1 2018-10-22 Martin Liska <mliska@suse.cz>
3 PR tree-optimization/87686
5 2018-08-29 Martin Liska <mliska@suse.cz>
7 * tree-switch-conversion.c (switch_conversion::expand):
8 Strenghten assumption about gswitch statements.
10 2018-10-22 Martin Liska <mliska@suse.cz>
12 * ipa-icf.c (sem_item::compare_attributes): Remove.
13 (sem_item::compare_referenced_symbol_properties): Use
14 attribute_list_equal instead.
15 (sem_function::equals_wpa): Likewise.
16 * ipa-icf.h: Remove compare_attributes.
18 2018-10-22 Richard Biener <rguenther@suse.de>
21 * mem-stats.h (mem_usage::operator==): Fix pasto.
23 2018-10-22 Richard Biener <rguenther@suse.de>
25 PR tree-optimization/87640
26 * tree-vrp.c (set_value_range_with_overflow): Decompose
28 (extract_range_from_binary_expr_1): Adjust.
30 2018-10-22 Martin Jambor <mjambor@suse.cz>
32 * tree-eh.h (stmt_could_throw_p): Add function parameter.
33 (stmt_can_throw_external): Likewise.
34 (stmt_can_throw_internal): Likewise.
35 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
36 (lower_eh_constructs_2): Likewise.
37 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
38 (stmt_can_throw_external): Likewise.
39 (stmt_can_throw_internal): Likewise.
40 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
41 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
42 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
43 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
44 (pass_lower_eh_dispatch::execute): Pass cfun to
45 stmt_can_throw_external.
46 (cleanup_empty_eh): Likewise.
47 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
48 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
49 stmt_can_throw_external instead of pushing it to cfun.
50 (symbol_table::create_edge): Likewise.
51 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
52 stmt_can_throw_internal.
53 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
54 to stmt_could_throw_p.
55 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
56 stmt_can_throw_internal.
57 (pass_store_merging::execute): Likewise.
58 * gimple-ssa-strength-reduction.c
59 (find_candidates_dom_walker::before_dom_children): Pass cfun to
61 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
62 stmt_can_throw_internal.
63 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
64 to stmt_can_throw_external.
65 (check_stmt): Pass cfun to stmt_could_throw_p.
66 (check_stmt): Pass cfun to stmt_can_throw_external.
67 (pass_nothrow::execute): Likewise.
68 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
69 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
70 stmt_can_throw_internal.
71 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
72 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
73 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
74 * tree-complex.c (expand_complex_libcall): Pass cfun to
75 stmt_could_throw_p and to stmt_can_throw_internal.
76 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
77 * tree-inline.c (copy_edges_for_bb): Likewise.
78 (maybe_move_debug_stmts_to_successors): Likewise.
79 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
81 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
82 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
83 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
84 stmt_can_throw_internal.
85 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
86 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
88 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
89 stmt_can_throw_internal.
90 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
91 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
93 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
94 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
95 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
96 (convert_mult_to_fma_1): Likewise.
97 (convert_to_divmod): Likewise.
98 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
99 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
100 * tree-ssa-propagate.c
101 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
102 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
103 (maybe_optimize_range_tests): Likewise.
104 (linearize_expr_tree): Likewise.
105 (reassociate_bb): Likewise.
106 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
107 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
108 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
109 (handle_char_store): Likewise.
110 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
111 stmt_can_throw_internal.
112 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
114 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
115 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
116 (vectorizable_simd_clone_call): Likewise.
117 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
118 (gimple_stringop_fixed_value): Likewise.
120 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
122 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
123 literal pool references.
124 (s390_check_qrst_address): Adapt to the new behavior of
125 s390_loadrelative_operand_p ().
127 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
130 * config/i386/sse.md (*andnot<mode>3_bcst): New.
132 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
135 * config/i386/sse.md (*<code><mode>3_bcst): New.
137 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
140 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
141 V4DI, V16SI and V8DI.
142 (*sub<mode>3<mask_name>_bcst): New.
143 (*add<mode>3<mask_name>_bcst): Likewise.
145 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
148 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
149 __builtin_ia32_vfnmsubpd512_mask.
150 (_mm512_mask_fnmsub_round_pd): Likewise.
151 (_mm512_fnmsub_pd): Likewise.
152 (_mm512_mask_fnmsub_pd): Likewise.
153 (_mm512_maskz_fnmsub_round_pd): Use
154 __builtin_ia32_vfnmsubpd512_maskz.
155 (_mm512_maskz_fnmsub_pd): Likewise.
156 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
157 (_mm512_mask_fnmsub_round_ps): Likewise.
158 (_mm512_fnmsub_ps): Likewise.
159 (_mm512_mask_fnmsub_ps): Likewise.
160 (_mm512_maskz_fnmsub_round_ps): Use
161 __builtin_ia32_vfnmsubps512_maskz.
162 (_mm512_maskz_fnmsub_ps): Likewise.
163 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
164 __builtin_ia32_vfnmsubpd256_mask.
165 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
166 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
167 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
168 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
169 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
170 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
171 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
172 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
173 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
174 __builtin_ia32_vfnmsubpd.
175 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
176 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
177 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
178 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
179 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
180 * config/i386/i386-builtin.def: Add
181 __builtin_ia32_vfnmsubpd256_mask,
182 __builtin_ia32_vfnmsubpd256_maskz,
183 __builtin_ia32_vfnmsubpd128_mask,
184 __builtin_ia32_vfnmsubpd128_maskz,
185 __builtin_ia32_vfnmsubps256_mask,
186 __builtin_ia32_vfnmsubps256_maskz,
187 __builtin_ia32_vfnmsubps128_mask,
188 __builtin_ia32_vfnmsubps128_maskz,
189 __builtin_ia32_vfnmsubpd512_mask,
190 __builtin_ia32_vfnmsubpd512_maskz,
191 __builtin_ia32_vfnmsubps512_mask,
192 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
193 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
194 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
195 __builtin_ia32_vfnmsubpd256.
196 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
197 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
198 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
200 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
202 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
204 (fmai_vmfnmsub_<mode><round_name>): Likewise.
206 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
209 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
210 __builtin_ia32_vfnmaddpd512_mask.
211 (_mm512_mask_fnmadd_round_pd): Likewise.
212 (_mm512_fnmadd_pd): Likewise.
213 (_mm512_mask_fnmadd_pd): Likewise.
214 (_mm512_maskz_fnmadd_round_pd): Use
215 __builtin_ia32_vfnmaddpd512_maskz.
216 (_mm512_maskz_fnmadd_pd): Likewise.
217 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
218 (_mm512_mask_fnmadd_round_ps): Likewise.
219 (_mm512_fnmadd_ps): Likewise.
220 (_mm512_mask_fnmadd_ps): Likewise.
221 (_mm512_maskz_fnmadd_round_ps): Use
222 __builtin_ia32_vfnmaddps512_maskz.
223 (_mm512_maskz_fnmadd_ps): Likewise.
224 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
225 __builtin_ia32_vfnmaddpd256_mask.
226 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
227 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
228 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
229 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
230 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
231 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
232 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
233 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
234 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
235 __builtin_ia32_vfnmaddpd.
236 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
237 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
238 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
239 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
240 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
241 * config/i386/i386-builtin.def: Add
242 __builtin_ia32_vfnmaddpd256_mask,
243 __builtin_ia32_vfnmaddpd256_maskz,
244 __builtin_ia32_vfnmaddpd128_mask,
245 __builtin_ia32_vfnmaddpd128_maskz,
246 __builtin_ia32_vfnmaddps256_mask,
247 __builtin_ia32_vfnmaddps256_maskz,
248 __builtin_ia32_vfnmaddps128_mask,
249 __builtin_ia32_vfnmaddps128_maskz,
250 __builtin_ia32_vfnmaddpd512_mask,
251 __builtin_ia32_vfnmaddpd512_maskz,
252 __builtin_ia32_vfnmaddps512_mask,
253 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
254 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
255 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
256 __builtin_ia32_vfnmaddpd256.
257 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
258 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
259 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
261 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
263 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
265 (fmai_vmfnmadd_<mode><round_name>): Likewise.
267 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
270 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
271 __builtin_ia32_vfmsubpd512_mask.
272 (_mm512_mask_fmsub_round_pd): Likewise.
273 (_mm512_fmsub_pd): Likewise.
274 (_mm512_mask_fmsub_pd): Likewise.
275 (_mm512_maskz_fmsub_round_pd): Use
276 __builtin_ia32_vfmsubpd512_maskz.
277 (_mm512_maskz_fmsub_pd): Likewise.
278 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
279 (_mm512_mask_fmsub_round_ps): Likewise.
280 (_mm512_fmsub_ps): Likewise.
281 (_mm512_mask_fmsub_ps): Likewise.
282 (_mm512_maskz_fmsub_round_ps): Use
283 __builtin_ia32_vfmsubps512_maskz.
284 (_mm512_maskz_fmsub_ps): Likewise.
285 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
286 __builtin_ia32_vfmsubpd256_mask.
287 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
288 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
289 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
290 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
291 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
292 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
293 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
294 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
295 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
296 __builtin_ia32_vfmsubpd.
297 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
298 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
299 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
300 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
301 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
302 * config/i386/i386-builtin.def: Add
303 __builtin_ia32_vfmsubpd256_mask,
304 __builtin_ia32_vfmsubpd256_maskz,
305 __builtin_ia32_vfmsubpd128_mask,
306 __builtin_ia32_vfmsubpd128_maskz,
307 __builtin_ia32_vfmsubps256_mask,
308 __builtin_ia32_vfmsubps256_maskz,
309 __builtin_ia32_vfmsubps128_mask,
310 __builtin_ia32_vfmsubps128_maskz,
311 __builtin_ia32_vfmsubpd512_mask,
312 __builtin_ia32_vfmsubpd512_maskz,
313 __builtin_ia32_vfmsubps512_mask,
314 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
315 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
316 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
317 __builtin_ia32_vfmsubpd256.
318 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
319 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
320 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
322 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
324 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
326 (fmai_vmfmsub_<mode><round_name>): Likewise.
328 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
330 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
331 Remove plus. Renamed to ...
332 (*sub<mode>3<mask_name>_bcst): This.
333 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
334 (*add<mode>3<mask_name>_bcst): This.
336 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
339 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
341 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
344 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
345 (_mm_or_epi32): Likewise.
346 (_mm256_xor_epi32): Likewise.
347 (_mm_xor_epi32): Likewise.
348 (_mm256_or_epi64): Likewise.
349 (_mm_or_epi64): Likewise.
350 (_mm256_xor_epi64): Likewise.
351 (_mm_xor_epi64): Likewise.
353 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
356 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
358 2018-10-20 Jakub Jelinek <jakub@redhat.com>
361 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
363 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
365 * doc/ux.texi: Move @section directly after @node.
367 2018-10-19 Jakub Jelinek <jakub@redhat.com>
371 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
372 depend closely nested inside of loop with ordered clause with
375 2018-10-19 David Malcolm <dmalcolm@redhat.com>
377 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
378 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
379 * doc/ux.texi: New file.
381 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
383 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
384 be the first CR field allocated.
386 2018-10-19 Richard Biener <rguenther@suse.de>
389 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
390 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
392 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
396 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
397 (*add<mode>3<mask_name>_bcst_2): Likewise.
399 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
402 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
403 Replace nonimmediate_operand with register_operand.
404 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
406 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
409 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
411 PR rtl-optimization/87596
412 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
413 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
414 for instructions in FROM..TO range.
416 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
418 * cfgexpand.c (expand_one_var): Use specific wording in error message
419 for non-local frame variables.
420 * stor-layout.c (layout_decl): Do not issue a warning for them.
422 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
424 * haifa-sched.c (priority): Add force_recompute parameter.
425 (apply_replacement): Call priority () with force_recompute = true.
426 (restore_pattern): Likewise.
428 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
430 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
431 HOST_BITS_PER_WIDE_INT.
432 (test_vector_ops_duplicate): Likewise.
434 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
437 * config/i386/sse.md (VF_AVX512): New.
438 (avx512bcst): Likewise.
439 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
441 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
443 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
446 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
448 * doc/invoke.texi (-dumpversion): Improve grammar.
449 (-dumpfullversion): Make more consistent with -dumpversion.
451 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
453 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
454 Set JUMP_LABEL to the jump insn.
455 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
456 Predict emitted jump and add label to jump insn.
458 2018-10-18 David Malcolm <dmalcolm@redhat.com>
460 PR tree-optimization/87562
461 * input.c (get_substring_ranges_for_loc): Use
462 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
463 getting the linemap for the endpoint. Verify that it's either
464 in the same linemap as the start point's spelling location, or
465 at least in the same file.
467 2018-10-18 Richard Biener <rguenther@suse.de>
469 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
470 feed width-specific load/store costs through ix86_vec_cost.
471 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
473 (bdver_cost): Likewise.
474 (znver1_cost): Likewise.
475 (btver1_cost): Likewise.
476 (btver2_cost): Likewise.
478 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
480 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
481 to simplify subreg of vec_merge.
483 2018-10-18 Richard Biener <rguenther@suse.de>
485 * config/i386/i386.c: Fix costing of vector FMA.
487 2018-10-18 Richard Biener <rguenther@suse.de>
489 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
491 (ix86_builtin_vectorization_cost): For vec_construct properly
492 cost insertion into SSE regs.
493 (...): Adjust calls to ix86_vec_cost.
495 2018-10-18 Richard Biener <rguenther@suse.de>
499 2018-02-07 Richard Biener <rguenther@suse.de>
501 PR tree-optimization/84204
502 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
505 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
508 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
510 (test_vector_ops_duplicate): Add test for a scalar subreg of a
511 VEC_MERGE of a VEC_DUPLICATE.
513 2018-10-17 Joseph Myers <joseph@codesourcery.com>
515 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
516 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
517 * doc/standards.texi (C Language): Document C2X.
518 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
519 (rl78_option_override): Handle "GNU C2X" language name.
521 2018-10-17 Joseph Myers <joseph@codesourcery.com>
523 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
524 Document C17 as published in 2018.
526 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
529 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
530 bail out if both sides do not have the same storage order.
532 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
534 * bitmap.c (bitmap_head::dump): New.
535 * bitmap.h (bitmap_head): Add dump().
536 * gimple-ssa-evrp-analyze.c
537 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
538 (evrp_range_analyzer::set_ssa_range_info): Same.
539 (evrp_range_analyzer::record_ranges_from_phis): Same.
540 (evrp_range_analyzer::record_ranges_from_stmt): Same.
541 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
542 * gimple-ssa-sprintf.c (get_int_range): Same.
543 (format_integer): Same.
544 (sprintf_dom_walker::handle_gimple_call): Same.
545 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
546 (ipcp_vr_lattice::top_p): Same.
547 (ipcp_vr_lattice::bottom_p): Same.
548 (ipcp_vr_lattice::set_to_bottom): Same.
549 (ipa_vr_operation_and_type_effects): Same.
550 (propagate_vr_across_jump_function): Same.
551 (ipcp_store_vr_results): Same.
552 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
553 (ipa_print_node_jump_functions_for_edge): Same.
554 (ipa_get_value_range): Same.
555 (ipa_compute_jump_functions_for_edge): Same.
556 (ipa_write_jump_function): Same.
557 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
558 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
560 * vr-values.c (set_value_range_to_nonnegative): Same.
561 (set_value_range_to_truthvalue): Same.
562 (vr_values::get_value_range): Same.
563 (vr_values::set_defs_to_varying): Same.
564 (vr_values::update_value_range): Same.
565 (symbolic_range_based_on_p): Same.
566 (vr_values::op_with_boolean_value_range_p): Same.
567 (vr_values::extract_range_for_var_from_comparison_expr): Same.
568 (vr_values::extract_range_from_ssa_name): Same.
569 (vr_values::extract_range_from_binary_expr): Same.
570 (vr_values::extract_range_from_unary_expr): Same.
571 (vr_values::extract_range_from_cond_expr): Same.
572 (vr_values::extract_range_from_comparison): Same.
573 (vr_values::check_for_binary_op_overflow): Same.
574 (vr_values::extract_range_basic): Same.
575 (vr_values::extract_range_from_assignment): Same.
576 (compare_ranges): Same.
577 (compare_range_with_value): Same.
578 (vr_values::adjust_range_with_scev): Same.
579 (vrp_valueize): Same.
580 (vrp_valueize_1): Same.
581 (vr_values::get_vr_for_comparison): Same.
582 (vr_values::compare_name_with_value): Same.
583 (vr_values::compare_names): Same.
584 (vr_values::vrp_evaluate_conditional): Same.
585 (find_case_label_ranges): Same.
586 (vr_values::vrp_visit_switch_stmt): Same.
587 (vr_values::extract_range_from_phi_node): Same.
588 (vr_values::simplify_div_or_mod_using_ranges): Same.
589 (vr_values::simplify_bit_ops_using_ranges): Same.
590 (test_for_singularity): Same.
591 (range_fits_type_p): Same.
592 (vr_values::simplify_cond_using_ranges_1): Same.
593 (vr_values::simplify_switch_using_ranges): Same.
594 (vr_values::simplify_float_conversion_using_ranges): Same.
595 (vr_values::two_valued_val_range_p): Same.
596 (vr_values::add_equivalence): Move to value_range::equiv_add.
597 * vr-values.h (vr_values::add_equivalence): Remove.
598 (VR_INITIALIZER): Remove.
599 * tree-vrp.c (value_range::set): New.
600 (value_range::equiv_add): New.
601 (value_range::value_range): New.
602 (value_range::deep_copy): New.
603 (value_range::check): New.
604 (value_range::equal_p): New.
605 (value_range::ignore_equivs_equal_p): New.
606 (value_range::operator==): New.
607 (value_range::operator!=): New.
608 (value_range::symbolic_p): New.
609 (value_range::numeric_p): New.
610 (value_range::set_undefined): New.
611 (value_range::set_varying): New.
612 (value_range::may_contain_p): New.
613 (value_range::equiv_clear): New.
614 (value_range::singleton_p): New.
615 (value_range::intersect): New.
616 (value_range::dump): New.
617 (value_range::set_and_canonicalize): New.
618 (set_value_range): Adjust for value_range API.
619 (set_value_range_to_undefined): Same.
620 (set_value_range_to_varying): Same.
621 (set_and_canonicalize_value_range): Same.
622 (set_value_range_to_nonnull): Same.
623 (set_value_range_to_null): Same.
624 (range_is_null): Same.
625 (range_is_nonnull): Same.
626 (range_int_cst_p): Same.
627 (range_int_cst_singleton_p): Same.
628 (symbolic_range_p): Same.
629 (range_includes_zero_p): Same.
630 (value_range_constant_singleton): Same.
631 (vrp_set_zero_nonzero_bits): Same.
632 (ranges_from_anti_range): Same.
633 (extract_range_into_wide_ints): Same.
634 (extract_range_from_multiplicative_op): Same.
635 (set_value_range_with_overflow): Same.
636 (extract_range_from_binary_expr_1): Same.
637 (extract_range_from_unary_expr): Same.
638 (dump_value_range): Same.
639 (debug_value_range): Same.
640 (vrp_prop::check_array_ref): Same.
641 (vrp_prop::check_mem_ref): Same.
642 (vrp_prop::vrp_initialize): Same.
643 (vrp_prop::visit_stmt): Same.
644 (intersect_ranges): Same.
645 (vrp_prop::visit_phi): Same.
646 (vrp_prop::vrp_finalize): Same.
647 (determine_value_range_1): Same.
648 (determine_value_range): Same.
649 (vrp_intersect_ranges_1): Rename to...
650 (vrp_intersect_1): this.
651 (vrp_intersect_ranges): Rename to...
652 (value_range::intersect_helper): ...this.
653 (vrp_meet_1): Rename to...
654 (value_range::union_helper): ...this.
655 (vrp_meet): Rename to...
656 (value_range::union_): ...this.
657 (copy_value_range): Remove.
658 * tree-vrp.h (struct value_range): Rewrite into a proper class.
659 (value_range::vrtype): New.
660 (value_range::type): New.
661 (value_range::equiv): New.
662 (value_range::min): New.
663 (value_range::max): New.
664 (value_range::varying_p): New.
665 (value_range::undefined_p): New.
666 (value_range::null_p): New.
667 (value_range::equiv_add): New.
668 (copy_value_range): Remove.
670 2018-10-17 David Malcolm <dmalcolm@redhat.com>
672 * Makefile.in (SELFTEST_TARGETS): New.
673 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
674 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
675 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
677 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
678 (selftest-c++-gdb, selftest-c++-valgrind): Move to
680 * configure: Regenerate.
681 * configure.ac (selftest_languages): New.
683 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
685 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
686 overflow wraps argument.
687 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
688 wide_int_range_multiplicative_op.
689 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
690 overflow wraps argument.
691 (wide_int_range_multiplicative_op): Same.
692 (wide_int_range_lshift): Same.
693 (wide_int_range_div): Same.
694 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
695 (wide_int_range_lshift): Same.
696 (wide_int_range_div): Same.
698 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
700 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
701 use sign as argument.
702 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
703 wide_int_range_shift_undefined_p.
705 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
707 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
709 (@despeculate_copy<ALLI_TI:mode>): ... This.
710 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
713 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
715 * config.gcc: Obsolete *-*-solaris2.10*.
716 * doc/install.texi (Specific, *-*-solaris2*): Document it.
718 2018-10-12 Jeff Law <law@redhat.com>
720 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
721 reg + sym +- const_int addressing modes.
723 2018-10-15 David Malcolm <dmalcolm@redhat.com>
725 * common.opt (fdiagnostics-minimum-margin-width=): New option.
726 * diagnostic-show-locus.c (layout::layout): Apply the minimum
728 (layout::start_annotation_line): Only print up to 3 of the
729 margin character, to avoid touching the left-hand side.
730 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
731 minimum margin width, as set by test_diagnostic_context's ctor.
732 (selftest::test_fixit_insert_containing_newline): Likewise.
733 (selftest::test_fixit_insert_containing_newline_2): Likewise.
734 (selftest::test_line_numbers_multiline_range): Clear
736 * diagnostic.c (diagnostic_initialize): Initialize
738 * diagnostic.h (struct diagnostic_context): Add field
740 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
741 * opts.c (common_handle_option): Handle
742 OPT_fdiagnostics_minimum_margin_width_.
743 * selftest-diagnostic.c
744 (selftest::test_diagnostic_context::test_diagnostic_context):
745 Initialize min_margin_width to 6.
746 * toplev.c (general_init): Initialize global_dc->min_margin_width.
748 2018-10-15 David Malcolm <dmalcolm@redhat.com>
750 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
751 Fix usage of "error_at_rich_loc" in the comment.
753 2018-10-15 Renlin Li <renlin.li@arm.com>
756 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
757 if-conversioned loop when it contains ifn with types not
758 supported by backend.
759 * internal-fn.c (expand_direct_optab_fn): Add an assert.
760 (direct_internal_fn_supported_p): New helper function.
761 * internal-fn.h (direct_internal_fn_supported_p): Declare.
763 2018-10-15 Jakub Jelinek <jakub@redhat.com>
766 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
767 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
768 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
770 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
772 PR tree-optimization/87022
773 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
774 bits in dist vector rather than the first one.
776 2018-10-15 Richard Biener <rguenther@suse.de>
779 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
780 (visit_loadstore): When a used restrict tag escaped verify that
781 the points-to solution of "other" pointers do not include
783 (compute_dependence_clique): If a used restrict tag escaped
784 communicated that down to visit_loadstore.
786 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
788 * config/s390/s390.c (s390_expand_vec_init): Force vector element
789 into reg if it isn't a general operand.
791 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
794 * config/i386/sse.md (*vec_dupv2di): Add register source to
797 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
800 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
801 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
802 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
803 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
805 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
807 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
808 (notice_args_size): Set it in the current trace if no insn that can
809 throw internally has been seen yet.
810 (connect_traces): When connecting args_size between traces, allow the
811 incoming values not to match if there is an insn setting it before the
812 first insn that can throw internally; in that case, force the creation
813 of a CFI note on this latter insn.
815 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
817 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
818 the base class, because using the injected-class-name was not clearly
819 specified until DR 176.
821 2018-10-12 Paul Koning <ni1d@arrl.net>
823 * config/pdp11/pdp11.md (doloop_end): New expander.
824 (doloop_end_insn): renamed from "doloop_end".
825 (addqi3): New pattern.
826 (subqi3): New pattern.
827 * config/pdp11/predicates.md (incdec_operand): New predicate.
829 2018-10-12 Yury Gribov <tetra2005@gmail.com>
832 * real.c (format_helper::can_represent_integral_type_p): New function
833 * real.h (format_helper::can_represent_integral_type_p): Ditto.
834 * match.pd: New pattern.
836 2018-10-12 Alexandre Oliva <oliva@adacore.com>
838 * configure.ac: Introduce --enable-large-address-aware
839 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
840 * doc/install.texi: Document it.
841 * configure, config.in: Rebuilt.
842 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
843 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
844 (LINK_SPEC): Insert it.
845 * config/i386/mingw-264.h: Likewise.
847 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
849 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
851 PR rtl-optimization/87600
852 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
854 2018-10-12 Paul Koning <ni1d@arrl.net>
856 * doc/md.texi (doloop_end): Document that the pattern code may
857 need to check operand mode.
859 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
861 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
862 to zero-extend between int and floating-point registers.
863 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
864 ldp into floating-point registers. Add type and arch attributes.
865 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
866 Use f_loads for type attribute.
868 2018-10-11 Martin Sebor <msebor@redhat.com>
870 * doc/extend.texi (attribute packed): Correct typos.
872 2018-10-11 Martin Sebor <msebor@redhat.com>
874 * doc/extend.texi (attribute flatten): Mention interaction with
877 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
880 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
881 Set new_decl virtual flag to zero.
883 2018-10-11 Martin Sebor <msebor@redhat.com>
886 * doc/extend.texi (attribute format_arg): Discuss using multiple
887 attributes on a single function.
889 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
891 PR tree-optimization/86829
892 * match.pd (sin (atan (x))): New simplification rules.
893 (cos (atan (x))): Likewise.
894 * real.c (build_sinatan_real): New function.
895 * real.h (build_sinatan_real): Prototype.
897 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
899 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
901 (fold_mergeeo_helper): New helper function.
902 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
903 intrinsics. Correct some whitespace indentation issues.
905 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
908 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
909 Use HOST_WIDE_INT_1U for shift.
911 2018-10-11 Doug Rupp <rupp@adacore.com>
912 Olivier Hainque <hainque@adacore.com>
914 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
915 Pass --relax to the linker for RTPs.
916 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
918 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
919 Jan Hubicka <jh@suse.cz>
920 Martin Jambor <mjambor@suse.cz>
922 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
923 the same elements are repeated rather than printing all of them.
924 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
926 * read-rtl-function.c (test_loading_repeat): New function.
927 (read_rtl_function_c_tests): Call test_loading_repeat.
928 * rtl-tests.c (test_dumping_repeat): New function.
929 (rtl_tests_c_tests): Call test_dumping_repeat.
931 2018-10-11 Richard Biener <rguenther@suse.de>
933 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
934 bdver?_cost): Unify to ...
935 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
936 * config/i386/i386.c (processor_cost_table): Adjust.
938 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
941 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
942 the thunk when expanding to GIMPLE.
944 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
946 * varasm.c (mergeable_string_section): Don't try to move zero-length
947 strings to the merge section.
949 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
952 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
954 2018-10-10 Jakub Jelinek <jakub@redhat.com>
957 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
960 2018-10-10 Richard Biener <rguenther@suse.de>
962 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
963 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
964 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
965 and recursing and pattern terminating the recursion on SSE
966 vector width using ix86_expand_reduc.
967 (reduc_sminmax_scal_<mode>): Split into part reducing to half
968 width and recursing and SSE2 vector variant doing the final
969 reduction with ix86_expand_reduc.
970 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
971 with terminating the recursion at AVX level, splitting that
974 2018-10-09 David Malcolm <dmalcolm@redhat.com>
976 * genmatch.c (error_cb): Rename to...
977 (diagnostic_cb): ...this, converting int params to enums.
978 (fatal_at): Update for renaming.
979 (warning_at): Likewise.
981 * input.c (selftest::ebcdic_execution_charset::apply):
982 Update for renaming of...
983 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
985 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
986 converting level and reason to enums.
987 (class selftest::lexer_error_sink): Rename to...
988 (class selftest::lexer_test_options): ...this, renaming field
989 "m_errors" to "m_diagnostics".
990 (selftest::lexer_test_options::apply): Update for renaming of...
991 (selftest::lexer_test_options::on_error): ...this, renaming to...
992 (selftest::lexer_test_options::on_diagnostic): ...this
993 converting level and reason to enums.
994 (selftest::test_lexer_string_locations_raw_string_unterminated):
995 Update for renamings.
996 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
999 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
1001 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1002 * config/rs6000/pmmintrin.h: New file.
1004 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
1006 PR tree-optimization/86659
1007 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1009 2018-10-09 Richard Biener <rguenther@suse.de>
1011 PR tree-optimization/63155
1012 * tree-ssa-structalias.c: Include tree-ssa.h.
1013 (get_constraint_for_ssa_var): For undefs return nothing_id.
1014 (find_func_aliases): Cleanup PHI handling.
1016 2018-10-09 Richard Biener <rguenther@suse.de>
1018 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1021 2018-10-09 Martin Liska <mliska@suse.cz>
1023 * asan.c (asan_emit_stack_protection): If a stack variable
1024 is located in a same file as current function, then emit
1025 line info into variable definition string.
1027 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
1029 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1032 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
1034 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1037 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
1039 PR tree-optimization/86659
1040 * gimple-match.h (struct gimple_match_op): Add reverse field.
1041 (gimple_match_op::set_op): New overloaded method.
1042 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1043 the REF_REVERSE_STORAGE_ORDER flag on the value.
1044 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
1045 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
1047 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
1050 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
1051 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
1053 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
1056 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
1057 Defined with __builtin_ia32_vfmaddsubpd512_mask.
1059 2018-10-08 Richard Biener <rguenther@suse.de>
1061 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
1062 cost the same as AVX128 ones.
1064 2018-10-08 Paul Koning <ni1d@arrl.net>
1066 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
1067 (expand_block_move): New function.
1068 * config/pdp11/pdp11.c (output_block_move): Remove.
1069 (expand_block_move): New function.
1070 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
1071 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
1072 (*movmemhi1): Remove.
1074 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
1076 * config/s390/2827.md: Increase latencies for some FP instructions.
1078 2018-10-08 Richard Biener <rguenther@suse.de>
1080 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1082 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
1083 * tree-vectorizer.h (dump_stmt_cost): Adjust.
1084 (add_stmt_cost): Dump return value of the hook.
1086 2018-10-08 Richard Biener <rguenther@suse.de>
1088 PR tree-optimization/63155
1089 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
1090 (ssa_propagation_engine::ssa_propagate): Remove redundant
1091 bitmap bit clearing.
1093 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
1095 PR rtl-optimization/86939
1096 PR rtl-optimization/87479
1097 * ira.h (non_conflicting_reg_copy_p): New prototype.
1098 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
1099 (make_hard_regno_dead): Don't add conflicts for register
1100 ignore_reg_for_conflicts.
1101 (make_object_dead): Likewise.
1102 (non_conflicting_reg_copy_p): New function.
1103 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
1104 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
1105 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
1106 (make_hard_regno_dead): Don't add conflicts for register
1107 ignore_reg_for_conflicts. Remove special conflict handling of
1108 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
1109 check_pic_pseudo_p and update callers.
1110 (mark_pseudo_dead): Don't add conflicts for register
1111 ignore_reg_for_conflicts.
1112 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
1114 2018-10-05 Andrew Waterman <andrew@sifive.com>
1115 Jim Wilson <jimw@sifive.com>
1117 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
1118 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
1119 new pattern using HONOR_SNANS that emits one extra instruction.
1121 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
1123 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
1124 patterns): Merge SI and DI patterns to a GPR pattern.
1125 (unnamed define_insn and define_split for record form of that): Merge
1126 to a single define_insn_and_split pattern.
1128 2018-10-05 David Malcolm <dmalcolm@redhat.com>
1131 * input.c (expand_location_to_spelling_point): Add param "aspect"
1132 and use rather than hardcoding LOCATION_ASPECT_CARET.
1133 (get_substring_ranges_for_loc): Handle the case of a single token
1134 within a macro expansion.
1135 * input.h (expand_location_to_spelling_point): Add "aspect" param,
1136 defaulting to LOCATION_ASPECT_CARET.
1138 2018-10-05 Paul Koning <ni1d@arrl.net>
1140 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
1141 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
1142 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1143 (pdp11_guard_type): New function.
1145 2018-10-05 Paul Koning <ni1d@arrl.net>
1147 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
1148 * config/pdp11/pdp11.opt (mfloat32): Remove.
1150 * doc/invoke.texi (pdp11 -mfloat32): Remove:
1151 (pdp11 -mfloat64): Remove.
1153 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
1155 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
1156 (*cmp<mode>_cc_i387): Ditto.
1157 (*cmpu<mode>_cc_i387): Ditto.
1158 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1159 * config/i386/i386.c (ix86_expand_fp_compare): Remove
1161 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
1162 Emit x86_sahf_1 pattern.
1163 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
1164 (ix86_expand_carry_flag_compare): Ditto.
1166 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
1168 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
1169 to reg_or_0_operand. Add "C" constraint.
1170 (*cmpxf_cc_i387): Ditto.
1171 (*cmp<mode>_i387): Change operand 2 predicate
1172 to nonimm_or_0_operand. Add "C" constraint.
1173 (*cmp<mode>_cc_i387): Ditto.
1174 (*cmp<mode>_0_i387): Remove insn pattern.
1175 (*cmp<mode>_0_cc_i387): Ditto.
1177 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
1179 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
1180 * config/i386/predicates.md (nonimm_or_0_operand): Rename
1181 from vector_move_operand. Update all uses.
1183 2018-10-05 Martin Sebor <msebor@redhat.com>
1185 PR tree-optimization/87490
1186 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
1189 2018-10-05 Richard Biener <rguenther@suse.de>
1191 PR tree-optimization/63155
1192 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
1193 vertical space in dumpfiles.
1194 * tree-ssa-propagate.h
1195 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
1196 * tree-ssa-propagate.c (cfg_blocks_back): New global.
1197 (ssa_edge_worklist_back): Likewise.
1198 (curr_order): Likewise.
1199 (cfg_blocks_get): Remove abstraction.
1200 (cfg_blocks_add): Likewise.
1201 (cfg_blocks_empty_p): Likewise.
1202 (add_ssa_edge): Add to current or next worklist based on
1204 (add_control_edge): Likewise.
1205 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
1207 (ssa_propagation_engine::ssa_propagate): ... here. Unify
1208 iteration from CFG and SSA edge worklist so we process
1209 everything in RPO order, prioritizing forward progress
1211 (ssa_prop_init): Allocate new worklists, do not dump
1213 (ssa_prop_fini): Free new worklists.
1215 2018-10-05 Richard Biener <rguenther@suse.de>
1217 * tree-core.h (tree_block::abstract_flag): Remove.
1218 (tree_block::block_num): Make full 32bits.
1219 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
1220 * tree.h (BLOCK_ABSTRACT): Remove.
1221 * dwarf2out.c (gen_lexical_block_die): Remove dead code
1222 resulting from BLOCK_ABSTRACT being always false.
1223 (gen_inlined_subroutine_die): Likewise.
1224 (gen_block_die): Likewise.
1225 * tree.c (block_ultimate_origin): Likewise.
1226 * tree-pretty-print.c (dump_block_node): Remove code dealing
1227 with BLOCK_ABSTRACT.
1228 * tree-ssa-live.c (dump_scope_block): Likewise.
1229 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
1230 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
1232 2018-10-05 Richard Biener <rguenther@suse.de>
1234 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
1235 is asked for initialize mode to the component mode of the
1238 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
1241 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
1242 assembler for -mavx.
1243 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
1245 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
1248 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
1250 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
1251 to be DImode. When using mffscrn, force the operand to a register.
1253 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
1255 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
1256 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
1257 X87MODEF mode iterator.
1258 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
1259 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
1260 X87MODEF mode iterator.
1262 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
1264 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
1265 -Wno-prio-ctor-dtor.
1267 2018-10-04 David Malcolm <dmalcolm@redhat.com>
1269 * Makefile.in (OBJS): Add opt-problem.o.
1270 * dump-context.h: Include "selftest.h.
1271 (selftest::temp_dump_context): New forward decl.
1272 (class dump_context): Make friend of class
1273 selftest::temp_dump_context.
1274 (dump_context::dump_loc_immediate): New decl.
1275 (class dump_pretty_printer): Move here from dumpfile.c.
1276 (class temp_dump_context): Move to namespace selftest.
1277 (temp_dump_context::temp_dump_context): Add param
1278 "forcibly_enable_dumping".
1279 (selftest::verify_dumped_text):
1280 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
1281 (selftest::verify_item):
1282 (ASSERT_IS_TEXT): Move here from dumpfile.c.
1283 (ASSERT_IS_TREE): Likewise.
1284 (ASSERT_IS_GIMPLE): Likewise.
1285 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
1287 (dump_context::dump_loc_immediate): ...this new function.
1288 (class dump_pretty_printer): Move to dump-context.h.
1289 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
1290 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
1291 (temp_dump_context::temp_dump_context): Move to "selftest"
1292 namespace. Add param "forcibly_enable_dumping", and use it to
1293 conditionalize the use of m_pp;
1294 (selftest::verify_dumped_text): Make non-static.
1295 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
1296 (selftest::verify_item): Make non-static.
1297 (ASSERT_IS_TEXT): Move to dump-context.h.
1298 (ASSERT_IS_TREE): Likewise.
1299 (ASSERT_IS_GIMPLE): Likewise.
1300 (selftest::test_capture_of_dump_calls): Pass "true" for new
1301 param of temp_dump_context.
1302 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
1303 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
1305 * opt-problem.cc: New file.
1306 * opt-problem.h: New file.
1307 * optinfo-emit-json.cc
1308 (selftest::test_building_json_from_dump_calls): Pass "true" for
1309 new param of temp_dump_context.
1310 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
1311 (optinfo::emit_for_opt_problem): New function.
1312 (optinfo::emit): Clarity which emit_item is used.
1313 * optinfo.h (optinfo::get_dump_location): New accessor.
1314 (optinfo::emit_for_opt_problem): New decl.
1315 (optinfo::emit): Make const.
1316 * selftest-run-tests.c (selftest::run_tests): Call
1317 selftest::opt_problem_cc_tests.
1318 * selftest.h (selftest::opt_problem_cc_tests): New decl.
