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
25517         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
25518         (VEC_ANY_DATA, VEC_STRUCT): New constants.
25519         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
25520         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
25521         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
25522         (aarch64_get_mask_mode): New functions.
25523         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
25524         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
25525         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
25526         predicate modes and predicate registers.  Explicitly restrict
25527         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
25528         to store a vector mode if it is recognized by
25529         aarch64_classify_vector_mode.
25530         (aarch64_regmode_natural_size): New function.
25531         (aarch64_hard_regno_caller_save_mode): Return the original mode
25532         for predicates.
25533         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
25534         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
25535         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
25536         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
25537         functions.
25538         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
25539         does not overlap dest if the function is frame-related.  Handle
25540         SVE constants.
25541         (aarch64_split_add_offset): New function.
25542         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
25543         them aarch64_add_offset.
25544         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
25545         and update call to aarch64_sub_sp.
25546         (aarch64_add_cfa_expression): New function.
25547         (aarch64_expand_prologue): Pass extra temporary registers to the
25548         functions above.  Handle the case in which we need to emit new
25549         DW_CFA_expressions for registers that were originally saved
25550         relative to the stack pointer, but now have to be expressed
25551         relative to the frame pointer.
25552         (aarch64_output_mi_thunk): Pass extra temporary registers to the
25553         functions above.
25554         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
25555         IP0 and IP1 values for SVE frames.
25556         (aarch64_expand_vec_series): New function.
25557         (aarch64_expand_sve_widened_duplicate): Likewise.
25558         (aarch64_expand_sve_const_vector): Likewise.
25559         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
25560         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
25561         into the register, rather than emitting a SET directly.
25562         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
25563         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
25564         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
25565         (offset_9bit_signed_scaled_p): New functions.
25566         (aarch64_replicate_bitmask_imm): New function.
25567         (aarch64_bitmask_imm): Use it.
25568         (aarch64_cannot_force_const_mem): Reject expressions involving
25569         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
25570         (aarch64_classify_index): Handle SVE indices, by requiring
25571         a plain register index with a scale that matches the element size.
25572         (aarch64_classify_address): Handle SVE addresses.  Assert that
25573         the mode of the address is VOIDmode or an integer mode.
25574         Update call to aarch64_classify_symbol.
25575         (aarch64_classify_symbolic_expression): Update call to
25576         aarch64_classify_symbol.
25577         (aarch64_const_vec_all_in_range_p): New function.
25578         (aarch64_print_vector_float_operand): Likewise.
25579         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
25580         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
25581         and the FP immediates 1.0 and 0.5.
25582         (aarch64_print_address_internal): Handle SVE addresses.
25583         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
25584         (aarch64_regno_regclass): Handle predicate registers.
25585         (aarch64_secondary_reload): Handle big-endian reloads of SVE
25586         data modes.
25587         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
25588         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
25589         (aarch64_convert_sve_vector_bits): New function.
25590         (aarch64_override_options): Use it to handle -msve-vector-bits=.
25591         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25592         rather than an rtx.
25593         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
25594         Handle SVE vector and predicate modes.  Accept VL-based constants
25595         that need only one temporary register, and VL offsets that require
25596         no temporary registers.
25597         (aarch64_conditional_register_usage): Mark the predicate registers
25598         as fixed if SVE isn't available.
25599         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
25600         Return true for SVE vector and predicate modes.
25601         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
25602         rather than an unsigned int.  Handle SVE modes.
25603         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
25604         SVE modes.
25605         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
25606         if SVE is enabled.
25607         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25608         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25609         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25610         (aarch64_sve_float_mul_immediate_p): New functions.
25611         (aarch64_sve_valid_immediate): New function.
25612         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
25613         Explicitly reject structure modes.  Check for INDEX constants.
25614         Handle PTRUE and PFALSE constants.
25615         (aarch64_check_zero_based_sve_index_immediate): New function.
25616         (aarch64_simd_imm_zero_p): Delete.
25617         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
25618         vector modes.  Accept constants in the range of CNT[BHWD].
25619         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
25620         ask for an Advanced SIMD mode.
25621         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
25622         (aarch64_simd_vector_alignment): Handle SVE predicates.
25623         (aarch64_vectorize_preferred_vector_alignment): New function.
25624         (aarch64_simd_vector_alignment_reachable): Use it instead of
25625         the vector size.
25626         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
25627         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
25628         functions.
25629         (MAX_VECT_LEN): Delete.
25630         (expand_vec_perm_d): Add a vec_flags field.
25631         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
25632         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
25633         (aarch64_evpc_ext): Don't apply a big-endian lane correction
25634         for SVE modes.
25635         (aarch64_evpc_rev): Rename to...
25636         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
25637         (aarch64_evpc_rev_global): New function.
25638         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
25639         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
25640         MAX_VECT_LEN.
25641         (aarch64_evpc_sve_tbl): New function.
25642         (aarch64_expand_vec_perm_const_1): Update after rename of
25643         aarch64_evpc_rev.  Handle SVE permutes too, trying
25644         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
25645         than aarch64_evpc_tbl.
25646         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
25647         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
25648         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
25649         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
25650         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
25651         (aarch64_expand_sve_vcond): New functions.
25652         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
25653         of aarch64_vector_mode_p.
25654         (aarch64_dwarf_poly_indeterminate_value): New function.
25655         (aarch64_compute_pressure_classes): Likewise.
25656         (aarch64_can_change_mode_class): Likewise.
25657         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
25658         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
25659         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
25660         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
25661         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
25662         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
25663         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
25664         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
25665         constraints.
25666         (Dn, Dl, Dr): Accept const as well as const_vector.
25667         (Dz): Likewise.  Compare against CONST0_RTX.
25668         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
25669         of "vector" where appropriate.
25670         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
25671         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
25672         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
25673         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
25674         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
25675         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
25676         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
25677         (v_int_equiv): Extend to SVE modes.
25678         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
25679         mode attributes.
25680         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
25681         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
25682         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
25683         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
25684         (SVE_COND_FP_CMP): New int iterators.
25685         (perm_hilo): Handle the new unpack unspecs.
25686         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
25687         attributes.
25688         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
25689         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
25690         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
25691         (aarch64_equality_operator, aarch64_constant_vector_operand)
25692         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
25693         (aarch64_sve_nonimmediate_operand): Likewise.
25694         (aarch64_sve_general_operand): Likewise.
25695         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
25696         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
25697         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
25698         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
25699         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
25700         (aarch64_sve_float_arith_immediate): Likewise.
25701         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
25702         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
25703         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
25704         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
25705         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
25706         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
25707         (aarch64_sve_float_arith_operand): Likewise.
25708         (aarch64_sve_float_arith_with_sub_operand): Likewise.
25709         (aarch64_sve_float_mul_operand): Likewise.
25710         (aarch64_sve_vec_perm_operand): Likewise.
25711         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
25712         (aarch64_mov_operand): Accept const_poly_int and const_vector.
25713         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
25714         as well as const_vector.
25715         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
25716         in file.  Use CONST0_RTX and CONSTM1_RTX.
25717         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
25718         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
25719         Use aarch64_simd_imm_zero.
25720         * config/aarch64/aarch64-sve.md: New file.
25721         * config/aarch64/aarch64.md: Include it.
25722         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
25723         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
25724         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
25725         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
25726         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
25727         (sve): New attribute.
25728         (enabled): Disable instructions with the sve attribute unless
25729         TARGET_SVE.
25730         (movqi, movhi): Pass CONST_POLY_INT operaneds through
25731         aarch64_expand_mov_immediate.
25732         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
25733         CNT[BHSD] immediates.
25734         (movti): Split CONST_POLY_INT moves into two halves.
25735         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
25736         Split additions that need a temporary here if the destination
25737         is the stack pointer.
25738         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
25739         (*add<mode>3_poly_1): New instruction.
25740         (set_clobber_cc): New expander.
25742 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25744         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
25745         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
25746         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
25747         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
25748         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
25749         Change innermode from fixed_mode_size to machine_mode.
25750         (simplify_subreg): Update call accordingly.  Handle a constant-sized
25751         subreg of a variable-length CONST_VECTOR.
25753 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25754             Alan Hayward  <alan.hayward@arm.com>
25755             David Sherwood  <david.sherwood@arm.com>
25757         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
25758         (add_offset_to_base): New function, split out from...
25759         (create_mem_ref): ...here.  When handling a scale other than 1,
25760         check first whether the address is valid without the offset.
25761         Add it into the base if so, leaving the index and scale as-is.
25763 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
25765         PR c++/83778
25766         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
25767         fold_for_warn before checking if arg2 is INTEGER_CST.
25769 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
25771         * config/rs6000/predicates.md (load_multiple_operation): Delete.
25772         (store_multiple_operation): Delete.
25773         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
25774         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
25775         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
25776         guarded by TARGET_STRING.
25777         (rs6000_output_load_multiple): Delete.
25778         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
25779         OPTION_MASK_STRING / TARGET_STRING handling.
25780         (print_operand) <'N', 'O'>: Add comment that these are unused now.
25781         (const rs6000_opt_masks) <"string">: Change mask to 0.
25782         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
25783         (MASK_STRING): Delete.
25784         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
25785         parts.  Simplify.
25786         (load_multiple): Delete.
25787         (*ldmsi8): Delete.
25788         (*ldmsi7): Delete.
25789         (*ldmsi6): Delete.
25790         (*ldmsi5): Delete.
25791         (*ldmsi4): Delete.
25792         (*ldmsi3): Delete.
25793         (store_multiple): Delete.
25794         (*stmsi8): Delete.
25795         (*stmsi7): Delete.
25796         (*stmsi6): Delete.
25797         (*stmsi5): Delete.
25798         (*stmsi4): Delete.
25799         (*stmsi3): Delete.
25800         (movmemsi_8reg): Delete.
25801         (corresponding unnamed define_insn): Delete.
25802         (movmemsi_6reg): Delete.
25803         (corresponding unnamed define_insn): Delete.
25804         (movmemsi_4reg): Delete.
25805         (corresponding unnamed define_insn): Delete.
25806         (movmemsi_2reg): Delete.
25807         (corresponding unnamed define_insn): Delete.
25808         (movmemsi_1reg): Delete.
25809         (corresponding unnamed define_insn): Delete.
25810         * config/rs6000/rs6000.opt (mno-string): New.
25811         (mstring): Replace by deprecation warning stub.
25812         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25814 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
25816         * regrename.c (regrename_do_replace): If replacing the same
25817         reg multiple times, try to reuse last created gen_raw_REG.
25819         PR debug/81155
25820         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25821         main to workaround a bug in GDB.
25823 2018-01-12  Tom de Vries  <tom@codesourcery.com>
25825         PR target/83737
25826         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25828 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
25830         PR rtl-optimization/80481
25831         * ira-color.c (get_cap_member): New function.
25832         (allocnos_conflict_by_live_ranges_p): Use it.
25833         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25834         (setup_slot_coalesced_allocno_live_ranges): Ditto.
25836 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
25838         PR target/83628
25839         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25840         (*saddl_se_1): Ditto.
25841         (*ssubsi_1): Ditto.
25842         (*ssubl_se_1): Ditto.
25844 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25846         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25847         rather than wi::to_widest for DR_INITs.
25848         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25849         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25850         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25851         INTEGER_CSTs.
25852         (vect_analyze_group_access_1): Note that here.
25854 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25856         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25857         polynomial type sizes.
25859 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
25861         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25862         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25863         (gimple_add_tmp_var): Likewise.
25865 2018-01-12  Martin Liska  <mliska@suse.cz>
25867         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25868         (gimple_alloc_sizes): Likewise.
25869         (dump_gimple_statistics): Use PRIu64 in printf format.
25870         * gimple.h: Change uint64_t to int.
25872 2018-01-12  Martin Liska  <mliska@suse.cz>
25874         * tree-core.h: Use uint64_t instead of int.
25875         * tree.c (tree_node_counts): Likewise.
25876         (tree_node_sizes): Likewise.
25877         (dump_tree_statistics): Use PRIu64 in printf format.
25879 2018-01-12  Martin Liska  <mliska@suse.cz>
25881         * Makefile.in: As qsort_chk is implemented in vec.c, add
25882         vec.o to linkage of gencfn-macros.
25883         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25884         passing the info to record_node_allocation_statistics.
25885         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25886         and pass the info.
25887         * ggc-common.c (struct ggc_usage): Add operator== and use
25888         it in operator< and compare function.
25889         * mem-stats.h (struct mem_usage): Likewise.
25890         * vec.c (struct vec_usage): Remove operator< and compare
25891         function. Can be simply inherited.
25893 2018-01-12  Martin Jambor  <mjambor@suse.cz>
25895         PR target/81616
25896         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
25897         * tree-ssa-math-opts.c: Include domwalk.h.
25898         (convert_mult_to_fma_1): New function.
25899         (fma_transformation_info): New type.
25900         (fma_deferring_state): Likewise.
25901         (cancel_fma_deferring): New function.
25902         (result_of_phi): Likewise.
25903         (last_fma_candidate_feeds_initial_phi): Likewise.
25904         (convert_mult_to_fma): Added deferring logic, split actual
25905         transformation to convert_mult_to_fma_1.
25906         (math_opts_dom_walker): New type.
25907         (math_opts_dom_walker::after_dom_children): New method, body moved
25908         here from pass_optimize_widening_mul::execute, added deferring logic
25909         bits.
25910         (pass_optimize_widening_mul::execute): Moved most of code to
25911         math_opts_dom_walker::after_dom_children.
25912         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
25913         * config/i386/i386.c (ix86_option_override_internal): Added
25914         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
25916 2018-01-12  Richard Biener  <rguenther@suse.de>
25918         PR debug/83157
25919         * dwarf2out.c (gen_variable_die): Do not reset old_die for
25920         inline instance vars.