1319 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
1320 bool to opt_result, converting fprintf messages to
1321 opt_result::failure_at calls. Add "stmt" param for use by the
1323 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
1324 (runtime_alias_check_p): Convert return type from bool to
1325 opt_result, converting dump_printf calls to
1326 opt_result::failure_at, using the statement DDR_A for their
1328 (find_data_references_in_stmt): Convert return type from bool to
1329 opt_result, converting "return false" to opt_result::failure_at
1331 * tree-data-ref.h: Include "opt-problem.h".
1332 (dr_analyze_innermost): Convert return type from bool to opt_result,
1333 and add a const gimple * param.
1334 (find_data_references_in_stmt): Convert return type from bool to
1336 (runtime_alias_check_p): Likewise.
1337 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
1338 dr_analyze_innermost.
1339 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
1340 Convert return type from bool to opt_result, adding a message for
1341 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
1342 (vect_analyze_data_ref_dependence): Convert return type from bool
1343 to opt_result. Change sense of return type from "false"
1344 effectively meaning "no problems" to "false" meaning a problem,
1345 so that "return false" becomes "return opt_result::success".
1346 Convert "return true" calls to opt_result::failure_at, using
1347 the location of statement A rather than vect_location.
1348 (vect_analyze_data_ref_dependences): Convert return type from bool
1350 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
1351 calls to opt_result::failure_at, using the stmt location rather
1353 (vect_verify_datarefs_alignment): Convert return type from bool
1355 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
1356 into multiple more-tightly-scoped copies.
1357 (vect_analyze_data_refs_alignment): Convert return type from bool
1359 (vect_analyze_data_ref_accesses): Likewise, converting a
1360 "return false" to a "return opt_result::failure_at", adding a
1362 (vect_prune_runtime_alias_test_list): Convert return type from
1363 bool to opt_result, converting dump_printf_loc to
1364 opt_result::failure_at. Add a %G to show the pertinent statement,
1365 and use the stmt's location rather than vect_location.
1366 (vect_find_stmt_data_reference): Convert return type from
1367 bool to opt_result, converting dump_printf_loc to
1368 opt_result::failure_at, using stmt's location.
1369 (vect_analyze_data_refs): Convert return type from bool to
1370 opt_result. Convert "return false" to "return
1371 opt_result::failure_at", adding messages as needed.
1372 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
1373 type from bool to opt_result.
1374 (vect_determine_vf_for_stmt): Likewise.
1375 (vect_determine_vectorization_factor): Likewise, converting
1376 dump_printf_loc to opt_result::failure_at, using location of phi
1377 rather than vect_location.
1378 (vect_analyze_loop_form_1): Convert return type from bool to
1379 opt_result, converting dump_printf_loc calls, retaining the use of
1381 (vect_analyze_loop_form): Convert return type from loop_vec_info
1382 to opt_loop_vec_info.
1383 (vect_analyze_loop_operations): Convert return type from bool to
1384 opt_result, converting dump_printf_loc calls, using the location
1385 of phi/stmt rather than vect_location where available. Convert
1386 various "return false" to "return opt_result::failure_at" with
1387 "unsupported phi" messages.
1388 (vect_get_datarefs_in_loop): Convert return type from bool to
1389 opt_result. Add a message for the
1390 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
1391 (vect_analyze_loop_2): Convert return type from bool to
1392 opt_result. Ensure "ok" is set to a opt_result::failure_at before
1393 each "goto again;", adding new messages where needed.
1394 Add "unsupported grouped {store|load}" messages.
1395 (vect_analyze_loop): Convert return type from loop_vec_info to
1397 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
1399 * tree-vect-stmts.c (process_use): Likewise, converting
1400 dump_printf_loc call and using stmt location, rather than
1402 (vect_mark_stmts_to_be_vectorized): Likeise.
1403 (vect_analyze_stmt): Likewise, adding a %G.
1404 (vect_get_vector_types_for_stmt): Convert return type from bool to
1405 opt_result, converting dump_printf_loc calls and using stmt
1406 location, rather than vect_location.
1407 (vect_get_mask_type_for_stmt): Convert return type from tree to
1408 opt_tree, converting dump_printf_loc calls and using stmt location.
1409 * tree-vectorizer.c: Include "opt-problem.h.
1410 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
1411 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
1412 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
1413 enabled, use it to report at the top level "couldn't vectorize
1414 loop" followed by the problem.
1415 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
1416 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
1418 (vect_analyze_stmt): Likewise.
1419 (vect_get_vector_types_for_stmt): Likewise.
1420 (tree vect_get_mask_type_for_stmt): Likewise.
1421 (vect_analyze_data_ref_dependences): Likewise.
1422 (vect_enhance_data_refs_alignment): Likewise.
1423 (vect_analyze_data_refs_alignment): Likewise.
1424 (vect_verify_datarefs_alignment): Likewise.
1425 (vect_analyze_data_ref_accesses): Likewise.
1426 (vect_prune_runtime_alias_test_list): Likewise.
1427 (vect_find_stmt_data_reference): Likewise.
1428 (vect_analyze_data_refs): Likewise.
1429 (vect_analyze_loop): Convert return type from loop_vec_info to
1431 (vect_analyze_loop_form): Likewise.
1432 (vect_analyze_slp): Convert return type from bool to opt_result.
1434 2018-10-04 David Malcolm <dmalcolm@redhat.com>
1436 * doc/invoke.texi (-fopt-info): Document new "internals"
1438 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
1439 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
1441 (optinfo_verbosity_options): Add "internals".
1442 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
1443 (dump_context::apply_dump_filter_p): New member function.
1444 (dump_context::dump_loc): Use apply_dump_filter_p rather than
1445 explicitly masking the dump_kind.
1446 (dump_context::begin_scope): Increment the scope depth first. Use
1447 apply_dump_filter_p rather than explicitly masking the dump_kind.
1448 (dump_context::emit_item): Use apply_dump_filter_p rather than
1449 explicitly masking the dump_kind.
1450 (dump_dec): Likewise.
1451 (dump_hex): Likewise.
1452 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
1453 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
1454 (opt_info_switch_p): Update handling of default
1455 MSG_OPTIMIZED_LOCATIONS to cope with default of
1456 MSG_PRIORITY_USER_FACING.
1457 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
1458 masking the dump_kind.
1459 (selftest::test_capture_of_dump_calls): Update test_dump_context
1460 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
1461 than MSG_ALL. Generalize scope test to be run at all four
1462 combinations of with/without MSG_PRIORITY_USER_FACING and
1463 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
1464 for each of the two values.
1465 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
1466 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
1467 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
1468 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
1469 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
1470 with MSG_PRIORITY_*.
1471 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
1472 dump messages as MSG_PRIORITY_USER_FACING.
1473 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
1474 about the interaction with MSG_PRIORITY_*.
1476 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1478 * varasm.c (output_constant): Add new parameter merge_strings.
1479 Make strings properly zero terminated in merge string sections.
1480 (mergeable_string_section): Don't fail if the last char is non-zero.
1481 (assemble_variable_contents): Handle merge string sections.
1482 (assemble_variable): Likewise.
1483 (assemble_constant_contents): Likewise.
1484 (output_constant_def_contents): Likewise.
1485 (output_constructor_array_range,
1486 output_constructor_regular_field): Adjust call to output_constant.
1487 (output_object_block): Adjust call to assemble_constant_contents
1488 and assemble_variable_contents.
1490 2018-10-04 Martin Liska <mliska@suse.cz>
1493 * cgraphunit.c (process_function_and_variable_attributes):
1494 Warn about a function with alias attribute and a body.
1496 2018-10-04 Martin Liska <mliska@suse.cz>
1499 * multiple_target.c (redirect_to_specific_clone): New function.
1500 (ipa_target_clone): Use it.
1501 * tree-inline.c: Fix comment.
1503 2018-10-04 David Malcolm <dmalcolm@redhat.com>
1505 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
1507 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
1508 (gcc::dump_manager::register_pass): New member function, adapted
1509 from loop body in gcc::pass_manager::register_pass, adding a
1510 call to update_dfi_for_opt_info.
1511 (gcc::dump_manager::opt_info_enable_passes): Store the
1512 -fopt-info options into the new fields. Move the loop
1514 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
1516 * dumpfile.h (struct opt_pass): New forward decl.
1517 (gcc::dump_manager::register_pass): New decl.
1518 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
1519 (class gcc::dump_manager): Add fields "m_optgroup_flags",
1520 "m_optinfo_flags", and "m_optinfo_filename".
1521 * passes.c (gcc::pass_manager::register_pass): Move all of the
1522 dump-handling code to gcc::dump_manager::register_pass.
1524 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
1526 PR rtl-optimization/87466
1527 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
1528 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
1529 * doc/tm.texi: Regenerate.
1530 * ira-lives.c (process_bb_node_lives): Use the new target hook.
1531 * lra-lives.c (process_bb_lives): Likewise.
1532 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
1535 2018-10-04 Tamar Christina <tamar.christina@arm.com>
1537 * params.c (add_params): Fix initialization.
1539 2018-10-04 Martin Liska <mliska@suse.cz>
1541 PR gcov-profile/84107
1542 * tree-profile.c (init_ic_make_global_vars):
1543 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
1544 Come up with new ic_tuple* variables. Emit
1545 __gcov_indirect_call{,_topn} variables.
1546 (gimple_gen_ic_profiler): Access the variable
1548 (gimple_gen_ic_func_profiler): Access
1549 __gcov_indirect_call.callee field.
1550 (gimple_init_gcov_profiler): Use ptr_type_node.
1551 * value-prof.c (gimple_ic): Use ptr_type_node.
1553 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1555 PR tree-optimization/85787
1556 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
1557 into this function and add support for detecting multiple phis.
1558 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
1560 2018-10-04 Martin Liska <mliska@suse.cz>
1563 * ipa-inline.c (inline_to_all_callers_1):
1564 Call ultimate_alias_target for node being inlined.
1566 2018-10-03 Jeff Law <law@redhat.com>
1568 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
1570 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
1571 * tree.h (get_typenode_from_name): Prototype.
1573 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
1575 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
1576 Change operand 2 predicate to nonimmediate_operand.
1577 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1579 2018-10-03 Martin Sebor <msebor@redhat.com>
1580 Jeff Law <law@redhat.com>
1582 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
1584 (get_string_length): Detect unterminated arrays.
1585 (format_string): Same.
1586 (format_directive): Warn about unterminated arrays.
1587 (handle_gimple_call): Mark statements with no_warning as needed.
1589 2018-10-03 Jim Wilson <jimw@sifive.com>
1591 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
1592 also define __riscv_abi_rve. Delete trailing white space.
1594 2018-10-03 Paul Koning <ni1d@arrl.net>
1596 Enable LRA register allocator for PDP11.
1597 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
1600 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
1601 * config/pdp11/pdp11.opt (-mlra): New option.
1602 * doc/invoke.texi (PDP-11 Options): Document -mlra.
1604 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
1606 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
1607 (*<absneg:code>extend<mode>xf2): Ditto.
1609 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
1611 PR tree-optimization/87415
1612 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
1615 2018-10-02 Jeff Law <law@redhat.com>
1617 * gimple-fold.c (get_range_strlen): Only set *nonstr when
1618 an unterminated string is discovered. Bubble up range
1619 even for unterminated strings.
1620 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
1621 indicates the string was not terminated via NONSTR.
1623 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
1625 * tree-vrp.c (extract_range_from_unary_expr): Special case all
1626 pointer conversions.
1627 Do not do anything special for anti-ranges.
1629 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
1631 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
1632 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
1634 2018-10-03 Martin Liska <mliska@suse.cz>
1636 PR gcov-profile/86109
1637 * coverage.c (coverage_begin_function): Do not
1638 mark lambdas as artificial.
1639 * tree-core.h (struct GTY): Remove tm_clone_flag
1640 and introduce new lambda_function.
1641 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
1643 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1646 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
1647 P8_VECTOR and VSX are enabled.
1649 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
1651 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
1652 0x3907 as CPU model number.
1654 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
1656 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
1657 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
1658 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
1659 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
1660 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
1661 * config/s390/s390.md: Likewise. Rename also the cpu attribute
1662 value from arch12 to z14.
1664 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
1666 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
1668 (isinf<mode>2): Ditto.
1670 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
1672 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
1673 before emitting fxam. Perform calculations in XFmode.
1675 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
1677 * match.pd (((X /[ex] A) +- B) * A): New transformation.
1679 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
1682 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
1683 * tree.c (signed_or_unsigned_type_for): Handle complex.
1685 2018-10-02 Jeff Law <law@redhat.com>
1687 * gimple-fold.c (get_range_strlen): Remove dead code.
1689 2018-10-02 Martin Sebor <msebor@redhat.com>
1690 Jeff Law <law@redhat.com>
1692 * builtins.c (unterminated_array): Add new arguments.
1693 If argument is not terminated, bubble up size and exact
1695 (expand_builtin_strnlen): Detect, avoid expanding
1696 and diagnose unterminated arrays.
1697 (c_strlen): Fill in offset of start of unterminated strings.
1698 * builtins.h (unterminated_array): Update prototype.
1700 2018-10-02 Richard Biener <rguenther@suse.de>
1702 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
1703 of haddv4df, first reduce to SSE width and exploit the fact
1704 that we only need element zero with the reduction result.
1705 (reduc_plus_scal_v2df): Likewise.
1707 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
1709 * dojump.h (do_jump): Delete.
1710 (do_jump_1): Likewise.
1711 (split_comparison): Move around.
1712 * dojump.c (do_jump): Make static.
1713 (do_jump_1): Likewise.
1714 (jumpifnot): Move around.
1715 (jumpifnot_1): Likewise.
1717 (jumpif_1): Likewise.
1718 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
1720 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
1722 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
1723 insns in the delay slot and add_insn_after for the jump insn.
1725 2018-10-02 Richard Biener <rguenther@suse.de>
1727 * tree-inline.c (expand_call_inline): Use the location of
1728 the callee declaration for the inline-entry marker.
1729 * final.c (notice_source_line): Remove special-casing of
1730 NOTE_INSN_INLINE_ENTRY.
1732 2018-10-01 Carl Love <cel@us.ibm.com>
1735 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
1736 (__builtin_mtfsb0): New.
1737 (__builtin_mtfsb1): New.
1738 ( __builtin_set_fpscr_rn): New.
1739 (__builtin_set_fpscr_drn): New.
1740 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
1741 (rs6000_expand_set_fpscr_rn_builtin): Add.
1742 (rs6000_expand_set_fpscr_drn_builtin): Add.
1743 (rs6000_expand_builtin): Add case statement entries for
1744 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
1745 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
1746 RS6000_BUILTIN_MFFSL.
1747 (rs6000_init_builtins): Add ftype initialization and def_builtin
1748 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1749 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1750 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1751 rs6000_mffscdrn): Add define_insn.
1752 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1753 * doc/extend.texi: Add documentation for the builtins.
1755 2018-10-01 Richard Biener <rguenther@suse.de>
1757 PR tree-optimization/87465
1758 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1759 causing branch miscounts.
1761 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1763 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1764 aarch64_option_default_param): New.
1765 (params.h): Include.
1766 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1767 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1768 stack-clash protection validation code.
1770 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1772 * params.c (validate_param): New.
1773 (add_params): Use it.
1774 (set_param_value): Refactor param validation into validate_param.
1775 (diagnostic.h): Include.
1776 * diagnostic.h (diagnostic_ready_p): New.
1778 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1780 * params.c (set_param_value):
1781 Add index of parameter being validated.
1782 * common/common-target.def (option_validate_param): New.
1783 * common/common-targhooks.h (default_option_validate_param): New.
1784 * common/common-targhooks.c (default_option_validate_param): New.
1785 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1786 * doc/tm.texi: Regenerate.
1788 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1791 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1792 Add validation for stack-clash parameters and set defaults.
1794 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1797 * configure.ac: Add stack-clash-protection-guard-size.
1798 * doc/install.texi: Document it.
1799 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1800 * params.def: Update comment for guard-size.
1801 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1802 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1803 * configure: Regenerate.
1805 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1808 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1809 STACK_DYNAMIC_OFFSET): New.
1810 * config/aarch64/aarch64.c (aarch64_layout_frame):
1811 Update outgoing args size.
1812 (aarch64_stack_clash_protection_alloca_probe_range,
1813 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1815 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1818 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1820 * target.def (stack_clash_protection_alloca_probe_range): New.
1821 (stack_clash_protection_final_dynamic_probe): Remove.
1822 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1823 (default_stack_clash_protection_final_dynamic_probe): Remove.
1824 * targhooks.c: Likewise.
1825 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1826 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1827 * doc/tm.texi: Regenerate.
1829 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1832 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1833 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
1834 aarch64_clamp_to_uimm12_shift): New.
1835 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
1836 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
1838 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1841 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
1843 2018-10-01 Jeff Law <law@redhat.com>
1844 Richard Sandiford <richard.sandiford@linaro.org>
1845 Tamar Christina <tamar.christina@arm.com>
1848 * config/aarch64/aarch64.md
1849 (probe_stack_range): Add k (SP) constraint.
1850 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
1851 STACK_CLASH_MAX_UNROLL_PAGES): New.
1852 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
1853 stack probes for stack clash.
1854 (aarch64_allocate_and_probe_stack_space): New.
1855 (aarch64_expand_prologue): Use it.
1856 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
1857 (aarch64_sub_sp): Add emit_move_imm optional param.
1859 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
1861 PR tree-optimization/87261
1862 * match.pd: Remove trailing whitespace.
1863 Add (x & y) | ~(x | y) -> ~(x ^ y),
1864 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
1866 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1868 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1870 (ashlsi3_insn): Update instruction constraints.
1871 (ashrsi3_insn): Likewise.
1872 (rotrsi3): Likewise.
1873 (add_shift): Likewise.
1874 * config/arc/constraints.md (Csz): New 32 bit constraint. It
1875 avoids placing in the limm field small constants which, otherwise,
1876 could end into a small instruction.
1878 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1880 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1881 destination register is not odd-even.
1882 (umaddsidi4_split): Likewise.
1884 2018-10-01 Richard Biener <rguenther@suse.de>
1886 * tree-inline.c (expand_call_inline): Store origin of fn
1887 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1888 * tree.c (block_ultimate_origin): Simplify and do some
1891 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
1893 * config/i386/mmx.md (EMMS): New int iterator.
1894 (emms): New int attribute.
1895 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1896 EMMS int iterator. Explicitly declare clobbers.
1897 (mmx_emms): Remove expander.
1899 * config/i386/predicates.md (emms_operation): Remove predicate.
1900 (vzeroall_pattern): New predicate.
1901 (vzeroupper_pattern): Rename from vzeroupper_operation.
1902 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1903 vzeroupper_pattern and vzeroall_pattern predicates.
1905 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
1907 PR rtl-optimization/86939
1908 * ira-lives.c (make_hard_regno_born): Rename from this...
1909 (make_hard_regno_live): ... to this. Remove update to conflict
1910 information. Update function comment.
1911 (make_hard_regno_dead): Add conflict information update. Update
1913 (make_object_born): Rename from this...
1914 (make_object_live): ... to this. Remove update to conflict information.
1915 Update function comment.
1916 (make_object_dead): Add conflict information update. Update function
1918 (mark_pseudo_regno_live): Call make_object_live.
1919 (mark_pseudo_regno_subword_live): Likewise.
1920 (mark_hard_reg_dead): Update function comment.
1921 (mark_hard_reg_live): Call make_hard_regno_live.
1922 (process_bb_node_lives): Likewise.
1923 * lra-lives.c (make_hard_regno_born): Rename from this...
1924 (make_hard_regno_live): ... to this. Remove update to conflict
1925 information. Remove now uneeded check_pic_pseudo_p argument.
1926 Update function comment.
1927 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1928 to conflict information. Update function comment.
1929 (mark_pseudo_live): Remove update to conflict information. Update
1931 (mark_pseudo_dead): Add conflict information update.
1932 (mark_regno_live): Call make_hard_regno_live.
1933 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1934 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1936 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
1939 * config/i386/i386.c (construct_container): Use TImode for
1940 BLKmode values in 2 integer registers.
1942 2018-09-29 Jeff Law <law@redhat.com>
1944 * builtins.c (unterminated_array): Pass in c_strlen_data * to
1945 c_strlen rather than just a tree *.
1946 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1947 Update recursive calls appropriately. If caller did not provide a
1948 suitable data pointer, create a local one. When a non-terminated
1949 string is discovered, bubble up information about the string via the
1950 c_strlen_data object.
1951 * builtins.h (c_strlen): Update prototype.
1952 (c_strlen_data): New structure.
1953 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1954 For a type 2 call, if c_strlen indicates a non-terminated string
1955 use the length of the non-terminated string.
1956 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1958 2018-09-29 Jakub Jelinek <jakub@redhat.com>
1961 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1962 __m512d type for __A argument rather than __m512.
1964 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
1966 * match.pd (simple_comparison): Don't optimize if either operand is
1967 a function pointer when target needs function pointer canonicalization.
1969 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1971 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1972 power5 .. power9 to remove indirection.
1973 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1974 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1975 ASM_CPU_476_SPEC): Delete.
1976 (ASM_CPU_SPEC): Adjust.
1977 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1978 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1980 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1982 * config.in: Delete HAVE_AS_DCI.
1983 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1984 * config/rs6000/rs6000.h: Ditto.
1985 * configure.ac: Delete HAVE_AS_DCI.
1986 * configure: Regenerate.
1988 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1990 * config.in (HAVE_AS_LWSYNC): Delete.
1991 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1992 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1994 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1995 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1997 * configure.ac: Delete HAVE_AS_LWSYNC.
1998 * configure: Regenerate.
2000 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
2001 Pierre-Marie de Rodat <derodat@adacore.com>
2003 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2004 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2005 (cgraph_node::create_thunk): Add indirect_offset parameter.
2006 (thunk_adjust): Likewise.
2007 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2008 and initialize the corresponding field with it.
2009 (cgraph_node::dump): Dump indirect_offset field.
2010 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2011 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2012 (thunk_adjust): Add indirect_offset parameter and deal with it.
2013 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2014 pass it to thunk_adjust. Do not call the target hook if it's non-zero
2015 or if the thunk is external or local. Fix formatting. Do not chain
2016 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
2017 if any, in the GIMPLE representation.
2018 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2019 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2020 (input_node): Read indirect_offset field.
2021 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2022 call to thunk_adjust.
2023 * tree-nested.c (struct nesting_info): Add thunk_p field.
2024 (create_nesting_tree): Set it.
2025 (convert_all_function_calls): Copy static chain from targets to thunks.
2026 (finalize_nesting_tree_1): Return early for thunks.
2027 (unnest_nesting_tree_1): Do not finalize thunks.
2028 (gimplify_all_functions): Do not gimplify thunks.
2030 2018-09-28 David Malcolm <dmalcolm@redhat.com>
2032 * opt-suggestions.c (option_proposer::build_option_suggestions):
2033 Release "option_values".
2035 2018-09-28 David Malcolm <dmalcolm@redhat.com>
2037 * coverage.c (get_coverage_counts): Convert problem-reporting dump
2038 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2039 * dumpfile.c (kind_as_string): New function.
2040 (dump_loc): Rather than a hardcoded prefix of "note: ", use
2041 kind_as_string to vary the prefix based on dump_kind.
2042 (selftest::test_capture_of_dump_calls): Update for above.
2044 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2046 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2047 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
2049 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2051 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
2052 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
2053 INVALID_REGNUM instead of FPSR_REG.
2054 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
2055 * config/i386/i386.md: Update comment of FP compares.
2056 (fldenv): Do not clobber FPSR_REG.
2058 2018-09-28 Richard Biener <rguenther@suse.de>
2060 * tree.h (BLOCK_ORIGIN): New.
2061 * omp-expand.c (grid_expand_target_grid_body): Assign
2062 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
2063 * tree-inline.c (remap_block): Likewise.
2064 * auto-profile.c (get_function_decl_from_block): Simplify
2065 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
2066 * langhooks.c (lhd_print_error_function): Likewise.
2067 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
2069 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2070 * tree.c (block_nonartificial_location): Likewise.
2071 (block_ultimate_origin): Likewise.
2072 * tree-pretty-print.c (percent_K_format): Likewise. Remove
2073 no longer needed LTO case.
2075 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
2076 Jan Hubicka <jh@suse.cz>
2077 Martin Jambor <mjambor@suse.cz>
2079 * simplify-rtx.c (simplify_merge_mask): New function.
2080 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
2081 same masks are used in op1 or op2.
2082 (test_vec_merge): New function.
2083 (test_vector_ops): Call test_vec_merge.
2085 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
2087 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
2088 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
2089 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
2090 (cypress_costs): Set it.
2091 (supersparc_costs): Likewise.
2092 (hypersparc_costs): Likewise.
2093 (leon_cost): Likewise.
2094 (leon3_costs): Likewise.
2095 (sparclet_costs): Likewise.
2096 (ultrasparc_costs): Likewise.
2097 (ultrasparc_costs): Likewise.
2098 (niagara_costs): Likewise.
2099 (niagara2_costs): Likewise.
2100 (niagara3_costs): Likewise.
2101 (niagara4_costs): Likewise.
2102 (niagara7_costs): Likewise.
2103 (m8_costs): Likewise.
2104 (TARGET_CAN_FOLLOW_JUMP): Define.
2105 (pass_work_around_errata::gate): Minor tweak.
2106 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
2107 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
2108 Automaitcally clear MASK_FSMULD mask for V7 processors.
2109 (sparc_can_follow_jump): New static function.
2110 (output_ubranch): Deal with CROSSING_JUMP_P.
2111 (sparc_use_sched_lookahead): Rewrite using switch statement.
2112 (sparc_issue_rate): Reorder.
2113 (sparc_branch_cost): New function.
2115 2018-09-27 Martin Sebor <msebor@redhat.com>
2117 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
2118 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
2119 (int_fits_type_p): Same.
2121 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
2123 * config/i386/i386.md (FPCR_REG): Remove.
2124 (UNSPEC_FLDCW): Remove.
2125 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
2126 (x86_fldcw_1): Remove insn pattern.
2127 (fnstenv): Do not clobber FPCR_REG.
2129 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
2130 (CALL_USED_REGISTERS): Ditto.
2131 (REG_ALLOC_ORDER): Ditto.
2132 (REG_CLASS_CONTENTS): Ditto.
2133 (HI_REGISTER_NAMES): Ditto.
2134 (ADDITIONAL_REGISTER_NAMES): Use defines instead
2135 of numerical constants.
2136 * config/i386/i386.c (regclass_map): Remove fpsr register.
2137 (dbx_register_map): Ditto.
2138 (dbx64_register_map): Ditto.
2139 (svr4_dbx_register_map): Ditto.
2140 (print_reg): Do not handle FPCR_REG.
2142 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
2145 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
2146 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
2147 Delete, always treat as true.
2148 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2149 Ditto. Simplify remaining code.
2150 * config/powerpcspe/powerpcspe.h: Ditto.
2151 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
2152 Simplify remaining code.
2153 (rs6000_expand_builtin): Ditto.
2154 * config/rs6000/rs6000.h: Ditto.
2155 * configure.ac: Ditto.
2156 * configure: Regenerate.
2158 2018-09-27 Martin Liska <mliska@suse.cz>
2160 * coverage.c (get_coverage_counts): Revert the formatting
2161 of missing profile opt info.
2163 2018-09-27 Richard Biener <rguenther@suse.de>
2167 * dwarf2out.c (set_block_origin_self): Do not mark outermost
2168 block as we do not output that.
2169 (gen_inlined_subroutine_die): Elide the originally outermost
2170 block, matching what we do for concrete instances.
2171 (decls_for_scope): Add parameter specifying whether to recurse
2174 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
2175 Tom de Vries <tom@codesourcery.com>
2179 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
2180 STORE_FLAG_VALUE == 1.
2182 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
2184 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
2185 constant definitions.
2186 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
2187 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
2188 ("speculation_barrier"): New expander definition.
2190 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
2192 PR gcov-profile/86957
2193 * common.opt: New warning option -Wmissing-profile.
2194 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
2195 * doc/invoke.texi: Document -Wmissing-profile.
2197 2018-09-26 Jim Wilson <jimw@sifive.com>
2199 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
2200 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
2202 2018-09-26 Martin Sebor <msebor@redhat.com>
2204 * tree.c (zerop): Change return type to bool.
2205 (integer_zerop, integer_onep, integer_each_onep): Same.
2206 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2207 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2208 (real_onep, real_minus_onep, chain_index): Same.
2209 (print_type_hash_statistics, type_list_equal): Same.
2210 * tree.h (zerop): Same.
2211 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
2212 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2213 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2214 (real_onep, real_minus_onep, chain_index): Same.
2215 (print_type_hash_statistics, type_list_equal): Same.
2217 2018-09-26 Jim Wilson <jimw@sifive.com>
2219 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
2221 2018-09-26 Jakub Jelinek <jakub@redhat.com>
2224 * config/i386/i386.c: Include debug.h and dwarf2out.h.
2225 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2228 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
2230 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
2232 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
2234 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
2235 and FP_SECOND_SSE_REGS.
2236 (REG_CLASS_NAMES): Ditto.
2237 (REG_CLASS_CONTENTS): Ditto.
2238 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
2239 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
2240 (ix86_preferred_output_reload_class): Ditto.
2241 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
2242 clobber constraint to "=&f".
2243 (fix_truncdi_i387): Ditto.
2244 (lrintxfdi2): Ditto.
2245 (fistdi2_<rounding>): Ditto.
2246 (fpremxf4_i387): Change "=u" constraint to "=f".
2247 (fprem1xf4_i387): Ditto.
2249 (fptanxf4_i387): Ditto.
2250 (fxtractxf3_i387): Ditto.
2251 (fscalexf4_i387): Ditto.
2252 (atan2xf3): Change "u" constraint to "f".
2253 (fyl2xxf3_i387): Ditto.
2254 (fyl2xp1xf3_i387): Ditto.
2256 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
2259 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
2260 for removed I387_MASK_PM entity.
2263 2018-09-26 Jeff Law <law@redhat.com>
2265 2018-09-26 Alexey Neyman <stilor@att.net>
2267 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2268 headers are no longer pulled in by <isl/val.h>.
2270 2018-09-26 Richard Biener <rguenther@suse.de>
2273 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
2274 or concrete instance DIE to the tree. Create abstract origin
2275 attributes also for concrete instances.
2277 2018-09-26 Alexey Neyman <stilor@att.net>
2279 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
2280 headers are no longer pulled in by <isl/val.h>.
2282 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
2284 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
2285 Use new helper functions.
2286 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
2287 Use new helper functions.
2288 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
2289 aarch_mm_needs_release): New declarations.
2290 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
2291 aarch_mm_needs_release): New.
2293 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
2295 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
2296 (arm32_output_mi_thunk): Deal with long calls.
2298 2018-09-26 Richard Biener <rguenther@suse.de>
2302 * tree-inline.c (expand_call_inline): When the location
2303 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2304 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2305 the inserted BLOCK to make inlined_function_outer_scope_p
2307 * dwarf2out.c (add_call_src_coords_attributes): Do not add
2308 coords for reserved locations.
2310 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
2312 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
2313 (*call_indirect_nonlocal_sysv<mode>): Ditto.
2314 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
2315 (*sibcall_nonlocal_sysv<mode>): Ditto.
2316 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
2317 (<bd>_<mode>): Ditto.
2318 (<bd>tf_<mode>): Ditto.
2320 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
2322 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
2323 control string as a list of templates instead of as C code.
2324 (*altivec_movti): Ditto.
2325 * config/rs6000/darwin.md (movdf_low_di): Ditto.
2327 2018-09-25 Jim Wilson <jimw@sifive.com>
2329 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
2330 when target symbol is weak.
2332 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
2335 * builtins.c (unterminated_array): Simplify.
2336 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
2337 where pointer arithmetic is safe.
2339 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
2342 * config/rs6000/altivec.md (altivec_vspltb): Use
2343 const_0_to_15_operand instead of u5bit_cint_operand.
2344 (*altivec_vspltb_internal): Ditto.
2345 (altivec_vspltb_direct): Ditto.
2346 (altivec_vsplth): Use const_0_to_7_operand instead of
2348 (*altivec_vsplth_internal): Ditto.
2349 (altivec_vsplth_direct): Ditto.
2350 (altivec_vspltw): Use const_0_to_3_operand instead of
2352 (*altivec_vspltw_internal): Ditto.
2353 (altivec_vspltw_direct): Ditto.
2354 (altivec_vspltsf): Ditto.
2355 (*altivec_vspltsf_internal): Ditto.
2356 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
2357 various splats with the proper size immediate. Reorder the various
2358 cases by ascending size of immediate, and put all such together.
2360 2018-09-25 Richard Biener <rguenther@suse.de>
2363 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
2365 (maybe_create_die_with_external_ref): Do not create
2366 DW_TAG_imported_unit here.
2367 (add_abstract_origin_attribute): Handle external BLOCK refs.
2368 (dwarf2out_abstract_function): Simplify LTO case.
2369 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
2370 rather than using maybe_create_die_with_external_ref.
2372 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
2375 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
2376 (frndintxf2_mask_pm_i387): Ditto.
2377 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
2378 Enable for !flag_trapping_math.
2379 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
2380 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
2381 Change operand 1 predicate to nonimmediate_operand.
2382 (attr "i387_cw"): Remove mask_pm.
2383 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
2384 (enum ix86_entity): Remove I387_MASK_PM.
2385 * config/i386/i386.c (ix86_i387_mode_needed): Do not
2386 handle I387_MASK_PM.
2387 (ix86_mode_needed): Ditto.
2388 (ix86_mode_after): Ditto.
2389 (ix86_mode_entry): Ditto.
2390 (ix86_mode_exit): Ditto.
2391 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
2393 2018-09-25 Jakub Jelinek <jakub@redhat.com>
2395 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
2396 to_update_switch_stmts to vNULL instead of calling create on them
2399 2018-09-25 Richard Biener <rguenther@suse.de>
2401 PR tree-optimization/87402
2402 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
2403 (visit_phi): Re-instantiate handling of supposed to be VARYING
2404 but non-VARYING backedge value.
2406 2018-09-25 Richard Biener <rguenther@suse.de>
2409 * dwarf2out.c (struct sym_off_pair): New.
2410 (external_die_map): New global.
2411 (lookup_decl_die): When in LTO create DIEs lazily from the
2413 (lookup_block_die): New function, create DIEs lazily in LTO.
2414 (equate_block_to_die): New function.
2415 (dwarf2out_die_ref_for_decl): During WPA get the association
2416 from the external DIE map.
2417 (dwarf2out_register_external_die): Record mapping into the
2419 (maybe_create_die_with_external_ref): New function split out from
2420 DIE generation part of old dwarf2out_register_external_die.
2421 (add_abstract_origin_attribute): Do not return the DIE. When
2422 in LTO reference externals directly.
2423 (dwarf2out_abstract_function): When in LTO ignore calls for
2424 decls with external DIEs (already present abstract instances).
2425 (gen_call_site_die): Adjust.
2426 (add_high_low_attributes): Likewise.
2427 (gen_lexical_block_die): Likewise.
2428 (gen_inlined_subroutine_die): Likewie.
2429 (gen_block_die): Likewise.
2430 (dwarf2out_inline_entry): Likewise.
2431 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
2434 2018-09-25 Martin Liska <mliska@suse.cz>
2436 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
2437 integers and not by a float value.
2439 2018-09-25 Martin Liska <mliska@suse.cz>
2442 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
2444 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
2445 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
2446 * godump.c (go_format_type): Remove extra parenthesis.
2448 2018-09-25 Martin Liska <mliska@suse.cz>
2450 * alias.c (set_dest_equal_p): Remove unused function.
2451 * config/i386/i386.c (def_builtin_pure2): Likewise.
2452 * diagnostic-show-locus.c (class layout): Remove
2454 (layout::layout): Likewise here.
2455 * dump-context.h (class temp_dump_context): Likewise.
2456 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
2457 (add_AT_loclistsptr): Likewise.
2458 (add_AT_offset): Likewise.
2459 (get_AT_hi_pc): Likewise.
2460 (is_comdat_die): Likewise.
2461 (type_is_enum): Likewise.
2462 (ceiling): Likewise.
2463 (add_AT_vms_delta): Likewise.
2464 (is_class_die): Likewise.
2465 * edit-context.c (class line_event): Remove unused field.
2466 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
2468 * ipa-cp.c (ipa_get_vr_lat): Likewise.
2469 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
2470 (ok_for_base_p_nonstrict): Likewise.
2471 * tree-chrec.c (is_not_constant_evolution): Likewise.
2472 (chrec_fold_poly_cst): Likewise.
2473 * tree-if-conv.c (has_pred_critical_p): Likewise.
2474 * tree-ssa-coalesce.c (print_exprs): Likewise.
2475 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
2476 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
2477 * tree-vrp.c (value_ranges_intersect_p): Likewise.
2478 (value_range_nonnegative_p): Likewise.
2480 2018-09-25 Martin Liska <mliska@suse.cz>
2482 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2483 Do not handle "GNU Pascal".
2484 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2486 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
2487 from documentation. Likewise.
2488 * dbxout.c (dbxout_range_type): Likewise.
2489 * doc/cpp.texi: Likewise.
2490 * doc/extend.texi: Likewise.
2491 * doc/frontends.texi: Likewise.
2492 * doc/invoke.texi: Remove Pascal entry.
2493 * tree.def (CLEANUP_POINT_EXPR): Likewise.
2494 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
2496 2018-09-25 Martin Liska <mliska@suse.cz>
2499 * doc/invoke.texi: Document all parameters and remove default
2502 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
2505 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
2506 contains HOST_WIDE_INTs when computing its size.
2508 2018-09-24 Jim Wilson <jimw@sifive.com>
2511 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
2513 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
2515 2018-09-24 Andrew Pinski <apinski@marvell.com>
2517 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2518 access prev before checking it for NULLness in the
2519 AARCH64_FUSE_CMP_BRANCH case.
2521 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
2524 * config/i386/i386.c (rest_of_insert_endbranch): Set
2525 endbr_queued_at_entrance to true and don't insert ENDBR if
2526 x86_function_profiler will be called.
2527 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2529 * config/i386/i386.h (machine_function): Add
2530 endbr_queued_at_entrance.
2532 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
2534 * genattrtab.c (mk_attr_alt): Use alternative_mask.
2535 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
2537 (check_attr_test): Use alternative_mask.
2538 (get_attr_value): Likewise.
2539 (compute_alternative_mask): Use alternative_mask and XWINT.
2540 (make_alternative_compare): Use alternative_mask.
2541 (attr_alt_subset_p): Use XWINT.
2542 (attr_alt_subset_of_compl_p): Likewise.
2543 (attr_alt_intersection): Use alternative_mask and XWINT.
2544 (attr_alt_union): Likewise.
2545 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
2546 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
2547 (simplify_test_exp): Use alternative_mask and XWINT.
2548 (write_test_expr): Use alternative_mask and XWINT, adjust bit
2549 number calculation to support 64 bits. Generate code that
2550 checks 64-bit masks.
2551 (main): Use alternative_mask.
2552 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
2554 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
2557 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
2558 RETURN+USE when returning via %r14.
2560 2018-09-24 Martin Liska <mliska@suse.cz>
2562 * gcov.c (output_lines): Print colorization legend
2563 for both flag_use_colors and flag_use_hotness_colors.
2566 2018-09-24 Martin Liska <mliska@suse.cz>
2568 * coverage.c (get_coverage_counts): Use warning_at
2569 with current_function_decl location. Use %qD in warning
2572 2018-09-24 Martin Liska <mliska@suse.cz>
2574 * memory-block.h (memory_block_pool::release): Annotate with
2575 valgrind that the memory is not accessible.
2577 2018-09-24 Martin Liska <mliska@suse.cz>
2580 * asan.c: Make asan_handled_variables extern.
2582 * cfgexpand.c (expand_stack_vars): Make sure
2583 a representative is unpoison if another
2584 variable in the partition is handled by
2585 use-after-scope sanitization.
2587 2018-09-24 Richard Biener <rguenther@suse.de>
2589 PR tree-optimization/63155
2590 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2591 the worklist when the edge of the respective argument isn't
2594 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
2596 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
2597 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
2598 (MASK_CLASS_P): Update for rename.
2599 (MAYBE_MASK_CLASS_P): Ditto.
2600 (REG_CLASS_NAMES): Update.
2601 (REG_CLASS_CONTENT): Update.
2602 * config/i386/i386.c (regclass_map): Update for MASK_REG
2603 and ALL_MASK_REGS rename.
2604 * config/i386/constraints.md (Yk): Update for rename.
2607 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
2609 * config/i386/i386.h (enum reg_class): Remove
2610 EVEX_SSE_REGS and MOD4_SSE_REGS.
2611 (REG_CLASS_NAMES): Update.
2612 (REG_CLASS_CONTENT): Update.
2613 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
2614 registers as ALL_SSE_REGS.
2615 (ix86_additional_allocno_class_p): Remove.
2616 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
2617 (ix86_register_priority): Lower priority of EVEX SSE registers.
2618 Use IN_RANGE macro where appropriate.
2619 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
2620 AVX-5124VNNIW checks.
2621 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
2622 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
2623 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
2624 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
2625 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
2626 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
2627 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
2628 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
2629 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
2630 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
2631 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
2632 * config/i386/constraints.md (Yh): Remove.
2634 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
2636 * config/i386/i386.c (regclass_map): Declare integer REX registers
2639 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
2641 * doc/service.texi (Service): Switch the fsf.org link to https.
2643 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
2646 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2647 Define to speculation_safe_value_not_needed.
2649 2018-09-21 Florian Weimer <fweimer@redhat.com>
2652 * doc/extend.texi (Common Function Attributes): Mention that
2653 noreturn suppresses tail call optimization.
2655 2018-09-21 Jeff Law <law@redhat.com>
2657 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
2658 vr_values::cleanup_edges_and_switches.
2659 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
2661 (identify_jump_threads): Remove EDGE_IGNORE handling.
2662 (execute_vrp): Move handling of to_remove_edges and
2663 to_update_switch_stmts into vr_values class member functions.
2664 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
2665 (to_update_switch_stmts): Likewise.
2666 * vr-values.c: Include cfghooks.h.
2667 (vr_values::vr_values): Initialize to_remove_edges and
2668 to_update_switch_stmts.
2669 (vr_values::~vr_values): Verify to_remove_edges and
2670 to_update_switch_stmts are empty.
2671 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
2672 (vr_values::cleanup_edges_and_switches): New member function.
2673 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
2674 function. Add new data members.
2676 2018-09-21 David Malcolm <dmalcolm@redhat.com>
2678 PR tree-optimization/87309
2679 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
2680 calls with pflags and alt_flags.
2681 (selftest::test_capture_of_dump_calls): Add test of interaction of
2682 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
2684 2018-09-21 Olivier Hainque <hainque@adacore.com>
2686 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
2688 2018-09-21 Olivier Hainque <hainque@adacore.com>
2690 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
2692 2018-09-21 Olivier Hainque <hainque@adacore.com>
2694 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
2695 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
2696 (VXWORKS_LIBS_RTP): Minor reordering.
2698 2018-09-21 Olivier Hainque <hainque@adacore.com>
2700 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
2701 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
2703 2018-09-21 Olivier Hainque <hainque@adacore.com>
2705 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
2706 (PTRDIFF_TYPE): Likewise.
2708 2018-09-21 Olivier Hainque <hainque@adacore.com>
2710 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
2711 triplet, similar to support for VxWorks7.
2712 * config/vxworks-dummy.h: Provide a default definition
2713 of TARGET_VXWORKS64 to 0.
2715 2018-09-21 Olivier Hainque <hainque@adacore.com>
2717 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
2718 * config/vxworks-dummy.h: here.
2720 2018-09-21 Olivier Hainque <hainque@adacore.com>
2722 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
2724 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
2725 Bo Zhou <zbo.zhou@hisilicon.com>
2727 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
2728 * config/aarch64/aarch64-tune.md: Regenerated.
2729 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
2730 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
2731 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
2733 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
2734 Julian Brown <julian@codesourcery.com>
2736 * builtins.c (get_builtin_sync_mem): Handle address spaces.
2738 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2740 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2741 if the call takes a static chain.
2743 2018-09-21 Martin Liska <mliska@suse.cz>
2745 * auto-profile.c (autofdo_source_profile::read): Do not
2747 (read_profile): Do not add working sets.
2748 (read_autofdo_file): Remove sum_all.
2749 (afdo_callsite_hot_enough_for_early_inline): Remove const
2751 * coverage.c (struct counts_entry): Remove gcov_summary.
2752 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2753 do not support GCOV_TAG_PROGRAM_SUMMARY.
2754 (get_coverage_counts): Remove summary and expected
2756 * coverage.h (get_coverage_counts): Likewise.
2757 * doc/gcov-dump.texi: Remove -w option.
2758 * gcov-dump.c (dump_working_sets): Remove.
2759 (main): Do not support '-w' option.
2760 (print_usage): Likewise.
2761 (tag_summary): Likewise.
2762 * gcov-io.c (gcov_write_summary): Do not dump
2764 (gcov_read_summary): Likewise.
2765 (gcov_histo_index): Remove.
2766 (gcov_histogram_merge): Likewise.
2767 (compute_working_sets): Likewise.
2768 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2770 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2771 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2772 (GCOV_HISTOGRAM_SIZE): Remove.
2773 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2774 (struct gcov_summary): Simplify rapidly just
2775 to runs and sum_max fields.
2776 (gcov_histo_index): Remove.
2777 (NUM_GCOV_WORKING_SETS): Likewise.
2778 (compute_working_sets): Likewise.
2779 * gcov-tool.c (print_overlap_usage_message): Remove
2780 trailing empty line.
2781 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2782 (output_lines): Remove program related line.
2783 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2784 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2786 (input_profile_summary): Do not read it.
2787 (merge_profile_summaries): And do not merge it.
2788 (input_symtab): Do not call removed function.
2789 * modulo-sched.c (sms_schedule): Do not print sum_max.
2790 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2791 removed when histogram method was invented.
2792 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2794 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2796 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2797 and dump selected value.
2798 * profile.c (add_working_set): Remove.
2799 (get_working_sets): Likewise.
2800 (find_working_set): Likewise.
2801 (get_exec_counts): Do not work with working sets.
2802 (read_profile_edge_counts): Do not inform as sum_max is removed.
2803 (compute_branch_probabilities): Likewise.
2804 (compute_value_histograms): Remove argument for call of
2805 get_coverage_counts.
2806 * profile.h: Do not make gcov_summary const.
2808 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
2810 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2812 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2814 PR tree-optimization/86990
2815 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2816 Check that the entire merged store group is made of constants only for
2819 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
2821 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2822 (VTABLE_VERIFICATION_SPEC): Likewise.
2823 (SANITIZER_EARLY_SPEC): Likewise.
2824 (SANITIZER_SPEC): Likewise.
2825 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2826 * doc/invoke.texi (Link Options): Document -r.
2828 2018-09-20 Richard Biener <rguenther@suse.de>
2831 * gimplify.c (gimplify_expr): Retain alignment of
2832 addressable lvalue in dereference.
2834 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
2837 * configure.ac: Check for .loc is_stmt support.
2838 * configure, config.in: Rebuilt.
2839 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2842 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
2844 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
2845 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
2848 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
2850 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2852 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2855 2018-09-20 Marek Polacek <polacek@redhat.com>
2857 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
2859 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2861 PR tree-optimization/87288
2862 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2863 into account when determining PEELING_FOR_NITERS.
2865 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2867 PR tree-optimization/86877
2868 * tree-vect-loop.c (vect_analyze_loop_2): Call
2869 vect_verify_datarefs_alignment.
2871 2018-09-19 Marek Polacek <polacek@redhat.com>
2873 * doc/invoke.texi: Document -Wclass-conversion.
2875 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
2877 * config/pa/pa.c (pa_adjust_priority): Delete.
2878 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2880 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2881 (atomic_storehi): Likewise.
2882 (atomic_storesi): Likewise.
2883 (atomic_loaddi): Restore compare and swap exchange loop code.
2885 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
2887 PR rtl-optimization/86902
2888 * combine.c (try_combine): When changing the CC mode used, don't change
2889 an unrelated mode in other_insn to that new CC mode.
2891 2018-09-19 David Malcolm <dmalcolm@redhat.com>
2893 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2894 with %T in place of calls to dump_generic_expr.
2895 (prune_runtime_alias_test_list): Likewise.
2896 (create_runtime_alias_checks): Likewise.
2897 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2898 (vect_analyze_data_ref_dependence): Likewise.
2899 (vect_slp_analyze_data_ref_dependence): Likewise.
2900 (vect_record_base_alignment): Likewise. Use %G in place of call
2901 to dump_gimple_stmt.
2902 (vect_compute_data_ref_alignment): Likewise.
2903 (verify_data_ref_alignment): Likewise.
2904 (vect_find_same_alignment_drs): Likewise.
2905 (vect_analyze_group_access_1): Likewise.
2906 (vect_analyze_data_ref_accesses): Likewise.
2907 (dependence_distance_ge_vf): Likewise.
2908 (dump_lower_bound): Likewise.
2909 (vect_prune_runtime_alias_test_list): Likewise.
2910 (vect_find_stmt_data_reference): Likewise.
2911 (vect_analyze_data_refs): Likewise.
2912 (vect_create_addr_base_for_vector_ref): Likewise.
2913 (vect_create_data_ref_ptr): Likewise.
2914 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2915 (vect_can_advance_ivs_p): Likewise.
2916 (vect_update_ivs_after_vectorizer): Likewise.
2917 (vect_gen_prolog_loop_niters): Likewise.
2918 (vect_prepare_for_masked_peels): Likewise.
2919 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2920 (vect_determine_vectorization_factor): Likewise.
2921 (vect_is_simple_iv_evolution): Likewise.
2922 (vect_analyze_scalar_cycles_1): Likewise.
2923 (vect_analyze_loop_operations): Likewise.
2924 (report_vect_op): Likewise.
2925 (vect_is_slp_reduction): Likewise.
2926 (check_reduction_path): Likewise.
2927 (vect_is_simple_reduction): Likewise.
2928 (vect_create_epilog_for_reduction): Likewise.
2929 (vect_finalize_reduction:): Likewise.
2930 (vectorizable_induction): Likewise.
2931 (vect_transform_loop_stmt): Likewise.
2932 (vect_transform_loop): Likewise.
2933 (optimize_mask_stores): Likewise.
2934 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2935 (vect_split_statement): Likewise.
2936 (vect_recog_over_widening_pattern): Likewise.
2937 (vect_recog_average_pattern): Likewise.
2938 (vect_determine_min_output_precision_1): Likewise.
2939 (vect_determine_precisions_from_range): Likewise.
2940 (vect_determine_precisions_from_users): Likewise.
2941 (vect_mark_pattern_stmts): Likewise.
2942 (vect_pattern_recog_1): Likewise.
2943 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2944 (vect_record_max_nunits): Likewise.
2945 (vect_build_slp_tree_1): Likewise.
2946 (vect_build_slp_tree_2): Likewise.
2947 (vect_print_slp_tree): Likewise.
2948 (vect_analyze_slp_instance): Likewise.
2949 (vect_detect_hybrid_slp_stmts): Likewise.
2950 (vect_detect_hybrid_slp_1): Likewise.
2951 (vect_slp_analyze_operations): Likewise.
2952 (vect_slp_analyze_bb_1): Likewise.
2953 (vect_transform_slp_perm_load): Likewise.
2954 (vect_schedule_slp_instance): Likewise.
2955 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2956 (vect_mark_stmts_to_be_vectorized): Likewise.
2957 (vect_init_vector_1): Likewise.
2958 (vect_get_vec_def_for_operand): Likewise.
2959 (vect_finish_stmt_generation_1): Likewise.
2960 (vect_check_load_store_mask): Likewise.
2961 (vectorizable_call): Likewise.
2962 (vectorizable_conversion): Likewise.
2963 (vectorizable_operation): Likewise.
2964 (vectorizable_load): Likewise.
2965 (vect_analyze_stmt): Likewise.
2966 (vect_is_simple_use): Likewise.
2967 (vect_get_vector_types_for_stmt): Likewise.
2968 (vect_get_mask_type_for_stmt): Likewise.
2969 * tree-vectorizer.c (increase_alignment): Likewise.
2971 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
2973 * doc/rtl.texi: Adjust vec_select description.
2974 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2975 non-constant selectors.
2977 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
2979 * config/aarch64/aarch64-protos.h
2980 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2981 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2982 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2983 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2984 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2985 (AARCH64_FL_PROFILE): Move index so flags are ordered.
2986 (AARCH64_ISA_RCPC8_4): New flag.
2987 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2988 to aarch64_offset_9bit_signed_unscaled_p.
2989 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2991 * config/aarch64/constraints.md (Ust): New constraint.
2992 * config/aarch64/predicates.md.
2993 (aarch64_9bit_offset_memory_operand): New predicate.
2994 (aarch64_rcpc_memory_operand): New predicate.
2996 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
2998 PR rtl-optimization/87361
2999 * rtlanal.c (nonzero_bits1): Revert accidental change.
3001 2018-09-19 Richard Biener <rguenther@suse.de>
3003 PR tree-optimization/87349
3004 PR tree-optimization/87342
3005 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3007 2018-09-18 Marek Polacek <polacek@redhat.com>
3009 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3010 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3012 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
3014 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3016 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
3018 PR rtl-optimization/86882
3019 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3021 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
3023 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3024 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3026 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
3029 * doc/invoke.texi (Link Options): Fix formatting and grammar.
3031 2018-09-18 Richard Biener <rguenther@suse.de>
3034 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3035 (compute_samebase_partition_bases): Likewise.
3036 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3037 (gimple_can_coalesce_p): Simplify.
3039 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
3041 Handle a library implementation of ffs calling __builtin_ffs.
3042 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3043 (mmix_init_libfuncs): New function: make __builtin_ffs expand
3046 2018-09-17 David Malcolm <dmalcolm@redhat.com>
3048 * diagnostic-show-locus.c (class layout_range): Add field
3050 (layout_range::layout_range): Add "original_idx" param and use it
3051 to initialize new field.
3052 (make_range): Use 0 for original_idx.
3053 (layout::layout): Pass in index to calls to
3054 maybe_add_location_range.
3055 (layout::maybe_add_location_range): Add param "original_idx" and
3056 pass it on to layout_range.
3057 (layout::print_any_labels): Pass on range->m_original_idx to
3059 (gcc_rich_location::add_location_if_nearby): Use 0 for
3061 * gcc-rich-location.h (text_range_label::get_text): Update for new
3063 (range_label_for_type_mismatch::get_text): Likewise.
3065 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
3067 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
3069 2018-09-17 David Malcolm <dmalcolm@redhat.com>
3071 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
3072 format_string_diagnostic_t.
3073 (fmtwarn_n): Likewise.
3074 * substring-locations.c
3075 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
3076 (format_warning_n_va): Convert to...
3077 (format_string_diagnostic_t::emit_warning_n_va): ...this.
3078 (format_warning_va): Convert to...
3079 (format_string_diagnostic_t::emit_warning_va): ...this.
3080 (format_warning_at_substring): Convert to...
3081 (format_string_diagnostic_t::emit_warning): ...this.
3082 (format_warning_at_substring_n): Convert to...
3083 (format_string_diagnostic_t::emit_warning_n): ...this.
3084 * substring-locations.h (class format_string_diagnostic_t): New
3086 (format_warning_va): Convert to
3087 format_string_diagnostic_t::emit_warning_va.
3088 (format_warning_n_va): Convert to
3089 format_string_diagnostic_t::emit_warning_n_va.
3090 (format_warning_at_substring): Convert to
3091 format_string_diagnostic_t::emit_warning.
3092 (format_warning_at_substring_n): Convert to
3093 format_string_diagnostic_t::emit_warning_n.
3095 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
3096 Bernd Schmidt <bernds_cb1@t-online.de>
3098 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
3101 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
3103 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
3104 operand 0 predicate to nonimmediate operand.
3105 (rint<mode>2_frndint): Remove insn pattern.
3106 (rint<mode>2): Change operand 1 predicate to general_operand.
3107 Extend operand 1 to XFmode and generate rintxf2 insn.
3108 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
3109 Do not use X87MODEF mode macro.
3110 (frndintxf2_<rounding>_i387): Rename from
3111 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
3112 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
3113 to XFmode and generate significandxf3 insn.
3115 2018-09-17 Richard Biener <rguenther@suse.de>
3117 PR tree-optimization/87328
3118 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
3119 visiting unexecutable backedges when not iterating.
3120 (do_rpo_vn): Mark all edges not executable even when not
3123 2018-09-17 Martin Jambor <mjambor@suse.cz>
3126 * doc/invoke.texi (Warning Options): Likewise.
3128 2018-09-17 Richard Biener <rguenther@suse.de>
3130 PR tree-optimization/87301
3131 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
3132 clean EH info from leftover copy assignments.
3134 2018-09-17 Martin Liska <mliska@suse.cz>
3136 PR gcov-profile/85871
3137 * gcov.c (output_intermediate_file): Fix out of bounds
3140 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
3142 * config/arc/arc.c: Object attributes for core4 not reflected
3144 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
3147 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
3149 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
3151 2018-09-17 Martin Liska <mliska@suse.cz>
3153 * doc/gcov.texi: Document new option --use-hotness-colors.
3154 * gcov.c (struct source_info): Declare new field.
3155 (source_info::source_info): Set default for maximum_count.
3156 (print_usage): Add new -q option.
3157 (process_args): Process it.
3158 (accumulate_line_info): Save src->maximum_count.
3159 (output_line_beginning): Make color line number if
3160 flag_use_hotness_colors is set.
3161 (output_line_details): Pass default argument value.
3162 (output_lines): Pass src->maximum_count.
3164 2018-09-17 Martin Liska <mliska@suse.cz>
3166 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
3167 Use processor_names table.
3168 * config/i386/i386.c (ix86_default_align): Use
3169 processor_cost_table for alignment values.
3170 (ix86_option_override_internal): Use processor_names.
3171 (ix86_function_specific_print): Likewise.
3172 * config/i386/i386.h (struct processor_costs):
3173 Add alignment values.
3174 (struct ptt): Remove and replace with const char *.
3175 * config/i386/x86-tune-costs.h (struct processor_costs):
3176 Declare default alignments for all costs.
3178 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
3180 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
3181 symbolics or VR_VARYING ranges for ABS_EXPR.
3182 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
3183 when range will wrap.
3185 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
3188 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
3189 before and after a JUMP_TABLE_DATA.
3191 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
3194 * dojump.c (do_compare_and_jump): Canonicalize function pointers
3195 when one operand is a function pointer. Use POINTER_TYPE_P and
3196 FUNC_OR_METHOD_TYPE_P.
3197 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3198 * fold-const.c (build_range_check): Likewise.
3199 * match.pd (simple_comparison): Likewise.
3201 2018-09-14 David Malcolm <dmalcolm@redhat.com>
3204 * spellcheck.c (get_edit_distance_cutoff): New function.
3205 (selftest::test_edit_distance_unit_test_oneway): Rename to...
3206 (selftest::test_get_edit_distance_one_way): ...this.
3207 (selftest::test_get_edit_distance_unit): Rename to...
3208 (selftest::test_get_edit_distance_both_ways): ...this.
3209 (selftest::test_edit_distances): Move tests to this new function,
3210 and test some more pairs of strings. Update for above renaming.
3211 (selftest::get_old_cutoff): New function.
3212 (selftest::test_get_edit_distance_cutoff): New function.
3213 (selftest::assert_suggested_for): New function.
3214 (ASSERT_SUGGESTED_FOR): New macro.
3215 (selftest::assert_not_suggested_for): New function.
3216 (ASSERT_NOT_SUGGESTED_FOR): New macro.
3217 (selftest::test_suggestions): New function.
3218 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
3219 tests to selftest::test_edit_distances and call it. Add calls to
3220 selftest::test_get_edit_distance_cutoff and
3221 selftest::test_suggestions.
3222 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
3223 (best_match::consider): Replace hard-coded cutoff calculation with
3225 (best_match::get_cutoff): New declaration.
3226 (best_match::get_best_meaningful_candidate): Likewise.
3228 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
3230 * builtins.c (fold_builtin_strlen): Remove TODO comment.
3232 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
3235 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3237 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3238 terminated string literal.
3240 2018-09-14 Martin Sebor <msebor@redhat.com>
3242 * builtins.c (unterminated_array): Handle ARRAY_REF.
3243 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
3244 * builtins.h (unterminated_array): Declare extern.
3245 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
3247 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
3250 2018-09-14 Martin Sebor <msebor@redhat.com>
3251 Jeff Law <law@redhat.com>
3253 * builtins.c (unterminated_array): New.
3254 (expand_builtin_strcpy): Adjust.
3255 (expand_builtin_strcpy_args): Detect unterminated arrays.
3256 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
3257 unterminated arrays.
3258 * gimple-fold.h (get_maxval_strlen): Add argument.
3259 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
3261 * gimple-fold.c (get_range_strlen): Add argument.
3262 (get_maxval_strlen): Adjust.
3263 * gimple-fold.h (get_range_strlen): Add argument.
3265 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
3267 * config/i386/movdirintrin.h: Fix copyright year.
3269 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
3271 * reg-stack.c: Include regs.h.
3272 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
3273 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
3274 FIRST_STACK_REG, not DFmode.
3275 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
3276 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
3277 (change stack): Default register mode to the reg_raw_mode of
3278 FIRST_STACK_REG, not DFmode.
3279 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
3280 (*swapxf): Rename from swapxf.
3282 2018-09-14 Carl Love <cel@us.ibm.com>
3284 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
3285 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
3287 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
3290 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
3293 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
3296 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
3298 2018-09-14 Jason Merrill <jason@redhat.com>
3300 Fix --enable-gather-detailed-mem-stats.
3301 * hash-table.c (hash_table_usage): Change from variable to function.
3302 * hash-table.h: Adjust.
3303 * Makefile.in: Add missing dependencies on hash-table.h.
3305 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3307 PR tree-optimization/87259
3309 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
3310 execute_cse_reciprocals_1 has tried transforming.
3312 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
3314 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
3315 VR_VARYING for PLUS/MINUS_EXPR.
3317 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
3319 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
3322 2018-09-14 Richard Biener <rguenther@suse.de>
3325 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
3326 bits for the merged partition.
3328 2018-09-13 Martin Sebor <msebor@redhat.com>
3329 Bernd Edlinger <bernd.edlinger@hotmail.de>
3331 * builtins.h (c_srlen): Add argument.
3332 * builtins.c (warn_string_no_nul): New function.
3333 (c_strlen): Add argument and use it. Update recursive calls.
3334 Pass DECL argument to string_constant to get info on non
3335 terminated strings. Update *NONSTR as needed.
3336 (fold_builtin_strlen): Add argument to calls to c_strlen.
3337 Warn for unterminated arrays.
3338 (warn_string_no_null): Add prototype.
3339 * expr.c (string_constant): Update arguments. Update recursive
3340 calls appropriately. Detect missing NUL terminator and outermost
3341 declaration its missing in.
3342 Improve checks for arrays with nonzero lower bound or elements
3343 that are not a single byte. Simplify offset computation.
3344 Simplify checks for non-NUL terminated strings.
3345 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
3346 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
3348 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
3350 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
3352 * fold-const.c (c_getstr): Fix function comment. Remove unused third
3353 argument. Fix range checks.
3354 * fold-const.h (c_getstr): Adjust protoype.
3355 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
3356 string is constant but contains no NUL byte.
3358 * expr.c (string_constant): Adjust function comment.
3359 Remove bogus check for zero termination.
3361 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
3363 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
3364 (get_constant_size): Don't make STRING_CSTs larger than they are.
3365 (check_string_literal): New check function for STRING_CSTs.
3366 (output_constant): Use it.
3368 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
3371 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
3373 2018-09-13 Richard Biener <rguenther@suse.de>
3375 PR tree-optimization/87263
3376 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
3377 (struct unwind_state): Add max_rpo field.
3378 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
3379 Compute max_rpo, the max RPO number a block can be backwards reached
3380 from. Re-write non-iterating mode to a RPO ordered worklist approach,
3381 separating it from the iterating mode.
3383 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
3385 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
3386 (rfs_decision): New scheduling decision.
3388 2018-09-13 Richard Biener <rguenther@suse.de>
3391 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
3392 (vn_nary_op_insert_pieces_predicated): Do not write useless
3393 valid_dominated_by_p entry outside of the allocated storage.
3395 2018-09-13 Omar Sandoval <osandov@osandov.com>
3396 Tom de Vries <tdevries@suse.de>
3399 * dwarf2out.c (is_c): New function.
3400 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
3402 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
3405 * config/aarch64/aarch64.md (*aarch64_bfxil):
3407 * config/aarch64/constraints.md (Ulc): Define.
3408 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
3410 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
3413 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
3415 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
3416 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
3417 aarch64_layout_frame call.
3418 (aarch64_expand_epilogue): Likewise.
3419 (aarch64_initial_elimination_offset): Likewise.
3420 (aarch64_get_separate_components): Likewise.
3421 (aarch64_use_return_insn_p): Likewise.
3422 (aarch64_layout_frame): Remove unneeded check.
3424 2018-09-13 Jakub Jelinek <jakub@redhat.com>
3426 * configure.ac: Only append
3427 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
3428 gcc_config_arguments if it was never reconfigured or last reconfigure
3429 was with different arguments.
3430 * configure: Regenerated.
3432 2018-09-13 Jakub Jelinek <jakub@redhat.com>
3433 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3436 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
3437 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
3439 2018-09-13 Jakub Jelinek <jakub@redhat.com>
3441 PR tree-optimization/87287
3442 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
3443 X % C == 0 to X % (unsigned) C == 0 optimization to ...
3444 * match.pd (X % C == 0): ... here. New optimization.
3446 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3449 * expr.h (maybe_optimize_mod_cmp): Declare.
3450 * expr.c (mod_inv): New function.
3451 (maybe_optimize_mod_cmp): New function.
3452 (do_store_flag): Use it.
3453 * cfgexpand.c (expand_gimple_cond): Likewise.
3455 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
3456 Julian Brown <julian@codesourcery.com>
3459 * gimplify.c (gimplify_scan_omp_clauses): Set
3460 target_firstprivatize_array_bases in OpenACC parallel and kernels
3461 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
3462 OpenACC data regions.
3464 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
3466 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
3467 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
3469 2018-09-12 Richard Biener <rguenther@suse.de>
3471 PR tree-optimization/87280
3472 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
3473 edge but unreachable target.
3474 (do_rpo_vn): For conservatively handling a PHI only mark
3475 the backedge executable but not the block reachable.
3477 2018-09-12 Richard Biener <rguenther@suse.de>
3479 PR tree-optimization/87266
3480 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
3483 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
3485 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
3487 ("trunc<BFP:mode><DFP_ALL:mode>2")
3488 ("trunc<DFP_ALL:mode><BFP:mode>2")
3489 ("extend<BFP:mode><DFP_ALL:mode>2")
3490 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
3491 according to the target operand type.
3493 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3494 Andreas Krebbel <krebbel@linux.ibm.com>
3496 PR tree-optimization/86844
3497 * gimple-ssa-store-merging.c
3498 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
3499 there are any overlapping stores in between them, make sure they are
3500 also coalesced or we give up completely.
3502 2018-09-12 Jakub Jelinek <jakub@redhat.com>
3505 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
3506 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
3508 2018-09-12 Tom de Vries <tdevries@suse.de>
3510 * common.opt (gdescribe-dies): Add option.
3511 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
3512 attribute for artifical and nameless decls.
3513 (dwarf2out_register_external_die): Add description attribute to
3514 external reference die.
3515 (add_desc_attribute): New functions.
3516 (gen_subprogram_die): Add description attribute to
3517 DW_TAG_call_site_parameter.
3518 * tree-pretty-print.c (print_generic_expr_to_str): New function.
3519 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
3520 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
3522 (@item -gdescribe-dies): Add.
3524 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
3526 * tree-vrp.c (vrp_shift_undefined_p): Remove.
3527 (extract_range_from_binary_expr_1: Call
3528 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
3529 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
3532 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
3534 * gimple-ssa-warn-alloca.c
3535 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
3536 field for ALLOCA_BOUND_*_LARGE.
3538 2018-09-11 Nathan Sidwell <nathan@acm.org>
3540 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
3542 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
3544 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
3545 for clobbers. Remove obsolete comment.
3547 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
3549 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
3550 mpxchk, mpxld and mpxst types.
3551 (define_attr length_immediate): Remove all processing of mpx types.
3552 (define_attr prefix_0f): Ditto.
3553 (define_attr memory): Ditto.
3555 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
3557 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
3558 (log<mode>2): Change operand 1 predicate to general_operand.
3559 Extend operand 1 to XFmode and generate logxf3 insn.
3560 (log10<mode>2): Change operand 1 predicate to general_operand.
3561 Extend operand 1 to XFmode and generate log10xf3 insn.
3562 (log2<mode>2): Change operand 1 predicate to general_operand.
3563 Extend operand 1 to XFmode and generate log2xf3 insn.
3564 (fyl2xp1_extend<mode>xf3_i387): Remove.
3565 (log1p<mode>2): Change operand 1 predicate to general_operand.
3566 Extend operand 1 to XFmode and generate log1pxf3 insn.
3567 (fxtract_extend<mode>xf3_i387): Remove.
3568 (logb<mode>2): Change operand 1 predicate to general_operand.
3569 Extend operand 1 to XFmode and generate logbxf3 insn.
3570 (ilogb<mode>2): Change operand 1 predicate to general_operand.
3571 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
3572 (significand<mode>2): Change operand 1 predicate to general_operand.
3573 Extend operand 1 to XFmode and generate significandxf3 insn.
3575 2018-09-11 Nathan Sidwell <nathan@acm.org>
3577 * gcc.c (perror_with_name, pfatal_with_name): Delete.
3578 (load_specs): Use fatal_error.
3579 (DELETE_IF_ORDINARY, process_command): Use error.
3580 (execute, run_attempt): Use fatal_error.
3582 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
3584 * diagnostic-core.h (sorry_at): New prototype.
3585 * diagnostic.c (sorry_at): New function.
3587 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
3589 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
3590 by zero as VR_UNDEFINED.
3592 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
3594 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
3595 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
3596 (<sincos>mode2): New expander.
3597 (sincos_extend<mode>xf3_i387): Remove insn pattern.
3598 (sincos -> sin, cos splitters): Remove splitter patterns.
3599 (sincos<mode>3): Change operand 2 predicate to general_operand.
3600 Extend operand 2 to XFmode and generate sincosxf3 insn.
3601 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
3602 Change operand 3 predicate to const1_operand.
3603 (fptan_extend<mode>xf4_i387): Remove insn pattern.
3604 (tanxf2): Update operands in the call to fptanxf4_i387.
3605 (tan<mode>2): Change operand 1 predicate to general_operand.
3606 Extend operand 1 to XFmode and generate tanxf3 insn.
3607 (atan2xf3): Rename from *fpatanxf3_i387.
3608 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
3609 (atan2xf3): Remove expander.
3610 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
3611 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
3612 (atan<mode>2): Change operand 1 predicate to general_operand.
3613 Extend operand 1 to XFmode and generate atanxf3 insn.
3615 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
3617 * config/i386/i386.md (x87/SSE constant load splitter): Use
3618 memory_operand instead of nonimmediate_operand for input operand
3621 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
3623 * config/i386/i386.md (float partial SSE register stall splitter): Move
3624 splitter near its instruction pattern.
3625 (float_extend partial SSE register stall splitter): Ditto.
3626 (float_truncate partial SSE register stall splitter): Ditto.
3628 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
3631 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3632 to speculation_safe_value_not_needed.
3635 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
3636 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
3637 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
3638 leaf_function_p, instead use has_hard_reg_initial_val.
3640 2018-09-09 Nathan Sidwell <nathan@acm.org>
3642 * gcc.h (pfatal_with_name): Don't declare here.
3643 * gcc.c (pfatal_with_name): Make static.
3645 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
3647 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
3650 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
3652 PR rtl-optimization/85458
3653 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
3654 priority hook to reduce the priority of EXPR.
3656 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
3658 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
3659 DImode for x87 on 32bit targets. Conditionally disable x87 modes
3660 with X87_ENABLE_FLOAT. Remove preparation code.
3661 (*float<SWI48:mode><MODEF:mode>2): Rename from
3662 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
3663 math using "enabled" attribute.
3664 (*floatdi<MODEF:mode>2_i387): Rename from
3665 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
3666 enable for 32bit targets only.
3667 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
3669 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
3670 as operand 1 predicate. Rewrite as define_insn_and_split.
3671 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
3673 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
3675 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
3676 to fallthru to FLOAT case.
3678 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3681 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
3682 around folding of vec_sl to handle out of range shift values.
3684 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
3686 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
3687 Update callers to gen_fix_trunc<mode>_i387_fisttp
3688 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
3689 nonimmediate_operand.
3690 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
3691 and corresponding splitters.
3692 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
3693 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
3694 (fix_truncdi_i387_with_temp): Remove insn pattern
3695 and corresponding splitters.
3696 (fix_trunc<mode>_i387): Change operand 0 predicate to
3697 nonimmediate_operand.
3698 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
3699 and corresponding splitters.
3700 (*fistdi2_1): Remove.
3702 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
3703 (lrintxfdi2): Remove expander. Reimplement as define_insn.
3704 (*fist<mode>2_1): Remove.
3705 (fist<mode>2): Ditto.
3706 (fist<mode>2_with_temp): Remove insn pattern and corresponding
3708 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
3709 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
3710 (fistdi2_<rounding>): Change operand 0 predicate to
3711 nonimmediate_operand.
3712 (fistdi2_<rounding>_with_temp): Remove insn pattern
3713 and corresponding splitters.
3714 (fist<mode>2_<rounding>): Change operand 0 predicate to
3715 nonimmediate_operand.
3716 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
3717 and corresponding splitters.
3719 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
3721 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
3723 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
3726 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3728 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
3729 early gimple folding of vec_splat().
3730 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
3731 * gimple-fold.h: Add an extern define for tree_vec_extract().
3733 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3735 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
3736 wrappers around TREE_TYPE comparisons.
3738 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3741 * config/s390/predicates.md: Add nonsym_memory_operand.
3742 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
3743 contains a SYMBOL_REF, load it into an intermediate pseudo.
3744 (s390_emit_compare_and_swap): Legitimize operand.
3745 * config/s390/s390.md: Use the new nonsym_memory_operand
3746 with UNSPECV_CAS patterns.
3748 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3751 * config/s390/s390-passes.def: New file.
3752 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3754 (make_pass_s390_early_mach): Add declaration.
3755 * config/s390/s390.c (make_pass_s390_early_mach):
3756 (s390_option_override): Remove dynamic registration.
3757 * config/s390/t-s390: Add s390-passes.def.
3759 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3761 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3762 Remove UNSPEC_LTREL_BASE check.
3763 (annotate_constant_pool_refs): Likewise.
3764 (find_constant_pool_ref): Likewise.
3765 (find_ltrel_base): Removed.
3766 (replace_ltrel_base): Removed.
3767 (s390_mainpool_finish): Remove replace_ltrel_base call.
3768 (s390_chunkify_start): Remove pending LTREL_BASE logic.
3769 (s390_chunkify_finish): Remove replace_ltrel_base call.
3770 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3772 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
3775 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3776 to speculation_safe_value_not_needed.
3778 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
3779 Bernd Schmidt <bernds_cb1@t-online.de>
3781 * config/nvptx/nvptx-opts.h: New file.
3782 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3783 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3785 (TARGET_SM35): New macro.
3786 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3788 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3790 (misa=): New option.
3791 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3793 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
3795 * config/i386/i386.md (truncdfsf2): Remove expander.
3796 (truncdfsf2_with_temp): Ditto.
3797 (truncxf<mode>2): Ditto.
3798 (*truncdfsf_fast_mixed): Remove insn pattern.
3799 (*truncdfsf_fast_i387): Ditto.
3800 (*truncdfsf_mixed): Ditto.
3801 (*truncdfsf_i387): Ditto.
3802 (*truncdfsf2_i387_1): Ditto.
3803 (*truncxfsf2_mixed): Ditto.
3804 (*truncxfdf2_mixed): Ditto.
3805 (*truncxf<mode>2_i387_noop): Ditto. Update callers
3806 to call gen_truncxf<mode>2 instead.
3807 (*truncxf<mode>2_i387): Remove.
3808 (reg->reg splitters): Remove splitter pattern.
3809 (reg->mem splitters): Ditto.
3811 (truncdfsf2): New insn pattern.
3812 (truncxf<mode>2): Ditto.
3814 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3816 * tree-ssa-math-opts.c (is_mult_by): New function.
3817 (is_square_of): Use the above.
3818 (optimize_recip_sqrt): New function.
3819 (pass_cse_reciprocals::execute): Use the above.
3821 2018-09-05 Richard Biener <rguenther@suse.de>
3824 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3825 to zero-init the emplaced vec.
3827 2018-09-05 Martin Liska <mliska@suse.cz>
3829 PR tree-optimization/87205
3830 * tree-switch-conversion.c (pass_lower_switch::execute):
3831 Group cases for switch statements.
3833 2018-09-05 Richard Biener <rguenther@suse.de>
3835 PR tree-optimization/87217
3836 * tree-ssa-sccvn.c (vuse_valueize): New.
3837 (vn_reference_lookup_pieces): Use it.
3838 (vn_reference_lookup): Likewise.