25922 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
25924         PR target/81819
25925         * config/rx/rx.c (rx_is_restricted_memory_address):
25926         Handle SUBREG case.
25928 2018-01-12  Richard Biener  <rguenther@suse.de>
25930         PR tree-optimization/80846
25931         * target.def (split_reduction): New target hook.
25932         * targhooks.c (default_split_reduction): New function.
25933         * targhooks.h (default_split_reduction): Declare.
25934         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25935         target requests first reduce vectors by combining low and high
25936         parts.
25937         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25938         (get_vectype_for_scalar_type_and_size): Export.
25939         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25940         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25941         * doc/tm.texi: Regenerate.
25942         * config/i386/i386.c (ix86_split_reduction): Implement
25943         TARGET_VECTORIZE_SPLIT_REDUCTION.
25945 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25947         PR target/83368
25948         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25949         in PIC mode except for TARGET_VXWORKS_RTP.
25950         * config/sparc/sparc.c: Include cfgrtl.h.
25951         (TARGET_INIT_PIC_REG): Define.
25952         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25953         (sparc_pic_register_p): New predicate.
25954         (sparc_legitimate_address_p): Use it.
25955         (sparc_legitimize_pic_address): Likewise.
25956         (sparc_delegitimize_address): Likewise.
25957         (sparc_mode_dependent_address_p): Likewise.
25958         (gen_load_pcrel_sym): Remove 4th parameter.
25959         (load_got_register): Adjust call to above.  Remove obsolete stuff.
25960         (sparc_expand_prologue): Do not call load_got_register here.
25961         (sparc_flat_expand_prologue): Likewise.
25962         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25963         (sparc_use_pseudo_pic_reg): New function.
25964         (sparc_init_pic_reg): Likewise.
25965         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25966         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25968 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
25970         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25971         Add item for branch_cost.
25973 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25975         PR rtl-optimization/83565
25976         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25977         not extend the result to a larger mode for rotate operations.
25978         (num_sign_bit_copies1): Likewise.
25980 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25982         PR target/40411
25983         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25984         -symbolic.
25985         Use values-Xc.o for -pedantic.
25986         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25988 2018-01-12  Martin Liska  <mliska@suse.cz>
25990         PR ipa/83054
25991         * ipa-devirt.c (final_warning_record::grow_type_warnings):
25992         New function.
25993         (possible_polymorphic_call_targets): Use it.
25994         (ipa_devirt): Likewise.
25996 2018-01-12  Martin Liska  <mliska@suse.cz>
25998         * profile-count.h (enum profile_quality): Use 0 as invalid
25999         enum value of profile_quality.
26001 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26003         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26004         -mext-string options.
26006 2018-01-12  Richard Biener  <rguenther@suse.de>
26008         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26009         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26010         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26011         Likewise.
26012         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26014 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
26016         * configure.ac (--with-long-double-format): Add support for the
26017         configuration option to change the default long double format on
26018         PowerPC systems.
26019         * config.gcc (powerpc*-linux*-*): Likewise.
26020         * configure: Regenerate.
26021         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26022         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26023         used without modification.
26025 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26027         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26028         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26029         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26030         MISC_BUILTIN_SPEC_BARRIER.
26031         (rs6000_init_builtins): Likewise.
26032         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26033         enum value.
26034         (speculation_barrier): New define_insn.
26035         * doc/extend.texi: Document __builtin_speculation_barrier.
26037 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26039         PR target/83203
26040         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26041         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26042         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26043         iterators.
26044         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
26045         integral modes instead of "ss" and "sd".
26046         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
26047         vectors with 32-bit and 64-bit elements.
26048         (vecdupssescalarmodesuffix): New mode attribute.
26049         (vec_dup<mode>): Use it.
26051 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
26053         PR target/83330
26054         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
26055         frame if argument is passed on stack.
26057 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26059         PR target/82682
26060         * ree.c (combine_reaching_defs): Optimize also
26061         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26062         reg2=any_extend(exp); reg1=reg2;, formatting fix.
26064 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26066         PR middle-end/83189
26067         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26069 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26071         PR middle-end/83718
26072         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26073         after they are computed.
26075 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
26077         PR tree-optimization/83695
26078         * gimple-loop-linterchange.cc
26079         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26080         reset cached scev information after interchange.
26081         (pass_linterchange::execute): Remove call to scev_reset_htab.
26083 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26085         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26086         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26087         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26088         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26089         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26090         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26091         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26092         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26093         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26094         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26095         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26096         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26097         (V_lane_reg): Likewise.
26098         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26099         New define_expand.
26100         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26101         (vfmal_lane_low<mode>_intrinsic,
26102         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26103         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
26104         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
26105         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
26106         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
26107         vfmsl_lane_high<mode>_intrinsic): New define_insns.
26109 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26111         * config/arm/arm-cpus.in (fp16fml): New feature.
26112         (ALL_SIMD): Add fp16fml.
26113         (armv8.2-a): Add fp16fml as an option.
26114         (armv8.3-a): Likewise.
26115         (armv8.4-a): Add fp16fml as part of fp16.
26116         * config/arm/arm.h (TARGET_FP16FML): Define.
26117         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
26118         when appropriate.
26119         * config/arm/arm-modes.def (V2HF): Define.
26120         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
26121         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
26122         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
26123         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
26124         vfmsl_low, vfmsl_high): New set of builtins.
26125         * config/arm/iterators.md (PLUSMINUS): New code iterator.
26126         (vfml_op): New code attribute.
26127         (VFMLHALVES): New int iterator.
26128         (VFML, VFMLSEL): New mode attributes.
26129         (V_reg): Define mapping for V2HF.
26130         (V_hi, V_lo): New mode attributes.
26131         (VF_constraint): Likewise.
26132         (vfml_half, vfml_half_selector): New int attributes.
26133         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
26134         define_expand.
26135         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
26136         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
26137         New define_insn.
26138         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
26139         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
26140         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
26141         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
26142         documentation.
26143         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
26144         Document new effective target and option set.
26146 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26148         * config/arm/arm-cpus.in (armv8_4): New feature.
26149         (ARMv8_4a): New fgroup.
26150         (armv8.4-a): New arch.
26151         * config/arm/arm-tables.opt: Regenerate.
26152         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
26153         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
26154         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
26155         Add matching rules for -march=armv8.4-a and extensions.
26156         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
26158 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
26160         PR target/81821
26161         * config/rx/rx.md (BW): New mode attribute.
26162         (sync_lock_test_and_setsi): Add mode suffix to insn output.
26164 2018-01-11  Richard Biener  <rguenther@suse.de>
26166         PR tree-optimization/83435
26167         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
26168         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
26169         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
26171 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26172             Alan Hayward  <alan.hayward@arm.com>
26173             David Sherwood  <david.sherwood@arm.com>
26175         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
26176         field.
26177         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
26178         (aarch64_print_address_internal): Use it to check for a zero offset.
26180 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26181             Alan Hayward  <alan.hayward@arm.com>
26182             David Sherwood  <david.sherwood@arm.com>
26184         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
26185         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
26186         Return a poly_int64 rather than a HOST_WIDE_INT.
26187         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
26188         rather than a HOST_WIDE_INT.
26189         * config/aarch64/aarch64.h (aarch64_frame): Protect with
26190         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
26191         hard_fp_offset, frame_size, initial_adjust, callee_offset and
26192         final_offset from HOST_WIDE_INT to poly_int64.
26193         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
26194         to_constant when getting the number of units in an Advanced SIMD
26195         mode.
26196         (aarch64_builtin_vectorized_function): Check for a constant number
26197         of units.
26198         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
26199         GET_MODE_SIZE.
26200         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
26201         attribute instead of GET_MODE_NUNITS.
26202         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
26203         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
26204         GET_MODE_SIZE for fixed-size registers.
26205         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
26206         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
26207         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
26208         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
26209         (aarch64_print_operand, aarch64_print_address_internal)
26210         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
26211         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
26212         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
26213         Handle polynomial GET_MODE_SIZE.
26214         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
26215         wider than SImode without modification.
26216         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
26217         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
26218         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
26219         passing and returning SVE modes.
26220         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
26221         rather than GEN_INT.
26222         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
26223         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
26224         (aarch64_allocate_and_probe_stack_space): Likewise.
26225         (aarch64_layout_frame): Cope with polynomial offsets.
26226         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
26227         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
26228         polynomial offsets.
26229         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
26230         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
26231         poly_int64 rather than a HOST_WIDE_INT.
26232         (aarch64_get_separate_components, aarch64_process_components)
26233         (aarch64_expand_prologue, aarch64_expand_epilogue)
26234         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
26235         (aarch64_anchor_offset): New function, split out from...
26236         (aarch64_legitimize_address): ...here.
26237         (aarch64_builtin_vectorization_cost): Handle polynomial
26238         TYPE_VECTOR_SUBPARTS.
26239         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
26240         GET_MODE_NUNITS.
26241         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
26242         number of elements from the PARALLEL rather than the mode.
26243         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
26244         rather than GET_MODE_BITSIZE.
26245         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
26246         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
26247         (aarch64_expand_vec_perm_const_1): Handle polynomial
26248         d->perm.length () and d->perm elements.
26249         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
26250         Apply to_constant to d->perm elements.
26251         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
26252         polynomial CONST_VECTOR_NUNITS.
26253         (aarch64_move_pointer): Take amount as a poly_int64 rather
26254         than an int.
26255         (aarch64_progress_pointer): Avoid temporary variable.
26256         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
26257         the mode attribute instead of GET_MODE.
26259 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26260             Alan Hayward  <alan.hayward@arm.com>
26261             David Sherwood  <david.sherwood@arm.com>
26263         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
26264         x exists before using it.
26265         (aarch64_add_constant_internal): Rename to...
26266         (aarch64_add_offset_1): ...this.  Replace regnum with separate
26267         src and dest rtxes.  Handle the case in which they're different,
26268         including when the offset is zero.  Replace scratchreg with an rtx.
26269         Use 2 additions if there is no spare register into which we can
26270         move a 16-bit constant.
26271         (aarch64_add_constant): Delete.
26272         (aarch64_add_offset): Replace reg with separate src and dest
26273         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
26274         Use aarch64_add_offset_1.
26275         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
26276         an rtx rather than an int.  Take the delta as a poly_int64
26277         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
26278         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
26279         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
26280         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
26281         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
26282         and aarch64_add_sp.
26283         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
26284         aarch64_add_constant.
26286 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26288         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
26289         Use scalar_float_mode.
26291 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26293         * config/aarch64/aarch64-simd.md
26294         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
26295         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
26296         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
26297         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
26298         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
26299         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
26300         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
26301         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
26302         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
26303         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
26305 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26307         PR target/83514
26308         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
26309         targ_options->x_arm_arch_string is non NULL.
26311 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
26313         * config/aarch64/aarch64.h
26314         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
26316 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
26318         PR target/82096
26319         * expmed.c (emit_store_flag_force): Swap if const op0
26320         and change VOIDmode to mode of op0.
26322 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26324         PR rtl-optimization/83761
26325         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
26326         than bytes to mode_for_size.
26328 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26330         PR middle-end/83189
26331         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
26332         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
26333         profile.
26335 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26337         PR middle-end/83575
26338         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
26339         when in layout mode.
26340         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
26341         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
26342         partition fixup.
26344 2018-01-10  Michael Collison  <michael.collison@arm.com>
26346         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
26347         * config/aarch64/aarch64-option-extension.def: Add
26348         AARCH64_OPT_EXTENSION of 'fp16fml'.
26349         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26350         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
26351         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
26352         * config/aarch64/constraints.md (Ui7): New constraint.
26353         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
26354         (VFMLA_SEL_W): Ditto.
26355         (f16quad): Ditto.
26356         (f16mac1): Ditto.
26357         (VFMLA16_LOW): New int iterator.
26358         (VFMLA16_HIGH): Ditto.
26359         (UNSPEC_FMLAL): New unspec.
26360         (UNSPEC_FMLSL): Ditto.
26361         (UNSPEC_FMLAL2): Ditto.
26362         (UNSPEC_FMLSL2): Ditto.
26363         (f16mac): New code attribute.
26364         * config/aarch64/aarch64-simd-builtins.def
26365         (aarch64_fmlal_lowv2sf): Ditto.
26366         (aarch64_fmlsl_lowv2sf): Ditto.
26367         (aarch64_fmlalq_lowv4sf): Ditto.
26368         (aarch64_fmlslq_lowv4sf): Ditto.
26369         (aarch64_fmlal_highv2sf): Ditto.
26370         (aarch64_fmlsl_highv2sf): Ditto.
26371         (aarch64_fmlalq_highv4sf): Ditto.
26372         (aarch64_fmlslq_highv4sf): Ditto.
26373         (aarch64_fmlal_lane_lowv2sf): Ditto.
26374         (aarch64_fmlsl_lane_lowv2sf): Ditto.
26375         (aarch64_fmlal_laneq_lowv2sf): Ditto.
26376         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
26377         (aarch64_fmlalq_lane_lowv4sf): Ditto.
26378         (aarch64_fmlsl_lane_lowv4sf): Ditto.
26379         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
26380         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
26381         (aarch64_fmlal_lane_highv2sf): Ditto.
26382         (aarch64_fmlsl_lane_highv2sf): Ditto.
26383         (aarch64_fmlal_laneq_highv2sf): Ditto.
26384         (aarch64_fmlsl_laneq_highv2sf): Ditto.
26385         (aarch64_fmlalq_lane_highv4sf): Ditto.
26386         (aarch64_fmlsl_lane_highv4sf): Ditto.
26387         (aarch64_fmlalq_laneq_highv4sf): Ditto.
26388         (aarch64_fmlsl_laneq_highv4sf): Ditto.
26389         * config/aarch64/aarch64-simd.md:
26390         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
26391         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26392         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
26393         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26394         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
26395         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
26396         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
26397         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
26398         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
26399         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
26400         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
26401         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
26402         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
26403         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
26404         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
26405         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
26406         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
26407         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
26408         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
26409         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
26410         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
26411         (vfmlsl_low_u32): Ditto.