3840 2018-09-05 Nathan Sidwell <nathan@acm.org>
3843 * stor-layout.c (place_field): Scan forwards to check last
3844 bitfield when ms_bitfield_placement is in effect.
3846 2018-09-05 Richard Biener <rguenther@suse.de>
3849 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3852 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
3853 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3855 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
3856 * config.gcc (extra_objs): Build it.
3857 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
3859 * config/aarch64/aarch64-passes.def
3860 (pass_tag_collision_avoidance): New pass.
3861 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
3862 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
3863 (aarch64_classify_address): Remove static qualifier.
3864 (aarch64_address_info, aarch64_address_type): Move to...
3865 * config/aarch64/aarch64-protos.h: ... here.
3866 (make_pass_tag_collision_avoidance): New function.
3867 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3870 2018-09-05 Martin Liska <mliska@suse.cz>
3872 * doc/gcov.texi: Update documentation of humar
3874 * gcov.c (format_count): Print one decimal place, it provides
3875 more fine number of situations like '1G' vs. '1.4G'.
3877 2018-09-05 Martin Liska <mliska@suse.cz>
3880 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3881 * optc-gen.awk: Allow 'Var' for Deprecated options in order
3882 to generate a MASK value.
3884 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
3887 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3888 if frame pointer isn't used.
3889 (compute_frame_pointer_to_fb_displacement): Likewise.
3891 2018-09-04 Jakub Jelinek <jakub@redhat.com>
3894 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3895 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3896 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3897 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3898 and OPTION_MASK_ISA_XSAVEC_UNSET.
3900 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
3902 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3903 XOR operations in NAND case.
3905 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3907 * wide-int-range.cc (wide_int_range_convert): New.
3908 * wide-int-range.h (wide_int_range_convert): New.
3909 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3910 code into wide_int_range_convert.
3911 (extract_range_into_wide_ints): Do not munge anti range constants
3912 into the entire domain. Just return the range back.
3914 2018-09-04 Martin Liska <mliska@suse.cz>
3916 * genmatch.c (output_line_directive): Add new argument
3918 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3920 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3922 * doc/invoke.texi (Option Summary): Add whitespace.
3924 * doc/invoke.texi (Option Summary): Add -Waligned-new.
3926 2018-09-04 Richard Biener <rguenther@suse.de>
3928 PR tree-optimization/87211
3929 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3930 backedge value we're supposed to treat as VARYING also number
3931 the PHI to VARYING in case it got a different value-number already.
3933 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3935 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3936 (extract_range_from_binary_expr_1): Do not call
3937 vrp_can_optimize_bit_op.
3938 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3940 (wide_int_range_get_mask_and_bounds): New.
3941 (wide_int_range_optimize_bit_op): New.
3942 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3943 (wide_int_range_bit_and): Same.
3944 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3945 (wide_int_range_optimize_bit_op): New.
3946 (wide_int_range_get_mask_and_bounds): New.
3948 2018-09-04 Richard Biener <rguenther@suse.de>
3950 PR tree-optimization/87176
3951 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3952 variable. When value-numbering a virtual PHI node make sure
3953 to not value-number to the backedge value.
3955 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3957 * doc/extend.texi (Long Long, Hex Floats): Document support for
3958 long long and hex floats in more recent versions of ISO C++.
3960 2018-09-03 Richard Biener <rguenther@suse.de>
3962 PR tree-optimization/87177
3963 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3966 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3968 * bb-reorder.c (edge_order): Convert to C-qsort-style
3969 tri-state comparator.
3970 (reorder_basic_blocks_simple): Change std::stable_sort to
3973 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3975 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3976 tri-state comparator.
3977 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3979 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3981 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
3982 (mergesort): ... here as maximum count for using netsort.
3983 (gcc_qsort): Set nlim to 3 if stable sort is requested.
3984 (gcc_stablesort): New.
3985 * system.h (gcc_stablesort): Declare.
3987 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3989 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3990 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
3991 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
3993 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3995 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3996 lxsdx and stxsdx alternatives.
3997 (*mov<mode>_hardfloat64): Ditto.
3998 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4000 2018-09-03 Richard Biener <rguenther@suse.de>
4002 PR tree-optimization/87200
4003 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4006 2018-09-03 Martin Liska <mliska@suse.cz>
4008 PR tree-optimization/87201
4009 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4010 Fix parenthesis in an expression.
4012 2018-09-03 Richard Biener <rguenther@suse.de>
4014 PR tree-optimization/87197
4015 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4016 visited. CSE the VN_INFO hashtable lookup.
4018 PR tree-optimization/87169
4019 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4020 iterating make sure there's no extra backedges from irreducible
4021 regions feeding the header. Mark the destination block
4024 2018-09-03 Martin Liska <mliska@suse.cz>
4027 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4028 * common/common-targhooks.c (default_get_valid_option_values):
4030 * common/common-targhooks.h (default_get_valid_option_values):
4032 * common/config/i386/i386-common.c: Move processor_target_table
4034 (ix86_get_valid_option_values): New function.
4035 (TARGET_GET_VALID_OPTION_VALUES): New macro.
4036 * config/i386/i386.c (struct ptt): Move to i386-common.c.
4037 (PTA_*): Move all defined masks into i386-common.c.
4038 (ix86_function_specific_restore): Use new processor_cost_table.
4039 * config/i386/i386.h (struct ptt): Moved from i386.c.
4040 (struct pta): Likewise.
4041 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4042 * doc/tm.texi.in: Likewise.
4043 * opt-suggestions.c (option_proposer::suggest_option):
4044 Pass prefix to build_option_suggestions.
4045 (option_proposer::get_completions): Likewise.
4046 (option_proposer::build_option_suggestions): Use the new target
4048 * opts.c (struct option_help_tuple): New struct.
4049 (print_filtered_help): Use the new target hook.
4051 2018-09-03 Martin Liska <mliska@suse.cz>
4054 * predict.c (set_even_probabilities): Add likely_edges
4055 argument and handle cases where we have precisely one
4057 (combine_predictions_for_bb): Catch also likely_edges.
4058 (tree_predict_by_opcode): Handle gswitch statements.
4059 * tree-cfg.h (find_case_label_for_value): New declaration.
4060 (find_taken_edge_switch_expr): Likewise.
4061 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4062 Find pivot in decision tree based on probabily, not by number of
4065 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
4067 * doc/standards.texi (Standards): Update Objective-C reference.
4069 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
4071 * doc/install.texi (Prerequisites): Update link for MPC.
4073 2018-09-01 Michael Matz <matz@suse.de>
4075 PR tree-optimization/87074
4076 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
4077 PHIs for outer-loop uses.
4079 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
4081 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
4082 * doc/invoke.texi (C Dialect Options): Ditto.
4084 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
4086 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
4088 2018-08-31 Richard Biener <rguenther@suse.de>
4090 PR tree-optimization/87168
4091 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
4092 (rpo_elim::eliminate_avail): When OP was not visited it must
4095 2018-08-31 David Malcolm <dmalcolm@redhat.com>
4097 * tree-vrp.c (copy_value_range): Convert param "from" from
4098 "value_range *" to "const value_range *".
4099 (range_is_null): Likewise for param "vr".
4100 (range_int_cst_p): Likewise.
4101 (range_int_cst_singleton_p): Likewise.
4102 (symbolic_range_p): Likewise.
4103 (value_ranges_intersect_p): Likewise for both params.
4104 (value_range_nonnegative_p): Likewise for param "vr".
4105 (value_range_constant_singleton): Likewise.
4106 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
4107 (extract_range_into_wide_ints): Likewise for param "vr".
4108 (extract_range_from_multiplicative_op): Likewise for params "vr0"
4110 (vrp_can_optimize_bit_op): Likewise.
4111 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
4113 (extract_range_from_unary_expr): Likewise.
4114 (debug_value_range): Likewise for param "vr".
4115 (value_range::dump): Add "const" qualifier.
4116 (vrp_prop::check_array_ref): Convert local "vr" from
4117 "value_range *" to "const value_range *".
4118 (vrp_prop::check_mem_ref): Likewise.
4119 (vrp_prop::visit_stmt): Likewise for local "old_vr".
4120 (vrp_intersect_ranges_1): Likewise for param "vr_1".
4121 (vrp_intersect_ranges): Likewise.
4122 (simplify_stmt_for_jump_threading): Likewise for local "vr".
4123 (vrp_prop::vrp_finalize): Likewise.
4124 * tree-vrp.h (value_range::dump): Add "const" qualifier.
4125 (vrp_intersect_ranges): Add "const" qualifier to params as above.
4126 (extract_range_from_unary_expr): Likewise.
4127 (value_range_constant_singleton): Likewise.
4128 (symbolic_range_p): Likewise.
4129 (copy_value_range): Likewise.
4130 (extract_range_from_binary_expr_1): Likewise.
4131 (range_int_cst_p): Likewise.
4132 (vrp_set_zero_nonzero_bits): Likewise.
4133 (range_int_cst_singleton_p): Likewise.
4135 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
4137 * config/aarch64/arm_neon.h (vabsd_s64): New.
4138 (vnegd_s64): Likewise.
4140 2018-08-31 Martin Jambor <mjambor@suse.cz>
4142 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
4144 2018-08-31 Martin Liska <mliska@suse.cz>
4146 * ipa-icf.c (sem_item::add_type): Use
4147 sem_item::m_type_hash_cache.
4148 * ipa-icf.h: Move the cache from sem_item_optimizer
4151 2018-08-31 Nathan Sidwell <nathan@acm.org>
4153 * doc/extend.texi (Backwards Compatibility): Remove implicit
4154 extern C leeway of () being (...).
4156 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4158 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
4160 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
4164 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
4166 2018-08-31 Jakub Jelinek <jakub@redhat.com>
4169 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
4170 gen_int_mode. Formatting fixes.
4172 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
4174 * target.def (custom_function_descriptors): Improve documentation.
4175 * doc/tm.texi.in (Trampolines): Expand discussion of function
4176 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
4177 beginning of the section.
4178 * doc/tm.texi: Regenerated.
4180 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
4182 * cfg.h (class auto_edge_flag): Spell out the template-id of the
4183 base class in the initializer list. This is a workaround for
4184 building with older GCC.
4185 (class auto_bb_flag): Likewise.
4187 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
4189 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
4190 (altivec_vcmpequ<VI_char>_p): Remove star.
4191 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
4193 (expand_strncmp_vec_sequence): New function.
4194 (emit_final_str_compare_vec): New function.
4195 (expand_strn_compare): Add support for vector strncmp.
4196 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
4197 length specification to bytes.
4198 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
4199 (vcmpnezb_p): New pattern.
4200 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
4201 for option -mstring-compare-inline-limit.
4203 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
4205 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
4206 (PTA_SKYLAKE): Add PTA_AES.
4207 (PTA_GOLDMONT): Likewise.
4209 2018-08-29 Jan Hubicka <jh@suse.cz>
4212 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
4213 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
4215 2018-08-29 Jan Hubicka <jh@suse.cz>
4217 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
4219 (hash_tree): Do not visit TYPE_STUB_DECL.
4220 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
4221 stream TYPE_STUB_DECL.
4222 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
4223 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
4224 after free_lang_data.
4225 (type_in_anonymous_namespace_p): Likewise.
4227 2018-08-29 Jan Hubicka <jh@suse.cz>
4229 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4230 comment that it has to be even number.
4231 (class sreal): Change m_sig type to int32_t.
4232 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4233 int64_t for temporary calculations.
4234 (sreal_verify_basics): Drop one bit from minimum and maximum.
4236 2018-08-30 Richard Biener <rguenther@suse.de>
4238 PR tree-optimization/87147
4239 * tree-ssa-sccvn.c (SSA_VISITED): New function.
4240 (visit_phi): When the degenerate result is from the backedge and
4241 we didn't visit its definition yet drop to VARYING.
4242 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
4244 2018-08-29 Jan Hubicka <jh@suse.cz>
4246 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
4248 (hash_tree): Likewise.
4250 2018-08-29 Jan Hubicka <jh@suse.cz>
4252 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
4253 and TYPE_NEXT_REF_TO.
4255 2018-08-29 Jan Hubicka <jh@suse.cz>
4257 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4258 comment that it has to be even number.
4259 (class sreal): Change m_sig type to int32_t.
4260 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4261 int64_t for temporary calculations.
4262 (sreal_verify_basics): Drop one bit from minimum and maximum.
4264 2018-08-30 Tamar Christina <tamar.christina@arm.com>
4266 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
4268 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
4271 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
4272 if to_add is negative.
4274 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
4277 * builtins.c (c_strlen): Improve range checks.
4279 2018-08-29 Martin Sebor <msebor@redhat.com>
4280 Jeff Law <law@redhat.com>
4282 PR tree-optimization/86714
4283 PR tree-optimization/86711
4284 * builtins.c (c_strlen): Add arguments to call to string_constant.
4285 * expr.c (string_constant): Add argument. Detect missing nul
4286 terminator and outermost declaration it's missing in.
4287 * expr.h (string_constant): Add argument.
4288 * fold-const.c (read_from_constant_string): Add arguments to call to
4290 (c_getstr): Likewise.
4291 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4293 * tree-ssa-strlen.c (get_stridx): Likewise.
4295 2018-08-29 Jan Hubicka <jh@suse.cz>
4297 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
4298 Do not stream DECL_VINDEX.
4299 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
4300 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
4301 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
4303 2018-08-29 Richard Biener <rguenther@suse.de>
4305 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
4306 virtual operands that are not default defs to honor region
4308 (rpo_vn_valueize): Remove ineffective code here.
4310 2018-08-29 Richard Biener <rguenther@suse.de>
4312 PR tree-optimization/87132
4313 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
4314 when skipping defs reachable over backedges.
4316 2018-08-29 Richard Biener <rguenther@suse.de>
4318 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
4319 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
4320 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
4321 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
4322 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
4323 (vn_reference_lookup_3): Remove use of const_parms.
4324 (free_rpo_vn): Do not free const_parms.
4325 (do_rpo_vn): Do not call init_const_parms.
4326 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
4327 SSA_NAME_POINTS_TO_READONLY_MEMORY.
4328 (call_may_clobber_ref_p_1): Likewise.
4330 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
4333 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
4335 (-ftree-scev-cprop): Document.
4337 2018-08-29 Jan Hubicka <jh@suse.cz>
4339 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
4341 (sreal constructor): Update.
4342 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
4343 sreal:operator/): Update.
4345 2018-08-29 Martin Liska <mliska@suse.cz>
4347 * tree-switch-conversion.c (switch_conversion::expand):
4348 Strenghten assumption about gswitch statements.
4350 2018-08-29 Richard Biener <rguenther@suse.de>
4352 PR tree-optimization/87117
4353 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
4354 re-value-number released SSA VDEFs.
4356 2018-08-29 Richard Biener <rguenther@suse.de>
4358 PR tree-optimization/87126
4359 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
4361 2018-08-28 Jim Wilson <jimw@sifive.com>
4363 * config/riscv/pic.md: Rewrite.
4364 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
4366 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
4367 (SOFTF, default_load, softload, softstore): New.
4369 2018-08-28 Jeff Law <law@redhat.com>
4371 * fold-const.c (fold_binary_loc): Remove recently added assert.
4373 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
4375 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
4376 to OP parmeter of generated function.
4378 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
4380 PR tree-optimization/87009
4381 * match.pd: Add boolean optimizations.
4383 2018-08-28 Martin Sebor <msebor@redhat.com>
4386 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
4387 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
4388 (pass_walloca::gate): Use it.
4389 (alloca_call_type): Same.
4390 (pass_walloca::execute): Same.
4391 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
4393 2018-08-28 David Malcolm <dmalcolm@redhat.com>
4395 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
4396 GCC_VERSION for usage of "__gcc_dump_printf__" format from
4399 2018-08-28 Richard Biener <rguenther@suse.de>
4401 PR tree-optimization/87124
4402 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
4403 constants before looking up avail.
4405 2018-08-28 Jakub Jelinek <jakub@redhat.com>
4408 * calls.c (maybe_warn_nonstring_arg): Punt early if
4409 warn_stringop_overflow is zero. Don't call get_range_strlen
4410 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
4411 Swap comparison operands to have constants on rhs. Only use
4412 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
4413 increment lenrng[0].
4415 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
4417 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
4418 use of tree_to_shwi. Remove duplicated test for the size being
4419 a whole number of bytes.
4421 2018-08-28 Richard Biener <rguenther@suse.de>
4423 PR tree-optimization/87117
4424 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
4425 Handle removed stmt without LHS (GIMPLE_NOP).
4427 2018-08-28 Richard Biener <rguenther@suse.de>
4429 PR tree-optimization/87117
4430 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
4431 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
4433 2018-08-28 Richard Biener <rguenther@suse.de>
4435 PR tree-optimization/87117
4436 * tree-ssa-pre.c (compute_avail): Do not make expressions
4437 with predicated values available.
4438 (get_expr_value_id): Assert we do not run into predicated value
4441 2018-08-28 Richard Biener <rguenther@suse.de>
4443 PR tree-optimization/87117
4444 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
4445 get virtual operands.
4446 (get_expr_operands): Handle STRING_CST like other decls.
4448 2018-08-28 Martin Liska <mliska@suse.cz>
4450 * tree.h: Update documentation of fndecl_built_in_p
4454 2018-08-27 Jeff Law <law@redhat.com>
4455 PR tree-optimization/87110
4456 * tree-ssa-dse.c (compute_trims): Handle non-constant
4459 2018-08-27 Martin Sebor <msebor@redhat.com>
4461 PR tree-optimization/86914
4462 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
4464 2018-08-27 Martin Sebor <msebor@redhat.com>
4466 PR tree-optimization/87112
4467 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
4468 the type of the bound argument.
4470 2018-08-27 Jeff Law <law@redhat.com>
4472 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
4473 type does not have a TYPE_SIZE_UNIT.
4475 2018-08-27 Steve Ellcey <sellcey@cavium.com>
4477 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
4478 with include of backend.h.
4480 2018-08-27 Richard Biener <rguenther@suse.de>
4482 PR tree-optimization/86927
4483 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
4484 use const cond reduction code.
4486 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
4488 PR tree-optimization/85758
4489 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
4491 2018-08-27 David Malcolm <dmalcolm@redhat.com>
4494 * diagnostic-show-locus.c (class layout_range): Update for
4495 conversion of show_caret_p to a tri-state.
4496 (layout_range::layout_range): Likewise.
4497 (make_range): Likewise.
4498 (layout::maybe_add_location_range): Likewise.
4499 (layout::should_print_annotation_line_p): Don't show annotation
4500 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
4501 (layout::get_state_at_point): Update for conversion of
4502 show_caret_p to a tri-state. Bail out early for
4503 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
4504 underlining or source colorization.
4505 (gcc_rich_location::add_location_if_nearby): Update for conversion
4506 of show_caret_p to a tri-state.
4507 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
4508 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
4510 (selftest::test_one_liner_labels): Likewise.
4511 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
4512 conversion of show_caret_p to a tri-state.
4513 * pretty-print.c (text_info::set_location): Likewise.
4514 * pretty-print.h (text_info::set_location): Likewise.
4515 * substring-locations.c (format_warning_n_va): Likewise.
4516 * tree-diagnostic.c (default_tree_printer): Likewise.
4517 * tree-pretty-print.c (newline_and_indent): Likewise.
4519 2018-08-27 David Malcolm <dmalcolm@redhat.com>
4522 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
4523 line above for line-insertion fix-it hints.
4524 (selftest::test_fixit_insert_containing_newline): Update the
4525 expected results, and add a test with line-numbering enabled.
4527 2018-08-27 Martin Liska <mliska@suse.cz>
4530 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
4531 params with DECL_HAS_VALUE_EXPR_P.
4533 2018-08-27 Martin Liska <mliska@suse.cz>
4535 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
4536 selected expansion strategy.
4538 2018-08-27 Martin Liska <mliska@suse.cz>
4540 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
4541 * builtins.c (is_builtin_fn): Likewise.
4542 * attribs.c (diag_attr_exclusions): Use new function
4543 fndecl_built_in_p and remove check for FUNCTION_DECL if
4545 (builtin_mathfn_code): Likewise.
4546 (fold_builtin_expect): Likewise.
4547 (fold_call_expr): Likewise.
4548 (fold_builtin_call_array): Likewise.
4549 (fold_call_stmt): Likewise.
4550 (set_builtin_user_assembler_name): Likewise.
4551 (is_simple_builtin): Likewise.
4552 * calls.c (gimple_alloca_call_p): Likewise.
4553 (maybe_warn_nonstring_arg): Likewise.
4554 * cfgexpand.c (expand_call_stmt): Likewise.
4555 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
4556 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
4557 (cgraph_node::verify_node): Likewise.
4558 * cgraphclones.c (build_function_decl_skip_args): Likewise.
4559 (cgraph_node::create_clone): Likewise.
4560 * config/arm/arm.c (arm_insert_attributes): Likewise.
4561 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
4562 * dse.c (scan_insn): Likewise.
4563 * expr.c (expand_expr_real_1): Likewise.
4564 * fold-const.c (operand_equal_p): Likewise.
4565 (fold_binary_loc): Likewise.
4566 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
4567 * gimple-low.c (lower_stmt): Likewise.
4568 * gimple-pretty-print.c (dump_gimple_call): Likewise.
4569 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
4571 * gimple.c (gimple_build_call_from_tree): Likewise.
4572 (gimple_call_builtin_p): Likewise.
4573 (gimple_call_combined_fn): Likewise.
4574 * gimplify.c (gimplify_call_expr): Likewise.
4575 (gimple_boolify): Likewise.
4576 (gimplify_modify_expr): Likewise.
4577 (gimplify_addr_expr): Likewise.
4578 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
4579 * ipa-cp.c (determine_versionability): Likewise.
4580 * ipa-fnsummary.c (compute_fn_summary): Likewise.
4581 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
4582 * ipa-split.c (visit_bb): Likewise.
4583 (split_function): Likewise.
4584 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4585 * lto-cgraph.c (input_node): Likewise.
4586 * lto-streamer-out.c (write_symbol): Likewise.
4587 * omp-low.c (setjmp_or_longjmp_p): Likewise.
4588 (lower_omp_1): Likewise.
4589 * predict.c (strip_predict_hints): Likewise.
4590 * print-tree.c (print_node): Likewise.
4591 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
4592 * trans-mem.c (is_tm_irrevocable): Likewise.
4593 (is_tm_load): Likewise.
4594 (is_tm_simple_load): Likewise.
4595 (is_tm_store): Likewise.
4596 (is_tm_simple_store): Likewise.
4597 (is_tm_abort): Likewise.
4598 (tm_region_init_1): Likewise.
4599 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
4600 * tree-cfg.c (verify_gimple_call): Likewise.
4601 (move_stmt_r): Likewise.
4602 (stmt_can_terminate_bb_p): Likewise.
4603 * tree-eh.c (lower_eh_constructs_2): Likewise.
4604 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
4605 * tree-inline.c (remap_gimple_stmt): Likewise.
4606 (copy_bb): Likewise.
4607 (estimate_num_insns): Likewise.
4608 (fold_marked_statements): Likewise.
4609 * tree-sra.c (scan_function): Likewise.
4610 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
4611 (optimize_stack_restore): Likewise.
4612 (pass_fold_builtins::execute): Likewise.
4613 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
4614 (mark_all_reaching_defs_necessary_1): Likewise.
4615 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
4616 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
4617 (pass_forwprop::execute): Likewise.
4618 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4619 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4620 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
4621 * tree-ssa-strlen.c (get_string_length): Likewise.
4622 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
4623 (find_func_aliases_for_call): Likewise.
4624 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
4625 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
4626 * tree-tailcall.c (find_tail_calls): Likewise.
4627 * tree.c (need_assembler_name_p): Likewise.
4628 (free_lang_data_in_decl): Likewise.
4629 (get_call_combined_fn): Likewise.
4630 * ubsan.c (is_ubsan_builtin_p): Likewise.
4631 * varasm.c (incorporeal_function_p): Likewise.
4632 * tree.h (DECL_BUILT_IN): Remove and replace with
4634 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
4635 (fndecl_built_in_p): New.
4637 2018-08-27 Martin Liska <mliska@suse.cz>
4639 PR tree-optimization/86847
4640 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
4641 Dump also subtree probability.
4642 (switch_decision_tree::do_jump_if_equal): New function.
4643 (switch_decision_tree::emit_case_nodes): Handle special
4644 situations in balanced tree that can be emitted much simpler.
4645 Fix calculation of probabilities that happen in tree expansion.
4646 * tree-switch-conversion.h (struct cluster): Add
4648 (struct simple_cluster): Likewise.
4649 (struct case_tree_node): Add new function has_child.
4650 (do_jump_if_equal): New.
4652 2018-08-27 Martin Liska <mliska@suse.cz>
4654 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
4655 Add new argument to bit_test_cluster constructor.
4656 (bit_test_cluster::emit): Set bits really number of values
4658 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
4659 probability argument.
4660 * tree-switch-conversion.h (struct bit_test_cluster):
4661 Add m_handles_entire_switch.
4663 2018-08-27 Martin Liska <mliska@suse.cz>
4665 PR tree-optimization/86702
4666 * tree-switch-conversion.c (jump_table_cluster::emit):
4667 Make probabilities even for values in jump table
4668 according to number of cases handled.
4669 (switch_decision_tree::compute_cases_per_edge): Pass
4670 argument to reset_out_edges_aux function.
4671 (switch_decision_tree::analyze_switch_statement): Likewise.
4672 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
4675 2018-08-27 Martin Liska <mliska@suse.cz>
4677 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
4678 cfun argument explicitly.
4679 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
4680 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
4681 function gimple_switch_default_bb.
4682 (convert_switch_statements):
4684 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
4685 * stmt.c (label_to_block_fn): Use label_to_block and pass
4686 cfun argument explicitly and use gimple_switch_label_bb.
4687 (expand_case): Likewise.
4688 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
4689 cfun argument explicitly. Likewise.
4690 (make_edges_bb): Likewise.
4691 (make_cond_expr_edges): Likewise.
4692 (get_cases_for_edge): Likewise.
4693 (make_gimple_switch_edges): Likewise.
4694 (label_to_block_fn): Likewise.
4695 (label_to_block): Likewise.
4696 (make_goto_expr_edges): Likewise.
4697 (make_gimple_asm_edges): Likewise.
4698 (main_block_label): Likewise.
4699 (group_case_labels_stmt): Likewise.
4700 (find_taken_edge_computed_goto): Likewise.
4701 (find_taken_edge_switch_expr): Likewise.
4702 (gimple_verify_flow_info): Likewise.
4703 (gimple_redirect_edge_and_branch): Likewise.
4704 (gimple_switch_label_bb): New function.
4705 (gimple_switch_default_bb): Likewise.
4706 (gimple_switch_edge): Likewise.
4707 (gimple_switch_default_edge): Likewise.
4708 * tree-cfg.h (label_to_block_fn): Remove and replace ...
4709 (label_to_block): ... with this.
4710 (gimple_switch_label_bb): New.
4711 (gimple_switch_default_bb): Likewise.
4712 (gimple_switch_edge): Likewise.
4713 (gimple_switch_default_edge): Likewise.
4714 * tree-cfgcleanup.c (convert_single_case_switch): Use
4715 new gimple functions and pass new argument to label_to_block.
4716 (cleanup_control_flow_bb):
4717 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
4718 cfun argument explicitly.
4719 (make_eh_edges): Likewise.
4720 (redirect_eh_dispatch_edge): Likewise.
4721 (lower_resx): Likewise.
4722 (lower_eh_dispatch): Likewise.
4723 (maybe_remove_unreachable_handlers): Likewise.
4724 (unsplit_eh): Likewise.
4725 (cleanup_empty_eh): Likewise.
4726 (verify_eh_edges): Likewise.
4727 (verify_eh_dispatch_edge): Likewise.
4728 * tree-ssa-dom.c (record_edge_info): Likewise.
4729 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
4730 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
4731 (thread_through_normal_block): Likewise.
4732 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
4733 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
4734 * tree-switch-conversion.c (switch_conversion::collect): Use new
4736 (switch_conversion::check_final_bb): Likewise.
4737 (switch_conversion::gather_default_values): Pass new argument
4739 (switch_conversion::build_constructors): Likewise.
4740 (switch_decision_tree::compute_cases_per_edge): Use new
4741 gimple_switch_edge function.
4742 (switch_decision_tree::analyze_switch_statement): Pass new argument
4744 (switch_decision_tree::try_switch_expansion): Use
4745 gimple_switch_default_edge.
4746 * tree-vrp.c (find_switch_asserts): Pass new argument
4748 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4749 (vr_values::simplify_switch_using_ranges): Likewise.
4751 2018-08-27 Richard Biener <rguenther@suse.de>
4753 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4754 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4756 * tree-ssa-sccvn.h (struct vn_pval): New structure.
4757 (struct vn_nary_op_s): Add unwind_to member. Add
4758 predicated_values flag and put result into a union together
4759 with a linked list of vn_pval.
4760 (struct vn_ssa_aux): Add name member to make maintaining
4761 a map of SSA name to vn_ssa_aux possible. Remove no longer
4762 needed info, dfsnum, low, visited, on_sccstack, use_processed
4763 and range_info_anti_range_p members.
4764 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4765 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4767 (vn_valueize): New global.
4768 (vn_context_bb): Likewise.
4769 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4770 VN_INFO_PTR_INFO): Remove.
4771 * tree-ssa-sccvn.c: ... (rewrite)
4772 (pass_fre::execute): For -O2+ initialize loops and run
4773 RPO VN in optimistic mode (iterating). For -O1 and -Og
4774 run RPO VN in non-optimistic mode.
4775 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4776 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4777 * doc/invoke.texi (sccvn-max-scc-size): Remove.
4778 (rpo-vn-max-loop-depth): Document.
4779 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4780 when valuezing the VUSE signals we walked out of the region.
4781 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4782 (phi_translate): Set VN context block to use for availability
4784 (compute_avail): Likewise.
4785 (pre_valueize): New function.
4786 (pass_pre::execute): Adjust to the RPO VN API.
4788 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4789 (propagate_constants_for_unrolling): Remove.
4790 (tree_unroll_loops_completely): Perform value-numbering
4791 on the unrolled bodies loop parent.
4793 2018-08-27 Richard Biener <rguenther@suse.de>
4795 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4796 for partial antic compute.
4798 2018-08-27 Jakub Jelinek <jakub@redhat.com>
4800 PR rtl-optimization/87065
4801 * combine.c (simplify_if_then_else): Formatting fix.
4802 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4804 (known_cond): Don't return const_true_rtx for vector modes. Use
4805 CONST0_RTX instead of const0_rtx. Formatting fixes.
4807 2018-08-27 Martin Liska <mliska@suse.cz>
4809 PR gcov-profile/87069
4810 * gcov.c (process_file): Record files already processed
4811 and warn about a file being processed multiple times.
4813 2018-08-27 Martin Liska <mliska@suse.cz>
4816 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4817 Set default values for x_aarch64_*_string strings.
4818 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4819 prefix. For -mabi do not print '=ABI' in help and use
4820 <option_value> format for -msve-vector-bits and -moverride
4823 2018-08-26 Jeff Law <law@redhat.com>
4825 * config/mips/frame-header-opt.c: Include "backend.h" rather than
4828 2018-08-26 Marek Polacek <polacek@redhat.com>
4830 PR c++/87029, Implement -Wredundant-move.
4831 * doc/invoke.texi: Document -Wredundant-move.
4833 2018-08-25 Martin Sebor <msebor@redhat.com>
4835 PR tree-optimization/87059
4836 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
4837 to the same type as the other.
4838 * fold-const.c (fold_binary_loc): Assert expectation.
4840 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
4842 * config/darwin.c (machopic_legitimize_pic_address): Clean up
4843 extraneous parentheses, dead code section and formatting.
4845 2018-08-24 David Malcolm <dmalcolm@redhat.com>
4848 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
4849 wide enough for jumps in the line-numbering to be visible.
4850 (layout::print_gap_in_line_numbering): New member function.
4851 (layout::calculate_line_spans): When using line numbering, merge
4852 line spans that are only 1 line apart.
4853 (diagnostic_show_locus): When printing line numbers, show gaps in
4854 line numbering directly, rather than printing headers.
4855 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
4856 line-numbering with multiple line spans.
4857 (selftest::test_fixit_insert_containing_newline_2): Add test of
4858 line-numbering, in which the spans are close enough to be merged.
4860 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4862 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
4863 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
4864 * tree-vrp.c (range_is_nonnull): Remove.
4865 (range_includes_zero_p): Accept value_range instead of min/max.
4866 (extract_range_from_binary_expr_1): Do not early bail on
4868 Use range_includes_zero_p instead of range_is_nonnull.
4869 (extract_range_from_unary_expr): Use range_includes_zero_p instead
4870 of range_is_nonnull.
4871 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
4872 special case VR_ANTI_RANGE.
4873 (vrp_finalize): Same.
4874 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4876 (range_is_nonnull): Remove.
4877 * vr-values.c (vrp_stmt_computes_nonzero): Use
4878 range_includes_zero_p instead of range_is_nonnull.
4879 (extract_range_basic): Pass value_range to range_includes_zero_p
4880 instead of range_is_nonnull.
4882 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
4884 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4885 * emit-rtl.h (rtl_data): Remove return_bnd.
4886 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4887 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4888 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4889 (POINTER_BOUNDS_MODE): Remove definition.
4890 (make_pointer_bounds_mode): Remove.
4891 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4892 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4893 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4894 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4895 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4896 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4897 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4899 * config/i386/i386-modes.def (BND32, BND64): Remove.
4900 * config/i386/i386.c (dbx_register_map): Remove bound registers.
4901 (dbx64_register_map): Ditto.
4902 (svr4_dbx_register_map): Ditto.
4903 (indirect_thunk_bnd_needed): Remove.
4904 (indirect_thunks_bnd_used): Ditto.
4905 (indirect_return_bnd_needed): Ditto.
4906 (indirect_return_via_cx_bnd): Ditto.
4907 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4908 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4909 (output_indirect_thunk): Ditto. Remove need_prefix argument.
4910 (output_indirect_thunk_function): Remove handling of
4911 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4912 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4913 (ix86_save_reg): Remove handling of crtl->return_bnd.
4914 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4915 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4916 and UNSPEC_BNDLX_ADDR.
4917 (ix86_output_indirect_branch_via_reg): Remove handling of
4918 indirect_thunk_prefix_bnd.
4919 (ix86_output_indirect_branch_via_push): Ditto.
4920 (ix86_output_function_return): Ditto.
4921 (ix86_output_indirect_function_return): Ditto.
4922 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4923 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4924 (CALL_USED_REGISTERS): Ditto.
4925 (REG_ALLOC_ORDER): Update for removal of bound registers.
4926 (HI_REGISTER_NAMES): Ditto.
4927 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4928 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4929 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4930 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4931 (FIRST_PSEUDO_REG): Update.
4932 (BND): Remove mode iterator.
4933 * config/i386/predicates.md (bnd_mem_operator): Remove.
4935 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4937 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4940 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4942 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4943 the case in which the permute needs only a single element and
4944 repeats for every vector of the result. Extend that case to
4945 handle variable-length vectors.
4946 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4948 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
4951 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4952 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4954 2018-08-24 Richard Biener <rguenther@suse.de>
4956 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4957 bb_flags_allocated members.
4958 (auto_flag): New RAII class for allocating flags.
4959 (auto_edge_flag): New RAII class for allocating edge flags.
4960 (auto_bb_flag): New RAII class for allocating bb flags.
4961 * cfgloop.c (verify_loop_structure): Allocate temporary edge
4963 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4964 in favor of temporarily allocated BB flag.
4965 * hsa-brig.c: Re-order includes.
4966 * hsa-dump.c: Likewise.
4967 * hsa-regalloc.c: Likewise.
4968 * print-rtl.c: Likewise.
4969 * profile-count.c: Likewise.
4971 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
4974 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4977 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4980 * wide-int-range.cc (wide_int_range_div): Do not ignore result
4981 from wide_int_range_multiplicative_op.
4983 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4985 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4988 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
4990 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4993 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
4995 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4996 full GENERIC expression used for replacement.
4998 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
5000 * tree-vrp.c (abs_extent_range): Remove.
5001 (extract_range_into_wide_ints): Pass wide ints by reference.
5002 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5003 Pass wide ints by reference in all calls to
5004 extract_range_into_wide_ints.
5005 * wide-int-range.cc (wide_int_range_div): New.
5006 * wide-int-range.h (wide_int_range_div): New.
5007 (wide_int_range_includes_zero_p): New.
5008 (wide_int_range_zero_p): New.
5010 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
5012 * config/aarch64/aarch64.md (arches): New enum.
5013 (arch): New enum attr.
5014 (arch_enabled): New attr.
5015 (enabled): Now uses arch_enabled only.
5016 (simd, sve, fp16): Removed attribute.
5017 (fp): Attr now defined in terms of 'arch'.
5018 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5019 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5020 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5021 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5022 attributes into 'arch'.
5023 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5024 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5025 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5026 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5027 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5028 'simd' attribute into 'arch'.
5029 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5030 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5031 Convert use of 'fp' attribute to 'arch'.
5032 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5033 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5035 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5036 (different modes) Merge 'fp' and 'simd' into 'arch'.
5037 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5040 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
5042 PR rtl-optimization/87026
5043 * expmed.c (canonicalize_comparison): If we can no longer create
5044 pseudoregisters, don't.
5046 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
5049 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
5051 * config/arm/arm.c (speculation_barrier_libfunc): New static
5053 (arm_init_libfuncs): Initialize it.
5054 (arm_emit_speculation_barrier): New function.
5055 * config/arm/arm.md (speculation_barrier): Call
5056 arm_emit_speculation_barrier for architectures that do not have
5058 (speculation_barrier_insn): Only match on Armv7 or later.
5060 2018-08-23 Richard Biener <rguenther@suse.de>
5063 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
5066 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
5068 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
5069 of single-vector TBLs.
5070 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
5073 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
5076 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
5077 aarch64_evpc_tbl guard.
5079 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
5081 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
5084 2018-08-22 Martin Sebor <msebor@redhat.com>
5087 * tree-pretty-print.c (pretty_print_string): Add argument.
5088 (dump_generic_node): Call to pretty_print_string with string size.
5090 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
5092 PR rtl-optimization/86771
5093 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
5094 of two SETs into those two SETs, one to be placed at i2, if that SETs
5095 destination is modified between i2 and i3.
5097 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5099 PR tree-optimization/86725
5100 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
5102 (vect_analyze_scalar_cycles_1): Check it.
5104 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5106 PR tree-optimization/86725
5107 * tree-vect-loop.c (vect_is_simple_reduction): When treating
5108 an outer loop phi as a double reduction, make sure that the
5109 single user of the phi result is an inner loop phi.