26412         (vfmlalq_low_u32): Ditto.
26413         (vfmlslq_low_u32): Ditto.
26414         (vfmlal_high_u32): Ditto.
26415         (vfmlsl_high_u32): Ditto.
26416         (vfmlalq_high_u32): Ditto.
26417         (vfmlslq_high_u32): Ditto.
26418         (vfmlal_lane_low_u32): Ditto.
26419         (vfmlsl_lane_low_u32): Ditto.
26420         (vfmlal_laneq_low_u32): Ditto.
26421         (vfmlsl_laneq_low_u32): Ditto.
26422         (vfmlalq_lane_low_u32): Ditto.
26423         (vfmlslq_lane_low_u32): Ditto.
26424         (vfmlalq_laneq_low_u32): Ditto.
26425         (vfmlslq_laneq_low_u32): Ditto.
26426         (vfmlal_lane_high_u32): Ditto.
26427         (vfmlsl_lane_high_u32): Ditto.
26428         (vfmlal_laneq_high_u32): Ditto.
26429         (vfmlsl_laneq_high_u32): Ditto.
26430         (vfmlalq_lane_high_u32): Ditto.
26431         (vfmlslq_lane_high_u32): Ditto.
26432         (vfmlalq_laneq_high_u32): Ditto.
26433         (vfmlslq_laneq_high_u32): Ditto.
26434         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
26435         (AARCH64_FL_FOR_ARCH8_4): New.
26436         (AARCH64_ISA_F16FML): New ISA flag.
26437         (TARGET_F16FML): New feature flag for fp16fml.
26438         (doc/invoke.texi): Document new fp16fml option.
26440 2018-01-10  Michael Collison  <michael.collison@arm.com>
26442         * config/aarch64/aarch64-builtins.c:
26443         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
26444         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26445         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
26446         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
26447         (AARCH64_ISA_SHA3): New ISA flag.
26448         (TARGET_SHA3): New feature flag for sha3.
26449         * config/aarch64/iterators.md (sha512_op): New int attribute.
26450         (CRYPTO_SHA512): New int iterator.
26451         (UNSPEC_SHA512H): New unspec.
26452         (UNSPEC_SHA512H2): Ditto.
26453         (UNSPEC_SHA512SU0): Ditto.
26454         (UNSPEC_SHA512SU1): Ditto.
26455         * config/aarch64/aarch64-simd-builtins.def
26456         (aarch64_crypto_sha512hqv2di): New builtin.
26457         (aarch64_crypto_sha512h2qv2di): Ditto.
26458         (aarch64_crypto_sha512su0qv2di): Ditto.
26459         (aarch64_crypto_sha512su1qv2di): Ditto.
26460         (aarch64_eor3qv8hi): Ditto.
26461         (aarch64_rax1qv2di): Ditto.
26462         (aarch64_xarqv2di): Ditto.
26463         (aarch64_bcaxqv8hi): Ditto.
26464         * config/aarch64/aarch64-simd.md:
26465         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
26466         (aarch64_crypto_sha512su0qv2di): Ditto.
26467         (aarch64_crypto_sha512su1qv2di): Ditto.
26468         (aarch64_eor3qv8hi): Ditto.
26469         (aarch64_rax1qv2di): Ditto.
26470         (aarch64_xarqv2di): Ditto.
26471         (aarch64_bcaxqv8hi): Ditto.
26472         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
26473         (vsha512h2q_u64): Ditto.
26474         (vsha512su0q_u64): Ditto.
26475         (vsha512su1q_u64): Ditto.
26476         (veor3q_u16): Ditto.
26477         (vrax1q_u64): Ditto.
26478         (vxarq_u64): Ditto.
26479         (vbcaxq_u16): Ditto.
26480         * config/arm/types.md (crypto_sha512): New type attribute.
26481         (crypto_sha3): Ditto.
26482         (doc/invoke.texi): Document new sha3 option.
26484 2018-01-10  Michael Collison  <michael.collison@arm.com>
26486         * config/aarch64/aarch64-builtins.c:
26487         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
26488         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26489         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
26490         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
26491         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
26492         (AARCH64_ISA_SM4): New ISA flag.
26493         (TARGET_SM4): New feature flag for sm4.
26494         * config/aarch64/aarch64-simd-builtins.def
26495         (aarch64_sm3ss1qv4si): Ditto.
26496         (aarch64_sm3tt1aq4si): Ditto.
26497         (aarch64_sm3tt1bq4si): Ditto.
26498         (aarch64_sm3tt2aq4si): Ditto.
26499         (aarch64_sm3tt2bq4si): Ditto.
26500         (aarch64_sm3partw1qv4si): Ditto.
26501         (aarch64_sm3partw2qv4si): Ditto.
26502         (aarch64_sm4eqv4si): Ditto.
26503         (aarch64_sm4ekeyqv4si): Ditto.
26504         * config/aarch64/aarch64-simd.md:
26505         (aarch64_sm3ss1qv4si): Ditto.
26506         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
26507         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
26508         (aarch64_sm4eqv4si): Ditto.
26509         (aarch64_sm4ekeyqv4si): Ditto.
26510         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
26511         (sm3part_op): Ditto.
26512         (CRYPTO_SM3TT): Ditto.
26513         (CRYPTO_SM3PART): Ditto.
26514         (UNSPEC_SM3SS1): New unspec.
26515         (UNSPEC_SM3TT1A): Ditto.
26516         (UNSPEC_SM3TT1B): Ditto.
26517         (UNSPEC_SM3TT2A): Ditto.
26518         (UNSPEC_SM3TT2B): Ditto.
26519         (UNSPEC_SM3PARTW1): Ditto.
26520         (UNSPEC_SM3PARTW2): Ditto.
26521         (UNSPEC_SM4E): Ditto.
26522         (UNSPEC_SM4EKEY): Ditto.
26523         * config/aarch64/constraints.md (Ui2): New constraint.
26524         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
26525         * config/arm/types.md (crypto_sm3): New type attribute.
26526         (crypto_sm4): Ditto.
26527         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
26528         (vsm3tt1aq_u32): Ditto.
26529         (vsm3tt1bq_u32): Ditto.
26530         (vsm3tt2aq_u32): Ditto.
26531         (vsm3tt2bq_u32): Ditto.
26532         (vsm3partw1q_u32): Ditto.
26533         (vsm3partw2q_u32): Ditto.
26534         (vsm4eq_u32): Ditto.
26535         (vsm4ekeyq_u32): Ditto.
26536         (doc/invoke.texi): Document new sm4 option.
26538 2018-01-10  Michael Collison  <michael.collison@arm.com>
26540         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
26541         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
26542         (AARCH64_FL_FOR_ARCH8_4): New.
26543         (AARCH64_FL_V8_4): New flag.
26544         (doc/invoke.texi): Document new armv8.4-a option.
26546 2018-01-10  Michael Collison  <michael.collison@arm.com>
26548         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26549         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
26550         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
26551         * config/aarch64/aarch64-option-extension.def: Add
26552         AARCH64_OPT_EXTENSION of 'sha2'.
26553         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
26554         (crypto): Disable sha2 and aes if crypto disabled.
26555         (crypto): Enable aes and sha2 if enabled.
26556         (simd): Disable sha2 and aes if simd disabled.
26557         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
26558         New flags.
26559         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
26560         (TARGET_SHA2): New feature flag for sha2.
26561         (TARGET_AES): New feature flag for aes.
26562         * config/aarch64/aarch64-simd.md:
26563         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
26564         conditional on TARGET_AES.
26565         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
26566         (aarch64_crypto_sha1hsi): Make pattern conditional
26567         on TARGET_SHA2.
26568         (aarch64_crypto_sha1hv4si): Ditto.
26569         (aarch64_be_crypto_sha1hv4si): Ditto.
26570         (aarch64_crypto_sha1su1v4si): Ditto.
26571         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
26572         (aarch64_crypto_sha1su0v4si): Ditto.
26573         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
26574         (aarch64_crypto_sha256su0v4si): Ditto.
26575         (aarch64_crypto_sha256su1v4si): Ditto.
26576         (doc/invoke.texi): Document new aes and sha2 options.
26578 2018-01-10  Martin Sebor  <msebor@redhat.com>
26580         PR tree-optimization/83781
26581         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
26582         as string arrays.
26584 2018-01-11  Martin Sebor  <msebor@gmail.com>
26585             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26587         PR tree-optimization/83501
26588         PR tree-optimization/81703
26590         * tree-ssa-strlen.c (get_string_cst): Rename...
26591         (get_string_len): ...to this.  Handle global constants.
26592         (handle_char_store): Adjust.
26594 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
26595             Jim Wilson  <jimw@sifive.com>
26597         * config/riscv/riscv-protos.h (riscv_output_return): New.
26598         * config/riscv/riscv.c (struct machine_function): New naked_p field.
26599         (riscv_attribute_table, riscv_output_return),
26600         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
26601         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
26602         (riscv_compute_frame_info): Only compute frame->mask if not a naked
26603         function.
26604         (riscv_expand_prologue): Add early return for naked function.
26605         (riscv_expand_epilogue): Likewise.
26606         (riscv_function_ok_for_sibcall): Return false for naked function.
26607         (riscv_set_current_function): New.
26608         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
26609         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
26610         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
26611         * doc/extend.texi (RISC-V Function Attributes): New.
26613 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
26615         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
26616         check for 128-bit long double before checking TCmode.
26617         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
26618         128-bit long doubles before checking TFmode or TCmode.
26619         (FLOAT128_IBM_P): Likewise.
26621 2018-01-10  Martin Sebor  <msebor@redhat.com>
26623         PR tree-optimization/83671
26624         * builtins.c (c_strlen): Unconditionally return zero for the empty
26625         string.
26626         Use -Warray-bounds for warnings.
26627         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
26628         for non-constant array indices with COMPONENT_REF, arrays of
26629         arrays, and pointers to arrays.
26630         (gimple_fold_builtin_strlen): Determine and set length range for
26631         non-constant character arrays.
26633 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
26635         PR middle-end/81897
26636         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
26637         empty blocks.
26639 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
26641         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
26643 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
26645         PR target/83399
26646         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
26647         VECTOR_MEM_ALTIVEC_OR_VSX_P.
26648         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
26649         indexed_or_indirect_operand predicate.
26650         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
26651         (*vsx_le_perm_load_v8hi): Likewise.
26652         (*vsx_le_perm_load_v16qi): Likewise.
26653         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
26654         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
26655         (*vsx_le_perm_store_v8hi): Likewise.
26656         (*vsx_le_perm_store_v16qi): Likewise.
26657         (eight unnamed splitters): Likewise.
26659 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
26661         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
26662         * config/rs6000/emmintrin.h: Likewise.
26663         * config/rs6000/mmintrin.h: Likewise.
26664         * config/rs6000/xmmintrin.h: Likewise.
26666 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
26668         PR c++/43486
26669         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
26670         "public_flag".
26671         * tree.c (tree_nop_conversion): Return true for location wrapper
26672         nodes.
26673         (maybe_wrap_with_location): New function.
26674         (selftest::check_strip_nops): New function.
26675         (selftest::test_location_wrappers): New function.
26676         (selftest::tree_c_tests): Call it.
26677         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
26678         (maybe_wrap_with_location): New decl.
26679         (EXPR_LOCATION_WRAPPER_P): New macro.
26680         (location_wrapper_p): New inline function.
26681         (tree_strip_any_location_wrapper): New inline function.
26683 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
26685         PR target/83735
26686         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
26687         stack_realign_offset for the largest alignment of stack slot
26688         actually used.
26689         (ix86_find_max_used_stack_alignment): New function.
26690         (ix86_finalize_stack_frame_flags): Use it.  Set
26691         max_used_stack_alignment if we don't realign stack.
26692         * config/i386/i386.h (machine_function): Add
26693         max_used_stack_alignment.
26695 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
26697         * config/arm/arm.opt (-mbranch-cost): New option.
26698         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
26699         account.
26701 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
26703         PR target/83629
26704         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
26705         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
26707 2018-01-10  Richard Biener  <rguenther@suse.de>
26709         PR debug/83765
26710         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
26711         early out so it also covers the case where we have a non-NULL
26712         origin.
26714 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
26716         PR tree-optimization/83753
26717         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
26718         for non-strided grouped accesses if the number of elements is 1.
26720 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
26722         PR target/81616
26723         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
26724         * i386.h (TARGET_USE_GATHER): Define.
26725         * x86-tune.def (X86_TUNE_USE_GATHER): New.
26727 2018-01-10  Martin Liska  <mliska@suse.cz>
26729         PR bootstrap/82831
26730         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
26731         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
26732         partitioning.
26733         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
26734         CLEANUP_NO_PARTITIONING is not set.
26736 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
26738         * doc/rtl.texi: Remove documentation of (const ...) wrappers
26739         for vectors, as a partial revert of r254296.
26740         * rtl.h (const_vec_p): Delete.
26741         (const_vec_duplicate_p): Don't test for vector CONSTs.
26742         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
26743         * expmed.c (make_tree): Likewise.
26745         Revert:
26746         * common.md (E, F): Use CONSTANT_P instead of checking for
26747         CONST_VECTOR.
26748         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
26749         checking for CONST_VECTOR.
26751 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26753         PR middle-end/83575
26754         * predict.c (force_edge_cold): Handle in more sane way edges
26755         with no prediction.
26757 2018-01-09  Carl Love  <cel@us.ibm.com>
26759         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
26760         V4SI, V4SF types.
26761         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
26762         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
26763         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
26764         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
26765         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
26766         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
26767         * config/rs6000/rs6000-protos.h: Add extern defition for
26768         rs6000_generate_float2_double_code.
26769         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
26770         function.
26771         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
26772         (float2_v2df): Add define_expand.
26774 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
26776         PR target/83628
26777         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
26778         op_mode in the force_to_mode call.
26780 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26782         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
26783         instead of checking each element individually.
26784         (aarch64_evpc_uzp): Likewise.
26785         (aarch64_evpc_zip): Likewise.
26786         (aarch64_evpc_ext): Likewise.
26787         (aarch64_evpc_rev): Likewise.
26788         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
26789         instead of checking each element individually.  Return true without
26790         generating rtl if
26791         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
26792         whether all selected elements come from the same input, instead of
26793         checking each element individually.  Remove calls to gen_rtx_REG,
26794         start_sequence and end_sequence and instead assert that no rtl is
26795         generated.
26797 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26799         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
26800         order of HIGH and CONST checks.
26802 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
26804         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
26805         if the destination isn't an SSA_NAME.
26807 2018-01-09  Richard Biener  <rguenther@suse.de>
26809         PR tree-optimization/83668
26810         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26811         move prologue...
26812         (canonicalize_loop_form): ... here, renamed from ...
26813         (canonicalize_loop_closed_ssa_form): ... this and amended to
26814         swap successor edges for loop exit blocks to make us use
26815         the RPO order we need for initial schedule generation.
26817 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
26819         PR tree-optimization/64811
26820         * match.pd: When optimizing comparisons with Inf, avoid
26821         introducing or losing exceptions from comparisons with NaN.
26823 2018-01-09  Martin Liska  <mliska@suse.cz>
26825         PR sanitizer/82517
26826         * asan.c (shadow_mem_size): Add gcc_assert.
26828 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
26830         Don't save registers in main().
26832         PR target/83738
26833         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26834         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26835         * config/avr/avr.c (avr_set_current_function): Don't error if
26836         naked, OS_task or OS_main are specified at the same time.
26837         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26838         OS_main.
26839         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26840         attribute.
26841         * common/config/avr/avr-common.c (avr_option_optimization_table):
26842         Switch on -mmain-is-OS_task for optimizing compilations.
26844 2018-01-09  Richard Biener  <rguenther@suse.de>
26846         PR tree-optimization/83572
26847         * graphite.c: Include cfganal.h.
26848         (graphite_transform_loops): Connect infinite loops to exit
26849         and remove fake edges at the end.
26851 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26853         * ipa-inline.c (edge_badness): Revert accidental checkin.
26855 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
26857         PR ipa/80763
26858         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26859         symbols; not inline clones.
26861 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
26863         PR target/83507
26864         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26865         hard registers.  Formatting fixes.
26867         PR preprocessor/83722
26868         * gcc.c (try_generate_repro): Pass
26869         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26870         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26871         do_report_bug.
26873 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
26874             Kito Cheng  <kito.cheng@gmail.com>
26876         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26877         (riscv_leaf_function_p): Delete.
26878         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26880 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26882         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26883         function.
26884         (do_ifelse): New function.
26885         (do_isel): New function.
26886         (do_sub3): New function.
26887         (do_add3): New function.
26888         (do_load_mask_compare): New function.
26889         (do_overlap_load_compare): New function.
26890         (expand_compare_loop): New function.
26891         (expand_block_compare): Call expand_compare_loop() when appropriate.
26892         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
26893         option description.
26894         (-mblock-compare-inline-loop-limit): New option.
26896 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26898         PR target/83677
26899         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
26900         Reverse order of second and third operands in first alternative.
26901         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
26902         of first and second elements in UNSPEC_VPERMR vector.
26903         (altivec_expand_vec_perm_le): Likewise.
26905 2018-01-08  Jeff Law  <law@redhat.com>
26907         PR rtl-optimizatin/81308
26908         * tree-switch-conversion.c (cfg_altered): New file scoped static.
26909         (process_switch): If group_case_labels makes a change, then set
26910         cfg_altered.
26911         (pass_convert_switch::execute): If a switch is converted, then
26912         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
26914         PR rtl-optimization/81308
26915         * recog.c (split_all_insns): Conditionally cleanup the CFG after
26916         splitting insns.
26918 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
26920         PR target/83663 - Revert r255946
26921         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
26922         generation for cases where splatting a value is not useful.
26923         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
26924         across a vec_duplicate and a paradoxical subreg forming a vector
26925         mode to a vec_concat.
26927 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26929         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26930         -march=armv8.3-a variants.
26931         * config/arm/t-multilib: Likewise.
26932         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
26934 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26936         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26937         to generate rtl.
26938         (cceq_ior_compare_complement): Give it a name so I can use it, and
26939         change boolean_or_operator predicate to boolean_operator so it can
26940         be used to generate a crand.
26941         (eqne): New code iterator.
26942         (bd/bd_neg): New code_attrs.
26943         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26944         a single define_insn.
26945         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26946         decrement (bdnzt/bdnzf/bdzt/bdzf).
26947         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26948         with the new names of the branch decrement patterns, and added the
26949         names of the branch decrement conditional patterns.
26951 2018-01-08  Richard Biener  <rguenther@suse.de>
26953         PR tree-optimization/83563
26954         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26955         cache.
26957 2018-01-08  Richard Biener  <rguenther@suse.de>
26959         PR middle-end/83713
26960         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26962 2018-01-08  Richard Biener  <rguenther@suse.de>
26964         PR tree-optimization/83685
26965         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26966         references to abnormals.
26968 2018-01-08  Richard Biener  <rguenther@suse.de>
26970         PR lto/83719
26971         * dwarf2out.c (output_indirect_strings): Handle empty
26972         skeleton_debug_str_hash.
26973         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26975 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26977         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26978         (emit_store_direct): Likewise.
26979         (arc_trampoline_adjust_address): Likewise.
26980         (arc_asm_trampoline_template): New function.
26981         (arc_initialize_trampoline): Use asm_trampoline_template.
26982         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26983         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26984         * config/arc/arc.md (flush_icache): Delete pattern.
26986 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
26988         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26989         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26990         munaligned-access.
26992 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
26994         PR target/83681
26995         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26996         by not USED_FOR_TARGET.
26997         (make_pass_resolve_sw_modes): Likewise.
26999 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27001         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27002         USED_FOR_TARGET.
27004 2018-01-08  Richard Biener  <rguenther@suse.de>
27006         PR middle-end/83580
27007         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27009 2018-01-08  Richard Biener  <rguenther@suse.de>
27011         PR middle-end/83517
27012         * match.pd ((t * 2) / 2) -> t): Add missing :c.
27014 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
27016         PR middle-end/81897
27017         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27018         basic blocks with a small number of successors.
27019         (convert_control_dep_chain_into_preds): Improve handling of
27020         forwarder blocks.
27021         (dump_predicates): Split apart into...
27022         (dump_pred_chain): ...here...
27023         (dump_pred_info): ...and here.
27024         (can_one_predicate_be_invalidated_p): Add debugging printfs.
27025         (can_chain_union_be_invalidated_p): Improve check for invalidation
27026         of paths.
27027         (uninit_uses_cannot_happen): Avoid unnecessary if
27028         convert_control_dep_chain_into_preds yielded nothing.
27030 2018-01-06  Martin Sebor  <msebor@redhat.com>
27032         PR tree-optimization/83640
27033         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27034         subtracting negative offset from size.
27035         (builtin_access::overlap): Adjust offset bounds of the access to fall
27036         within the size of the object if possible.
27038 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
27040         PR rtl-optimization/83699
27041         * expmed.c (extract_bit_field_1): Restrict the vector usage of
27042         extract_bit_field_as_subreg to cases in which the extracted
27043         value is also a vector.
27045         * lra-constraints.c (process_alt_operands): Test for the equivalence
27046         substitutions when detecting a possible reload cycle.
27048 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
27050         PR debug/83480
27051         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
27052         by default if flag_selective_schedling{,2}.  Formatting fixes.
27054         PR rtl-optimization/83682
27055         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
27056         if it has non-VECTOR_MODE element mode.
27057         (vec_duplicate_p): Likewise.
27059         PR middle-end/83694
27060         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27061         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27063 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27065         PR target/83604
27066         * config/i386/i386-builtin.def
27067         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27068         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27069         Require also OPTION_MASK_ISA_AVX512F in addition to
27070         OPTION_MASK_ISA_GFNI.
27071         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27072         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27073         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27074         to OPTION_MASK_ISA_GFNI.
27075         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27076         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27077         OPTION_MASK_ISA_AVX512BW.
27078         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27079         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27080         addition to OPTION_MASK_ISA_GFNI.
27081         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27082         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27083         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27084         to OPTION_MASK_ISA_GFNI.
27085         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27086         a requirement for all ISAs rather than any of them with a few
27087         exceptions.
27088         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27089         processing.
27090         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27091         bitmasks to be enabled with 3 exceptions, instead of requiring any
27092         enabled ISA with lots of exceptions.
27093         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27094         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27095         Change avx512bw in isa attribute to avx512f.
27096         * config/i386/sgxintrin.h: Add license boilerplate.
27097         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
27098         to __AVX512F__ and __AVX512VL to __AVX512VL__.
27099         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27100         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27101         defined.
27102         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27103         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
27104         temporarily sse2 rather than sse if not enabled already.
27106         PR target/83604
27107         * config/i386/sse.md (VI248_VLBW): Rename to ...
27108         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
27109         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
27110         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
27111         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
27112         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
27113         mode iterator instead of VI248_VLBW.
27115 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
27117         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
27118         (record_modified): Skip clobbers; add debug output.
27119         (param_change_prob): Use sreal frequencies.
27121 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27123         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
27124         punt for user-aligned variables.
27126 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27128         * tree-chrec.c (chrec_contains_symbols): Return true for
27129         POLY_INT_CST.
27131 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
27133         PR target/82439
27134         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
27135         of (x|y) == x for BICS pattern.
27137 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27139         PR tree-optimization/83605
27140         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
27141         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
27142         can throw.
27144 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27146         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
27147         * config/epiphany/rtems.h: New file.
27149 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27150             Uros Bizjak  <ubizjak@gmail.com>
27152         PR target/83554
27153         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
27154         QIreg_operand instead of register_operand predicate.
27155         * config/i386/i386.c (ix86_rop_should_change_byte_p,
27156         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
27157         comments instead of -fmitigate[-_]rop.
27159 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27161         PR bootstrap/81926
27162         * cgraphunit.c (symbol_table::compile): Switch to text_section
27163         before calling assembly_start debug hook.
27164         * run-rtl-passes.c (run_rtl_passes): Likewise.
27165         Include output.h.
27167 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27169         * tree-vrp.c (extract_range_from_binary_expr_1): Check
27170         range_int_cst_p rather than !symbolic_range_p before calling
27171         extract_range_from_multiplicative_op_1.
27173 2018-01-04  Jeff Law  <law@redhat.com>
27175         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
27176         redundant test in assertion.
27178 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27180         * doc/rtl.texi: Document machine_mode wrapper classes.
27182 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27184         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
27185         using tree_to_uhwi.
27187 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27189         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
27190         the VEC_PERM_EXPR fold to fail.
27192 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27194         PR debug/83585
27195         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
27196         to switched_sections.
27198 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27200         PR target/83680
27201         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
27202         test for d.testing.
27204 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
27206         PR target/83387
27207         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
27208         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
27210 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27212         PR debug/83666
27213         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
27214         is BLKmode and bitpos not zero or mode change is needed.
27216 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27218         PR target/83675
27219         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
27220         TARGET_VIS2.
27222 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
27224         PR target/83628
27225         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
27226         instead of MULT rtx.  Update all corresponding splitters.
27227         (*saddl_se): Ditto.
27228         (*ssub<modesuffix>): Ditto.
27229         (*ssubl_se): Ditto.
27230         (*cmp_sadd_di): Update split patterns.
27231         (*cmp_sadd_si): Ditto.
27232         (*cmp_sadd_sidi): Ditto.
27233         (*cmp_ssub_di): Ditto.
27234         (*cmp_ssub_si): Ditto.
27235         (*cmp_ssub_sidi): Ditto.
27236         * config/alpha/predicates.md (const23_operand): New predicate.
27237         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
27238         Look for ASHIFT, not MULT inner operand.
27239         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
27241 2018-01-04  Martin Liska  <mliska@suse.cz>
27243         PR gcov-profile/83669
27244         * gcov.c (output_intermediate_file): Add version to intermediate
27245         gcov file.
27246         * doc/gcov.texi: Document new field 'version' in intermediate
27247         file format. Fix location of '-k' option of gcov command.
27249 2018-01-04  Martin Liska  <mliska@suse.cz>
27251         PR ipa/82352
27252         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
27254 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27256         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
27258 2018-01-03  Martin Sebor  <msebor@redhat.com>
27260         PR tree-optimization/83655
27261         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
27262         checking calls with invalid arguments.
27264 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27266         * tree-vect-stmts.c (vect_get_store_rhs): New function.
27267         (vectorizable_mask_load_store): Delete.
27268         (vectorizable_call): Return false for masked loads and stores.
27269         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
27270         instead of gimple_assign_rhs1.
27271         (vectorizable_load): Handle IFN_MASK_LOAD.
27272         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
27274 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27276         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
27277         split out from..,
27278         (vectorizable_mask_load_store): ...here.
27279         (vectorizable_load): ...and here.
27281 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27283         * tree-vect-stmts.c (vect_build_all_ones_mask)
27284         (vect_build_zero_merge_argument): New functions, split out from...
27285         (vectorizable_load): ...here.
27287 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27289         * tree-vect-stmts.c (vect_check_store_rhs): New function,
27290         split out from...
27291         (vectorizable_mask_load_store): ...here.
27292         (vectorizable_store): ...and here.
27294 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27296         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
27297         split out from...
27298         (vectorizable_mask_load_store): ...here.
27300 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27302         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
27303         (vect_model_store_cost): Take a vec_load_store_type instead of a
27304         vect_def_type.