5111 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5113 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
5114 grouped stores with gaps to a strided group.
5116 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
5118 * tree-vect-stmts.c (get_group_load_store_type)
5119 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
5120 first statement in a group.
5122 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5124 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
5125 the sequence used in gcc/gcc.c.
5127 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5130 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
5133 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5135 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
5136 Darwin10-specific unwinder-shim.
5137 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
5138 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
5139 New to cater for Darwin10 Rosetta.
5141 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5143 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
5144 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
5146 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
5151 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
5152 (dwarf2out_switch_text_section): Generate a local label for the
5153 second function sub-section and apply it as the second FDE start
5155 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
5156 second sub-section start.
5158 2018-08-22 Richard Biener <rguenther@suse.de>
5160 PR tree-optimization/86988
5161 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
5163 2018-08-22 Richard Biener <rguenther@suse.de>
5165 PR tree-optimization/86945
5166 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
5168 2018-08-22 Alexandre Oliva <oliva@adacore.com>
5170 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
5171 a comment about how uses of r2 for .sdata2 come about.
5173 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
5175 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
5177 2018-08-21 Marek Polacek <polacek@redhat.com>
5179 PR c++/86981, Implement -Wpessimizing-move.
5180 * doc/invoke.texi: Document -Wpessimizing-move.
5182 2018-08-21 Jan Hubicka <jh@suse.cz>
5184 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
5185 * tree.h (is_redundant_typedef): Remove.
5186 * dwarf2out.c (is_redundant_typedef): Turn into static function.
5188 2018-08-21 Jan Hubicka <jh@suse.cz>
5190 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
5193 2018-08-21 Tamar Christina <tamar.christina@arm.com>
5195 * expmed.c (extract_low_bits): Reject invalid subregs early.
5197 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
5200 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
5203 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5205 * config/vxworks.h: Guard vxworks_asm_out_constructor and
5206 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
5207 * config/vxworks.c: Likewise.
5209 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5211 * config/vxworks.c: Set targetm.have_ctors_dtors
5212 if HAVE_INITFINI_ARRAY_SUPPORT.
5213 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
5214 if HAVE_INITFINI_ARRAY_SUPPORT.
5216 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5218 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
5219 default search path for VxWorks < 7.
5221 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
5223 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
5224 (get_format_string): Refer to c_getstr.
5226 2018-08-21 Tom de Vries <tdevries@suse.de>
5228 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5229 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5230 (debuginfo_early_stop): Declare.
5231 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5232 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5233 (debuginfo_early_stop): New function.
5234 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
5235 and debuginfo_early_stop.
5236 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
5237 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
5238 (general_init): Call debuginfo_early_init.
5239 (finalize): Call debuginfo_fini.
5240 (do_compile): Call debuginfo_init.
5241 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
5243 (@item -fdump-debug, @item -fdump-earlydebug): Add.
5245 2018-08-21 Tom de Vries <tdevries@suse.de>
5247 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
5248 flag_dump_noaddr and flag_dump_unnumbered.
5250 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
5252 * wide-int-range.cc (wide_int_range_abs): New.
5253 (wide_int_range_order_set): Rename from wide_int_range_min_max.
5254 * wide-int-range.h (wide_int_range_abs): New.
5255 (wide_int_range_min_max): New.
5256 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
5257 case to call wide_int_range_abs.
5258 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
5259 (extract_range_from_abs_expr): Delete.
5261 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
5264 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
5265 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
5268 2018-08-20 Nathan Sidwell <nathan@acm.org>
5269 Jeff Law <law@redhat.com>
5271 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
5272 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
5274 2018-08-20 David Malcolm <dmalcolm@redhat.com>
5277 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
5278 (decl_attributes): Likewise.
5279 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
5281 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5282 * diagnostic-core.h (class auto_diagnostic_group): New class.
5283 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
5284 (diagnostic_report_diagnostic): Handle the first diagnostics within
5286 (emit_diagnostic): Add auto_diagnostic_group instance.
5288 (inform_n): Likewise.
5289 (warning): Likewise.
5290 (warning_at): Likewise.
5291 (warning_n): Likewise.
5292 (pedwarn): Likewise.
5293 (permerror): Likewise.
5295 (error_n): Likewise.
5296 (error_at): Likewise.
5298 (fatal_error): Likewise.
5299 (internal_error): Likewise.
5300 (internal_error_no_backtrace): Likewise.
5301 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
5302 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
5303 * diagnostic.h (struct diagnostic_context): Add fields
5304 "diagnostic_group_nesting_depth",
5305 "diagnostic_group_emission_count", "begin_group_cb",
5307 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
5308 Add auto_diagnostic_group instance(s).
5309 (find_explicit_erroneous_behavior): Likewise.
5310 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
5311 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
5312 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
5313 (gimplify_va_arg_expr): Likewise.
5314 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
5315 (HSA_SORRY_AT): Likewise.
5316 * ipa-devirt.c (compare_virtual_tables): Likewise.
5317 (warn_odr): Likewise.
5318 * multiple_target.c (expand_target_clones): Likewise.
5319 * opts-common.c (cmdline_handle_error): Likewise.
5320 * reginfo.c (globalize_reg): Likewise.
5321 * substring-locations.c (format_warning_n_va): Likewise.
5322 * tree-inline.c (expand_call_inline): Likewise.
5323 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
5324 * tree-ssa-loop-niter.c
5325 (do_warn_aggressive_loop_optimizations): Likewise.
5326 * tree-ssa-uninit.c (warn_uninit): Likewise.
5327 * tree.c (warn_deprecated_use): Likewise.
5329 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
5332 * config/i386/i386.md (eh_return): Always update EH return
5333 address in word_mode.
5335 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
5337 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
5338 TARGET_SPLIT_COMPLEX_ARG is defined.
5340 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
5342 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
5344 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
5347 * expr.c (expand_assignment): Assert that bitpos is positive.
5348 (store_field): Likewise
5349 (expand_expr_real_1): Make sure that bitpos is positive.
5350 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
5353 2018-08-20 Nathan Sidwell <nathan@acm.org>
5355 * Makefile.in (CPP_ID_DATA_H): Delete.
5356 (CPP_INTERNAL_H): Don't add it.
5357 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
5358 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
5360 2018-08-20 Richard Biener <rguenther@suse.de>
5362 PR tree-optimization/78655
5363 * tree-vrp.c (extract_range_from_binary_expr_1): Make
5364 pointer + offset nonnull if either operand is nonnull work.
5366 2018-08-20 Tom de Vries <tdevries@suse.de>
5368 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
5369 unless the referenced die describes the added property using
5370 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
5371 Otherwise, add a DW_AT_location to the referenced die.
5373 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
5376 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
5377 register_operand when calling ix86_set_reg_reg_cost.
5378 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
5379 Set *total to 0 for operands that satisfy x86_64_immediate_operand
5380 predicate and to 1 otherwise.
5382 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
5384 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
5385 emit a diagnostic that it is not supported and reset the option.
5386 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
5387 supported and consume the option. (ASM_FINAL_SPEC): New.
5389 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
5391 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
5394 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
5396 C-SKY port: Documentation
5398 * doc/extend.texi (C-SKY Function Attributes): New section.
5399 * doc/invoke.texi (Option Summary): Add C-SKY options.
5400 (C-SKY Options): New section.
5401 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
5403 2018-08-17 Jojo <jijie_rong@c-sky.com>
5404 Huibin Wang <huibin_wang@c-sky.com>
5405 Sandra Loosemore <sandra@codesourcery.com>
5406 Chung-Lin Tang <cltang@codesourcery.com>
5408 C-SKY port: Backend implementation
5410 * config/csky/*: New.
5411 * common/config/csky/*: New.
5413 2018-08-17 Jojo <jijie_rong@c-sky.com>
5414 Huibin Wang <huibin_wang@c-sky.com>
5415 Sandra Loosemore <sandra@codesourcery.com>
5416 Chung-Lin Tang <cltang@codesourcery.com>
5417 Andrew Jenner <andrew@codesourcery.com>
5419 C-SKY port: Configury
5421 * config.gcc (csky-*-*): New.
5422 * configure.ac: Add csky to targets for dwarf2 debug_line support.
5423 * configure: Regenerated.
5425 2018-08-17 David Malcolm <dmalcolm@redhat.com>
5427 * dump-context.h: Include "dumpfile.h".
5428 (dump_context::dump_printf_va): Convert final param from va_list
5429 to va_list *. Convert from ATTRIBUTE_PRINTF to
5430 ATTRIBUTE_GCC_DUMP_PRINTF.
5431 (dump_context::dump_printf_loc_va): Likewise.
5432 * dumpfile.c: Include "stringpool.h".
5433 (make_item_for_dump_printf_va): Delete.
5434 (make_item_for_dump_printf): Delete.
5435 (class dump_pretty_printer): New class.
5436 (dump_pretty_printer::dump_pretty_printer): New ctor.
5437 (dump_pretty_printer::emit_items): New member function.
5438 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
5440 (dump_pretty_printer::emit_item): New member function.
5441 (dump_pretty_printer::stash_item): New member function.
5442 (dump_pretty_printer::format_decoder_cb): New member function.
5443 (dump_pretty_printer::decode_format): New member function.
5444 (dump_context::dump_printf_va): Reimplement in terms of
5445 dump_pretty_printer.
5446 (dump_context::dump_printf_loc_va): Convert final param from va_list
5448 (dump_context::begin_scope): Reimplement call to
5449 make_item_for_dump_printf.
5450 (dump_printf): Update for change to dump_printf_va.
5451 (dump_printf_loc): Likewise.
5452 (selftest::test_capture_of_dump_calls): Convert "stmt" from
5453 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
5454 with %T, %E, and %G.
5455 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
5456 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
5457 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
5458 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
5459 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
5460 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
5461 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
5462 within a dump_printf_loc call to "%wu".
5463 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
5464 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
5465 missing space after "=".
5466 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
5467 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
5468 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
5469 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
5470 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
5471 duplicate "vectorized" from message.
5473 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
5475 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
5476 polyNxK_t element's TYPE_STRING_FLAG.
5478 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
5480 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
5481 (they were unnamed before). Fix comments.
5483 2018-08-17 Nathan Sidwell <nathan@acm.org>
5485 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
5487 2018-08-17 Richard Biener <rguenther@suse.de>
5489 PR tree-optimization/86841
5490 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
5492 2018-08-17 Martin Liska <mliska@suse.cz>
5494 * common.opt: Remove Warn, Init and Report for options with
5495 Ignore/Deprecated flag. Warning is done automatically for
5497 * config/i386/i386.opt: Likewise.
5498 * config/ia64/ia64.opt: Likewise.
5499 * config/rs6000/rs6000.opt: Likewise.
5500 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
5501 Remove usage of flag_check_pointer_bounds.
5502 * lto-wrapper.c (merge_and_complain): Do not handle
5503 OPT_fcheck_pointer_bounds.
5504 (append_compiler_options): Likewise.
5505 * opt-functions.awk: Do not handle Deprecated.
5506 * optc-gen.awk: Check that Var, Report and Init are not
5507 used for an option with Ignore/Deprecated flag.
5508 * opts-common.c (decode_cmdline_option): Do not report
5510 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
5512 * opts.h (struct cl_option): Remove cl_deprecated flag.
5513 (CL_ERR_DEPRECATED): Remove error enum value.
5515 2018-08-17 Richard Biener <rguenther@suse.de>
5518 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
5519 across a va-arg-pack using call adjust its return value accordingly.
5521 2018-08-16 Martin Sebor <msebor@redhat.com>
5523 PR tree-optimization/86853
5524 * gimple-ssa-sprintf.c (struct format_result): Rename member.
5525 (struct fmtresult): Add member and initialize it in ctors.
5526 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
5527 (format_string): Handle %S the same as %ls. Set MAYFAIL.
5528 (format_directive): Set POSUNDER4K when MAYFAIL is set.
5529 (parse_directive): Handle %C same as %c.
5530 (sprintf_dom_walker::compute_format_length): Adjust.
5531 (is_call_safe): Adjust.
5533 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5535 * builtins.c (c_strlen): Add new parameter eltsize. Use it
5536 for determining how to count the elements.
5537 * builtins.h (c_strlen): Adjust prototype.
5538 * expr.c (string_constant): Add new parameter mem_size.
5539 Set *mem_size appropriately.
5540 * expr.h (string_constant): Adjust protoype.
5541 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
5542 * gimple-fold.h (get_range_strlen): Adjust prototype.
5543 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
5544 (format_string): Call get_string_length with eltsize.
5546 2018-08-16 David Malcolm <dmalcolm@redhat.com>
5548 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
5549 to emit the span, rather than setting it as the prefix.
5551 2018-08-16 David Malcolm <dmalcolm@redhat.com>
5553 * diagnostic-show-locus.c (layout::start_annotation_line): Add
5554 "margin_char" parameter, defaulting to space. Use it in place
5555 of pp_space for the initial part of the margin.
5556 (layout::print_leading_fixits): Use '+' when filling the margin
5557 of line-insertion fix-it hints.
5559 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5561 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
5564 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5566 * config/rs6000/altivec.md: Don't set length attribute to the default
5568 * config/rs6000/darwin.md: Ditto.
5569 * config/rs6000/dfp.md: Ditto.
5570 * config/rs6000/htm.md: Ditto.
5571 * config/rs6000/rs6000.md: Ditto.
5572 * config/rs6000/sync.md: Ditto.
5573 * config/rs6000/vsx.md: Ditto.
5575 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5577 * config/rs6000/altivec.md: Don't set length attribute to the default
5578 value, for branch instructions.
5579 * config/rs6000/darwin.md: Ditto.
5580 * config/rs6000/rs6000.md: Ditto.
5582 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
5584 * config/rs6000/rs6000.md (length): Always define as const_int 4.
5585 (unnamed conditional branch define_insn): Set length to 4 or 8
5586 depending on offset.
5587 (<bd>_<mode>): Similar, for alternative 0.
5588 (<bd>tf_<mode>): Ditto.
5590 2018-08-16 Tamar Christina <tamar.christina@arm.com>
5592 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
5594 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
5596 * doc/rtl.texi: Replace old RTX class names with new names.
5599 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
5601 * expmed.h (canonicalize_comparison): New declaration.
5602 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
5603 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
5604 * optabs.c (prepare_cmp_insn): Likewise.
5605 * rtl.h (unsigned_condition_p): New function which checks if a
5606 comparison operator is unsigned.
5608 2018-08-16 Nathan Sidwell <nathan@acm.org>
5610 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
5611 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
5613 2018-08-16 Tamar Christina <tamar.christina@arm.com>
5616 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
5617 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
5618 (mov<mov>): ..this and enable unconditionally.
5620 2018-08-16 Tamar Christina <tamar.christina@arm.com>
5622 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
5624 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
5626 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
5627 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
5628 "Common" with "Target".
5630 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
5632 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
5633 * doc/invoke.texi (mmitigate-rop): Remove.
5634 * config/i386/i386.c: Do not include "regrename.h".
5635 (ix86_rop_should_change_byte_p, reg_encoded_number)
5636 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
5638 (ix86_reorg): Remove call to ix86_mitigate_rop.
5639 * config/i386/i386.md (attr "modrm_class"): Remove.
5640 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
5641 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
5642 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
5644 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
5646 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
5647 allow folding of mergeh() and mergel() for the float and double types.
5648 (fold_mergehl_helper): Rework to handle building a permute tree
5651 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
5653 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
5656 2018-08-15 David Malcolm <dmalcolm@redhat.com>
5658 * common.opt (fdiagnostics-show-labels): New option.
5659 * diagnostic-show-locus.c (class layout_range): Add field
5661 (class layout): Add field "m_show_labels_p".
5662 (layout_range::layout_range): Add param "label" and use it to
5664 (make_range): Pass in NULL for new "label" param of layout_range's
5666 (layout::layout): Initialize m_show_labels_p.
5667 (layout::maybe_add_location_range): Pass in loc_range->m_label
5668 when constructing layout_range instances.
5669 (struct line_label): New struct.
5670 (layout::print_any_labels): New member function.
5671 (layout::print_line): Call it if label-printing is enabled.
5672 (selftest::test_one_liner_labels): New test.
5673 (selftest::test_diagnostic_show_locus_one_liner): Call it.
5674 * diagnostic.c (diagnostic_initialize): Initialize
5675 context->show_labels_p.
5676 * diagnostic.h (struct diagnostic_context): Add field
5678 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5679 -fno-diagnostics-show-labels.
5680 * dwarf2out.c (gen_producer_string): Add
5681 OPT_fdiagnostics_show_labels to the ignored options.
5682 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
5684 (gcc_rich_location::maybe_add_expr): Likewise.
5685 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
5686 label" param, defaulting to NULL.
5687 (gcc_rich_location::add_expr): Add "label" param.
5688 (gcc_rich_location::maybe_add_expr): Likewise.
5689 (class text_range_label): New class.
5690 (class range_label_for_type_mismatch): New class.
5691 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
5692 of format_warning_va.
5693 (fmtwarn_n): Likewise for new params of format_warning_n_va.
5694 * lto-wrapper.c (merge_and_complain): Add
5695 OPT_fdiagnostics_show_labels to the "pick one setting" options.
5696 (append_compiler_options): Likewise to the dropped options.
5697 (append_diag_options): Likewise to the passed-on options.
5698 * opts.c (common_handle_option): Handle the new option.
5699 * selftest-diagnostic.c
5700 (test_diagnostic_context::test_diagnostic_context): Enable
5702 * substring-locations.c: Include "gcc-rich-location.h".
5703 (format_warning_n_va): Add "fmt_label" and "param_label" params
5704 and use them as appropriate.
5705 (format_warning_va): Add "fmt_label" and "param_label" params,
5706 passing them on to format_warning_n_va.
5707 (format_warning_at_substring): Likewise.
5708 (format_warning_at_substring_n): Likewise.
5709 * substring-locations.h (format_warning_va): Add "fmt_label" and
5710 "param_label" params.
5711 (format_warning_n_va): Likewise.
5712 (format_warning_at_substring): Likewise.
5713 (format_warning_at_substring_n): Likewise.
5714 * toplev.c (general_init): Initialize global_dc->show_labels_p.
5716 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
5719 * builtins.c (expand_builtin_memcmp): Do not expand the call
5720 when overflow is detected.
5722 2018-08-15 Martin Sebor <msebor@redhat.com>
5724 PR tree-optimization/71625
5725 * config/aarch64/aarch64-builtins.c
5726 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
5728 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
5730 * config/s390/s390.c (s390_reorg): Remove loop.
5732 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
5735 (darwin_function_switched_text_sections): Delete.
5736 * gcc/config/darwin.h
5737 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
5739 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
5742 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
5743 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
5744 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
5746 2018-08-15 Martin Liska <mliska@suse.cz>
5748 PR tree-optimization/86925
5749 * predict.c (expr_expected_value_1): When taking
5750 later predictor, assign also probability.
5751 Use fold_build2_initializer_loc in order to fold
5752 the expression in -frounding-math.
5754 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
5756 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5758 (expand_vec_perm_1): Try the new method.
5760 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
5763 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5764 Check whether lra_live_max_point is 0 before dividing.
5766 2018-08-14 Martin Sebor <msebor@redhat.com>
5768 PR tree-optimization/86650
5769 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5770 (vrp_prop::check_mem_ref): Same.
5772 2018-08-13 Liu Hao <lh_mouse@126.com>
5774 * pretty-print.c (eat_esc_sequence): Swap the foreground and
5775 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5776 and clear it thereafter, as it only works for DBCS.
5778 2018-08-13 Liu Hao <lh_mouse@126.com>
5780 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5781 handle returned by _get_osf_handle().
5783 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5785 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5786 for folding vec_perm.
5788 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5790 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5791 Add support for gimple-folding of vec_pack() and vec_unpack()
5794 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5796 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5797 vec_xst variants to the list.
5798 (rs6000_gimple_fold_builtin): Add support for folding unaligned
5799 vector loads and stores.
5801 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
5803 * config.gcc (rs6000-ibm-aix4.x): Delete.
5804 (rs6000-ibm-aix5.1): Delete.
5805 (rs6000-ibm-aix5.2): Delete.
5806 (rs6000-ibm-aix5.3): Delete.
5807 * config/rs6000/aix43.h: Delete.
5808 * config/rs6000/aix51.h: Delete.
5809 * config/rs6000/aix52.h: Delete.
5810 * config/rs6000/t-aix43: Delete.
5812 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
5814 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5816 (s390_decompose_address): Factor out constant pool ref
5819 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5821 * config/nds32/nds32-predicates.c
5822 (nds32_can_use_bclr_p): Change return type as bool.
5823 (nds32_can_use_bset_p): Ditto.
5824 (nds32_can_use_btgl_p): Ditto.
5825 (nds32_can_use_bitci_p): Ditto.
5826 * config/nds32/nds32-protos.h
5827 (nds32_can_use_bclr_p): Change declaration.
5828 (nds32_can_use_bset_p): Ditto.
5829 (nds32_can_use_btgl_p): Ditto.
5830 (nds32_can_use_bitci_p): Ditto.
5832 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5834 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
5835 Support -msched-prolog-epilog option.
5836 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
5838 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5840 * common/config/nds32/nds32-common.c
5841 (nds32_option_optimization_table): Enalbe -malways-align.
5843 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5845 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
5847 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
5848 OPT_misr_secure_ case.
5849 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
5850 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
5851 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
5853 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
5854 (nds32_isr_info): New field security_level.
5855 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
5856 * config/nds32/nds32.md (return_internal): Consider critical attribute.
5857 * config/nds32/nds32.opt (misr-secure): New option.
5858 * config/nds32/nds32_init.inc: New file.
5859 * config/nds32/nds32_isr.h: New file.
5861 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
5863 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
5864 Update comment for atomic instructions.
5865 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
5866 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5868 (atomic_loaddi): Revise fence expansion to only emit fence prior to
5869 load for __ATOMIC_SEQ_CST model.
5870 (atomic_loaddi_1): Remove float register target.
5871 (atomic_storedi): Handle CONST_INT values.
5872 (atomic_storedi_1): Remove float register source. Add special case
5874 (memory_barrier): New expander and insn.
5876 2018-08-11 Jakub Jelinek <jakub@redhat.com>
5878 PR tree-optimization/86835
5879 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5880 new_stmt after def_gsi, make sure to insert new_square_stmt after
5881 that stmt, not 2 stmts before it.
5883 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
5886 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5889 2018-08-10 Martin Liska <mliska@suse.cz>
5892 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5894 * builtins.c (expand_builtin_expect_with_probability):
5896 (expand_builtin_expect_with_probability): New function.
5897 (build_builtin_expect_predicate): Add new argumnet probability
5898 for BUILT_IN_EXPECT_WITH_PROBABILITY.
5899 (fold_builtin_expect):
5902 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5903 * builtins.h (fold_builtin_expect): Set new argument.
5904 * doc/extend.texi: Document __builtin_expect_with_probability.
5905 * doc/invoke.texi: Likewise.
5906 * gimple-fold.c (gimple_fold_call): Pass new argument.
5907 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5908 also BUILT_IN_EXPECT_WITH_PROBABILITY.
5909 * predict.c (get_predictor_value): New function.
5910 (expr_expected_value): Add new argument probability. Assume
5911 that predictor and probability are always non-null.
5912 (expr_expected_value_1): Likewise. For __builtin_expect and
5913 __builtin_expect_with_probability set probability. Handle
5914 combination in binary expressions.
5915 (tree_predict_by_opcode): Simplify code by simply calling
5916 get_predictor_value.
5917 (pass_strip_predict_hints::execute): Add handling of
5918 BUILT_IN_EXPECT_WITH_PROBABILITY.
5919 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5921 * tree.h (DECL_BUILT_IN_P): New function.
5923 2018-08-10 Martin Liska <mliska@suse.cz>
5925 PR tree-optimization/85799
5926 * passes.def: Add argument for pass_strip_predict_hints.
5927 * predict.c (class pass_strip_predict_hints): Add new argument
5929 (strip_predictor_early): New function.
5930 (pass_strip_predict_hints::execute): Call the function to
5932 (strip_predict_hints): New function.
5933 * predict.def: Fix comment.
5935 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
5937 * Makefile.in: Clarify which tm.texi to copy over to assert the
5938 right to grant a GFDL license for all.
5940 2018-08-09 Jeff Law <law@redhat.com>
5942 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5945 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
5947 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5950 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5952 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5953 reductions for variable-length vectors.
5955 2018-08-09 David Malcolm <dmalcolm@redhat.com>
5958 * common.opt (fdiagnostics-show-line-numbers): New option.
5959 * diagnostic-show-locus.c (class layout): Add fields
5960 "m_show_line_numbers_p" and "m_linenum_width";
5961 (num_digits): New function.
5962 (test_num_digits): New function.
5963 (layout::layout): Initialize new fields. Update m_x_offset
5964 logic to handle any left margin.
5965 (layout::print_source_line): Print line number when requested.
5966 (layout::start_annotation_line): New member function.
5967 (layout::print_annotation_line): Call it.
5968 (layout::print_leading_fixits): Likewise.
5969 (layout::print_trailing_fixits): Likewise. Update calls to
5970 move_to_column for new parameter.
5971 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5972 it to potentially call start_annotation_line.
5973 (layout::show_ruler): Call start_annotation_line.
5974 (selftest::test_line_numbers_multiline_range): New selftest.
5975 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5976 and selftest::test_line_numbers_multiline_range.
5977 * diagnostic.c (diagnostic_initialize): Initialize
5978 show_line_numbers_p.
5979 * diagnostic.h (struct diagnostic_context): Add field
5980 "show_line_numbers_p".
5981 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5982 -fno-diagnostics-show-line-numbers.
5983 * dwarf2out.c (gen_producer_string): Add
5984 OPT_fdiagnostics_show_line_numbers to the ignored options.
5985 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5986 one setting" options.
5987 (append_compiler_options): Likewise to the dropped options.
5988 (append_diag_options): Likewise to the passed-on options.
5989 * opts.c (common_handle_option): Handle the new option.
5990 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5992 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
5994 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5995 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
5996 third argument of type "const signed char" to descriptions of
5997 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5998 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5999 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6000 __builtin_bcdsub_ov functions.
6002 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
6004 PR tree-optimization/86858
6005 * tree-vect-loop.c (vect_is_simple_reduction): Restore
6006 flow_bb_inside_loop_p calls.
6008 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
6010 PR tree-optimization/86871
6011 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6012 instead of gimple_assign_lhs.
6014 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
6017 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6018 register constraint to operand 0.
6019 (add<mode>3_carryinC): Likewise.
6020 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6022 2018-08-09 Martin Liska <mliska@suse.cz>
6025 * common.opt: Remove extra line.
6027 2018-08-09 Martin Liska <mliska@suse.cz>
6029 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6030 at the end of a line, make first letter capital and end up
6031 a sentence with a dot.
6032 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6033 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6034 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6035 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6036 (PARAM_MAX_ISL_OPERATIONS): Likewise.
6037 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6038 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6039 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6040 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6041 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6042 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6043 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6044 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
6045 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
6046 (PARAM_TREE_REASSOC_WIDTH): Likewise.
6047 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
6048 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
6049 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
6051 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
6054 * config/s390/s390.c (s390_option_override_internal): Reduce the
6055 stack-clash-protection-probe-interval param if it would be too big
6058 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
6061 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
6062 * config/m68k/m68k.c (handle_move_double): Don't call
6063 m68k_final_prescan_insn.
6064 (m68k_adjust_decorated_operand): Renamed from
6065 m68k_final_prescan_insn, remove first and third operand and
6067 (print_operand): Call it.
6068 (print_operand_address): Call it.
6070 2018-08-08 Nathan Sidwell <nathan@acm.org>
6072 * diagnostic.c (diagnostic_report_current_module): Use
6073 linemap_included_from & linemap_included_from_linemap.
6075 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
6077 * config/aarch64/aarch64-cores.def: Add phecda core.
6078 * config/aarch64/aarch64-tune.md: Regenerate.
6079 * doc/invoke.texi: Add phecda core.
6081 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
6084 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
6086 * config/s390/s390.md ("movti"): Add more alternatives for
6087 constant to GPR copies.
6089 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
6091 * config/s390/s390.c: Fix whitespace damage throughout the file.
6092 * config/s390/s390.h: Likewise.
6093 * config/s390/tpf.h: Likewise.
6095 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
6097 * config/s390/s390.c (s390_loadrelative_operand_p):
6098 Remove TARGET_CPU_ZARCH usages.
6099 (s390_rtx_costs): Likewise.
6100 (s390_legitimate_constant_p): Likewise.
6101 (s390_cannot_force_const_mem): Likewise.
6102 (legitimate_reload_constant_p): Likewise.
6103 (s390_preferred_reload_class): Likewise.
6104 (legitimize_pic_address): Likewise.
6105 (legitimize_tls_address): Likewise.
6106 (s390_split_branches): Removed.
6107 (s390_add_execute): Removed.
6108 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
6109 (s390_mainpool_start): Likewise.
6110 (s390_mainpool_finish): Likewise.
6111 (s390_mainpool_cancel): Removed.
6112 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
6113 (s390_chunkify_cancel): Likewise.
6114 (s390_return_addr_rtx): Likewise.
6115 (s390_register_info): Remove split_branches_pending_p uages.
6116 (s390_optimize_register_info): Likewise.
6117 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
6118 split_branches_pending_p usages.
6119 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
6120 (s390_load_got): Likewise.
6121 (s390_expand_split_stack_prologue): Likewise.
6122 (output_asm_nops): Likewise.
6123 (s390_function_profiler): Likewise.
6124 (s390_emit_call): Likewise.
6125 (s390_conditional_register_usage): Likewise.
6126 (s390_optimize_prologue): Likewise.
6127 (s390_reorg): Remove TARGET_CPU_ZARCH and
6128 split_branches_pending_p usages.
6129 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
6131 (s390_output_indirect_thunk_function): Likewise.
6132 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
6133 (TARGET_CPU_ZARCH_P): Removed.
6134 (struct machine_function): Remove split_branches_pending_p.
6135 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
6137 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
6139 * common/config/s390/s390-common.c (processor_flags_table):
6141 * config.gcc: Remove with_arch/with_tune support.
6142 * config/s390/2064.md: Remove cpu attribute comparisons.
6143 * config/s390/driver-native.c (s390_host_detect_local_cpu):
6145 * config/s390/linux.h (ASM_SPEC):
6146 Remove -march support.
6147 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
6148 Use a table to get an arch level.
6149 * config/s390/s390-opts.h (enum processor_type):
6151 * config/s390/s390.c
6152 (processor_table): Remove entries, add arch_level values.
6153 (s390_issue_rate): Remove cases.
6154 (s390_option_override): Adjust
6155 s390_option_override_internal() call.
6156 (s390_option_override_internal): Remove deprecation warning.
6157 (s390_valid_target_attribute_tree): Adjust
6158 s390_option_override_internal() call.
6159 * config/s390/s390.h (struct s390_processor):
6160 Share with s390-c.c, add arch_level field.
6161 * config/s390/s390.md:
6162 Remove occurrences in cpu attribute.
6163 * config/s390/s390.opt: Remove -march/-mtune support.
6164 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
6165 * doc/invoke.texi: Remove deprecation warning.
6167 2018-08-08 Luis Machado <luis.machado@linaro.org>
6169 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
6171 (qdf24xx_tunings): Set vector cost structure to
6172 qdf24xx_vector_cost.
6174 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
6175 <register_sextend>: Set to 3.
6177 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
6180 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
6181 * config/aarch64/aarch64-simd.md
6182 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
6183 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
6184 (aarch64_frecpx<mode>): ...this new pattern.
6185 * config/aarch64/aarch64-simd-builtins.def: Remove comment
6186 about aarch64_frecp<FRECP:frecp_suffix><mode>.
6188 2018-08-07 Martin Liska <mliska@suse.cz>
6191 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
6192 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
6193 * predict.def (PRED_MALLOC_NONNULL): New predictor.
6194 * doc/extend.texi: Document that malloc attribute adds
6197 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
6200 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6201 Define to speculation_safe_value_not_needed.
6203 2018-08-06 Jeff Law <law@redhat.com>
6205 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
6206 the vr_values instance to cprop_into_stmt.
6207 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
6208 (cprop_operand): Also query EVRP to determine if OP is a constant.
6210 2018-08-06 Nathan Sidwell <nathan@acm.org>
6212 * diagnostic.c (diagnostic_report_current_module): Reroll
6213 included-at loop. Translate text.
6215 2018-08-06 David Malcolm <dmalcolm@redhat.com>
6217 * function-tests.c (selftest::test_expansion_to_rtl): Call
6218 free_after_compilation.
6220 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6222 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
6224 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
6226 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
6227 loops with memory block operations from getting unrolled.
6229 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
6232 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6233 Define to speculation_safe_value_not_needed.
6235 2018-08-06 Jeff Law <law@redhat.com>
6237 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
6240 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6243 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
6244 with all enabled __intN types.
6246 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
6248 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6250 * alias.c (record_set): Check for clobber high.
6251 * cfgexpand.c (expand_gimple_stmt): Likewise.
6252 * combine-stack-adj.c (single_set_for_csa): Likewise.
6253 * combine.c (find_single_use_1): Likewise.
6254 (set_nonzero_bits_and_sign_copies): Likewise.
6255 (get_combine_src_dest): Likewise.
6256 (is_parallel_of_n_reg_sets): Likewise.
6257 (try_combine): Likewise.
6258 (record_dead_and_set_regs_1): Likewise.
6259 (reg_dead_at_p_1): Likewise.
6260 (reg_dead_at_p): Likewise.
6261 * dce.c (deletable_insn_p): Likewise.
6262 (mark_nonreg_stores_1): Likewise.
6263 (mark_nonreg_stores_2): Likewise.
6264 * df-scan.c (df_find_hard_reg_defs): Likewise.
6265 (df_uses_record): Likewise.
6266 (df_get_call_refs): Likewise.
6267 * dwarf2out.c (mem_loc_descriptor): Likewise.
6268 * haifa-sched.c (haifa_classify_rtx): Likewise.
6269 * ira-build.c (create_insn_allocnos): Likewise.
6270 * ira-costs.c (scan_one_insn): Likewise.
6271 * ira.c (equiv_init_movable_p): Likewise.
6272 (rtx_moveable_p): Likewise.
6273 (interesting_dest_for_shprep): Likewise.
6274 * jump.c (mark_jump_label_1): Likewise.
6275 * postreload-gcse.c (record_opr_changes): Likewise.
6276 * postreload.c (reload_cse_simplify): Likewise.
6277 (struct reg_use): Add source expr.
6278 (reload_combine): Check for clobber high.
6279 (reload_combine_note_use): Likewise.
6280 (reload_cse_move2add): Likewise.
6281 (move2add_note_store): Likewise.
6282 * print-rtl.c (print_pattern): Likewise.
6283 * recog.c (decode_asm_operands): Likewise.
6284 (store_data_bypass_p): Likewise.
6285 (if_test_bypass_p): Likewise.
6286 * regcprop.c (kill_clobbered_value): Likewise.
6287 (kill_set_value): Likewise.
6288 * reginfo.c (reg_scan_mark_refs): Likewise.
6289 * reload1.c (maybe_fix_stack_asms): Likewise.
6290 (eliminate_regs_1): Likewise.
6291 (elimination_effects): Likewise.
6292 (mark_not_eliminable): Likewise.
6293 (scan_paradoxical_subregs): Likewise.
6294 (forget_old_reloads_1): Likewise.
6295 * reorg.c (find_end_label): Likewise.
6296 (try_merge_delay_insns): Likewise.
6297 (redundant_insn): Likewise.
6298 (own_thread_p): Likewise.
6299 (fill_simple_delay_slots): Likewise.
6300 (fill_slots_from_thread): Likewise.
6301 (dbr_schedule): Likewise.
6302 * resource.c (update_live_status): Likewise.
6303 (mark_referenced_resources): Likewise.
6304 (mark_set_resources): Likewise.
6305 * rtl.c (copy_rtx): Likewise.
6306 * rtlanal.c (reg_referenced_p): Likewise.
6307 (single_set_2): Likewise.
6308 (noop_move_p): Likewise.
6309 (note_stores): Likewise.
6310 * sched-deps.c (sched_analyze_reg): Likewise.
6311 (sched_analyze_insn): Likewise.
6313 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6315 * cse.c (invalidate_reg): New function extracted from...
6316 (invalidate): ...here.
6317 (canonicalize_insn): Check for clobber high.
6318 (invalidate_from_clobbers): invalidate clobber highs.
6319 (invalidate_from_sets_and_clobbers): Likewise.
6320 (count_reg_usage): Check for clobber high.
6321 (insn_live_p): Likewise.
6322 * cselib.c (cselib_expand_value_rtx_1):Likewise.
6323 (cselib_invalidate_regno): Check for clobber in setter.
6324 (cselib_invalidate_rtx): Pass through setter.
6325 (cselib_invalidate_rtx_note_stores):
6326 (cselib_process_insn): Check for clobber high.
6327 * cselib.h (cselib_invalidate_rtx): Add operand.
6329 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6331 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
6332 (mark_not_eliminable): Likewise.
6333 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
6334 * lra-lives.c (process_bb_lives): Check for clobber high.
6335 * lra.c (new_insn_reg): Remember clobber highs.
6336 (collect_non_operand_hard_regs): Check for clobber high.
6337 (lra_set_insn_recog_data): Likewise.
6338 (add_regs_to_insn_regno_info): Likewise.
6339 (lra_update_insn_regno_info): Likewise.
6341 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6343 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
6344 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
6346 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6348 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
6349 (copy_insn_1): Likewise.
6350 (gen_hard_reg_clobber_high): New gen function.
6351 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
6352 * genemit.c (gen_exp): Likewise.
6353 (gen_emit_seq): Pass through info.
6354 (gen_insn): Check for CLOBBER_HIGH.
6355 (gen_expand): Pass through info.
6356 (gen_split): Likewise.
6357 (output_add_clobbers): Likewise.
6358 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
6359 (remove_clobbers): Likewise.
6360 * rtl.h (gen_hard_reg_clobber_high): New declaration.
6362 2018-08-06 Alan Hayward <alan.hayward@arm.com>
6364 * doc/rtl.texi (clobber_high): Add.
6365 (parallel): Add in clobber high
6366 * rtl.c (rtl_check_failed_code3): Add function.
6367 * rtl.def (CLOBBER_HIGH): Add expression.
6368 * rtl.h (RTL_CHECKC3): Add macro.
6369 (rtl_check_failed_code3): Add declaration.
6370 (XC3EXP): Add macro.
6372 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
6375 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
6376 cfun->machine->max_used_stack_alignment if needed.