27305         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
27306         (vect_model_store_cost): Take a vec_load_store_type instead of a
27307         vect_def_type.
27308         (vectorizable_mask_load_store): Update accordingly.
27309         (vectorizable_store): Likewise.
27310         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
27312 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27314         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
27315         IFN_MASK_LOAD calls here rather than...
27316         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
27318 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27319             Alan Hayward  <alan.hayward@arm.com>
27320             David Sherwood  <david.sherwood@arm.com>
27322         * expmed.c (extract_bit_field_1): For vector extracts,
27323         fall back to extract_bit_field_as_subreg if vec_extract
27324         isn't available.
27326 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27327             Alan Hayward  <alan.hayward@arm.com>
27328             David Sherwood  <david.sherwood@arm.com>
27330         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
27331         they are variable or constant sized.
27332         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
27333         slots for constant-sized data.
27335 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27336             Alan Hayward  <alan.hayward@arm.com>
27337             David Sherwood  <david.sherwood@arm.com>
27339         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
27340         handling COND_EXPRs with boolean comparisons, try to find a better
27341         basis for the mask type than the boolean itself.
27343 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27345         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
27346         is calculated and how it can be overridden.
27347         * genmodes.c (max_bitsize_mode_any_mode): New variable.
27348         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
27349         if defined.
27350         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
27351         if nonzero.
27353 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27354             Alan Hayward  <alan.hayward@arm.com>
27355             David Sherwood  <david.sherwood@arm.com>
27357         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
27358         Remove the mode argument.
27359         (aarch64_simd_valid_immediate): Remove the mode and inverse
27360         arguments.
27361         * config/aarch64/iterators.md (bitsize): New iterator.
27362         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
27363         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
27364         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
27365         aarch64_simd_valid_immediate.
27366         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
27367         (aarch64_reg_or_bic_imm): Likewise.
27368         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
27369         with an insn_type enum and msl with a modifier_type enum.
27370         Replace element_width with a scalar_mode.  Change the shift
27371         to unsigned int.  Add constructors for scalar_float_mode and
27372         scalar_int_mode elements.
27373         (aarch64_vect_float_const_representable_p): Delete.
27374         (aarch64_can_const_movi_rtx_p)
27375         (aarch64_simd_scalar_immediate_valid_for_move)
27376         (aarch64_simd_make_constant): Update call to
27377         aarch64_simd_valid_immediate.
27378         (aarch64_advsimd_valid_immediate_hs): New function.
27379         (aarch64_advsimd_valid_immediate): Likewise.
27380         (aarch64_simd_valid_immediate): Remove mode and inverse
27381         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
27382         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
27383         and aarch64_float_const_representable_p on the result.
27384         (aarch64_output_simd_mov_immediate): Remove mode argument.
27385         Update call to aarch64_simd_valid_immediate and use of
27386         simd_immediate_info.
27387         (aarch64_output_scalar_simd_mov_immediate): Update call
27388         accordingly.
27390 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27391             Alan Hayward  <alan.hayward@arm.com>
27392             David Sherwood  <david.sherwood@arm.com>
27394         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
27395         (mode_nunits): Likewise CONST_MODE_NUNITS.
27396         * machmode.def (ADJUST_NUNITS): Document.
27397         * genmodes.c (mode_data::need_nunits_adj): New field.
27398         (blank_mode): Update accordingly.
27399         (adj_nunits): New variable.
27400         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
27401         parameter.
27402         (emit_mode_size_inline): Set need_bytesize_adj for all modes
27403         listed in adj_nunits.
27404         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
27405         listed in adj_nunits.  Don't emit case statements for such modes.
27406         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
27407         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
27408         nothing if adj_nunits is nonnull.
27409         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
27410         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
27411         (emit_mode_fbit): Update use of print_maybe_const_decl.
27412         (emit_move_size): Likewise.  Treat the array as non-const
27413         if adj_nunits.
27414         (emit_mode_adjustments): Handle adj_nunits.
27416 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27418         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
27419         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
27420         (VECTOR_MODES): Use it.
27421         (make_vector_modes): Take the prefix as an argument.
27423 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27424             Alan Hayward  <alan.hayward@arm.com>
27425             David Sherwood  <david.sherwood@arm.com>
27427         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
27428         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
27429         for MODE_VECTOR_BOOL.
27430         * machmode.def (VECTOR_BOOL_MODE): Document.
27431         * genmodes.c (VECTOR_BOOL_MODE): New macro.
27432         (make_vector_bool_mode): New function.
27433         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
27434         MODE_VECTOR_BOOL.
27435         * lto-streamer-in.c (lto_input_mode_table): Likewise.
27436         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
27437         Likewise.
27438         * stor-layout.c (int_mode_for_mode): Likewise.
27439         * tree.c (build_vector_type_for_mode): Likewise.
27440         * varasm.c (output_constant_pool_2): Likewise.
27441         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
27442         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
27443         for MODE_VECTOR_BOOL.
27444         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
27445         of mode class checks.
27446         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
27447         instead of a list of mode class checks.
27448         (expand_vector_scalar_condition): Likewise.
27449         (type_for_widest_vector_mode): Handle BImode as an inner mode.
27451 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27452             Alan Hayward  <alan.hayward@arm.com>
27453             David Sherwood  <david.sherwood@arm.com>
27455         * machmode.h (mode_size): Change from unsigned short to
27456         poly_uint16_pod.
27457         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
27458         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27459         or if measurement_type is not polynomial.
27460         (fixed_size_mode::includes_p): Check for constant-sized modes.
27461         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
27462         return a poly_uint16 rather than an unsigned short.
27463         (emit_mode_size): Change the type of mode_size from unsigned short
27464         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
27465         (emit_mode_adjustments): Cope with polynomial vector sizes.
27466         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27467         for GET_MODE_SIZE.
27468         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27469         for GET_MODE_SIZE.
27470         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
27471         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
27472         * caller-save.c (setup_save_areas): Likewise.
27473         (replace_reg_with_saved_mem): Likewise.
27474         * calls.c (emit_library_call_value_1): Likewise.
27475         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
27476         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
27477         (gen_lowpart_for_combine): Likewise.
27478         * convert.c (convert_to_integer_1): Likewise.
27479         * cse.c (equiv_constant, cse_insn): Likewise.
27480         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
27481         (cselib_subst_to_values): Likewise.
27482         * dce.c (word_dce_process_block): Likewise.
27483         * df-problems.c (df_word_lr_mark_ref): Likewise.
27484         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
27485         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
27486         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
27487         (rtl_for_decl_location): Likewise.
27488         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
27489         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
27490         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
27491         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
27492         (expand_expr_real_1): Likewise.
27493         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
27494         (pad_below): Likewise.
27495         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27496         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
27497         * ira.c (get_subreg_tracking_sizes): Likewise.
27498         * ira-build.c (ira_create_allocno_objects): Likewise.
27499         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
27500         (ira_sort_regnos_for_alter_reg): Likewise.
27501         * ira-costs.c (record_operand_costs): Likewise.
27502         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
27503         (resolve_simple_move): Likewise.
27504         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
27505         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
27506         (lra_constraints): Likewise.
27507         (CONST_POOL_OK_P): Reject variable-sized modes.
27508         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
27509         (add_pseudo_to_slot, lra_spill): Likewise.
27510         * omp-low.c (omp_clause_aligned_alignment): Likewise.
27511         * optabs-query.c (get_best_extraction_insn): Likewise.
27512         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27513         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
27514         (expand_mult_highpart, valid_multiword_target_p): Likewise.
27515         * recog.c (offsettable_address_addr_space_p): Likewise.
27516         * regcprop.c (maybe_mode_change): Likewise.
27517         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
27518         * regrename.c (build_def_use): Likewise.
27519         * regstat.c (dump_reg_info): Likewise.
27520         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
27521         (find_reloads, find_reloads_subreg_address): Likewise.
27522         * reload1.c (eliminate_regs_1): Likewise.
27523         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
27524         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
27525         (simplify_binary_operation_1, simplify_subreg): Likewise.
27526         * targhooks.c (default_function_arg_padding): Likewise.
27527         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
27528         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
27529         (verify_gimple_assign_ternary): Likewise.
27530         * tree-inline.c (estimate_move_cost): Likewise.
27531         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27532         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
27533         (get_address_cost_ainc): Likewise.
27534         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
27535         (vect_supportable_dr_alignment): Likewise.
27536         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27537         (vectorizable_reduction): Likewise.
27538         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
27539         (vectorizable_operation, vectorizable_load): Likewise.
27540         * tree.c (build_same_sized_truth_vector_type): Likewise.
27541         * valtrack.c (cleanup_auto_inc_dec): Likewise.
27542         * var-tracking.c (emit_note_insn_var_location): Likewise.
27543         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
27544         (ADDR_VEC_ALIGN): Likewise.
27546 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27547             Alan Hayward  <alan.hayward@arm.com>
27548             David Sherwood  <david.sherwood@arm.com>
27550         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
27551         unsigned short.
27552         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27553         or if measurement_type is polynomial.
27554         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
27555         * combine.c (make_extraction): Likewise.
27556         * dse.c (find_shift_sequence): Likewise.
27557         * dwarf2out.c (mem_loc_descriptor): Likewise.
27558         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
27559         (extract_bit_field, extract_low_bits): Likewise.
27560         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
27561         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
27562         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
27563         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
27564         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27565         * reload.c (find_reloads): Likewise.
27566         * reload1.c (alter_reg): Likewise.
27567         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
27568         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
27569         * tree-if-conv.c (predicate_mem_writes): Likewise.
27570         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
27571         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
27572         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
27573         * valtrack.c (dead_debug_insert_temp): Likewise.
27574         * varasm.c (mergeable_constant_section): Likewise.
27575         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
27577 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27578             Alan Hayward  <alan.hayward@arm.com>
27579             David Sherwood  <david.sherwood@arm.com>
27581         * expr.c (expand_assignment): Cope with polynomial mode sizes
27582         when assigning to a CONCAT.
27584 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27585             Alan Hayward  <alan.hayward@arm.com>
27586             David Sherwood  <david.sherwood@arm.com>
27588         * machmode.h (mode_precision): Change from unsigned short to
27589         poly_uint16_pod.
27590         (mode_to_precision): Return a poly_uint16 rather than an unsigned
27591         short.
27592         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
27593         or if measurement_type is not polynomial.
27594         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
27595         in which the mode is already known to be a scalar_int_mode.
27596         * genmodes.c (emit_mode_precision): Change the type of mode_precision
27597         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
27598         initializer.
27599         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27600         for GET_MODE_PRECISION.
27601         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27602         for GET_MODE_PRECISION.
27603         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
27604         as polynomial.
27605         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
27606         (expand_field_assignment, make_extraction): Likewise.
27607         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
27608         (get_last_value): Likewise.
27609         * convert.c (convert_to_integer_1): Likewise.
27610         * cse.c (cse_insn): Likewise.
27611         * expr.c (expand_expr_real_1): Likewise.
27612         * lra-constraints.c (simplify_operand_subreg): Likewise.
27613         * optabs-query.c (can_atomic_load_p): Likewise.
27614         * optabs.c (expand_atomic_load): Likewise.
27615         (expand_atomic_store): Likewise.
27616         * ree.c (combine_reaching_defs): Likewise.
27617         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
27618         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
27619         * tree.h (type_has_mode_precision_p): Likewise.
27620         * ubsan.c (instrument_si_overflow): Likewise.
27622 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27623             Alan Hayward  <alan.hayward@arm.com>
27624             David Sherwood  <david.sherwood@arm.com>
27626         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
27627         polynomial numbers of units.
27628         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
27629         (valid_vector_subparts_p): New function.
27630         (build_vector_type): Remove temporary shim and take the number
27631         of units as a poly_uint64 rather than an int.
27632         (build_opaque_vector_type): Take the number of units as a
27633         poly_uint64 rather than an int.
27634         * tree.c (build_vector_from_ctor): Handle polynomial
27635         TYPE_VECTOR_SUBPARTS.
27636         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
27637         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
27638         (build_vector_from_val): If the number of units is variable,
27639         use build_vec_duplicate_cst for constant operands and
27640         VEC_DUPLICATE_EXPR otherwise.
27641         (make_vector_type): Remove temporary is_constant ().
27642         (build_vector_type, build_opaque_vector_type): Take the number of
27643         units as a poly_uint64 rather than an int.
27644         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
27645         VECTOR_CST_NELTS.
27646         * cfgexpand.c (expand_debug_expr): Likewise.
27647         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
27648         (store_constructor, expand_expr_real_1): Likewise.
27649         (const_scalar_mask_from_tree): Likewise.
27650         * fold-const-call.c (fold_const_reduction): Likewise.
27651         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
27652         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
27653         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
27654         (fold_relational_const): Likewise.
27655         (native_interpret_vector): Likewise.  Change the size from an
27656         int to an unsigned int.
27657         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
27658         TYPE_VECTOR_SUBPARTS.
27659         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
27660         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
27661         duplicating a non-constant operand into a variable-length vector.
27662         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
27663         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
27664         * ipa-icf.c (sem_variable::equals): Likewise.
27665         * match.pd: Likewise.
27666         * omp-simd-clone.c (simd_clone_subparts): Likewise.
27667         * print-tree.c (print_node): Likewise.
27668         * stor-layout.c (layout_type): Likewise.
27669         * targhooks.c (default_builtin_vectorization_cost): Likewise.
27670         * tree-cfg.c (verify_gimple_comparison): Likewise.
27671         (verify_gimple_assign_binary): Likewise.
27672         (verify_gimple_assign_ternary): Likewise.
27673         (verify_gimple_assign_single): Likewise.
27674         * tree-pretty-print.c (dump_generic_node): Likewise.
27675         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27676         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
27677         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
27678         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
27679         (vect_shift_permute_load_chain): Likewise.
27680         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
27681         (expand_vector_condition, optimize_vector_constructor): Likewise.
27682         (lower_vec_perm, get_compute_type): Likewise.
27683         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27684         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
27685         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
27686         (vect_recog_mask_conversion_pattern): Likewise.
27687         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
27688         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
27689         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27690         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
27691         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
27692         (vectorizable_shift, vectorizable_operation, vectorizable_store)
27693         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
27694         (supportable_widening_operation): Likewise.
27695         (supportable_narrowing_operation): Likewise.
27696         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
27697         Likewise.
27698         * varasm.c (output_constant): Likewise.
27700 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27701             Alan Hayward  <alan.hayward@arm.com>
27702             David Sherwood  <david.sherwood@arm.com>
27704         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
27705         so that both the length == 3 and length != 3 cases set up their
27706         own permute vectors.  Add comments explaining why we know the
27707         number of elements is constant.
27708         (vect_permute_load_chain): Likewise.
27710 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27711             Alan Hayward  <alan.hayward@arm.com>
27712             David Sherwood  <david.sherwood@arm.com>
27714         * machmode.h (mode_nunits): Change from unsigned char to
27715         poly_uint16_pod.
27716         (ONLY_FIXED_SIZE_MODES): New macro.
27717         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
27718         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
27719         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
27720         New typedefs.
27721         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
27722         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
27723         or if measurement_type is not polynomial.
27724         * genmodes.c (ZERO_COEFFS): New macro.
27725         (emit_mode_nunits_inline): Make mode_nunits_inline return a
27726         poly_uint16.
27727         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
27728         Use ZERO_COEFFS when emitting initializers.
27729         * data-streamer.h (bp_pack_poly_value): New function.
27730         (bp_unpack_poly_value): Likewise.
27731         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27732         for GET_MODE_NUNITS.
27733         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27734         for GET_MODE_NUNITS.
27735         * tree.c (make_vector_type): Remove temporary shim and make
27736         the real function take the number of units as a poly_uint64
27737         rather than an int.
27738         (build_vector_type_for_mode): Handle polynomial nunits.
27739         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
27740         * emit-rtl.c (const_vec_series_p_1): Likewise.
27741         (gen_rtx_CONST_VECTOR): Likewise.
27742         * fold-const.c (test_vec_duplicate_folding): Likewise.
27743         * genrecog.c (validate_pattern): Likewise.
27744         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
27745         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27746         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
27747         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
27748         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
27749         * rtlanal.c (subreg_get_info): Likewise.
27750         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27751         (vect_grouped_load_supported): Likewise.
27752         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
27753         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
27754         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
27755         (simplify_const_unary_operation, simplify_binary_operation_1)
27756         (simplify_const_binary_operation, simplify_ternary_operation)
27757         (test_vector_ops_duplicate, test_vector_ops): Likewise.
27758         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
27759         instead of CONST_VECTOR_NUNITS.
27760         * varasm.c (output_constant_pool_2): Likewise.
27761         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
27762         explicit-encoded elements in the XVEC for variable-length vectors.
27764 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27766         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
27768 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27769             Alan Hayward  <alan.hayward@arm.com>
27770             David Sherwood  <david.sherwood@arm.com>
27772         * coretypes.h (fixed_size_mode): Declare.
27773         (fixed_size_mode_pod): New typedef.
27774         * builtins.h (target_builtins::x_apply_args_mode)
27775         (target_builtins::x_apply_result_mode): Change type to
27776         fixed_size_mode_pod.
27777         * builtins.c (apply_args_size, apply_result_size, result_vector)
27778         (expand_builtin_apply_args_1, expand_builtin_apply)
27779         (expand_builtin_return): Update accordingly.
27781 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27783         * cse.c (hash_rtx_cb): Hash only the encoded elements.
27784         * cselib.c (cselib_hash_rtx): Likewise.
27785         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
27786         CONST_VECTOR encoding.
27788 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27789             Jeff Law  <law@redhat.com>
27791         PR target/83641
27792         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
27793         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
27794         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
27795         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
27797         PR target/83641
27798         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
27799         explicitly probe *sp in a noreturn function if there were any callee
27800         register saves or frame pointer is needed.
27802 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27804         PR debug/83621
27805         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27806         BLKmode for ternary, binary or unary expressions.
27808         PR debug/83645
27809         * var-tracking.c (delete_vta_debug_insn): New inline function.
27810         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27811         insns from get_insns () to NULL instead of each bb separately.
27812         Use delete_vta_debug_insn.  No longer static.
27813         (vt_debug_insns_local, variable_tracking_main_1): Adjust
27814         delete_vta_debug_insns callers.
27815         * rtl.h (delete_vta_debug_insns): Declare.
27816         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27817         instead of variable_tracking_main.
27819 2018-01-03  Martin Sebor  <msebor@redhat.com>
27821         PR tree-optimization/83603
27822         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27823         arguments past the endof the argument list in functions declared
27824         without a prototype.
27825         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27826         Avoid checking when arguments are null.
27828 2018-01-03  Martin Sebor  <msebor@redhat.com>
27830         PR c/83559
27831         * doc/extend.texi (attribute const): Fix a typo.
27832         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27833         issuing -Wsuggest-attribute for void functions.
27835 2018-01-03  Martin Sebor  <msebor@redhat.com>
27837         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27838         offset_int::from instead of wide_int::to_shwi.
27839         (maybe_diag_overlap): Remove assertion.
27840         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27841         * gimple-ssa-sprintf.c (format_directive): Same.
27842         (parse_directive): Same.
27843         (sprintf_dom_walker::compute_format_length): Same.
27844         (try_substitute_return_value): Same.
27846 2018-01-03  Jeff Law  <law@redhat.com>
27848         PR middle-end/83654
27849         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27850         non-constant residual for zero at runtime and avoid probing in
27851         that case.  Reorganize code for trailing problem to mirror handling
27852         of the residual.
27854 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27856         PR tree-optimization/83501
27857         * tree-ssa-strlen.c (get_string_cst): New.
27858         (handle_char_store): Call get_string_cst.
27860 2018-01-03  Martin Liska  <mliska@suse.cz>
27862         PR tree-optimization/83593
27863         * tree-ssa-strlen.c: Include tree-cfg.h.
27864         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27865         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27866         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27867         to false.
27868         (strlen_dom_walker::before_dom_children): Call
27869         gimple_purge_dead_eh_edges. Dump tranformation with details
27870         dump flags.
27871         (strlen_dom_walker::before_dom_children): Update call by adding
27872         new argument cleanup_eh.
27873         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27875 2018-01-03  Martin Liska  <mliska@suse.cz>
27877         PR ipa/83549
27878         * cif-code.def (VARIADIC_THUNK): New enum value.
27879         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27880         thunks.
27882 2018-01-03  Jan Beulich  <jbeulich@suse.com>
27884         * sse.md (mov<mode>_internal): Tighten condition for when to use
27885         vmovdqu<ssescalarsize> for TI and OI modes.
27887 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27889         Update copyright years.
27891 2018-01-03  Martin Liska  <mliska@suse.cz>
27893         PR ipa/83594
27894         * ipa-visibility.c (function_and_variable_visibility): Skip
27895         functions with noipa attribure.
27897 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
27899         * gcc.c (process_command): Update copyright notice dates.
27900         * gcov-dump.c (print_version): Ditto.
27901         * gcov.c (print_version): Ditto.
27902         * gcov-tool.c (print_version): Ditto.
27903         * gengtype.c (create_file): Ditto.
27904         * doc/cpp.texi: Bump @copying's copyright year.
27905         * doc/cppinternals.texi: Ditto.
27906         * doc/gcc.texi: Ditto.
27907         * doc/gccint.texi: Ditto.
27908         * doc/gcov.texi: Ditto.
27909         * doc/install.texi: Ditto.
27910         * doc/invoke.texi: Ditto.
27912 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27914         * vector-builder.h (vector_builder::m_full_nelts): Change from
27915         unsigned int to poly_uint64.
27916         (vector_builder::full_nelts): Update prototype accordingly.
27917         (vector_builder::new_vector): Likewise.
27918         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
27919         (vector_builder::operator ==): Likewise.
27920         (vector_builder::finalize): Likewise.
27921         * int-vector-builder.h (int_vector_builder::int_vector_builder):
27922         Take the number of elements as a poly_uint64 rather than an
27923         unsigned int.
27924         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
27925         from unsigned int to poly_uint64.
27926         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27927         (vec_perm_indices::new_vector): Likewise.
27928         (vec_perm_indices::length): Likewise.
27929         (vec_perm_indices::nelts_per_input): Likewise.
27930         (vec_perm_indices::input_nelts): Likewise.
27931         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27932         number of elements per input as a poly_uint64 rather than an
27933         unsigned int.  Use the original encoding for variable-length
27934         vectors, rather than clamping each individual element.
27935         For the second and subsequent elements in each pattern,
27936         clamp the step and base before clamping their sum.
27937         (vec_perm_indices::series_p): Handle polynomial element counts.
27938         (vec_perm_indices::all_in_range_p): Likewise.
27939         (vec_perm_indices_to_tree): Likewise.
27940         (vec_perm_indices_to_rtx): Likewise.
27941         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27942         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27943         (tree_vector_builder::new_binary_operation): Handle polynomial
27944         element counts.  Return false if we need to know the number
27945         of elements at compile time.
27946         * fold-const.c (fold_vec_perm): Punt if the number of elements
27947         isn't known at compile time.
27949 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27951         * vec-perm-indices.h (vec_perm_builder): Change element type
27952         from HOST_WIDE_INT to poly_int64.
27953         (vec_perm_indices::element_type): Update accordingly.
27954         (vec_perm_indices::clamp): Handle polynomial element_types.
27955         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27956         (vec_perm_indices::all_in_range_p): Likewise.
27957         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27958         than shwi trees.
27959         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27960         polynomial vec_perm_indices element types.
27961         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27962         * fold-const.c (fold_vec_perm): Likewise.
27963         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27964         * tree-vect-generic.c (lower_vec_perm): Likewise.
27965         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27966         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27967         element type to HOST_WIDE_INT.
27969 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27970             Alan Hayward  <alan.hayward@arm.com>
27971             David Sherwood  <david.sherwood@arm.com>
27973         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27974         rather than an int.  Use plus_constant.
27975         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27976         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27978 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27979             Alan Hayward  <alan.hayward@arm.com>
27980             David Sherwood  <david.sherwood@arm.com>
27982         * calls.c (emit_call_1, expand_call): Change struct_value_size from
27983         a HOST_WIDE_INT to a poly_int64.
27985 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27986             Alan Hayward  <alan.hayward@arm.com>
27987             David Sherwood  <david.sherwood@arm.com>
27989         * calls.c (load_register_parameters): Cope with polynomial
27990         mode sizes.  Require a constant size for BLKmode parameters
27991         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
27992         forces a parameter to be padded at the lsb end in order to
27993         fill a complete number of words, require the parameter size
27994         to be ordered wrt UNITS_PER_WORD.
27996 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27997             Alan Hayward  <alan.hayward@arm.com>
27998             David Sherwood  <david.sherwood@arm.com>
28000         * reload1.c (spill_stack_slot_width): Change element type
28001         from unsigned int to poly_uint64_pod.
28002         (alter_reg): Treat mode sizes as polynomial.
28004 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28005             Alan Hayward  <alan.hayward@arm.com>
28006             David Sherwood  <david.sherwood@arm.com>
28008         * reload.c (complex_word_subreg_p): New function.
28009         (reload_inner_reg_of_subreg, push_reload): Use it.
28011 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28012             Alan Hayward  <alan.hayward@arm.com>
28013             David Sherwood  <david.sherwood@arm.com>
28015         * lra-constraints.c (process_alt_operands): Reject matched
28016         operands whose sizes aren't ordered.
28017         (match_reload): Refer to this check here.
28019 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28020             Alan Hayward  <alan.hayward@arm.com>
28021             David Sherwood  <david.sherwood@arm.com>
28023         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28024         that the mode size is in the set {1, 2, 4, 8, 16}.
28026 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28027             Alan Hayward  <alan.hayward@arm.com>
28028             David Sherwood  <david.sherwood@arm.com>
28030         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28031         Use plus_constant instead of gen_rtx_PLUS.
28033 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28034             Alan Hayward  <alan.hayward@arm.com>
28035             David Sherwood  <david.sherwood@arm.com>
28037         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28038         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28039         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28040         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28041         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28042         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28043         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28044         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28045         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28046         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
28047         a poly_int64.
28048         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
28049         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
28050         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
28051         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
28052         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
28053         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
28054         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
28055         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
28056         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
28057         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28058         function.
28059         * expr.c (emit_move_resolve_push): Treat the input and result
28060         of PUSH_ROUNDING as a poly_int64.
28061         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28062         (emit_push_insn): Likewise.
28063         * lra-eliminations.c (mark_not_eliminable): Likewise.
28064         * recog.c (push_operand): Likewise.
28065         * reload1.c (elimination_effects): Likewise.
28066         * rtlanal.c (nonzero_bits1): Likewise.
28067         * calls.c (store_one_arg): Likewise.  Require the padding to be
28068         known at compile time.
28070 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28071             Alan Hayward  <alan.hayward@arm.com>
28072             David Sherwood  <david.sherwood@arm.com>
28074         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28075         Use plus_constant instead of gen_rtx_PLUS.
28077 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28078             Alan Hayward  <alan.hayward@arm.com>
28079             David Sherwood  <david.sherwood@arm.com>
28081         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28082         rather than an int.
28084 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28085             Alan Hayward  <alan.hayward@arm.com>
28086             David Sherwood  <david.sherwood@arm.com>
28088         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28089         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28090         via stack temporaries.  Treat the mode size as polynomial too.