6378 2018-08-04 Martin Sebor <msebor@redhat.com>
6380 PR tree-optimization/86571
6381 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
6384 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
6387 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6390 2018-08-03 Jeff Law <law@redhat.com>
6393 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6394 Define to speculation_safe_value_not_needed.
6396 2018-08-03 David Malcolm <dmalcolm@redhat.com>
6398 * doc/gcov.texi (-x): Remove duplicate "to".
6399 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
6400 (-Wif-not-aligned): Remove duplicate "is".
6401 (-flto): Remove duplicate "the".
6402 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
6403 duplicate "v5.00.b".
6404 (MSP430 Options): Remove duplicate "and" from the description
6405 of "-mgprel-sec=regexp".
6406 (x86 Options): Remove duplicate copies of "vmldLog102" and
6407 vmlsLog104 from description of "-mveclibabi=type".
6409 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
6411 * internal-fn.h (first_commutative_argument): Declare.
6412 * internal-fn.c (first_commutative_argument): New function.
6413 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
6414 restrictions for pattern statements. Use first_commutative_argument
6415 to look for commutative operands in calls to internal functions.
6417 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
6419 * Makefile.in (wide-int-range.o): New.
6420 * tree-vrp.c: Move all the wide_int_* functions to...
6421 * wide-int-range.cc: ...here.
6422 * tree-vrp.h: Move all the wide_int_* prototypes to...
6423 * wide-int-range.h: ...here.
6425 2018-08-03 Tom de Vries <tdevries@suse.de>
6427 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
6429 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
6430 * except.c (output_function_exception_table): Do early exit if
6431 targetm_common.except_unwind_info (&global_options) == UI_NONE.
6433 2018-08-03 Martin Liska <mliska@suse.cz>
6435 * predict.c (dump_prediction): Change to 2 digits
6438 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
6440 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
6442 (falkor_am_1_gtov_gtov): ... a new insn reservation.
6444 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
6446 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
6447 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
6448 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
6449 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
6450 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
6452 2018-08-02 David Malcolm <dmalcolm@redhat.com>
6454 * diagnostic-show-locus.c (diagnostic_show_locus): Use
6455 pp_take_prefix when saving the existing prefix.
6456 * diagnostic.c (diagnostic_append_note): Likewise.
6457 * langhooks.c (lhd_print_error_function): Likewise.
6458 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
6459 param's type. Free the existing prefix.
6460 (pp_take_prefix): New function.
6461 (pretty_printer::pretty_printer): Drop the prefix parameter.
6462 Rename the length parameter to match the comment.
6463 (pretty_printer::~pretty_printer): Free the prefix.
6464 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
6466 (struct pretty_printer): Drop the "const" from "prefix" field's
6467 type and clarify memory management.
6468 (pp_set_prefix): Drop the "const" from the 2nd param.
6469 (pp_take_prefix): New decl.
6471 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
6473 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
6474 for word_mode_ok here instead of passing as argument.
6475 (expand_block_compare): Change select_block_compare_mode() call.
6476 (expand_strncmp_gpr_sequence): New function.
6477 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
6479 2018-08-02 Jeff Law <law@redhat.com>
6482 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6483 Define to speculation_safe_value_not_needed.
6486 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6487 Define to speculation_safe_value_not_needed.
6489 2018-08-02 Tom de Vries <tdevries@suse.de>
6492 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
6493 function. Return UI_TARGET unconditionally.
6494 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
6495 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
6497 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
6499 * genemit.c (print_overload_test): Fix typo.
6501 2018-08-02 Richard Biener <rguenther@suse.de>
6503 PR tree-optimization/86816
6504 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
6505 which checks for value availability before querying it.
6506 (gvn_uses_equal): Use it.
6507 (same_succ_hash): Likewise.
6508 (gimple_equal_p): Likewise.
6510 2018-08-02 Nick Clifton <nickc@redhat.com>
6513 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6514 Define to speculation_safe_value_not_needed.
6517 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6518 Define to speculation_safe_value_not_needed.
6521 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6522 Define to speculation_safe_value_not_needed.
6525 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6526 Define to speculation_safe_value_not_needed.
6529 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6530 Define to speculation_safe_value_not_needed.
6533 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6534 Define to speculation_safe_value_not_needed.
6537 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6538 Define to speculation_safe_value_not_needed.
6541 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6542 Define to speculation_safe_value_not_needed.
6545 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6546 speculation_safe_value_not_needed.
6549 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
6550 to speculation_safe_value_not_needed.
6552 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
6554 * doc/md.texi: Expand the documentation of instruction names
6555 to mention port-local uses. Document '@' in pattern names.
6556 * read-md.h (overloaded_instance, overloaded_name): New structs.
6558 (md_reader::handle_overloaded_name): New member function.
6559 (md_reader::get_overloads): Likewise.
6560 (md_reader::m_first_overload): New member variable.
6561 (md_reader::m_next_overload_ptr): Likewise.
6562 (md_reader::m_overloads_htab): Likewise.
6563 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
6564 m_next_overload_ptr and m_overloads_htab.
6565 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
6566 (get_mode_token, get_code_token, get_int_token): New functions.
6567 (map_attr_string): Add an optional argument that passes back
6568 the associated iterator.
6569 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
6570 (md_reader::handle_overloaded_name, add_overload_instance): New
6572 (apply_iterators): Handle '@' names. Report an error if '@'
6573 is used without iterators.
6574 (initialize_iterators): Initialize the new iterator_group fields.
6575 * genopinit.c (handle_overloaded_code_for)
6576 (handle_overloaded_gen): New functions.
6577 (main): Use them to print declarations of maybe_code_for_* and
6578 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
6579 * genemit.c (print_overload_arguments, print_overload_test)
6580 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
6581 (main): Use it to print definitions of maybe_code_for_* and
6582 maybe_gen_* functions.
6583 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
6584 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
6585 instead of explicit mode checks.
6586 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
6587 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
6588 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
6589 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
6590 (aarch64_expand_compare_and_swap): Likewise
6591 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
6592 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
6593 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
6594 (aarch64_constant_pool_reload_icode): Delete.
6595 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
6596 instead of aarch64_constant_pool_reload_icode. Use
6597 code_for_aarch64_reload_mov instead of explicit mode checks.
6598 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
6599 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
6600 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
6601 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
6602 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
6603 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
6604 (aarch64_atomic_load_op_code): Delete.
6605 (aarch64_emit_atomic_load_op): Likewise.
6606 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
6607 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
6608 instead of aarch64_emit_atomic_load_op.
6609 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
6610 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
6611 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
6612 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
6613 character before the pattern name.
6614 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
6615 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
6616 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
6617 (aarch64_frecps<mode>): Likewise.
6618 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
6619 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
6620 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
6621 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
6622 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
6624 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
6626 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
6627 Allow HFmode constants if TARGET_FP_F16INST.
6629 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
6632 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
6633 No longer check last store for clobber of address register.
6635 2018-08-02 Martin Liska <mliska@suse.cz>
6637 PR gcov-profile/86817
6638 * gcov.c (process_all_functions): New function.
6640 (process_file): Move functions processing to
6641 process_all_functions.
6643 2018-08-02 David Malcolm <dmalcolm@redhat.com>
6645 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
6646 "const" to the "gimple *" and "rtx_insn *" parameters.
6647 * dumpfile.h (dump_user_location_t::dump_user_location_t):
6649 (dump_location_t::dump_location_t): Likewise.
6651 2018-08-01 Martin Sebor <msebor@redhat.com>
6653 PR tree-optimization/86650
6654 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
6655 rather than a "gcall *". Directly pass the data of interest
6656 to percent_K_format, rather than building a temporary CALL_EXPR
6658 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
6659 (gimple_fold_builtin_strncat): Adjust.
6660 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
6661 gcall* argument with gimple*.
6662 * gimple-ssa-warn-restrict.c (check_call): Same.
6663 (wrestrict_dom_walker::before_dom_children): Same.
6664 (builtin_access::builtin_access): Same.
6665 (check_bounds_or_overlap): Same
6666 (maybe_diag_overlap): Same.
6667 (maybe_diag_offset_bounds): Same.
6668 * tree-diagnostic.c (default_tree_printer): Move usage of
6669 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
6671 * tree-pretty-print.c (percent_K_format): Add argument.
6672 * tree-pretty-print.h: Add argument.
6673 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
6674 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
6675 (maybe_diag_stxncpy_trunc): Same.
6676 (handle_builtin_stxncpy): Same.
6677 (handle_builtin_strcat): Same.
6679 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6681 * match.pd: Optimise pointer range checks.
6683 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6685 PR tree-optimization/86758
6686 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
6687 to remove pattern statements.
6689 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6691 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
6692 result of dfs_enumerate_from when constructing stmt_vec_infos,
6693 instead of additionally calling get_loop_body.
6695 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6697 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
6699 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
6700 When creating an iv, assert that the step is not known to be zero.
6701 (vect_setup_realignment): Update call accordingly.
6702 * tree-vect-stmts.c (vectorizable_store): Likewise.
6703 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
6705 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6707 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
6708 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
6709 (vectorizable_reduction): Likewise.
6710 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6711 (vect_detect_hybrid_slp_stmts): Likewise.
6712 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6714 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
6716 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
6717 (wide_int_set_zero_nonzero_bits): ...this.
6718 (zero_nonzero_bits_from_vr): Rename to...
6719 (vrp_set_zero_nonzero_bits): ...this.
6720 (extract_range_from_multiplicative_op_1): Abstract wide int
6722 (wide_int_range_multiplicative_op): ...here.
6723 (extract_range_from_binary_expr_1): Extract wide int binary
6724 operations into their own functions.
6725 (wide_int_range_lshift): New.
6726 (wide_int_range_can_optimize_bit_op): New.
6727 (wide_int_range_shift_undefined_p): New.
6728 (wide_int_range_bit_xor): New.
6729 (wide_int_range_bit_ior): New.
6730 (wide_int_range_bit_and): New.
6731 (wide_int_range_trunc_mod): New.
6732 (extract_range_into_wide_ints): New.
6733 (vrp_shift_undefined_p): New.
6734 (extract_range_from_multiplicative_op): New.
6735 (vrp_can_optimize_bit_op): New.
6736 * tree-vrp.h (value_range::dump): New.
6737 (wide_int_range_multiplicative_op): New.
6738 (wide_int_range_lshift):New.
6739 (wide_int_range_shift_undefined_p): New.
6740 (wide_int_range_bit_xor): New.
6741 (wide_int_range_bit_ior): New.
6742 (wide_int_range_bit_and): New.
6743 (wide_int_range_trunc_mod): New.
6744 (zero_nonzero_bits_from_bounds): Rename to...
6745 (wide_int_set_zero_nonzero_bits): ...this.
6746 (zero_nonzero_bits_from_vr): Rename to...
6747 (vrp_set_zero_nonzero_bits): ...this.
6748 (range_easy_mask_min_max): Rename to...
6749 (wide_int_range_can_optimize_bit_op): this.
6750 * vr-values.c (simplify_bit_ops_using_ranges): Rename
6751 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6753 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6755 * tree-vectorizer.h (vect_orig_stmt): New function.
6756 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6757 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6758 (vect_create_epilog_for_reduction): Likewise.
6759 (vectorizable_live_operation): Likewise.
6760 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6761 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6762 * tree-vect-stmts.c (vectorizable_call): Likewise.
6763 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6765 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6767 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6769 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6770 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6771 (vect_transform_loop): Likewise.
6772 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6774 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6776 * tree-vectorizer.h (vect_schedule_slp): Return void.
6777 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6778 (vect_schedule_slp): Likewise.
6780 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6782 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6784 (vect_transform_loop): Update calls accordingly. Schedule SLP
6785 instances before the main loop, if any exist.
6787 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6789 PR tree-optimization/86749
6790 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6791 If the lhs is used in a COND_EXPR, check that it is being used
6792 as the "then" or "else" value.
6794 2018-08-01 Tom de Vries <tdevries@suse.de>
6797 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6798 speculation_safe_value_not_needed.
6800 2018-08-01 Richard Biener <rguenther@suse.de>
6802 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6805 2018-08-01 Martin Liska <mliska@suse.cz>
6807 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6808 format how successful transformation is dumped.
6809 (gimple_mod_pow2_value_transform): Likewise.
6810 (gimple_mod_subtract_transform): Likewise.
6811 (gimple_stringops_transform): Likewise.
6813 2018-08-01 Martin Liska <mliska@suse.cz>
6816 * value-prof.c (interesting_stringop_to_profile_p):
6817 Simplify the code and add BUILT_IN_MEMMOVE.
6818 (gimple_stringops_transform): Likewise.
6820 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
6822 * config/aarch64/aarch64-simd.md
6823 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6824 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6825 use GPI iterator instead of SI mode.
6827 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6829 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6830 rs6000_speculation_barrier.
6831 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6832 new barrier pattern name.
6834 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6836 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
6837 (speculation_barrier): New insn.
6839 2018-08-01 Richard Biener <rguenther@suse.de>
6842 * graphite.h: Include isl/id.h and isl/space.h to allow build
6845 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
6848 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6849 mode for COFF targets.
6850 * defaults.h (TARGET_COFF): Define.
6851 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
6852 TARGET_COFF): Define.
6853 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
6854 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
6855 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
6857 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6859 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
6860 * gimple-streamer-out.c (output_bb): Save it.
6861 * lto-streamer-in.c (input_struct_function_base): Restore
6862 instance discriminator if available. Create map on demand.
6863 * lto-streamer-out.c (output_struct_function_base): Save it if
6865 * final.c (decl_to_instance_map): Document LTO strategy.
6867 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6868 Olivier Hainque <hainque@adacore.com>
6870 * debug.h (decl_to_instance_map_t): New type.
6871 (decl_to_instance_map): Declare.
6872 (maybe_create_decl_to_instance_map): New inline function.
6873 * final.c (bb_discriminator, last_bb_discriminator): New statics,
6874 to track basic block discriminators.
6875 (final_start_function_1): Initialize them.
6876 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6878 (decl_to_instance_map): New variable.
6879 (map_decl_to_instance, maybe_set_discriminator): New functions.
6880 (notice_source_line): Set discriminator.
6882 2018-07-31 Ian Lance Taylor <iant@golang.org>
6884 * targhooks.c (default_have_speculation_safe_value): Add
6887 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6889 * dump-context.h: Include "pretty-print.h".
6890 (dump_context::refresh_dumps_are_enabled): New decl.
6891 (dump_context::emit_item): New decl.
6892 (class dump_context): Add fields "m_test_pp" and
6894 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6895 (temp_dump_context::get_dumped_text): New decl.
6896 (class temp_dump_context): Add field "m_pp".
6897 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6898 (dump_context::refresh_dumps_are_enabled): ...and add a test for
6900 (set_dump_file): Update for above change.
6901 (set_alt_dump_file): Likewise.
6902 (dump_loc): New overload, taking a pretty_printer *.
6903 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
6904 to any test pretty-printer.
6905 (make_item_for_dump_gimple_stmt): New function, adapted from
6906 optinfo::add_gimple_stmt.
6907 (dump_context::dump_gimple_stmt): Call it, and use the result,
6908 eliminating the direct usage of dump_file and alt_dump_file in
6909 favor of indirectly using them via emit_item.
6910 (make_item_for_dump_gimple_expr): New function, adapted from
6911 optinfo::add_gimple_expr.
6912 (dump_context::dump_gimple_expr): Call it, and use the result,
6913 eliminating the direct usage of dump_file and alt_dump_file in
6914 favor of indirectly using them via emit_item.
6915 (make_item_for_dump_generic_expr): New function, adapted from
6917 (dump_context::dump_generic_expr): Call it, and use the result,
6918 eliminating the direct usage of dump_file and alt_dump_file in
6919 favor of indirectly using them via emit_item.
6920 (make_item_for_dump_printf_va): New function, adapted from
6921 optinfo::add_printf_va.
6922 (make_item_for_dump_printf): New function.
6923 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6924 and use the result, eliminating the direct usage of dump_file and
6925 alt_dump_file in favor of indirectly using them via emit_item.
6926 (make_item_for_dump_dec): New function.
6927 (dump_context::dump_dec): Call it, and use the result,
6928 eliminating the direct usage of dump_file and alt_dump_file in
6929 favor of indirectly using them via emit_item.
6930 (make_item_for_dump_symtab_node): New function, adapted from
6931 optinfo::add_symtab_node.
6932 (dump_context::dump_symtab_node): Call it, and use the result,
6933 eliminating the direct usage of dump_file and alt_dump_file in
6934 favor of indirectly using them via emit_item.
6935 (dump_context::begin_scope): Reimplement, avoiding direct usage
6936 of dump_file and alt_dump_file in favor of indirectly using them
6938 (dump_context::emit_item): New member function.
6939 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6940 Set up test pretty-printer on the underlying context. Call
6941 refresh_dumps_are_enabled.
6942 (temp_dump_context::~temp_dump_context): Call
6943 refresh_dumps_are_enabled.
6944 (temp_dump_context::get_dumped_text): New member function.
6945 (selftest::verify_dumped_text): New function.
6946 (ASSERT_DUMPED_TEXT_EQ): New macro.
6947 (selftest::test_capture_of_dump_calls): Run all tests twice, with
6948 and then without optinfo enabled. Add uses of
6949 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
6950 * dumpfile.h: Update comment for the dump_* API.
6951 * optinfo-emit-json.cc
6952 (selftest::test_building_json_from_dump_calls): Update for new
6953 param for temp_dump_context ctor.
6954 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6955 and "m_owned" field.
6956 (optinfo_item::~optinfo_item): Likewise.
6957 (optinfo::add_item): New member function.
6958 (optinfo::emit): Update comment.
6959 (optinfo::add_string): Delete.
6960 (optinfo::add_printf): Delete.
6961 (optinfo::add_printf_va): Delete.
6962 (optinfo::add_gimple_stmt): Delete.
6963 (optinfo::add_gimple_expr): Delete.
6964 (optinfo::add_tree): Delete.
6965 (optinfo::add_symtab_node): Delete.
6966 (optinfo::add_dec): Delete.
6967 * optinfo.h (class dump_context): New forward decl.
6968 (optinfo::add_item): New decl.
6969 (optinfo::add_string): Delete.
6970 (optinfo::add_printf): Delete.
6971 (optinfo::add_printf_va): Delete.
6972 (optinfo::add_gimple_stmt): Delete.
6973 (optinfo::add_gimple_expr): Delete.
6974 (optinfo::add_tree): Delete.
6975 (optinfo::add_symtab_node): Delete.
6976 (optinfo::add_dec): Delete.
6977 (optinfo::add_poly_int): Delete.
6978 (optinfo_item::optinfo_item): Remove "owned" param.
6979 (class optinfo_item): Remove field "m_owned".
6981 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6984 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6985 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6986 requested variable alignment.
6987 (expand_one_ssa_partition): Likewise.
6988 (expand_one_var): Likewise.
6990 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6992 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6993 to speculation_safe_value_not_needed.
6995 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6997 * targhooks.h (speculation_safe_value_not_needed): New prototype.
6998 * targhooks.c (speculation_safe_value_not_needed): New function.
6999 * target.def (have_speculation_safe_value): Update documentation.
7000 * doc/tm.texi: Regenerated.
7002 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7004 * config/aarch64/iterators.md (ALLI_TI): New iterator.
7005 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7007 (despeculate_copy<ALLI:mode>_insn): New insn.
7008 (despeculate_copyti_insn): New insn.
7009 (despeculate_simple<ALLI:mode>): New insn
7010 (despeculate_simpleti): New insn.
7011 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7013 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7014 aarch64_speculation_safe_value.
7015 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7017 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7019 * config/aarch64/aarch64-speculation.cc: New file.
7020 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7021 before pass_reorder_blocks.
7022 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7024 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7025 X14 and X15 when tracking speculation.
7026 * config/aarch64/aarch64.md (register name constants): Add
7027 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7028 (unspec): Add UNSPEC_SPECULATION_TRACKER.
7029 (speculation_barrier): New insn attribute.
7030 (cmp<mode>): Allow SP in comparisons.
7031 (speculation_tracker): New insn.
7032 (speculation_barrier): Add speculation_barrier attribute.
7033 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7034 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7035 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7037 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7039 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7040 aarch64_track_speculation is true.
7041 (tb<optab><mode>1): Likewise.
7042 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7043 generate CB[N]Z when tracking speculation.
7044 (aarch64_split_compare_and_swap): Likewise.
7045 (aarch64_split_atomic_op): Likewise.
7047 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7049 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
7051 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7053 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
7054 (speculation_barrier): New insn.
7056 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7058 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
7059 * config/arm/arm.md (speculation_barrier): New expand.
7060 (speculation_barrier_insn): New pattern.
7062 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
7064 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
7065 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
7066 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
7067 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
7069 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
7070 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
7071 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
7072 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
7073 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
7074 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
7075 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
7076 * builtins.c (expand_speculation_safe_value): New function.
7077 (expand_builtin): Call it.
7078 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
7079 * doc/extend.texi: Document __builtin_speculation_safe_value.
7080 * doc/md.texi: Document "speculation_barrier" pattern.
7081 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
7082 TARGET_HAVE_SPECULATION_SAFE_VALUE.
7083 * doc/tm.texi: Regenerated.
7084 * target.def (have_speculation_safe_value, speculation_safe_value): New
7086 * targhooks.c (default_have_speculation_safe_value): New function.
7087 (default_speculation_safe_value): New function.
7088 * targhooks.h (default_have_speculation_safe_value): Add prototype.
7089 (default_speculation_safe_value): Add prototype.
7091 2018-07-31 David Malcolm <dmalcolm@redhat.com>
7093 * dump-context.h (dump_context::dump_loc): New decl.
7094 * dumpfile.c (dump_context::dump_loc): New member function.
7095 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
7096 and dump_gimple_stmt.
7097 (dump_context::dump_gimple_expr_loc): Likewise, using
7099 (dump_context::dump_generic_expr_loc): Likewise, using
7101 (dump_context::dump_printf_loc_va): Likewise, using
7103 (dump_context::begin_scope): Explicitly using the global function
7104 "dump_loc", rather than the member function.
7106 2018-07-31 Martin Sebor <msebor@redhat.com>
7108 PR tree-optimization/86741
7109 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
7111 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
7113 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
7114 depend on whether prefetch instructions will be emitted or not.
7115 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
7116 will be emitted or not.
7117 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
7118 (TARGET_SETMEM_PFD): New macros.
7120 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7122 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
7123 (NULL_STMT_VEC_INFO): Delete.
7124 (stmt_vec_info::operator*): Likewise.
7125 (stmt_vec_info::operator gimple *): Likewise.
7126 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
7127 of NULL_STMT_VEC_INFO.
7128 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7129 (vect_reassociating_reduction_p): Likewise.
7130 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7131 (vectorizable_store): Likewise.
7132 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
7133 (vec_info::free_stmt_vec_infos): Likewise.
7135 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7137 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
7138 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
7140 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7142 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
7143 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7144 (vec_info::free_stmt_vec_info): New private member functions.
7145 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
7146 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
7147 * tree-parloops.c (gather_scalar_reductions): Remove calls to
7148 set_stmt_vec_info_vec and free_stmt_vec_infos.
7149 * tree-vect-loop.c (_loop_vec_info): Remove call to
7150 set_stmt_vec_info_vec.
7151 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
7152 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
7153 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
7154 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7155 (vec_info::free_stmt_vec_info): ...these new functions. Remove
7156 assignments in {vec_info::,}new_stmt_vec_info that are redundant
7157 with the clearing in the xcalloc.
7158 (stmt_vec_info_vec): Delete.
7159 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
7160 (vectorize_loops): Likewise.
7161 (vec_info::~vec_info): Remove argument from call to
7162 free_stmt_vec_infos.
7163 (vec_info::add_stmt): Remove vinfo argument from call to
7166 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7168 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
7169 rather than a gimple stmt.
7170 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
7171 information for pattern statements when passed the original
7172 statement; instead wait to be passed the pattern statement itself.
7173 Don't call set_vinfo_for_stmt here.
7174 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
7175 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
7176 stmt_vec_infos here.
7177 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
7178 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
7179 stmt_vec_infos entry.
7181 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7183 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
7184 * tree-vectorizer.c (vec_info::replace_stmt): New function.
7185 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
7186 * tree-vect-stmts.c (vectorizable_call): Likewise.
7187 (vectorizable_simd_clone_call): Likewise.
7189 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7191 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
7192 * tree-vectorizer.c (vec_info::remove_stmt): New function.
7193 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
7194 * tree-vect-loop.c (vect_transform_loop): Likewise.
7195 * tree-vect-slp.c (vect_schedule_slp): Likewise.
7196 * tree-vect-stmts.c (vect_remove_stores): Likewise.
7198 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7200 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
7201 (vect_dr_stmt): Delete.
7202 * tree-vectorizer.c (vec_info::lookup_dr): New function.
7203 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
7205 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7206 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
7207 (vect_verify_datarefs_alignment, vect_peeling_supportable)
7208 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7209 (vect_analyze_data_refs): Likewise.
7210 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
7212 (vect_find_same_alignment_drs): Likewise.
7213 (vect_slp_analyze_node_dependences): Update calls accordingly.
7214 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
7215 instead of DR_VECT_AUX.
7216 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
7217 of a vector data references. Use vec_info::lookup_dr instead of
7219 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
7220 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
7221 instead of DR_VECT_AUX.
7223 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7225 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
7227 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7229 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7230 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7231 (vect_gen_prolog_loop_niters): Likewise.
7233 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7235 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
7236 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
7237 (vect_known_alignment_in_bytes, vect_dr_behavior)
7238 (vect_get_scalar_dr_size): Take references as dr_vec_infos
7239 instead of data_references. Update calls to other routines for
7240 which the same change has been made.
7241 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
7242 dr_vec_infos instead of stmt_vec_infos.
7243 (vect_analyze_data_ref_dependence): Update call accordingly.
7244 (vect_slp_analyze_data_ref_dependence)
7245 (vect_record_base_alignments): Use DR_VECT_AUX.
7246 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
7247 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
7248 (vector_alignment_reachable_p, vect_get_data_access_cost)
7249 (vect_peeling_supportable, vect_analyze_group_access_1)
7250 (vect_analyze_group_access, vect_analyze_data_ref_access)
7251 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
7252 (vect_compile_time_alias, vect_small_gap_p)
7253 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
7254 (vect_supportable_dr_alignment): Take references as dr_vec_infos
7255 instead of data_references. Update calls to other routines for
7256 which the same change has been made.
7257 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
7258 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7259 (vect_slp_analyze_and_verify_node_alignment)
7260 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7261 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7262 (vect_setup_realignment): Use dr_vec_infos. Update calls after
7264 (_vect_peel_info::dr): Replace with...
7265 (_vect_peel_info::dr_info): ...this new field.
7266 (vect_peeling_hash_get_most_frequent)
7267 (vect_peeling_hash_choose_best_peeling): Update accordingly.
7268 (vect_peeling_hash_get_lowest_cost):
7269 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
7270 routines for which the same change has been made.
7271 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
7273 * tree-vect-loop-manip.c (get_misalign_in_elems)
7274 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
7276 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7277 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
7278 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
7279 (get_group_load_store_type, get_negative_load_store_type)
7280 (vect_get_data_ptr_increment, vectorizable_store)
7281 (vectorizable_load): Likewise.
7282 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
7283 Update calls to other routines for which the same change has been made.
7285 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7287 * tree-vectorizer.h (vec_info::move_dr): New member function.
7288 (dataref_aux): Rename to...
7289 (dr_vec_info): ...this and add "dr" and "stmt" fields.
7290 (_stmt_vec_info::dr_aux): Update accordingly.
7291 (_stmt_vec_info::data_ref_info): Delete.
7292 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
7293 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
7294 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
7295 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
7297 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
7298 (STMT_VINFO_DR_INFO): New macro.
7299 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
7300 (set_dr_misalignment): Update after rename of dataref_aux.
7301 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
7302 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
7303 initialization of STMT_VINFO_DATA_REF.
7304 * tree-vectorizer.c (vec_info::move_dr): New function.
7305 * tree-vect-patterns.c (vect_recog_bool_pattern)
7306 (vect_recog_mask_conversion_pattern)
7307 (vect_recog_gather_scatter_pattern): Use it.
7308 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
7309 the "dr" and "stmt" fields of dr_vec_info instead of
7310 STMT_VINFO_DATA_REF.
7312 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7314 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
7315 (is_pattern_stmt_p): Use it.
7316 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
7317 on pattern statements.
7319 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7321 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
7322 original stmt as a stmt_vec_info rather than a gimple stmt.
7323 (vect_pattern_recog_1): Take the statement directly as a
7324 stmt_vec_info, rather than via a gimple_stmt_iterator.
7325 Update call to vect_mark_pattern_stmts.
7326 (vect_pattern_recog): Update calls accordingly.
7328 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7330 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
7331 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
7332 a vect_def_type for the first argument.
7333 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
7334 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
7335 operand if it isn't defined by a vectorized statement.
7336 (vect_build_gather_load_calls): Remove the mask_dt argument and
7337 update calls to vect_get_vec_def_for_stmt_copy.
7338 (vectorizable_bswap): Likewise the dt argument.
7339 (vectorizable_call): Update calls to vectorizable_bswap and
7340 vect_get_vec_def_for_stmt_copy.
7341 (vectorizable_simd_clone_call, vectorizable_assignment)
7342 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
7343 (vectorizable_comparison): Update calls to
7344 vect_get_vec_def_for_stmt_copy.
7345 (vectorizable_store): Likewise. Remove now-unnecessary calls to
7347 (vect_get_loop_based_defs): Remove dt argument and update call
7348 to vect_get_vec_def_for_stmt_copy.
7349 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
7350 and vect_get_vec_def_for_stmt_copy.
7351 (vectorizable_load): Update calls to vect_build_gather_load_calls
7352 and vect_get_vec_def_for_stmt_copy.
7353 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7354 (vectorizable_reduction, vectorizable_live_operation): Update calls
7355 to vect_get_vec_def_for_stmt_copy.
7357 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7359 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
7360 and gimple stmt arguments with a stmt_vec_info.
7361 (vect_record_base_alignments): Update calls accordingly.
7362 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
7363 and gimple stmt arguments with a stmt_vec_info.
7364 (vect_build_slp_tree_1): Remove vinfo argument and update call
7365 to vect_record_max_nunits.
7366 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
7367 and vect_record_max_nunits.
7369 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7371 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
7372 file and take a stmt_vec_info instead of a gimple stmt.
7373 (supportable_widening_operation, vect_finish_replace_stmt)
7374 (vect_finish_stmt_generation, vect_get_store_rhs)
7375 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
7376 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
7377 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
7378 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
7379 (vect_create_data_ref_ptr, bump_vector_ptr)
7380 (vect_permute_store_chain, vect_setup_realignment)
7381 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
7382 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
7383 (vectorizable_reduction, vectorizable_induction)
7384 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
7385 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
7386 than gimple stmts as arguments.
7387 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
7388 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
7389 (can_group_stmts_p, vect_check_gather_scatter)
7390 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
7391 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
7392 (vect_permute_load_chain, vect_shift_permute_load_chain)
7393 (vect_transform_grouped_load)
7394 (vect_record_grouped_load_vectors): Likewise.
7395 * tree-vect-loop.c (vect_fixup_reduc_chain)
7396 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
7397 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
7398 (vectorizable_reduction, vectorizable_induction)
7399 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
7400 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
7401 (vect_get_load_store_mask): Likewise.
7402 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
7403 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
7404 * tree-vect-stmts.c (vect_mark_relevant)
7405 (is_simple_and_all_uses_invariant)
7406 (exist_non_indexing_operands_for_use_p, process_use)
7407 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
7408 (vect_get_vec_def_for_operand, vect_get_vec_defs)
7409 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
7410 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
7411 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
7412 (get_negative_load_store_type, get_load_store_type)
7413 (vect_check_load_store_mask, vect_check_store_rhs)
7414 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
7415 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7416 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7417 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7418 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
7419 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
7420 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
7421 (supportable_widening_operation): Likewise.
7423 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7425 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
7426 a stmt_vec_info instead of a gcall.
7427 (vect_check_gather_scatter): Update call accordingly.
7428 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
7430 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
7431 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
7432 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
7433 instead of a gimple stmt.
7434 (vect_transform_loop): Update calls accordingly.
7435 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
7436 stmt_vec_infos instead of gimple stmts.
7437 (vect_analyze_slp_instance): Update use accordingly.
7438 * tree-vect-stmts.c (read_vector_array, write_vector_array)
7439 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
7440 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
7441 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
7442 (vect_gen_widened_results_half, vect_get_loop_based_defs)
7443 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
7444 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
7445 down to subroutines.
7447 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7449 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
7450 of the worklist from a vector of gimple stmts to a vector of
7452 * tree-vect-stmts.c (vect_mark_relevant, process_use)
7453 (vect_mark_stmts_to_be_vectorized): Likewise
7455 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7457 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
7458 statement before passing it to vect_analyze_stmt.
7459 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
7460 the chain of phi vector definitions. Track the exit phi via its
7462 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
7463 STMT_VINFO_REDUC_DEF.
7464 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
7465 stmt_vec_infos to handle the statement chains.
7466 (vect_get_slp_defs): Record the first statement in the node
7467 using a stmt_vec_info.
7468 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
7469 statements here and pass their stmt_vec_info down to subroutines.
7470 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
7471 down to vect_finish_stmt_generation.
7472 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
7473 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
7474 stmt_vec_infos to subroutines.
7475 (vect_remove_stores): Use stmt_vec_infos to handle the statement
7478 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7480 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
7481 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
7482 (vect_permute_store_chain, vect_setup_realignment)
7483 (vect_permute_load_chain, vect_shift_permute_load_chain)
7484 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
7485 stmts internally, and when passing values to other vectorizer routines.
7486 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7487 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
7488 (vect_analyze_loop_operations, get_initial_def_for_reduction)
7489 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
7490 (vectorizable_reduction, vectorizable_induction)
7491 (vectorizable_live_operation, vect_transform_loop_stmt)
7492 (vect_transform_loop): Likewise.
7493 * tree-vect-patterns.c (vect_reassociating_reduction_p)
7494 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
7495 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
7496 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7497 (vect_slp_analyze_node_operations_1): Likewise.
7498 * tree-vect-stmts.c (vect_mark_relevant, process_use)
7499 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
7500 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
7501 (vect_finish_stmt_generation_1, get_group_load_store_type)
7502 (get_load_store_type, vect_build_gather_load_calls)
7503 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7504 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7505 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7506 (vectorizable_store, vectorizable_load, vectorizable_condition)
7507 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
7508 (supportable_widening_operation): Likewise.
7509 (vect_get_vector_types_for_stmt): Likewise.
7510 * tree-vectorizer.h (vect_dr_behavior): Likewise.
7512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7514 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7515 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
7516 (vect_permute_store_chain, vect_permute_load_chain)
7517 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
7518 repeated stmt_vec_info lookups.
7519 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
7520 (vect_update_ivs_after_vectorizer): Likewise.
7521 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
7522 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7523 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
7524 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7525 (vect_bb_slp_scalar_cost): Likewise.
7526 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
7528 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7530 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
7531 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
7532 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
7533 to get gassigns and gcalls, rather than operating on generc gimple
7535 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
7536 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
7537 (vectorizable_load, vect_analyze_stmt): Likewise.
7538 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
7540 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7542 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
7543 return stmt_vec_infos rather than gimple stmts. Do not accept
7545 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
7547 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7548 Update use of get_later_stmt.
7549 (vect_get_constant_vectors): Update call accordingly.
7550 (vect_schedule_slp_instance): Likewise
7551 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
7552 (vect_slp_analyze_instance_dependence): Likewise.
7553 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
7555 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7557 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
7558 (stmt_info_for_cost::stmt_info): ...this new field.
7559 (add_stmt_costs): Update accordingly.
7560 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7561 (vect_get_known_peeling_cost): Likewise.
7562 (vect_estimate_min_profitable_iters): Likewise.
7563 * tree-vect-stmts.c (record_stmt_cost): Likewise.
7565 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7567 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
7568 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7569 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7571 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
7573 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7575 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
7576 a gimple stmt to a stmt_vec_info.
7577 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7579 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7581 * tree-vectorizer.h (vec_info::grouped_stores): Change from
7582 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7583 (_loop_vec_info::reduction_chains): Likewise.
7584 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
7586 * tree-vect-slp.c (vect_analyze_slp): Likewise.
7588 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7590 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
7591 a gimple stmt to a stmt_vec_info.
7592 (_stmt_vec_info::next_element): Likewise.
7593 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
7594 (vect_slp_analyze_and_verify_node_alignment)
7595 (vect_analyze_group_access_1, vect_analyze_group_access)
7596 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
7597 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
7598 (vect_supportable_dr_alignment): Update accordingly.
7599 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
7600 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
7601 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
7602 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7603 * tree-vect-slp.c (vect_build_slp_tree_1)
7604 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
7605 (vect_split_slp_store_group, vect_analyze_slp_instance)
7606 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
7607 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
7608 (get_group_load_store_type, get_load_store_type)
7609 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
7610 (vect_transform_stmt, vect_remove_stores): Likewise.
7612 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7614 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
7616 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7617 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
7618 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
7619 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
7620 (vector_alignment_reachable_p, vect_get_data_access_cost)
7621 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
7622 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
7623 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
7624 (vect_analyze_group_access_1, vect_analyze_group_access)
7625 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
7626 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
7627 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
7628 result of vect_dr_stmt and use the stmt_vec_info instead of
7629 the associated gimple stmt.
7630 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7631 (vect_gen_prolog_loop_niters): Likewise.
7632 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7634 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7636 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
7637 to a vec<stmt_vec_info>.
7638 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
7639 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
7641 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
7642 to a vec<stmt_vec_info>.
7643 (bst_traits::value_type, bst_traits::value_type): Likewise.
7644 (bst_traits::hash): Update accordingly.
7645 (vect_get_and_check_slp_defs): Change the stmts parameter from
7646 a vec<gimple *> to a vec<stmt_vec_info>.
7647 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
7648 (vect_build_slp_tree): Likewise.
7649 (vect_build_slp_tree_2): Likewise. Update uses of
7650 SLP_TREE_SCALAR_STMTS.
7651 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
7652 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
7653 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
7654 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
7655 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
7656 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
7657 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
7658 (vect_get_constant_vectors, vect_get_slp_defs)
7659 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
7660 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
7661 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
7662 instead of gimple stmts.
7663 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
7664 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
7665 (vect_slp_analyze_instance_dependence): Update uses of
7666 SLP_TREE_SCALAR_STMTS.
7667 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7668 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
7669 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
7670 (get_initial_defs_for_reduction): Likewise.
7671 (vect_create_epilog_for_reduction): Likewise.
7672 (vectorize_fold_left_reduction): Likewise.
7673 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
7674 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
7675 (can_vectorize_live_stmts): Likewise.