28092 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28093             Alan Hayward  <alan.hayward@arm.com>
28094             David Sherwood  <david.sherwood@arm.com>
28096         * expr.c (expand_expr_real_2): When handling conversions involving
28097         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28098         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
28099         as a poly_uint64 too.
28101 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28102             Alan Hayward  <alan.hayward@arm.com>
28103             David Sherwood  <david.sherwood@arm.com>
28105         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
28107 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28108             Alan Hayward  <alan.hayward@arm.com>
28109             David Sherwood  <david.sherwood@arm.com>
28111         * combine.c (can_change_dest_mode): Handle polynomial
28112         REGMODE_NATURAL_SIZE.
28113         * expmed.c (store_bit_field_1): Likewise.
28114         * expr.c (store_constructor): Likewise.
28115         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
28116         and polynomial REGMODE_NATURAL_SIZE.
28117         (gen_lowpart_common): Likewise.
28118         * reginfo.c (record_subregs_of_mode): Likewise.
28119         * rtlanal.c (read_modify_subreg_p): Likewise.
28121 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28122             Alan Hayward  <alan.hayward@arm.com>
28123             David Sherwood  <david.sherwood@arm.com>
28125         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
28126         numbers of elements.
28128 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28129             Alan Hayward  <alan.hayward@arm.com>
28130             David Sherwood  <david.sherwood@arm.com>
28132         * match.pd: Cope with polynomial numbers of vector elements.
28134 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28135             Alan Hayward  <alan.hayward@arm.com>
28136             David Sherwood  <david.sherwood@arm.com>
28138         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
28139         in a POINTER_PLUS_EXPR.
28141 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28142             Alan Hayward  <alan.hayward@arm.com>
28143             David Sherwood  <david.sherwood@arm.com>
28145         * omp-simd-clone.c (simd_clone_subparts): New function.
28146         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
28147         (ipa_simd_modify_function_body): Likewise.
28149 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28150             Alan Hayward  <alan.hayward@arm.com>
28151             David Sherwood  <david.sherwood@arm.com>
28153         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
28154         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
28155         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
28156         (expand_vector_condition, vector_element): Likewise.
28157         (subparts_gt): New function.
28158         (get_compute_type): Use subparts_gt.
28159         (count_type_subparts): Delete.
28160         (expand_vector_operations_1): Use subparts_gt instead of
28161         count_type_subparts.
28163 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28164             Alan Hayward  <alan.hayward@arm.com>
28165             David Sherwood  <david.sherwood@arm.com>
28167         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
28168         (vect_compile_time_alias): ...this new function.  Do the calculation
28169         on poly_ints rather than trees.
28170         (vect_prune_runtime_alias_test_list): Update call accordingly.
28172 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28173             Alan Hayward  <alan.hayward@arm.com>
28174             David Sherwood  <david.sherwood@arm.com>
28176         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
28177         numbers of units.
28178         (vect_schedule_slp_instance): Likewise.
28180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28181             Alan Hayward  <alan.hayward@arm.com>
28182             David Sherwood  <david.sherwood@arm.com>
28184         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
28185         constant and extern definitions for variable-length vectors.
28186         (vect_get_constant_vectors): Note that the number of units
28187         is known to be constant.
28189 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28190             Alan Hayward  <alan.hayward@arm.com>
28191             David Sherwood  <david.sherwood@arm.com>
28193         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
28194         of units as polynomial.  Choose between WIDE and NARROW based
28195         on multiple_p.
28197 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28198             Alan Hayward  <alan.hayward@arm.com>
28199             David Sherwood  <david.sherwood@arm.com>
28201         * tree-vect-stmts.c (simd_clone_subparts): New function.
28202         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
28204 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28205             Alan Hayward  <alan.hayward@arm.com>
28206             David Sherwood  <david.sherwood@arm.com>
28208         * tree-vect-stmts.c (vectorizable_call): Treat the number of
28209         vectors as polynomial.  Use build_index_vector for
28210         IFN_GOMP_SIMD_LANE.
28212 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28213             Alan Hayward  <alan.hayward@arm.com>
28214             David Sherwood  <david.sherwood@arm.com>
28216         * tree-vect-stmts.c (get_load_store_type): Treat the number of
28217         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
28218         for variable-length vectors.
28219         (vectorizable_mask_load_store): Treat the number of units as
28220         polynomial, asserting that it is constant if the condition has
28221         already been enforced.
28222         (vectorizable_store, vectorizable_load): Likewise.
28224 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28225             Alan Hayward  <alan.hayward@arm.com>
28226             David Sherwood  <david.sherwood@arm.com>
28228         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
28229         of units as polynomial.  Punt if we can't tell at compile time
28230         which vector contains the final result.
28232 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28233             Alan Hayward  <alan.hayward@arm.com>
28234             David Sherwood  <david.sherwood@arm.com>
28236         * tree-vect-loop.c (vectorizable_induction): Treat the number
28237         of units as polynomial.  Punt on SLP inductions.  Use an integer
28238         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
28239         cast of such a series for variable-length floating-point
28240         reductions.
28242 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28243             Alan Hayward  <alan.hayward@arm.com>
28244             David Sherwood  <david.sherwood@arm.com>
28246         * tree.h (build_index_vector): Declare.
28247         * tree.c (build_index_vector): New function.
28248         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
28249         of units as polynomial, forcibly converting it to a constant if
28250         vectorizable_reduction has already enforced the condition.
28251         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
28252         to create a {1,2,3,...} vector.
28253         (vectorizable_reduction): Treat the number of units as polynomial.
28254         Choose vectype_in based on the largest scalar element size rather
28255         than the smallest number of units.  Enforce the restrictions
28256         relied on above.
28258 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28259             Alan Hayward  <alan.hayward@arm.com>
28260             David Sherwood  <david.sherwood@arm.com>
28262         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
28263         number of units as polynomial.
28265 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28266             Alan Hayward  <alan.hayward@arm.com>
28267             David Sherwood  <david.sherwood@arm.com>
28269         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
28270         * target.def (autovectorize_vector_sizes): Return the vector sizes
28271         by pointer, using vector_sizes rather than a bitmask.
28272         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
28273         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
28274         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
28275         Likewise.
28276         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
28277         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
28278         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
28279         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
28280         * omp-general.c (omp_max_vf): Likewise.
28281         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28282         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
28283         * tree-vect-loop.c (vect_analyze_loop): Likewise.
28284         * tree-vect-slp.c (vect_slp_bb): Likewise.
28285         * doc/tm.texi: Regenerate.
28286         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
28287         to a poly_uint64.
28288         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
28289         the vector size as a poly_uint64 rather than an unsigned int.
28290         (current_vector_size): Change from an unsigned int to a poly_uint64.
28291         (get_vectype_for_scalar_type): Update accordingly.
28292         * tree.h (build_truth_vector_type): Take the size and number of
28293         units as a poly_uint64 rather than an unsigned int.
28294         (build_vector_type): Add a temporary overload that takes
28295         the number of units as a poly_uint64 rather than an unsigned int.
28296         * tree.c (make_vector_type): Likewise.
28297         (build_truth_vector_type): Take the number of units as a poly_uint64
28298         rather than an unsigned int.
28300 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28301             Alan Hayward  <alan.hayward@arm.com>
28302             David Sherwood  <david.sherwood@arm.com>
28304         * target.def (get_mask_mode): Take the number of units and length
28305         as poly_uint64s rather than unsigned ints.
28306         * targhooks.h (default_get_mask_mode): Update accordingly.
28307         * targhooks.c (default_get_mask_mode): Likewise.
28308         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
28309         * doc/tm.texi: Regenerate.
28311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28312             Alan Hayward  <alan.hayward@arm.com>
28313             David Sherwood  <david.sherwood@arm.com>
28315         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
28316         * omp-general.c (omp_max_vf): Likewise.
28317         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
28318         (expand_omp_simd): Handle polynomial safelen.
28319         * omp-low.c (omplow_simd_context): Add a default constructor.
28320         (omplow_simd_context::max_vf): Change from int to poly_uint64.
28321         (lower_rec_simd_input_clauses): Update accordingly.
28322         (lower_rec_input_clauses): Likewise.
28324 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28325             Alan Hayward  <alan.hayward@arm.com>
28326             David Sherwood  <david.sherwood@arm.com>
28328         * tree-vectorizer.h (vect_nunits_for_cost): New function.
28329         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
28330         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
28331         (vect_analyze_slp_cost): Likewise.
28332         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
28333         (vect_model_load_cost): Likewise.
28335 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28336             Alan Hayward  <alan.hayward@arm.com>
28337             David Sherwood  <david.sherwood@arm.com>
28339         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
28340         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
28341         from an unsigned int * to a poly_uint64_pod *.
28342         (calculate_unrolling_factor): New function.
28343         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
28345 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28346             Alan Hayward  <alan.hayward@arm.com>
28347             David Sherwood  <david.sherwood@arm.com>
28349         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
28350         from an unsigned int to a poly_uint64.
28351         (_loop_vec_info::slp_unrolling_factor): Likewise.
28352         (_loop_vec_info::vectorization_factor): Change from an int
28353         to a poly_uint64.
28354         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
28355         (vect_get_num_vectors): New function.
28356         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
28357         (vect_get_num_copies): Use vect_get_num_vectors.
28358         (vect_analyze_data_ref_dependences): Change max_vf from an int *
28359         to an unsigned int *.
28360         (vect_analyze_data_refs): Change min_vf from an int * to a
28361         poly_uint64 *.
28362         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28363         than an unsigned HOST_WIDE_INT.
28364         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
28365         (vect_analyze_data_ref_dependence): Change max_vf from an int *
28366         to an unsigned int *.
28367         (vect_analyze_data_ref_dependences): Likewise.
28368         (vect_compute_data_ref_alignment): Handle polynomial vf.
28369         (vect_enhance_data_refs_alignment): Likewise.
28370         (vect_prune_runtime_alias_test_list): Likewise.
28371         (vect_shift_permute_load_chain): Likewise.
28372         (vect_supportable_dr_alignment): Likewise.
28373         (dependence_distance_ge_vf): Take the vectorization factor as a
28374         poly_uint64 rather than an unsigned HOST_WIDE_INT.
28375         (vect_analyze_data_refs): Change min_vf from an int * to a
28376         poly_uint64 *.
28377         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
28378         vfm1 as a poly_uint64 rather than an int.  Make the same change
28379         for the returned bound_scalar.
28380         (vect_gen_vector_loop_niters): Handle polynomial vf.
28381         (vect_do_peeling): Likewise.  Update call to
28382         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
28383         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
28384         be constant.
28385         * tree-vect-loop.c (vect_determine_vectorization_factor)
28386         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
28387         (vect_get_known_peeling_cost): Likewise.
28388         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
28389         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
28390         (vect_transform_loop): Likewise.  Use the lowest possible VF when
28391         updating the upper bounds of the loop.
28392         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
28393         rather than an int.
28394         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
28395         polynomial unroll factors.
28396         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
28397         (vect_make_slp_decision): Likewise.
28398         (vect_supported_load_permutation_p): Likewise, and polynomial
28399         vf too.
28400         (vect_analyze_slp_cost): Handle polynomial vf.
28401         (vect_slp_analyze_node_operations): Likewise.
28402         (vect_slp_analyze_bb_1): Likewise.
28403         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28404         than an unsigned HOST_WIDE_INT.
28405         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
28406         (vectorizable_load): Handle polynomial vf.
28407         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
28408         a poly_uint64.
28409         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
28411 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28412             Alan Hayward  <alan.hayward@arm.com>
28413             David Sherwood  <david.sherwood@arm.com>
28415         * match.pd: Handle bit operations involving three constants
28416         and try to fold one pair.
28418 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28420         * tree-vect-loop-manip.c: Include gimple-fold.h.
28421         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
28422         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
28423         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
28424         Add a path that uses a step of VF instead of 1, but disable it
28425         for now.
28426         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
28427         and niters_no_overflow parameters.  Update calls to
28428         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
28429         Create a new SSA name if the latter choses to use a ste other
28430         than zero, and return it via niters_vector_mult_vf_var.
28431         * tree-vect-loop.c (vect_transform_loop): Update calls to
28432         vect_do_peeling, vect_gen_vector_loop_niters and
28433         slpeel_make_loop_iterate_ntimes.
28434         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
28435         (vect_gen_vector_loop_niters): Update declarations after above changes.
28437 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
28439         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
28440         128-bit round to integer instructions.
28441         (ceil<mode>2): Likewise.
28442         (btrunc<mode>2): Likewise.
28443         (round<mode>2): Likewise.
28445 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28447         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
28448         unaligned VSX load/store on P8/P9.
28449         (expand_block_clear): Allow the use of unaligned VSX
28450         load/store on P8/P9.
28452 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28454         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
28455         New function.
28456         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
28457         swap associated with both a load and a store.
28459 2018-01-02  Andrew Waterman  <andrew@sifive.com>
28461         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
28462         * config/riscv/riscv.md (clear_cache): Use it.
28464 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
28466         * web.c: Remove out-of-date comment.
28468 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28470         * expr.c (fixup_args_size_notes): Check that any existing
28471         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
28472         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
28473         (emit_single_push_insn): ...here.
28475 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28477         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
28478         (const_vector_encoded_nelts): New function.
28479         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
28480         (const_vector_int_elt, const_vector_elt): Declare.
28481         * emit-rtl.c (const_vector_int_elt_1): New function.
28482         (const_vector_elt): Likewise.
28483         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
28484         of CONST_VECTOR_ELT.
28486 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28488         * expr.c: Include rtx-vector-builder.h.
28489         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
28490         directly on the tree encoding.
28491         (const_vector_from_tree): Likewise.
28492         * optabs.c: Include rtx-vector-builder.h.
28493         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
28494         sequence of "u" values.
28495         * vec-perm-indices.c: Include rtx-vector-builder.h.
28496         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
28497         directly on the vec_perm_indices encoding.
28499 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28501         * doc/rtl.texi (const_vector): Describe new encoding scheme.
28502         * Makefile.in (OBJS): Add rtx-vector-builder.o.
28503         * rtx-vector-builder.h: New file.
28504         * rtx-vector-builder.c: Likewise.
28505         * rtl.h (rtx_def::u2): Add a const_vector field.
28506         (CONST_VECTOR_NPATTERNS): New macro.
28507         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
28508         (CONST_VECTOR_DUPLICATE_P): Likewise.
28509         (CONST_VECTOR_STEPPED_P): Likewise.
28510         (CONST_VECTOR_ENCODED_ELT): Likewise.
28511         (const_vec_duplicate_p): Check for a duplicated vector encoding.
28512         (unwrap_const_vec_duplicate): Likewise.
28513         (const_vec_series_p): Check for a non-duplicated vector encoding.
28514         Say that the function only returns true for integer vectors.
28515         * emit-rtl.c: Include rtx-vector-builder.h.
28516         (gen_const_vec_duplicate_1): Delete.
28517         (gen_const_vector): Call gen_const_vec_duplicate instead of
28518         gen_const_vec_duplicate_1.
28519         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
28520         (gen_const_vec_duplicate): Use rtx_vector_builder.
28521         (gen_const_vec_series): Likewise.
28522         (gen_rtx_CONST_VECTOR): Likewise.
28523         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
28524         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28525         Build a new vector rather than modifying a CONST_VECTOR in-place.
28526         (handle_special_swappables): Update call accordingly.
28527         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
28528         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28529         Build a new vector rather than modifying a CONST_VECTOR in-place.
28530         (handle_special_swappables): Update call accordingly.
28532 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28534         * simplify-rtx.c (simplify_const_binary_operation): Use
28535         CONST_VECTOR_ELT instead of XVECEXP.
28537 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28539         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
28540         the selector elements to be different from the data elements
28541         if the selector is a VECTOR_CST.
28542         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
28543         ssizetype for the selector.
28545 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28547         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
28548         before testing each element individually.
28549         * tree-vect-generic.c (lower_vec_perm): Likewise.
28551 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28553         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
28554         * selftest-run-tests.c (selftest::run_tests): Call it.
28555         * vector-builder.h (vector_builder::operator ==): New function.
28556         (vector_builder::operator !=): Likewise.
28557         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
28558         (vec_perm_indices::all_from_input_p): New function.
28559         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28560         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
28561         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
28562         instead of reading the VECTOR_CST directly.  Detect whether both
28563         vector inputs are the same before constructing the vec_perm_indices,
28564         and update the number of inputs argument accordingly.  Use the
28565         utility functions added above.  Only construct sel2 if we need to.
28567 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28569         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
28570         the broadcast of the low byte.
28571         (expand_mult_highpart): Use an explicit encoding for the permutes.
28572         * optabs-query.c (can_mult_highpart_p): Likewise.
28573         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
28574         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28575         (vectorizable_bswap): Likewise.
28576         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
28577         explicit encoding for the power-of-2 permutes.
28578         (vect_permute_store_chain): Likewise.
28579         (vect_grouped_load_supported): Likewise.
28580         (vect_permute_load_chain): Likewise.
28582 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28584         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
28585         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
28586         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
28587         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28588         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
28589         (vect_gen_perm_mask_any): Likewise.
28591 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28593         * int-vector-builder.h: New file.
28594         * vec-perm-indices.h: Include int-vector-builder.h.
28595         (vec_perm_indices): Redefine as an int_vector_builder.
28596         (auto_vec_perm_indices): Delete.
28597         (vec_perm_builder): Redefine as a stand-alone class.
28598         (vec_perm_indices::vec_perm_indices): New function.
28599         (vec_perm_indices::clamp): Likewise.
28600         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
28601         (vec_perm_indices::new_vector): New function.
28602         (vec_perm_indices::new_expanded_vector): Update for new
28603         vec_perm_indices class.
28604         (vec_perm_indices::rotate_inputs): New function.
28605         (vec_perm_indices::all_in_range_p): Operate directly on the
28606         encoded form, without computing elided elements.
28607         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
28608         encoding.  Update for new vec_perm_indices class.
28609         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
28610         the given vec_perm_builder.
28611         (expand_vec_perm_var): Update vec_perm_builder constructor.
28612         (expand_mult_highpart): Use vec_perm_builder instead of
28613         auto_vec_perm_indices.
28614         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
28615         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
28616         or double series encoding as appropriate.
28617         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
28618         vec_perm_indices instead of auto_vec_perm_indices.
28619         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28620         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28621         (vect_permute_store_chain): Likewise.
28622         (vect_grouped_load_supported): Likewise.
28623         (vect_permute_load_chain): Likewise.
28624         (vect_shift_permute_load_chain): Likewise.
28625         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28626         (vect_transform_slp_perm_load): Likewise.
28627         (vect_schedule_slp_instance): Likewise.
28628         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28629         (vectorizable_mask_load_store): Likewise.
28630         (vectorizable_bswap): Likewise.
28631         (vectorizable_store): Likewise.
28632         (vectorizable_load): Likewise.
28633         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
28634         vec_perm_indices instead of auto_vec_perm_indices.  Use
28635         tree_to_vec_perm_builder to read the vector from a tree.
28636         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
28637         vec_perm_builder instead of a vec_perm_indices.
28638         (have_whole_vector_shift): Use vec_perm_builder and
28639         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
28640         truncation to calc_vec_perm_mask_for_shift.
28641         (vect_create_epilog_for_reduction): Likewise.
28642         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
28643         from auto_vec_perm_indices to vec_perm_indices.
28644         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28645         instead of changing individual elements.
28646         (aarch64_vectorize_vec_perm_const): Use new_vector to install
28647         the vector in d.perm.
28648         * config/arm/arm.c (expand_vec_perm_d::perm): Change
28649         from auto_vec_perm_indices to vec_perm_indices.
28650         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28651         instead of changing individual elements.
28652         (arm_vectorize_vec_perm_const): Use new_vector to install
28653         the vector in d.perm.
28654         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
28655         Update vec_perm_builder constructor.
28656         (rs6000_expand_interleave): Likewise.
28657         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
28658         (rs6000_expand_interleave): Likewise.
28660 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28662         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
28663         to qimode could truncate the indices.
28664         * optabs.c (expand_vec_perm_var): Likewise.
28666 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28668         * Makefile.in (OBJS): Add vec-perm-indices.o.
28669         * vec-perm-indices.h: New file.
28670         * vec-perm-indices.c: Likewise.
28671         * target.h (vec_perm_indices): Replace with a forward class
28672         declaration.
28673         (auto_vec_perm_indices): Move to vec-perm-indices.h.
28674         * optabs.h: Include vec-perm-indices.h.
28675         (expand_vec_perm): Delete.
28676         (selector_fits_mode_p, expand_vec_perm_var): Declare.
28677         (expand_vec_perm_const): Declare.
28678         * target.def (vec_perm_const_ok): Replace with...
28679         (vec_perm_const): ...this new hook.
28680         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
28681         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
28682         * doc/tm.texi: Regenerate.
28683         * optabs.def (vec_perm_const): Delete.
28684         * doc/md.texi (vec_perm_const): Likewise.
28685         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
28686         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
28687         expand_vec_perm for constant permutation vectors.  Assert that
28688         the mode of variable permutation vectors is the integer equivalent
28689         of the mode that is being permuted.
28690         * optabs-query.h (selector_fits_mode_p): Declare.
28691         * optabs-query.c: Include vec-perm-indices.h.
28692         (selector_fits_mode_p): New function.
28693         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
28694         is defined, instead of checking whether the vec_perm_const_optab
28695         exists.  Use targetm.vectorize.vec_perm_const instead of
28696         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
28697         fit in the vector mode before using a variable permute.
28698         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
28699         vec_perm_indices instead of an rtx.
28700         (expand_vec_perm): Replace with...
28701         (expand_vec_perm_const): ...this new function.  Take the selector
28702         as a vec_perm_indices rather than an rtx.  Also take the mode of
28703         the selector.  Update call to shift_amt_for_vec_perm_mask.
28704         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
28705         Use vec_perm_indices::new_expanded_vector to expand the original
28706         selector into bytes.  Check whether the indices fit in the vector
28707         mode before using a variable permute.
28708         (expand_vec_perm_var): Make global.
28709         (expand_mult_highpart): Use expand_vec_perm_const.
28710         * fold-const.c: Includes vec-perm-indices.h.
28711         * tree-ssa-forwprop.c: Likewise.
28712         * tree-vect-data-refs.c: Likewise.
28713         * tree-vect-generic.c: Likewise.
28714         * tree-vect-loop.c: Likewise.
28715         * tree-vect-slp.c: Likewise.
28716         * tree-vect-stmts.c: Likewise.
28717         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
28718         Delete.
28719         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
28720         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
28721         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
28722         (aarch64_vectorize_vec_perm_const): ...this new function.
28723         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28724         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28725         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
28726         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
28727         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28728         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28729         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
28730         into...
28731         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
28732         check for NEON modes.
28733         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
28734         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
28735         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
28736         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
28737         into...
28738         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
28739         the old VEC_PERM_CONST conditions.
28740         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
28741         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
28742         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
28743         (ia64_vectorize_vec_perm_const_ok): Merge into...
28744         (ia64_vectorize_vec_perm_const): ...this new function.
28745         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
28746         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
28747         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
28748         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
28749         * config/mips/mips.c (mips_expand_vec_perm_const)
28750         (mips_vectorize_vec_perm_const_ok): Merge into...
28751         (mips_vectorize_vec_perm_const): ...this new function.
28752         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
28753         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
28754         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
28755         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
28756         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
28757         (rs6000_expand_vec_perm_const): Delete.
28758         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
28759         Delete.
28760         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28761         (altivec_expand_vec_perm_const_le): Take each operand individually.
28762         Operate on constant selectors rather than rtxes.
28763         (altivec_expand_vec_perm_const): Likewise.  Update call to
28764         altivec_expand_vec_perm_const_le.
28765         (rs6000_expand_vec_perm_const): Delete.
28766         (rs6000_vectorize_vec_perm_const_ok): Delete.
28767         (rs6000_vectorize_vec_perm_const): New function.
28768         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28769         an element count and rtx array.
28770         (rs6000_expand_extract_even): Update call accordingly.
28771         (rs6000_expand_interleave): Likewise.
28772         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
28773         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
28774         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
28775         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
28776         (rs6000_expand_vec_perm_const): Delete.
28777         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28778         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28779         (altivec_expand_vec_perm_const_le): Take each operand individually.
28780         Operate on constant selectors rather than rtxes.
28781         (altivec_expand_vec_perm_const): Likewise.  Update call to
28782         altivec_expand_vec_perm_const_le.
28783         (rs6000_expand_vec_perm_const): Delete.
28784         (rs6000_vectorize_vec_perm_const_ok): Delete.
28785         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
28786         reference to the SPE evmerge intructions.
28787         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28788         an element count and rtx array.
28789         (rs6000_expand_extract_even): Update call accordingly.
28790         (rs6000_expand_interleave): Likewise.
28791         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
28792         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
28793         new function.
28794         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28796 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28798         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
28799         vector mode and that that mode matches the mode of the data
28800         being permuted.
28801         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
28802         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
28803         directly using expand_vec_perm_1 when forcing selectors into
28804         registers.
28805         (expand_vec_perm_var): New function, split out from expand_vec_perm.
28807 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28809         * optabs-query.h (can_vec_perm_p): Delete.
28810         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28811         * optabs-query.c (can_vec_perm_p): Split into...
28812         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28813         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28814         particular selector is valid.
28815         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28816         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28817         (vect_grouped_load_supported): Likewise.
28818         (vect_shift_permute_load_chain): Likewise.
28819         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28820         (vect_transform_slp_perm_load): Likewise.
28821         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28822         (vectorizable_bswap): Likewise.
28823         (vect_gen_perm_mask_checked): Likewise.
28824         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
28825         implementations of variable permutation vectors into account
28826         when deciding which selector to use.
28827         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28828         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28829         with a false third argument.
28830         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28831         to test whether the constant selector is valid and can_vec_perm_var_p
28832         to test whether a variable selector is valid.
28834 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28836         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28837         * optabs-query.c (can_vec_perm_p): Likewise.
28838         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28839         instead of vec_perm_indices.
28840         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28841         (vect_gen_perm_mask_checked): Likewise,
28842         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28843         (vect_gen_perm_mask_checked): Likewise,
28845 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
28847         * optabs-query.h (qimode_for_vec_perm): Declare.
28848         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28849         (qimode_for_vec_perm): ...this new function.
28850         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28852 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28854         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28855         does not have a conditional at the top.
28857 2018-01-02  Richard Biener  <rguenther@suse.de>
28859         * ipa-inline.c (big_speedup_p): Fix expression.
28861 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28863         PR target/81616
28864         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28865         for generic 4->6.
28867 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
28869         PR target/81616
28870         Generic tuning.
28871         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28872         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28873         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28874         cond_taken_branch_cost 3->4.
28876 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
28878         PR tree-optimization/83581
28879         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28880         TODO_cleanup_cfg if any changes have been made.
28882         PR middle-end/83608
28883         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28884         convert_modes if target mode has the right side, but different mode
28885         class.
28887         PR middle-end/83609
28888         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28889         last argument when extracting from CONCAT.  If either from_real or
28890         from_imag is NULL, use expansion through memory.  If result is not
28891         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28892         the parts directly to inner mode, if even that fails, use expansion
28893         through memory.
28895         PR middle-end/83623
28896         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
28897         check for bswap in mode rather than HImode and use that in expand_unop
28898         too.
28900 Copyright (C) 2018 Free Software Foundation, Inc.
28902 Copying and distribution of this file, with or without modification,
28903 are permitted in any medium without royalty provided the copyright
28904 notice and this notice are preserved.