7677 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7679 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
7680 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7681 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
7683 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
7685 (gather_scalar_reductions): Update after above interface changes.
7686 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
7687 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
7689 (vect_force_simple_reduction): Likewise.
7690 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
7691 LOOP_VINFO_REDUCTIONS.
7692 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7694 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7696 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
7697 a gimple stmt to a stmt_vec_info.
7698 * tree-vect-loop.c (vect_active_double_reduction_p)
7699 (vect_force_simple_reduction, vectorizable_reduction): Update
7702 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7704 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
7705 vec<gimple *> to a vec<stmt_vec_info>.
7706 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
7707 the reduction_phis argument from a vec<gimple *> to a
7709 (vectorizable_reduction): Likewise the phis local variable that
7710 is passed to vect_create_epilog_for_reduction. Update for new type
7711 of SLP_TREE_VEC_STMTS.
7712 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
7713 (vectorizable_live_operation): Likewise.
7714 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
7715 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
7717 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7719 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
7720 a gimple stmt to a stmt_vec_info.
7721 (vectorizable_condition, vectorizable_live_operation)
7722 (vectorizable_reduction, vectorizable_induction): Pass back the
7723 vectorized statement as a stmt_vec_info.
7724 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
7725 use of STMT_VINFO_VEC_STMT.
7726 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
7727 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
7728 as stmt_vec_infos rather than gimple stmts.
7729 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
7731 (vectorizable_live_operation): Likewise.
7732 (vectorizable_reduction, vectorizable_induction): Likewise,
7733 updating use of STMT_VINFO_VEC_STMT.
7734 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
7735 of STMT_VINFO_VEC_STMT.
7736 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
7737 (vectorizable_simd_clone_call, vectorizable_conversion)
7738 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7739 (vectorizable_store, vectorizable_load, vectorizable_condition)
7740 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
7741 from a gimple stmt to a stmt_vec_info.
7742 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
7743 pointer to a stmt_vec_info to the vectorizable_* routines.
7745 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7747 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7748 a gimple stmt to a stmt_vec_info.
7749 (is_pattern_stmt_p): Update accordingly.
7750 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7751 (vect_record_grouped_load_vectors): Likewise.
7752 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7753 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7754 (vect_model_reduction_cost): Likewise.
7755 (vect_create_epilog_for_reduction): Likewise.
7756 (vectorizable_reduction, vectorizable_induction): Likewise.
7757 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7758 Return the stmt_vec_info for the pattern statement.
7759 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7760 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7761 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7762 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7763 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7764 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7765 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7766 (free_stmt_vec_info, vect_is_simple_use): Likewise.
7768 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7770 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7771 (vect_finish_stmt_generation): Likewise.
7772 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7773 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7774 (vect_build_gather_load_calls): Use the return value of the above
7775 functions instead of a separate call to vinfo_for_stmt. Use narrow
7776 scopes for the input gimple stmt and wider scopes for the associated
7777 stmt_vec_info. Use vec_info::lookup_def when setting these
7778 stmt_vec_infos from an SSA_NAME definition.
7779 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7780 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7781 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7782 (vectorizable_store, vectorizable_load, vectorizable_condition)
7783 (vectorizable_comparison): Likewise.
7784 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7785 (vectorizable_reduction): Likewise.
7787 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7789 * tree-vectorizer.h (vect_is_simple_use): Add an optional
7790 stmt_vec_info * parameter before the optional gimple **.
7791 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7792 (process_use, vect_get_vec_def_for_operand_1): Update callers.
7793 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7794 * tree-vect-loop.c (vectorizable_reduction): Likewise.
7795 (vectorizable_live_operation): Likewise.
7796 * tree-vect-patterns.c (type_conversion_p): Likewise.
7797 (vect_look_through_possible_promotion): Likewise.
7798 (vect_recog_rotate_pattern): Likewise.
7799 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7801 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7803 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7804 a typedef to a wrapper class.
7805 (NULL_STMT_VEC_INFO): New macro.
7806 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7807 (stmt_vec_info::operator*): New function.
7808 (stmt_vec_info::operator gimple *): Likewise.
7809 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7810 (add_stmt_costs): Likewise.
7811 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7812 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7813 (vect_get_known_peeling_cost): Likewise.
7814 (vect_estimate_min_profitable_iters): Likewise.
7815 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7816 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7817 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7818 (vectorizable_store, free_stmt_vec_infos): Likewise.
7819 (new_stmt_vec_info): Change return type of xcalloc to
7822 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7824 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7825 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7826 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7827 a single_imm_use-based sequence.
7828 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7830 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7832 * tree-vectorizer.h (vec_info::lookup_def): Declare.
7833 * tree-vectorizer.c (vec_info::lookup_def): New function.
7834 * tree-vect-patterns.c (vect_get_internal_def): Use it.
7835 (vect_widened_op_tree): Likewise.
7836 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7837 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
7838 (vectorizable_reduction): Likewise.
7839 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
7841 (vect_is_slp_reduction): Update calls accordingly. Use
7842 vec_info::lookup_def.
7843 (vect_is_simple_reduction): Likewise
7844 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
7846 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7848 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
7849 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
7850 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
7852 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
7853 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
7854 (vect_update_vf_for_slp, vect_analyze_loop_operations)
7855 (vect_is_slp_reduction, vectorizable_induction)
7856 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
7857 * tree-vect-patterns.c (vect_init_pattern_stmt):
7858 (vect_determine_min_output_precision_1, vect_determine_precisions)
7859 (vect_pattern_recog): Likewise.
7860 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
7861 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
7862 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
7863 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7864 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
7865 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
7866 info field from a loop to a loop_vec_info.
7868 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7870 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7871 (vec_info::add_stmt): Declare.
7872 * tree-vectorizer.c (vec_info::add_stmt): New function.
7873 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7874 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7875 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7876 (vectorizable_induction): Likewise.
7877 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7878 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7879 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7880 (vectorizable_load): Likewise.
7881 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7882 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7883 (vect_recog_gather_scatter_pattern): Likewise.
7884 (append_pattern_def_seq): Likewise. Remove a check that is
7885 performed by add_stmt itself.
7887 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7889 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7890 which make_ssa_name was called with new_stmt before new_stmt
7893 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7895 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7897 (vect_is_slp_reduction): ...here...
7898 (vect_is_simple_reduction): ...and here. Remove repetition of tests
7899 that are already known to be false.
7901 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7903 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7904 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
7905 STMT_VINFO_NUM_SLP_USES when it's true.
7906 (vect_free_slp_instance): Add a final_p parameter and pass it to
7908 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7909 (vect_analyze_slp_instance): Likewise.
7910 (vect_slp_analyze_operations): Likewise.
7911 (vect_slp_analyze_bb_1): Likewise.
7912 * tree-vectorizer.c (vec_info): Likewise.
7913 * tree-vect-loop.c (vect_transform_loop): Likewise.
7915 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7917 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7918 function is not called for second and subsequent members of
7921 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7923 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7924 cases for nested loops from here to ...
7925 (vect_create_epilog_for_reduction): ...here. Only call
7926 vect_is_simple_use for inner-loop reductions.
7928 2018-07-31 Martin Liska <mliska@suse.cz>
7930 PR gcov-profile/85338
7931 PR gcov-profile/85350
7932 PR gcov-profile/85372
7933 * profile.c (struct location_triplet): New.
7934 (struct location_triplet_hash): Likewise.
7935 (output_location): Do not output a BB that
7936 is already recorded for a line.
7937 (branch_prob): Use streamed_locations.
7939 2018-07-31 Martin Liska <mliska@suse.cz>
7941 PR gcov-profile/85370
7942 * coverage.c (coverage_begin_function): Do not mark target
7943 clones as artificial functions.
7945 2018-07-31 Martin Liska <mliska@suse.cz>
7947 PR gcov-profile/83813
7948 PR gcov-profile/84758
7949 PR gcov-profile/85217
7950 PR gcov-profile/85332
7951 * profile.c (branch_prob): Do not record GOTO expressions
7952 for GIMPLE statements which locations are already streamed.
7954 2018-07-31 Olivier Hainque <hainque@adacore.com>
7956 * gcc.c (handle_spec_function): Accept a soft_matched_part
7957 argument, as do_spec_1. Pass it down to ...
7958 (eval_spec_function): Accept a soft_matched_part argument,
7959 and pass it down to ...
7960 (do_spec_2): Accept a soft_matched_part argument, and pass
7961 it down to do_spec_1.
7962 (do_spec_1): Pass soft_matched_part to handle_spec_function.
7963 (handle_braces): Update call to handle_spec_function.
7964 (driver::set_up_specs): Update calls to do_spec_2.
7965 (compare_debug_dump_opt_spec_function): Likewise.
7966 (compare_debug_self_opt_spec_function): Likewise.
7968 2018-07-31 Olivier Hainque <hainque@adacore.com>
7970 * common.opt (nolibc): New option.
7971 * doc/invoke.texi (Link Options): Document it.
7972 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7973 * config/alpha/linux.h: Likewise.
7974 * config/arc/elf.h: Likewise.
7975 * config/arm/uclinux-elf.h: Likewise.
7976 * config/arm/unknown-elf.h: Likewise.
7977 * config/avr/avrlibc.h: Likewise.
7978 * config/bfin/bfin.h: Likewise.
7979 * config/bfin/linux.h: Likewise.
7980 * config/bfin/uclinux.h: Likewise.
7981 * config/darwin.h: Likewise.
7982 * config/darwin10.h: Likewise.
7983 * config/darwin12.h: Likewise.
7984 * config/gnu-user.h: Likewise.
7985 * config/lm32/uclinux-elf.h: Likewise.
7986 * config/pa/pa-hpux11.h: Likewise.
7987 * config/pa/pa64-hpux.h: Likewise.
7988 * config/sparc/sparc.h: Likewise.
7990 2018-07-31 Olivier Hainque <hainque@adacore.com>
7992 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7993 undefined variables.
7995 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7998 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8001 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8003 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8004 terminated string literal.
8006 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
8008 PR rtl-optimization/85160
8009 * combine.c (is_just_move): New function.
8010 (try_combine): Allow combining two instructions into two if neither of
8011 the original instructions was a move.
8013 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
8016 * doc/extend.texi (Global Register Variables): Discourage use of type
8018 (Local Register Variables): Likewise.
8020 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
8022 PR tree-optimization/86506
8023 * hwint.h (ceil_log2): Resync with hwint.c implementation.
8025 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
8028 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8029 hard_regno, make sure no insn between `from` and `to` clobbers it.
8031 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
8032 Tom de Vries <tdevries@suse.de>
8034 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8035 (PTX_DEFAULT_RUNTIME_DIM): ... this.
8036 (nvptx_goacc_validate_dims): Set default worker and gang dims to
8037 PTX_DEFAULT_RUNTIME_DIM.
8038 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8040 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
8042 * config/pa/pa.c (pa_output_addr_vec): Align address table.
8043 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8044 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
8046 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
8048 * config/rs6000/constraints.md (wG constraint): Delete, no longer
8050 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
8051 predicate to reflect toc fusion has been deleted.
8052 (toc_fusion_mem_raw): Delete, no longer used.
8053 (toc_fusion_mem_wrapped): Likewise.
8054 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
8056 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
8057 Delete, no longer used.
8058 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
8059 meant to be used for toc fusion.
8060 (rs6000_debug_print_mode): Delete toc fusion debugging.
8061 (rs6000_debug_reg_global): Likewise.
8062 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
8063 fusion and secondary reload support that were never used.
8064 (rs6000_option_override_internal): Delete TOC fusion, that was only
8065 partially defined, and it did not work unless you also used the
8067 (rs6000_legitimate_address_p): Delete TOC fusion support.
8068 (rs6000_opt_masks): Likewise.
8069 (fusion_wrap_memory_address): Delete function, no longer used.
8070 (fusion_split_address); Delete TOC fusion support.
8071 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
8072 longer used with toc fusion being deleted.
8073 (TARGET_TOC_FUSION_FP): Likewise.
8074 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
8076 (toc fusion spliter): Delete TOC fusion support.
8077 (toc_fusionload_<mode>): Likewise.
8078 (toc_fusionload_di): Likewise.
8079 (fusion_gpr_load_<mode>): Delete generator function, this insn no
8080 longer needs to be named. Rename predicate to delete TOC fusion.
8081 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8082 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8083 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8084 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8085 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
8087 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
8089 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
8090 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
8091 __int128 in built-in function prototypes.
8092 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
8093 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
8095 2018-07-27 Martin Sebor <msebor@redhat.com>
8097 PR tree-optimization/86696
8098 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
8099 types, including enums.
8100 (handle_char_store): Be prepared for the above function to fail.
8102 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
8104 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
8105 when optimization level is lower than 2 or optimize for size.
8107 2018-07-26 Martin Sebor <msebor@redhat.com>
8109 PR tree-optimization/86043
8110 PR tree-optimization/86042
8111 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
8112 (get_string_cst_length): Rename...
8113 (get_min_string_length): ...to this. Add argument.
8114 (handle_char_store): Extend to handle multi-character stores by
8116 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
8117 * tree.h (initializer_zerop): Add argument.
8119 2018-07-26 Jakub Jelinek <jakub@redhat.com>
8122 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
8123 declare target to variables if they have always,{to,from,tofrom} map
8126 2018-07-26 Martin Liska <mliska@suse.cz>
8129 * lto-wrapper.c: Add linker_output as prefix
8130 for ltrans_output_file.
8132 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
8134 PR rtl-optimization/85805
8135 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
8136 value for hard registers if that was written in the same mode.
8138 2018-07-26 Martin Liska <mliska@suse.cz>
8140 PR gcov-profile/86536
8141 * gcov.c (format_gcov): Use printf format %.*f directly
8142 and do not handle special values.
8144 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8146 * common/config/arc/arc-common.c (arc_option_optimization_table):
8147 Update default optimizations for size.
8149 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8151 * config/arc/arc.md (movsf_insn): Add short instruction selection.
8152 * config/arc/constraints.md (CfZ): New constraint.
8153 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
8154 (subsf3_fpu): Likewise.
8155 (cmpsf_fpu): Likewise.
8156 (cmpsf_fpu_uneq): Likewise.
8158 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8160 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
8162 (arc_is_uncached_mem_p): uncached applies to both the variable and
8165 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
8167 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
8170 2018-07-25 David Malcolm <dmalcolm@redhat.com>
8172 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
8173 field "m_scopes" from vec to auto_vec.
8175 2018-07-25 Martin Liska <mliska@suse.cz>
8177 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
8180 2018-07-25 Richard Biener <rguenther@suse.de>
8183 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
8184 special wrt context_die late.
8185 (gen_subprogram_die): Re-use DIEs in local scope.
8187 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
8189 PR tree-optimization/86644
8190 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
8192 2018-07-25 Martin Liska <mliska@suse.cz>
8195 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
8196 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
8198 2018-07-25 Martin Liska <mliska@suse.cz>
8201 * params.def: Explain ASan abbreviation and provide
8202 a documentation link.
8204 2018-07-24 Martin Sebor <msebor@redhat.com>
8206 PR tree-optimization/86622
8207 PR tree-optimization/86532
8208 * builtins.h (string_length): Declare.
8209 * builtins.c (c_strlen): Correct handling of non-constant offsets.
8210 (check_access): Be prepared for non-constant length ranges.
8211 (string_length): Make extern.
8212 * expr.c (string_constant): Only handle the minor non-constant
8213 array index. Use string_constant to compute the length of
8214 a generic string constant.
8216 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
8218 PR tree-optimization/86618
8219 * tree-vect-stmts.c (vectorizable_call): Don't take the address
8220 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
8222 2018-07-24 David Malcolm <dmalcolm@redhat.com>
8224 PR tree-optimization/86636
8225 * json.cc (json::object::set): Fix comment. Add assertions.
8226 (json::array::append): Move here from json.h. Add comment and an
8228 (json::string::string): Likewise.
8229 * json.h (json::array::append): Move to json.cc.
8230 (json::string::string): Likewise.
8231 * optinfo-emit-json.cc
8232 (optrecord_json_writer::impl_location_to_json): Assert that we
8233 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
8234 wrapper around it. Expand the location once, rather than three
8236 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
8237 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
8239 (optrecord_json_writer::optinfo_to_json): Likewise, in four
8240 places. Fix some overlong lines.
8242 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
8244 * config/aarch64/aarch64-simd.md
8245 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
8246 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
8247 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
8248 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
8249 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
8250 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
8251 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
8252 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
8253 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
8255 2018-07-24 Jakub Jelinek <jakub@redhat.com>
8258 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
8259 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
8260 and abs_d == d, do the power of two handling if profitable.
8262 2018-07-24 Richard Biener <rguenther@suse.de>
8264 * match.pd: Add BIT_FIELD_REF canonicalizations.
8266 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
8269 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
8271 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
8273 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
8274 terminated STRING_CST object.
8276 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
8278 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
8280 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
8282 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
8283 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
8284 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
8285 the elements into a register.
8286 (rs6000_split_v4si_init_di_reg): Delete.
8287 (rs6000_split_v4si_init): Delete.
8288 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
8289 (vsx_init_v4si): Rewrite as a define_expand.
8291 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
8293 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
8294 zero_extend argument from memory): New.
8296 2018-07-22 Martin Sebor <msebor@redhat.com>
8299 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
8300 diagnosing calls with unknown arguments unless -Walloca-larger-than
8301 is restricted to less than PTRDIFF_MAX bytes.
8303 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
8305 * doc/gcov.texi (Invoking Gcov): Editorial changes.
8307 2018-07-20 David Malcolm <dmalcolm@redhat.com>
8309 * pretty-print.c (text_info::set_location): Remove redundant
8310 "line_table" parameter from call to rich_location::set_range.
8312 2018-07-20 Martin Sebor <msebor@redhat.com>
8315 * builtins.c (expand_builtin_alloca): Adjust.
8316 * calls.c (alloc_max_size): Simplify.
8317 * cgraphunit.c (cgraph_node::expand): Adjust.
8318 * common.opt (larger_than_size, warn_frame_larger_than): Remove
8320 (frame_larger_than_size): Same.
8321 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
8322 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
8324 * doc/invoke.texi (GCC Command Options): Document option arguments.
8325 Explain byte-size arguments and suffixes.
8326 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
8327 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
8328 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
8329 * doc/options.texi (UInteger): Expand.
8330 (Host_Wide_Int, ByteSize): Document new properties.
8331 * final.c (final_start_function_1): Include sizes in an error message.
8332 * function.c (frame_offset_overflow): Same.
8333 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
8334 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
8335 Diagnose unbounded alloca calls only for limits of less than
8337 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
8338 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
8340 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
8341 only for limits of less than PTRDIFF_MAX.
8342 * langhooks-def.h (lhd_handle_option): Change function argument
8344 * langhooks.c (lhd_handle_option): Same.
8345 * langhooks.h (handle_option): Same.
8346 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
8348 (var_type, var_type_struct): Same.
8349 (var_set): Handle ByteSize flag.
8350 * optc-gen.awk: Add comments to output to ease debugging. Make
8351 use of HOST_WIDE_INT where appropriate.
8352 * opts-gen-save.awk: Use %lx to format unsigned long.
8353 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
8354 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
8355 arguments. Parse bytes-size suffixes.
8356 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
8357 (enum_value_to_arg): Same.
8358 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
8359 (handle_option): Adjust.
8360 (generate_option): Change function argument to HOST_WIDE_INT.
8361 (cmdline_handle_error): Adjust.
8362 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
8363 (set_option): Change function argument to HOST_WIDE_INT.
8364 (option_enabled): Handle cl_host_wide_int.
8365 (get_option_state): Handle CLVC_SIZE.
8366 (control_warning_option): Same.
8367 * opts.c (common_handle_option): Change function argument to
8368 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
8369 OPT_Wvla_larger_than_.
8370 * opts.h (enum cl_var_type): Add an enumerator.
8371 * stor-layout.c (layout_decl): Print a more meaningful warning.
8372 * toplev.c (output_stack_usage): Adjust.
8374 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
8376 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
8377 call to inline_expand_builtin_string_cmp.
8378 (expand_builtin_strcmp): Likewise.
8379 (expand_builtin_strncmp): Likewise.
8380 (inline_string_cmp): Delete the last parameter, change char_type_node
8381 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
8383 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
8384 the inlining expansion on target where the type of the call has same or
8385 narrower precision than unsigned char.
8387 2018-07-20 David Malcolm <dmalcolm@redhat.com>
8389 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
8390 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
8391 * common.opt (fsave-optimization-record): New option.
8392 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
8393 * doc/invoke.texi (-fsave-optimization-record): New option.
8394 * dumpfile.c: Include "optinfo-emit-json.h".
8395 (struct kv_pair): Move to coretypes.h.
8396 (optgroup_options): Make non-static.
8397 (dump_context::end_scope): Call
8398 optimization_records_maybe_pop_dump_scope.
8399 * dumpfile.h (optgroup_options): New decl.
8400 * json.cc: New file.
8402 * optinfo-emit-json.cc: New file.
8403 * optinfo-emit-json.h: New file.
8404 * optinfo.cc: Include "optinfo-emit-json.h".
8405 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
8406 (optinfo_enabled_p): Check optimization_records_enabled_p.
8407 (optinfo_wants_inlining_info_p): Likewise.
8408 * optinfo.h: Update comment.
8409 * profile-count.c (profile_quality_as_string): New function.
8410 * profile-count.h (profile_quality_as_string): New decl.
8411 (profile_count::quality): New accessor.
8412 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
8413 and optinfo_emit_json_cc_tests.
8414 * selftest.h (selftest::json_cc_tests): New decl.
8415 (selftest::optinfo_emit_json_cc_tests): New decl.
8416 * toplev.c: Include "optinfo-emit-json.h".
8417 (compile_file): Call optimization_records_finish.
8418 (do_compile): Call optimization_records_start.
8419 * tree-ssa-live.c: Include optinfo.h.
8420 (remove_unused_scope_block_p): Retain inlining information if
8421 optinfo_wants_inlining_info_p returns true.
8423 2018-07-20 Richard Biener <rguenther@suse.de>
8426 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
8427 to cover -flto-partition=none.
8429 2018-07-20 Martin Liska <mliska@suse.cz>
8431 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
8432 (get_decl_source_range): Remove unused function.
8434 2018-07-20 Richard Biener <rguenther@suse.de>
8436 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
8437 (struct vn_phi_s): Likewise.
8438 (struct vn_reference_s): Likewise.
8439 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
8440 for searching the slot of an entry known to be in the hash itself.
8441 (vn_phi_hasher::equal): Likewise.
8442 (vn_reference_hasher::equal): Likewise.
8443 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
8445 (optimistic_info, current_info): Remove, keeping only valid_info.
8446 (vn_reference_lookup_1): Remove fallback lookup.
8447 (vn_reference_lookup_2): Likewise.
8448 (vn_nary_op_lookup_1): Likewise.
8449 (vn_phi_lookup): Likewise.
8450 (vn_nary_build_or_lookup_1): Make sure to not chain the built
8452 (vn_reference_insert): Adjust, chain the inserted hash element
8453 at last_inserted_ref.
8454 (vn_reference_insert_pieces): Likewise.
8455 (visit_reference_op_call): Likewise.
8456 (vn_nary_op_insert_into): Chain the inserted hash element at
8458 (vn_nary_op_insert_pieces): Adjust.
8459 (vn_nary_op_insert): Likewise.
8460 (vn_nary_op_insert_stmt): Likewise.
8461 (vn_phi_insert): Adjust, chain the inserted hash element at
8463 (process_scc): Remove clearing and copying the optimistic
8464 table. Instead remove elements inserted during an optimistic
8465 iteration from the single table we maintain.
8466 (init_scc_vn): Adjust.
8467 (free_scc_vn): Likewise.
8468 (sccvn_dom_walker::record_cond): Likewise.
8469 (sccvn_dom_walker::after_dom_children): Likewise.
8471 2018-07-19 Martin Sebor <msebor@redhat.com>
8473 PR tree-optimization/84047
8474 PR tree-optimization/83776
8475 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
8476 (check_array_bounds): Call it.
8478 2018-07-19 Martin Sebor <msebor@redhat.com>
8480 * align.h (align_flags): Use member initialization.
8482 2018-07-19 David Malcolm <dmalcolm@redhat.com>
8484 * Makefile.in (OBJS): Add optinfo.o.
8485 * coretypes.h (class symtab_node): New forward decl.
8486 (struct cgraph_node): New forward decl.
8487 (class varpool_node): New forward decl.
8488 * dump-context.h: New file.
8489 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
8491 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
8492 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
8493 (set_alt_dump_file): Likewise.
8494 (dump_context::~dump_context): New dtor.
8495 (dump_gimple_stmt): Move implementation to...
8496 (dump_context::dump_gimple_stmt): ...this new member function.
8497 Add the stmt to any pending optinfo, creating one if need be.
8498 (dump_gimple_stmt_loc): Move implementation to...
8499 (dump_context::dump_gimple_stmt_loc): ...this new member function.
8500 Start a new optinfo and add the stmt to it.
8501 (dump_gimple_expr): Move implementation to...
8502 (dump_context::dump_gimple_expr): ...this new member function.
8503 Add the stmt to any pending optinfo, creating one if need be.
8504 (dump_gimple_expr_loc): Move implementation to...
8505 (dump_context::dump_gimple_expr_loc): ...this new member function.
8506 Start a new optinfo and add the stmt to it.
8507 (dump_generic_expr): Move implementation to...
8508 (dump_context::dump_generic_expr): ...this new member function.
8509 Add the tree to any pending optinfo, creating one if need be.
8510 (dump_generic_expr_loc): Move implementation to...
8511 (dump_context::dump_generic_expr_loc): ...this new member
8512 function. Add the tree to any pending optinfo, creating one if
8514 (dump_printf): Move implementation to...
8515 (dump_context::dump_printf_va): ...this new member function. Add
8516 the text to any pending optinfo, creating one if need be.
8517 (dump_printf_loc): Move implementation to...
8518 (dump_context::dump_printf_loc_va): ...this new member function.
8519 Start a new optinfo and add the stmt to it.
8520 (dump_dec): Move implementation to...
8521 (dump_context::dump_dec): ...this new member function. Add the
8522 value to any pending optinfo, creating one if need be.
8523 (dump_context::dump_symtab_node): New member function.
8524 (dump_context::get_scope_depth): New member function.
8525 (dump_context::begin_scope): New member function.
8526 (dump_context::end_scope): New member function.
8527 (dump_context::ensure_pending_optinfo): New member function.
8528 (dump_context::begin_next_optinfo): New member function.
8529 (dump_context::end_any_optinfo): New member function.
8530 (dump_context::s_current): New global.
8531 (dump_context::s_default): New global.
8532 (dump_scope_depth): Delete global.
8533 (dumpfile_ensure_any_optinfo_are_flushed): New function.
8534 (dump_symtab_node): New function.
8535 (get_dump_scope_depth): Reimplement in terms of dump_context.
8536 (dump_begin_scope): Likewise.
8537 (dump_end_scope): Likewise.
8538 (selftest::temp_dump_context::temp_dump_context): New ctor.
8539 (selftest::temp_dump_context::~temp_dump_context): New dtor.
8540 (selftest::verify_item): New function.
8541 (ASSERT_IS_TEXT): New macro.
8542 (ASSERT_IS_TREE): New macro.
8543 (ASSERT_IS_GIMPLE): New macro.
8544 (selftest::test_capture_of_dump_calls): New test.
8545 (selftest::dumpfile_c_tests): Call it.
8546 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
8547 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
8548 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
8549 descriptive comment.
8550 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
8551 (dump_node, dump_bb): Move these unrelated decls.
8552 (class dump_manager): Add leading comment.
8553 * optinfo.cc: New file.
8554 * optinfo.h: New file.
8556 2018-07-19 Michael Collison <michael.collison@arm.com>
8557 Richard Henderson <rth@redhat.com>
8559 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
8560 (subti): Handle op1 zero.
8561 (subvti4, usub4ti4): New.
8562 (*sub<GPI>3_compare1_imm): New.
8563 (sub<GPI>3_carryinCV): New.
8564 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
8565 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
8567 2018-07-19 Michael Collison <michael.collison@arm.com>
8568 Richard Henderson <rth@redhat.com>
8570 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
8571 (addti3): Create simpler code if low part is already known to be 0.
8572 (addvti4, uaddvti4): New.
8573 (*add<GPI>3_compareC_cconly_imm): New.
8574 (*add<GPI>3_compareC_cconly): New.
8575 (*add<GPI>3_compareC_imm): New.
8576 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
8577 handle constants within this pattern..
8578 (*add<GPI>3_compareV_cconly_imm): New.
8579 (*add<GPI>3_compareV_cconly): New.
8580 (*add<GPI>3_compareV_imm): New.
8581 (add<GPI>3_compareV): New.
8582 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
8583 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
8584 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
8585 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
8586 with 'comparison' operator.
8587 (*add<GPI>3_compareV_cconly_imm): Ditto.
8588 (*add<GPI>3_compareV_cconly): Ditto.
8589 (*add<GPI>3_compareV_imm): Ditto.
8590 (add<GPI>3_compareV): Ditto.
8591 (add<mode>3_carryinC): Ditto.
8592 (*add<mode>3_carryinC_zero): Ditto.
8593 (*add<mode>3_carryinC): Ditto.
8594 (add<mode>3_carryinV): Ditto.
8595 (*add<mode>3_carryinV_zero): Ditto.
8596 (*add<mode>3_carryinV): Ditto.
8598 2018-07-19 Michael Collison <michael.collison@arm.com>
8599 Richard Henderson <rth@redhat.com>
8601 * config/aarch64/aarch64-modes.def (CC_V): New.
8602 * config/aarch64/aarch64-protos.h
8603 (aarch64_addti_scratch_regs): Declare
8604 (aarch64_subvti_scratch_regs): Declare.
8605 (aarch64_expand_subvti): Declare.
8606 (aarch64_gen_unlikely_cbranch): Declare
8607 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
8608 for signed overflow using CC_Vmode.
8609 (aarch64_get_condition_code_1): Handle CC_Vmode.
8610 (aarch64_gen_unlikely_cbranch): New function.
8611 (aarch64_addti_scratch_regs): New function.
8612 (aarch64_subvti_scratch_regs): New function.
8613 (aarch64_expand_subvti): New function.
8615 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
8617 * config/aarch64/aarch64-option-extensions.def: New entry for profile
8619 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
8620 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
8623 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
8626 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8627 address check not strict.
8629 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
8631 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
8633 (store_pair_lanes<mode>): Likewise.
8634 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
8635 enum value 'ADDR_QUERY_LDP_STP_N'.
8636 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
8637 (aarch64_print_address_internal): Add declaration.
8638 (aarch64_print_ldpstp_address): Remove.
8639 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
8640 (aarch64_print_operand): Change printing of 'y'.
8641 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
8642 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
8643 'true' rather than '1'.
8644 * config/aarch64/constraints.md (Uml): Likewise.
8645 (Uml): Rename to Umn.
8648 2018-07-19 Richard Biener <rguenther@suse.de>
8650 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
8652 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
8653 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
8654 (vn_reference_hasher): Likewise.
8655 (struct vn_tables_s): Remove obstack and alloc-pool members.
8656 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
8657 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
8658 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
8659 (vn_reference_insert_pieces): Likewise.
8660 (alloc_vn_nary_op_noinit): Adjust.
8661 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
8662 (vn_phi_eq): Adjust.
8663 (shared_lookup_phiargs): Remove.
8664 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
8665 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
8666 (visit_reference_op_call): Likewise.
8667 (copy_nary, copy_phi, copy_reference): Remove.
8668 (process_scc): Rewind the obstack when iterating. Do not
8669 copy the elements to valid_info but just move them from one
8670 hashtable to the other.
8671 (allocate_vn_table): Adjust.
8672 (free_vn_table): Likewise.
8673 (init_scc_vn): Likewise.
8674 (free_scc_vn): Likewise.
8676 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
8679 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
8680 indirect_return as function type attribute.
8681 (ix86_attribute_table): Change indirect_return to function
8683 * doc/extend.texi: Update indirect_return attribute.
8685 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
8687 * wide-int.h (widest2_int): New.
8688 * gimple-fold.c (arith_overflowed_p): Use it.
8689 * tree.h (widest2_int_cst): New.
8690 * tree-vrp.c (wide_int_binop_overflow): Rename from
8691 vrp_int_const_binop.
8692 Rewrite to work on trees.
8693 (extract_range_from_multiplicative_op_1): Abstract code to...
8694 (wide_int_range_min_max): ...here.
8695 (wide_int_range_cross_product): ...and here.
8696 (extract_range_from_binary_expr_1): Abstract overflow code to...
8697 (wide_int_range_mult_wrapping): ...here.
8698 * tree-vrp.h (wide_int_range_cross_product): New.
8699 (wide_int_range_mult_wrapping): New.
8701 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
8702 Julia Koval <julia.koval@intel.com>
8704 * config/i386/x86-tune-costs.h (skylake_memcpy,
8705 skylake_memset): Replace rep_prefix with unrolling for size 512.
8707 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
8710 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
8711 comparision with EQ_EXPR in last stmt.
8713 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
8715 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
8716 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
8717 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
8718 previously known as "PowerPC AltiVec Built-in Functions". Move
8719 some material to new subsubsections "PowerPC AltiVec Built-in
8720 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
8722 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
8723 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
8724 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
8725 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
8727 2018-07-18 Richard Biener <rguenther@suse.de>
8729 PR tree-optimization/86557
8730 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
8733 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
8735 * config/s390/s390.c (s390_function_profiler): Generate CFI.
8737 2018-07-17 Jeff Law <law@redhat.com>
8739 * config/arm/arm.c (get_label_padding): Update for recent
8740 changes to label_to_alignment.
8742 PR tree-optimization/86010
8743 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
8745 * config/mips/mips.c (vr4130_align_insns): Update for recent
8746 changes to label_to_alignment.
8748 * config/frv/frv.c (frv_label_align): Update for recent changes
8749 to label_to_alignment.
8751 * config/nios2/nios2.c (nios2_label_align): Update for recent
8752 changes which dropped ALIGN_LABELS_LOG.
8754 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
8756 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8757 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8759 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
8761 * config/arc/arc.c (arc_label_align): Use align_labels instead of
8762 deprecated align_labels_log.
8764 2018-07-17 Richard Biener <rguenther@suse.de>
8767 * dwarf2out.c (init_sections_and_labels): Always generate
8768 a debug_line_str_section for early LTO debug.
8769 (dwarf2out_finish): Reset debug_line_str_hash output early.
8770 Bump counter for extra dwarf5 .debug_loc labels to not conflict
8772 (dwarf2out_early_finish): Output debug_line_str.
8774 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8776 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8777 index register on z196 or later.
8779 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8781 * config/s390/s390.c (s390_default_align): Set default function
8783 (s390_override_options_after_change): Call s390_default align.
8784 (s390_option_override_internal): Call s390_default_align.
8785 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8787 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8790 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8791 to _looptemp_ clauses, other than the first two.
8793 2018-07-17 Martin Liska <mliska@suse.cz>
8795 * opts.c: Do not enable OPT_falign_* for -Os.
8797 2018-07-17 Martin Liska <mliska@suse.cz>
8799 * align.h (MAX_CODE_ALIGN): New.
8800 (MAX_CODE_ALIGN_VALUE): New.
8801 * common/config/i386/i386-common.c (ix86_handle_option):
8802 (MAX_CODE_ALIGN): Moved to align.h.
8803 * final.c (MAX_CODE_ALIGN): Likewise.
8804 * opts.c (parse_and_check_align_values):
8805 (MAX_CODE_ALIGN): Likewise.
8806 (MAX_CODE_ALIGN_VALUE): Likewise.
8808 2018-07-17 Martin Liska <mliska@suse.cz>
8810 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8811 in order to fulfil coding style.
8812 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8813 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8814 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8815 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8816 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8817 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8818 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8819 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8820 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8822 2018-07-17 Martin Liska <mliska@suse.cz>
8824 * align.h: New file.
8825 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8827 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8828 align_flags of label_to_alignment.
8829 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8831 * config/m68k/m68k.c: Do not use removed align_labels_value and
8833 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
8834 (LOOP_ALIGN): Likewise.
8835 (LABEL_ALIGN): Likewise.
8836 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8837 Remove not used macro.
8838 (rs6000_loop_align): Change return type to align_flags.
8839 (rs6000_loop_align_max_skip): Remove.
8840 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
8841 Change return type to align_flags.
8842 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8843 Remove not used macro.
8844 (rs6000_loop_align): Change return type to align_flags.
8845 (rs6000_loop_align_max_skip): Remove.
8846 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
8847 * config/rx/rx-protos.h (rx_align_for_label): Make it
8849 * config/rx/rx.c (rx_align_for_label): Change return type
8851 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8853 into align_flags class.
8854 (LABEL_ALIGN): Likewise.
8855 (LOOP_ALIGN): Likewise.
8856 * config/s390/s390.c (s390_label_align): Use align_flags
8858 (s390_asm_output_function_label): Likewise.
8859 * config/sh/sh.c (sh_override_options_after_change):
8860 Use align_flags class directly without macros.
8861 (find_barrier): Likewise.
8862 (barrier_align): Likewise.
8863 (sh_loop_align): Likewise.
8864 * config/spu/spu.c (spu_option_override):
8865 Use align_flags_tuple::get_value instead of removed macros.
8866 (spu_sched_init): Likewise.
8867 * config/spu/spu.h (GTY): Likewise.
8868 * config/visium/visium.c (visium_option_override):
8869 Set "8" as default secondary alignment.
8870 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8871 in order to guarantee secondary alignment of 8.
8872 * coretypes.h: Include align.h header file.
8873 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8874 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8875 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8876 * doc/tm.texi.in: Likewise.
8877 * final.c (struct label_alignment): Remove not used structure.
8878 (LABEL_ALIGN): Change type to align_flags.
8879 (LOOP_ALIGN): Likewise.
8880 (JUMP_ALIGN): Likewise.
8881 (default_loop_align_max_skip): Remove.
8882 (default_label_align_max_skip): Likewise.
8883 (default_jump_align_max_skip): Likewise.
8884 (default_label_align_after_barrier_max_skip):
8885 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8886 (LABEL_TO_MAX_SKIP): Remove.
8887 (label_to_alignment): Return align_flags type instead of integer.
8888 (label_to_max_skip): Remove.
8889 (align_fuzz): Use align_flags type.
8890 (compute_alignments): Use align_flags type and use align_flags::max
8891 to combine multiple alignments.
8892 (grow_label_align): Grow vec instead of C array.
8893 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8894 (shorten_branches): Use align_flags type and use align_flags::max
8895 to combine multiple alignments.
8896 (final_scan_insn_1): Remove usage of secondary alignment that comes
8897 from label alignment, but instead use proper secondary alignment
8898 which is computed in grow_label_align.
8899 * flags.h (struct align_flags_tuple): Move to align.h.
8900 (struct align_flags): Likewise.
8901 (state_align_loops): Rename to align_loops.
8902 (state_align_jumps): Rename to align_jumps.
8903 (state_align_labels): Rename to align_labels.
8904 (state_align_functions): Rename to align_functions.
8905 (align_loops_log): Remove.
8906 (align_jumps_log): Remove.
8907 (align_labels_log): Remove.
8908 (align_functions_log): Remove.
8909 (align_loops_max_skip): Remove.
8910 (align_jumps_max_skip): Remove.
8911 (align_labels_max_skip): Remove.
8912 (align_functions_max_skip): Remove.
8913 (align_loops_value): Remove.
8914 (align_jumps_value): Remove.
8915 (align_labels_value): Remove.
8916 (align_functions_value): Remove.
8917 * output.h (label_to_alignment): Change return type to align_flags.
8918 (label_to_max_skip): Remove.
8919 * target.def: Remove loop_align_max_skip, label_align_max_skip,
8920 jump_align_max_skip macros.
8921 * targhooks.h (default_loop_align_max_skip): Remove.
8922 (default_label_align_max_skip): Likewise.
8923 (default_jump_align_max_skip): Likewise.
8924 (default_label_align_after_barrier_max_skip): Remove.
8925 * toplev.c (read_log_maxskip): Use ::normalize function.
8926 (parse_N_M): Remove not used argument and also call ::normalize.
8927 (parse_alignment_opts): Do not pass unused arguments.
8928 * varasm.c (assemble_start_function): Use directly align_functions
8929 instead of removed macros.
8930 * system.h: Do not poison removed macros.
8932 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8935 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8936 and cond temporaries don't have reference type if iterator has
8937 pointer type. For init use &for_pre_body instead of pre_p if
8938 for_pre_body is non-empty.
8940 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8942 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8943 double-double modes to SFmode directly directly.
8944 (trunc<mode>sf2_fprs): Delete.
8946 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8948 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8949 for conversions between IFmode and the decimal floating point modes.
8950 (init_float128_ieee): Use the correct names for conversions between
8951 KFmode and the decimal floating point modes.
8953 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8955 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8956 for the conversions between TDmode and IFmode.
8957 (init_float128_ieee): Use more correct names for the conversions
8958 between TDmode and KFmode.
8960 2018-07-16 Jakub Jelinek <jakub@redhat.com>
8962 PR tree-optimization/86526
8963 * builtins.c (expand_builtin_memcmp): Formatting fixes.
8964 (inline_expand_builtin_string_cmp): Likewise.
8965 (inline_string_cmp): Likewise. Use c_readstr instead of
8966 builtin_memcpy_read_str. Add unit_mode temporary.
8968 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
8971 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8972 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8974 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8976 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8977 Alphabetize prototypes of built-in functions, separating out
8978 built-in functions that are listed in this section but should be
8979 described elsewhere.
8981 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
8984 * expmed.c (emit_store_flag): Do not emit setcc followed by a
8985 conditional move when trapping comparison was split to a
8986 non-trapping one (and vice versa).
8988 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8990 * config/s390/s390.c (s390_function_profiler): Generate nops
8991 instead of profiler call sequences.
8992 * config/s390/s390.opt: Add the new option.
8994 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8996 * config/s390/s390.c (s390_function_profiler): Generate
8997 __mcount_loc section.
8998 * config/s390/s390.opt: Add the new option.
9000 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9002 * common.opt: Add the new warning.
9003 * config/s390/s390.c (s390_function_profiler): Emit "brasl
9004 %r0,__fentry__" when -mfentry is specified.
9005 (s390_option_override_internal): Disallow -mfentry for 31-bit
9007 * config/s390/s390.opt: Add the new option.
9009 2018-07-16 Richard Biener <rguenther@suse.de>
9012 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9013 for function-local FUNCTION_DECL and RESULT_DECL immediately.
9015 2018-07-16 Martin Liska <mliska@suse.cz>
9018 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9021 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
9023 * config/arc/arcHS.md: Update ARCHS scheduling rules.
9025 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
9027 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9029 * config/arc/arc-cpus.def (hs4x): New cpu.
9031 * config/arc/arc-tables.opt: Regenerate.
9032 * config/arc/arc.c (arc_sched_issue_rate): New function.
9033 (TARGET_SCHED_ISSUE_RATE): Define.
9034 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9035 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9037 (attr tune): Add ARCHS4x tune values.
9038 (attr tune_dspmpy): Define.
9039 (*tst): Correct instruction type.
9040 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9041 * config/arc/arcHS4x.md: New file.
9042 * config/arc/fpu.md: Update instruction type attributes.
9043 * config/arc/t-multilib: Regenerate.
9045 2018-07-16 Tom de Vries <tdevries@suse.de>
9048 * var-tracking.c (vt_initialize): Fix pre_dec handling.
9050 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9052 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
9055 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
9057 PR tree-optimization/86514
9058 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
9059 conversion to a boolean type from a type with greater precision.
9061 2018-07-16 Tom de Vries <tdevries@suse.de>
9063 * var-tracking.c (vt_initialize): Print adjusted insn slim if
9064 dump_flags request TDF_SLIM.
9066 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
9068 * fold-const.c (int_const_binop_1): Abstract...
9069 (wide_int_binop): ...wide int code here.
9070 (poly_int_binop): ...poly int code here.
9071 Abstract the rest of int_const_binop_1 into int_const_binop.
9072 * fold-const.h (wide_int_binop): New.
9073 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
9074 Remove useless PLUS/MINUS_EXPR case.
9075 (zero_nonzero_bits_from_vr): Move wide int code...
9076 (zero_nonzero_bits_from_bounds): ...here.
9077 (extract_range_from_binary_expr_1): Move mask optimization code...
9078 (range_easy_mask_min_max): ...here.
9079 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
9080 (range_easy_mask_min_max): New.
9082 2018-07-15 Jeff Law <law@redhat.com>
9085 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
9088 2018-07-14 Jim Wilson <jimw@sifive.com>
9090 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
9092 2018-07-14 Paul Koning <ni1d@arrl.net>
9094 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
9096 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
9098 * lto-streamer-out.c (copy_function_or_variable): Dump info about
9101 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
9102 Steve Munroe <munroesj52@gmail.com>
9104 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
9105 (_mm_andnot_si128): Likewise.
9106 (_mm_or_si128): Likewise.
9107 (_mm_xor_si128): Likewise.
9109 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
9112 * builtins.c (expand_builtin_memcmp): Inline the calls first
9113 when result_eq is false.
9114 (expand_builtin_strcmp): Inline the calls first.
9115 (expand_builtin_strncmp): Likewise.
9116 (inline_string_cmp): New routine. Expand a string compare
9117 call by using a sequence of char comparison.
9118 (inline_expand_builtin_string_cmp): New routine. Inline expansion
9119 a call to str(n)cmp/memcmp.
9120 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
9122 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
9124 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
9126 * config/arm/driver-arm.c: Include arm-native.h.
9127 (host_detect_local_cpu): Use auto-generated data tables.
9128 (vendors, arm_cpu_table): Delete. Move part information to ...
9129 * config/arm/arm-cpus.in: ... here.
9130 * config/arm/parsecpu.awk (gen_native): New function.
9131 (vendor, part): New CPU fields.
9132 (END): Add support for building the native CPU detection tables.
9133 * config/arm/t-arm (arm-native.h): Add build rule.
9134 (driver-arm.o): Add dependency on arm-native.h.
9136 2018-07-13 Richard Biener <rguenther@suse.de>
9139 * match.pd (addr1 - addr2): Allow either of the operand to
9142 2018-07-13 Tom de Vries <tdevries@suse.de>
9144 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
9147 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
9149 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
9150 arrays instead of numbered variables.
9152 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
9154 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
9155 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
9157 2018-07-13 Richard Biener <rguenther@suse.de>
9160 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
9161 instead of get_context_die.
9163 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
9164 Richard Biener <rguenther@suse.de>
9167 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
9168 that the loop latch destination where phi is defined.
9170 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
9172 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
9173 (riscv_expand_epilogue): Add assertion to check interrupt mode.
9174 (riscv_set_current_function): Extract getting interrupt type to new
9176 (riscv_get_interrupt_type): New function.
9177 (riscv_merge_decl_attributes): New function, checking interrupt type is
9179 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
9181 2018-07-12 Paul Koning <ni1d@arrl.net>
9183 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
9186 2018-07-12 Paul Koning <ni1d@arrl.net>
9188 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
9189 zero reference, add doloop_end instead.
9190 * doc/md.texi (decrement_and_branch_until_zero): Remove.
9191 (Looping patterns): Remove decrement_and_branch_until_zero. Add
9192 detail for doloop_end.
9194 2018-07-12 Martin Sebor <msebor@redhat.com>
9197 * attribs.c (decl_attributes): Reject conflicting attributes before
9198 calling attribute handlers.
9200 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
9202 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
9204 (gcc::dump_manager::get_dump_file_name): likewise.
9205 (dump_begin): Likewise.
9206 * dumpfile.h (dump_begin): Update prototype.
9207 (gcc::dump_manager::get_dump_file_name,
9208 gcc::dump_manager::get_dump_file_name): Update prototype.
9210 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9212 * internal-fn.h (vectorizable_internal_fn_p): New function.
9213 * tree-vect-slp.c (compatible_calls_p): Likewise.
9214 (vect_build_slp_tree_1): Remove nops argument. Handle calls
9215 to internal functions.
9216 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
9218 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9220 * fold-const.h (inverse_conditions_p): Declare.
9221 * fold-const.c (inverse_conditions_p): New function.
9222 * match.pd: Use inverse_conditions_p. Add folds of view_converts
9223 that test the inverse condition of a conditional internal function.
9224 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
9225 * internal-fn.c (internal_fn_mask_index): Handle conditional
9227 (vectorized_internal_fn_supported_p): New function.
9228 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
9229 (any_pred_load_store): Replace with...
9230 (need_to_predicate): ...this new variable.
9231 (redundant_ssa_names): New variable.
9232 (ifcvt_can_use_mask_load_store): Move initial checks to...
9233 (ifcvt_can_predicate): ...this new function. Handle tree codes
9234 for which a conditional internal function exists.
9235 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
9236 instead of ifcvt_can_use_mask_load_store. Update after variable
9238 (predicate_load_or_store): New function, split out from
9239 predicate_mem_writes.
9240 (check_redundant_cond_expr): New function.
9241 (value_available_p): Likewise.
9242 (predicate_rhs_code): Likewise.
9243 (predicate_mem_writes): Rename to...
9244 (predicate_statements): ...this. Use predicate_load_or_store
9245 and predicate_rhs_code.
9246 (combine_blocks, tree_if_conversion): Update after above name changes.
9247 (ifcvt_local_dce): Handle redundant_ssa_names.
9248 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
9249 general conditional functions.
9250 * tree-vect-stmts.c (vectorizable_call): Likewise.
9252 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9253 Alan Hayward <alan.hayward@arm.com>
9254 David Sherwood <david.sherwood@arm.com>
9256 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
9257 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
9258 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
9259 plus and minus and convert them into IFN_COND_FMA-based sequences.
9260 (convert_mult_to_fma): Handle conditional plus and minus.
9262 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9264 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
9265 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
9266 (cond_fnms_optab): New optabs.
9267 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
9269 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
9270 * internal-fn.h (get_conditional_internal_fn): Declare.
9271 (get_unconditional_internal_fn): Likewise.
9272 * internal-fn.c (cond_ternary_direct): New macro.
9273 (expand_cond_ternary_optab_fn): Likewise.
9274 (direct_cond_ternary_optab_supported_p): Likewise.
9275 (FOR_EACH_COND_FN_PAIR): Likewise.
9276 (get_conditional_internal_fn): New function.
9277 (get_unconditional_internal_fn): Likewise.
9278 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
9279 (gimple_match_op::gimple_match_op): Add a new overload for 5
9281 (gimple_match_op::set_op): Likewise.
9282 (gimple_resimplify5): Declare.
9283 * genmatch.c (decision_tree::gen): Generate simplifications for
9285 * gimple-match-head.c (gimple_simplify): Define an overload for
9286 5 operands. Handle calls with 5 arguments in the top-level overload.
9287 (convert_conditional_op): Handle conversions from unconditional
9288 internal functions to conditional ones.
9289 (gimple_resimplify5): New function.
9290 (build_call_internal): Pass a fifth operand.
9291 (maybe_push_res_to_seq): Likewise.
9292 (try_conditional_simplification): Try converting conditional
9293 internal functions to unconditional internal functions.
9294 Handle 3-operand unconditional forms.
9295 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
9296 Define ternary equivalents of the current rules for binary conditional
9298 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
9300 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
9301 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
9302 (optab): Handle them.
9303 (SVE_COND_FP_TERNARY): New int iterator.
9304 (sve_fmla_op, sve_fmad_op): New int attributes.
9305 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
9306 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
9307 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
9309 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
9311 * target.def (preferred_else_value): New target hook.
9312 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
9313 * doc/tm.texi: Regenerate.
9314 * targhooks.h (default_preferred_else_value): Declare.
9315 * targhooks.c (default_preferred_else_value): New function.
9316 * internal-fn.h (conditional_internal_fn_code): Declare.
9317 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
9318 (get_conditional_internal_fn): Use it.
9319 (conditional_internal_fn_code): New function.
9320 * gimple-match.h (gimple_match_cond): New struct.
9321 (gimple_match_op): Add a cond member function.
9322 (gimple_match_op::gimple_match_op): Update all forms to take a
9324 * genmatch.c (expr::gen_transform): Use the same condition as res_op
9325 for the suboperation, but don't specify a particular else_value.
9326 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
9327 (visit_nary_op, visit_reference_op_load): Pass
9328 gimple_match_cond::UNCOND to the gimple_match_op constructor.
9329 * gimple-match-head.c: Include tree-eh.h
9330 (convert_conditional_op): New function.
9331 (maybe_resimplify_conditional_op): Likewise.
9332 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
9333 (gimple_resimplify2): Likewise.
9334 (gimple_resimplify3): Likewise.
9335 (gimple_resimplify4): Likewise.
9336 (maybe_push_res_to_seq): Return null for conditional operations.
9337 (try_conditional_simplification): New function.
9338 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
9340 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
9342 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
9344 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
9346 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
9348 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9350 (hash_tree): Do not hash DECL_FCONTEXT
9351 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
9352 Do not stream DECL_FCONTEXT.
9353 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
9354 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
9356 2018-07-12 Richard Biener <rguenther@suse.de>
9359 * dwarf2out.c (gen_block_die): Only output blocks when they have
9360 at least one !DECL_IGNORED_P variable.
9362 2018-07-12 Richard Biener <rguenther@suse.de>
9365 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
9366 Remove -mieee-fp handling.
9368 2018-07-12 Richard Biener <rguenther@suse.de>
9370 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
9371 left-over from last patch.
9373 2018-07-12 Jakub Jelinek <jakub@redhat.com>
9375 PR tree-optimization/86492
9376 * gimple-ssa-store-merging.c
9377 (imm_store_chain_info::coalesce_immediate_stores): Call
9378 check_no_overlap even for the merge_overlapping case. Formatting fix.
9380 2018-07-12 Richard Biener <rguenther@suse.de>
9383 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
9384 move possibly trapping operations into the conditional.
9386 2018-07-12 Richard Biener <rguenther@suse.de>
9388 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
9389 (vn_lookup_simplify_result): Remove recursion limit applied
9391 (vn_nary_build_or_lookup_1): Adjust.
9392 (try_to_simplify): Likewise.
9393 * gimple-match-head.c (gimple_resimplify1): Instead apply one
9395 (gimple_resimplify2): Likewise.
9396 (gimple_resimplify3): Likewise.
9397 (gimple_resimplify4): Likewise.
9399 2018-07-11 Jakub Jelinek <jakub@redhat.com>
9401 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
9402 Use __mmask64 type instead of __mmask8 for __M argument.
9403 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
9404 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
9406 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
9407 __mmask16 for __M argument.
9408 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
9409 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
9410 to __mmask16 instead of __mmask8.
9411 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
9412 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
9413 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
9414 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
9415 instead of __mmask16 for __U argument.
9416 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
9417 __mmask16 instead of __mmask8 for __U argument.
9418 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
9420 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
9422 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
9424 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
9426 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
9428 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
9430 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
9432 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
9434 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
9436 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
9438 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
9439 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
9440 return type as well as __M argument type and all casts from __mmask8
9442 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
9443 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
9444 return type as well as __M argument type and all casts from __mmask8
9446 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
9447 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
9448 return type as well as __M argument type and all casts from __mmask8
9450 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
9451 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
9452 return type as well as __M argument type and all casts from __mmask8
9454 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
9455 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
9458 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
9460 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
9461 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
9464 2018-07-11 Paul Koning <ni1d@arrl.net>
9466 * doc/md.texi (define_subst): Document how multiple occurrences of
9467 the same argument in the replacement pattern are handled.
9469 2018-07-11 Paul Koning <ni1d@arrl.net>
9471 * doc/extend.texi (Common Variable Attributes): Move "mode" into
9473 (Common Type Attributes): Add "mode" attribute.
9475 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
9477 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
9478 stream DECL_ORIGINAL_TYPE.
9479 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
9480 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
9481 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
9482 Do not walk original type.
9483 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
9485 (write_ts_decl_non_common_tree_pointers): Do not stream
9487 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
9488 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
9490 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
9492 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
9493 thread twice from the same starting edge.
9495 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
9497 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
9498 * gimple.c (gimple_call_nonnull_result_p): ...here...
9499 (gimple_call_nonnull_arg): ...and here.
9500 * gimple.h (gimple_call_nonnull_result_p): New.
9501 (gimple_call_nonnull_arg): New.
9503 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
9505 * config/arm/arm-cpus.in: Move information from fpu field of each
9506 cpu definition to the isa field.
9507 * config/arm/parsecpu.awk (fpu): Delete match rule.
9508 (gen_comm_data): Don't add bits from the CPU's FPU entry.
9510 2018-07-11 Richard Biener <rguenther@suse.de>
9513 * dwarf2out.c (init_sections_and_labels): Use
9514 output_asm_line_debug_info consistently.
9515 (dwarf2out_early_finish): Likewise.
9516 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
9519 2018-07-11 Richard Biener <rguenther@suse.de>
9521 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
9522 Rework father_bb setting in a way to avoid propagating constants
9523 multiple times on a loop body.
9525 2018-07-10 Mark Wielaard <mark@klomp.org>
9528 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
9530 2018-07-10 Richard Biener <rguenther@suse.de>
9532 * hash-map.h (hash_map::iterator::operator*): Return
9533 references to key and value.
9535 2018-07-10 Jakub Jelinek <jakub@redhat.com>
9538 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
9539 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
9540 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
9541 (gimplify_omp_for): For composite loops, move outer
9542 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
9543 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
9544 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
9545 TREE_LIST for both the original class iterator and the "last" helper
9546 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
9547 loop, remember has_decl_expr from outer composite loops for the
9548 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
9550 2018-07-09 Martin Sebor <msebor@redhat.com>
9554 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
9555 accessing implicitly initialized array elements.
9556 * expr.c (string_constant): Handle string initializers of
9557 character arrays within aggregates.
9558 * gimple-fold.c (fold_array_ctor_reference): Add argument.
9559 Store element offset. As a special case, handle zero size.
9560 (fold_nonarray_ctor_reference): Same.
9561 (fold_ctor_reference): Add argument. Store subobject offset.
9562 * gimple-fold.h (fold_ctor_reference): Add argument.
9564 2018-07-09 Paul Koning <ni1d@arrl.net>
9566 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
9567 (pdp11_insn_cost): New function.
9568 (pdp11_md_asm_adjust): New function.
9569 (TARGET_INVALID_WITHIN_DOLOOP): Define.
9570 (pdp11_rtx_costs): Update to match machine better.
9571 (output_addr_const_pdp11): Correct format mismatch warnings.
9572 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
9573 * config/pdp11/pdp11.md: General change to add base_cost and/or
9574 length attributes for use by new pdp11_insn_cost function.
9575 (MIN_BRANCH): Correct definition.
9577 (doloop_end): Use standard pattern name for looping pattern.
9578 (doloop_end_nocc): New.
9579 (movsf): Add another constraint alternative.
9580 (zero_extendqihi2): Add constraint alternatives for not in place
9582 (zero_extendhisi2): Remove.
9583 (shift patterns): Add CC handling variants.
9587 (define_peephole2): New peephole to recognize mov that sets CC for
9590 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9592 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
9593 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
9596 2018-07-09 Jakub Jelinek <jakub@redhat.com>
9599 * real.c (real_nextafter): Return true if result is denormal.
9601 2018-07-09 Martin Liska <mliska@suse.cz>
9603 * common.opt: Add back wrongly removed attribute.
9605 2018-07-09 Richard Biener <rguenther@suse.de>
9608 * dwarf2out.c (gen_block_die): For an early generated DIE
9609 always output high/low PC attributes.
9611 2018-07-09 Tom de Vries <tdevries@suse.de>
9613 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
9614 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
9615 onto VAR_DECL with abstract origin.
9617 2018-07-07 Jim Wilson <jimw@sifive.com>
9619 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
9621 2018-07-07 Tom de Vries <tdevries@suse.de>
9623 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
9625 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
9627 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
9629 (combine_bound): Use wide-int overflow calculation instead of
9631 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
9633 * fold-const.c (int_const_binop_2): Same.
9634 (extract_muldiv_1): Same.
9635 (fold_div_compare): Same.
9636 (fold_abs_const): Same.
9638 * poly-int.h (add): Same.
9642 * predict.c (predict_iv_comparison): Same.
9643 * profile-count.c (slow_safe_scale_64bit): Same.
9644 * simplify-rtx.c (simplify_const_binary_operation): Same.
9645 * tree-chrec.c (tree_fold_binomial): Same.
9646 * tree-data-ref.c (split_constant_offset_1): Same.
9647 * tree-if-conv.c (idx_within_array_bound): Same.
9648 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
9649 * tree-ssa-phiopt.c (minmax_replacement): Same.
9650 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
9651 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
9652 * vr-values.c (vr_values::adjust_range_with_scev): Same.
9653 * wide-int.cc (wi::add_large): Same.
9654 (wi::mul_internal): Same.
9655 (wi::sub_large): Same.
9656 (wi::divmod_internal): Same.
9657 * wide-int.h: Change overflow type to overflow_type for neg, add,
9658 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
9659 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
9660 mul_internal, divmod_internal.
9661 (overflow_type): New enum.
9662 (accumulate_overflow): New.
9664 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
9666 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
9667 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
9669 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
9671 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
9672 argument is checked for zero before entering loop, avoid checking again.
9674 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
9676 * gimplify.h (generic_expr_could_trap_p): Set as global function.
9677 * gimplify.h (generic_expr_could_trap_p): Likwise.
9678 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
9680 2018-07-06 Jakub Jelinek <jakub@redhat.com>
9682 PR tree-optimization/86401
9683 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
9684 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
9685 (fold_bit_and_mask): ... here. New helper function for match.pd.
9686 * fold-const.h (fold_bit_and_mask): Declare.
9687 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
9689 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
9692 * target.def (translate_mode_attribute): New hook.
9693 * targhooks.h (default_translate_mode_attribute): Declare.
9694 * targhooks.c (default_translate_mode_attribute): New function.
9695 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
9696 * doc/tm.texi: Regenerate.
9697 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
9698 (rs6000_translate_mode_attribute): New function.
9700 2018-07-06 Paul Koning <ni1d@arrl.net>
9702 * doc/md.texi (define_split): Document DONE and FAIL.
9703 (define_peephole2): Ditto.
9705 2018-07-05 Jeff Law <law@redhat.com>
9707 PR tree-optimization/86010
9708 * tree-ssa-dse.c (compute_trims): More aggressively trim at
9709 both the head and tail of mem* and str* calls.
9711 2018-07-05 Jim Wilson <jimw@sifive.com>
9713 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
9715 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
9717 * config/aarch64/aarch64-simd.md: correct flags text for
9718 MIN_EXPR replacement.
9720 2018-07-05 James Clarke <jrtc27@jrtc27.com>
9722 * configure: Regenerated.
9724 2018-07-05 Carl Love <cel@us.ibm.com>
9726 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
9727 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9728 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
9729 VSX_BUILTIN_DOUBLEL_V4SF.
9731 2018-07-05 Martin Sebor <msebor@redhat.com>
9734 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
9735 than its domain to compute its the upper bound of a char array.
9737 2018-07-05 Nathan Sidwell <nathan@acm.org>
9739 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
9740 * doc/cpp.texi: Update comment.
9741 * doc/tm.texi: Rebuilt.
9742 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
9743 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
9744 * doc/extend.texi (Backwards Compatibility): Clarify it is system
9745 headers affected by extern "C".
9746 * system.h: Poison NO_IMPLICIT_EXTERN_C.
9747 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9748 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9749 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9750 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9751 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9752 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9753 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9754 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9755 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9756 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9757 config/sparc/sp64-elf.h, config/spu/spu.h,
9758 config/stormy16/stormy16.h, config/v850/v850.h,
9759 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9760 define NO_IMPLICIT_EXTERN_C.
9761 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9763 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9766 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9767 instead of GET_MODE_SIZE when comparing Units.
9769 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9772 * rtlanal.c (set_noop_p): Constrain on mode change,
9773 include hard-reg-set.h
9775 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9777 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9779 2018-07-05 Jakub Jelinek <jakub@redhat.com>
9782 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9785 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9787 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9789 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9792 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9794 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9796 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
9799 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9801 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
9803 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9804 neg_*_op* variables.
9806 2018-07-04 Martin Liska <mliska@suse.cz>
9808 * tree-switch-conversion.c: Define
9809 max_ratio_for_speed and max_ratio_for_size constants.
9811 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9812 Martin Liska <mliska@suse.cz>
9817 * common.opt: Rename align options with 'str_' prefix.
9818 * common/config/i386/i386-common.c (set_malign_value): New
9820 (ix86_handle_option): Use it to set -falign-* options/
9821 * config/aarch64/aarch64-protos.h (struct tune_params): Change
9822 type from int to string.
9823 * config/aarch64/aarch64.c: Update default values from int
9825 * config/alpha/alpha.c (alpha_override_options_after_change):
9827 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9828 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9829 max skip conditionally.
9830 * config/i386/freebsd.h (SUBALIGN_LOG): New.
9831 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9832 max skip conditionally.
9833 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9834 max skip conditionally.
9835 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
9836 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9837 max skip conditionally.
9838 * config/i386/i386.c (struct ptt): Change type from int to
9840 (ix86_default_align): Set default values.
9841 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
9842 max skip conditionally.
9843 * config/i386/iamcu.h (SUBALIGN_LOG): New.
9844 (ASM_OUTPUT_MAX_SKIP_ALIGN):
9845 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
9846 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9847 max skip conditionally.
9848 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
9849 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
9850 * config/i386/x86-64.h (SUBALIGN_LOG): New.
9851 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9852 max skip conditionally.
9853 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9854 * config/ia64/ia64.c (ia64_option_override): Set default values
9855 for alignment options.
9856 * config/m68k/m68k.c: Handle new str_align_* options.
9857 * config/mips/mips.c (mips_set_compression_mode): Change
9859 (mips_option_override): Set default values for options.
9860 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9862 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9864 * config/rx/rx.c (rx_option_override): Likewise.
9865 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
9866 (LABEL_ALIGN): Use align_labels_log.
9867 (LOOP_ALIGN): Use align_loops_align.
9868 * config/s390/s390.c (s390_asm_output_function_label): Use new
9870 * config/sh/sh.c (sh_override_options_after_change):
9871 Change type of constants.
9872 * config/spu/spu.c (spu_sched_init): Likewise.
9873 * config/sparc/sparc.c (sparc_option_override): Set default
9875 * config/visium/visium.c (visium_option_override): Likewise.
9876 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9877 emit p2align format with last argument if it's not needed.
9878 * doc/invoke.texi: Document extended format of -falign-*.
9879 * final.c: Use align_labels alignment.
9880 * flags.h (struct target_flag_state): Change type to use
9882 (struct align_flags_tuple): New.
9883 (struct align_flags): Likewise.
9884 (align_loops_log): Redefine macro to use new types.
9885 (align_loops_max_skip): Redefine macro to use new types.
9886 (align_jumps_log): Redefine macro to use new types.
9887 (align_jumps_max_skip): Redefine macro to use new types.
9888 (align_labels_log): Redefine macro to use new types.
9889 (align_labels_max_skip): Redefine macro to use new types.
9890 (align_functions_log): Redefine macro to use new types.
9891 (align_loops): Redefine macro to use new types.
9892 (align_jumps): Redefine macro to use new types.
9893 (align_labels): Redefine macro to use new types.
9894 (align_functions): Redefine macro to use new types.
9895 (align_functions_max_skip): Redefine macro to use new types.
9896 (align_loops_value): New macro.
9897 (align_jumps_value): New macro.
9898 (align_labels_value): New macro.
9899 (align_functions_value): New macro.
9900 * function.c (invoke_set_current_function_hook): Propagate
9901 alignment values from flags to global variables default in
9903 * ipa-icf.c (sem_function::equals_wpa): Use
9904 cl_optimization_option_eq instead of memcmp.
9905 * lto-streamer.h (cl_optimization_stream_out): Support streaming
9907 (cl_optimization_stream_in): Likewise.
9908 * optc-save-gen.awk: Support strings in cl_optimization.
9909 * opth-gen.awk: Likewise.
9910 * opts.c (finish_options): Remove error checking of invalid
9912 (MAX_CODE_ALIGN): Remove.
9913 (MAX_CODE_ALIGN_VALUE): Likewise.
9914 (parse_and_check_align_values): New function.
9915 (check_alignment_argument): Likewise.
9916 (common_handle_option): Use check_alignment_argument.
9917 * opts.h (parse_and_check_align_values): Declare.
9918 * toplev.c (init_alignments): Remove.
9919 (read_log_maxskip): New.
9920 (parse_N_M): Likewise.
9921 (parse_alignment_opts): Likewise.
9922 (backend_init_target): Remove usage of init_alignments.
9923 * toplev.h (parse_alignment_opts): Declare.
9924 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9926 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9927 * tree.c (cl_option_hasher::equal): New.
9928 * varasm.c: Use new global macros.
9930 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9932 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9933 Use a simpler align directive also if MAXSKIP = ALIGN-1.
9934 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9935 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9936 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9937 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9938 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9939 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9940 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9941 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9942 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9943 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9944 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9946 2018-07-04 Martin Liska <mliska@suse.cz>
9947 Jonathan Wakely <jwakely@redhat.com>
9949 * coverage.c: Use correct type.
9950 * doc/invoke.texi: Language correction.
9952 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
9955 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9956 ENDBRANCH for non-tail call which may return via indirect branch.
9957 * doc/extend.texi: Document indirect_return attribute.
9959 2018-07-03 Martin Sebor <msebor@redhat.com>
9961 PR tree-optimization/86274
9962 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9964 (format_floating): Correct handling of infinities and NaNs.
9966 2018-07-03 Martin Sebor <msebor@redhat.com>
9968 * print-tree.c (print_real_cst): New function.
9969 (print_node_brief): Call it.
9970 (print_node): Ditto.
9972 2018-07-03 Jeff Law <law@redhat.com>
9974 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9975 into a single pattern.
9977 * config/h8300/h8300.md (ors code_iterator): New.
9978 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9979 a single pattern and single splitter.
9980 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9981 (iorqi3_1, xorqi3_1): Likewise.
9982 (iorqi3, xorqi3 expanders): Similarly.
9984 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9985 (movmd_internal) into a single pattern using the P mode iterator.
9986 (movmd splitters): Similarly.
9987 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9988 (movsd splitters): Similarly.
9990 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9991 ADDB, ADDW and ADDL into a single ADD attribute which selects the
9992 right table based on the size of the operand.
9993 * config/h8300/h8300.md (length_table): Corresponding changes. All
9994 references to "addb", "addw" and "addl" changed to "add".
9995 (btst patterns): Merge two variants into a single pattern.
9996 (tstqi, tsthi): Likewise.
9997 (addhi3_incdec, addsi3_incdec): Likewise.
9998 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9999 (mulhi3, mulsi3): Likewise.
10000 (udivhi3, udivsi3): Likewise.
10001 (divhi3, divsi3): Likewise.
10002 (andorqi3, andorhi3, andorsi3): Likewise.
10004 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
10007 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10008 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10010 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10012 PR tree-optimization/85694
10013 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10014 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10016 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10017 (<u>avg<mode>3_ceil): New patterns.
10019 2018-07-03 David Malcolm <dmalcolm@redhat.com>
10021 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10022 scan-tree-dump directive.
10023 * gcc.dg/vect/slp-perm-2.c: Likewise.
10024 * gcc.dg/vect/slp-perm-3.c: Likewise.
10025 * gcc.dg/vect/slp-perm-5.c: Likewise.
10026 * gcc.dg/vect/slp-perm-6.c: Likewise.
10027 * gcc.dg/vect/slp-perm-7.c: Likewise.
10028 * gcc.dg/vect/slp-perm-8.c: Likewise.
10030 2018-07-03 Marek Polacek <polacek@redhat.com>
10032 PR middle-end/86202
10033 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10035 2018-07-03 Richard Biener <rguenther@suse.de>
10038 * tree-ssa-structalias.c (find_func_clobbers): Properly
10039 handle indirect calls.
10041 2018-07-03 Jeff Law <law@redhat.com>
10043 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10044 (shifts): New code iterator.
10045 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
10046 expander. Fix HImode handling on H8/SX.
10047 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
10048 (subqi3, subhi3, subsi3 expanders): Likewise.
10049 (andqi3, andhi3, andsi3 expanders): Likewise.
10050 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
10051 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
10052 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
10053 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
10054 (zero_extendqihi2, zero_extendqisi2): Likewise.
10055 (extendqihi2, extendqisi2): Likewise.
10056 (rotlqi3, rotlhi3, rotlsi3): Likewise.
10057 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
10058 (rotlqi3_1, rotlhi3_1): Likewise.
10059 (logicalhi3_sn, logicalsi3_sn): Likewise.
10060 (logicalhi3, logicalsi3): Likewise.
10062 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10064 * tree-vect-patterns.c (vect_recog_rotate_pattern)
10065 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10066 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
10067 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
10068 type to append_pattern_def_seq instead of creating a stmt_vec_info
10070 (build_mask_conversion): Likewise. Remove vinfo argument.
10071 (vect_add_conversion_to_patterm): Likewise, renaming to...
10072 (vect_add_conversion_to_pattern): ...this.
10073 (vect_recog_mask_conversion_pattern): Update call to
10074 build_mask_conversion. Pass the vector type to
10075 append_pattern_def_seq here too.
10076 (vect_recog_gather_scatter_pattern): Update call to
10077 vect_add_conversion_to_pattern.
10079 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10081 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
10082 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10083 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
10084 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
10085 STMT_VINFO_PATTERN_DEF_SEQ to null here.
10086 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
10087 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10088 append_pattern_def_seq instead of new_pattern_def_seq.
10089 (vect_recog_divmod_pattern): Do both of the above.
10090 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
10093 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10095 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10096 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
10097 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
10098 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
10099 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
10100 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10101 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
10102 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10103 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10104 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
10105 parameter with a single stmt_vec_info.
10106 (vect_recog_func_ptr): Likewise.
10107 (vect_recog_gather_scatter_pattern): Likewise, folding in...
10108 (vect_try_gather_scatter_pattern): ...this.
10109 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
10110 the stmt_vec_info of the statement to be matched. Don't clear
10111 STMT_VINFO_RELATED_STMT.
10112 (vect_pattern_recog): Update call accordingly.
10114 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10116 PR tree-optimization/85694
10117 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
10118 (uavgM3_ceil): Document new optabs.
10119 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
10120 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
10122 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
10123 (savg_ceil_optab): New optabs.
10124 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
10125 (vect_vect_recog_func_ptrs): Add it.
10126 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
10127 constant directly from the associated lhs.
10129 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10131 * tree-vect-patterns.c (vect_split_statement): New function.
10132 (vect_convert_input): Use it to try to split an existing cast.
10134 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10136 * poly-int.h (print_hex): New function.
10137 * dumpfile.h (dump_dec, dump_hex): Declare.
10138 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
10139 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
10140 min_input_precision, operation_precision and operation_sign.
10141 * tree-vect-patterns.c (vect_get_range_info): New function.
10142 (vect_same_loop_or_bb_p, vect_single_imm_use)
10143 (vect_operation_fits_smaller_type): Delete.
10144 (vect_look_through_possible_promotion): Add an optional
10145 single_use_p parameter.
10146 (vect_recog_over_widening_pattern): Rewrite to use new
10147 stmt_vec_info infomration. Handle one operation at a time.
10148 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
10149 (vect_truncatable_operation_p, vect_set_operation_type)
10150 (vect_set_min_input_precision): New functions.
10151 (vect_determine_min_output_precision_1): Likewise.
10152 (vect_determine_min_output_precision): Likewise.
10153 (vect_determine_precisions_from_range): Likewise.
10154 (vect_determine_precisions_from_users): Likewise.
10155 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
10156 (vect_vect_recog_func_ptrs): Put over_widening first.
10158 (vect_pattern_recog): Call vect_determine_precisions.
10160 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
10162 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
10163 statements that have been replaced by further pattern statements.
10164 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
10166 2018-07-03 Richard Biener <rguenther@suse.de>
10168 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10169 always set *dt. Dump vectype in vectype overload.
10170 * dumpfile.h (dump_gimple_expr): New function.
10171 (dump_gimple_expr_loc): Likewise.
10172 * dumpfile.c (dump_gimple_expr): New function.
10173 (dump_gimple_expr_loc): Likewise.
10175 2018-07-02 Jeff Law <law@redhat.com>
10177 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
10178 the H8/300, H8/300H and H8/S variants into a single pattern.
10179 (movhi_h8300, movqi_h8300hs): Similarly.
10180 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
10181 (QHI mode iterator): New.
10183 * config/h8300/h8300.md: Remove trailing whitespace.
10185 2018-07-02 Jim Wilson <jimw@sifive.com>
10187 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
10188 instead of emit_insn for interrupt returns.
10189 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
10190 (riscv_sret, riscv_uret): Likewise.
10192 2018-07-02 David Malcolm <dmalcolm@redhat.com>
10194 * pretty-print.c (selftest::test_pp_format): Move save and restore
10195 of quotes to class auto_fix_quotes, and add an instance.
10196 * selftest.c: Include "intl.h".
10197 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
10198 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
10199 * selftest.h (selftest::auto_fix_quotes): New class.
10201 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
10203 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
10204 (aarch64_sve_prepare_conditional_op): Remove.
10205 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
10206 Allow aarch64_simd_reg_or_zero as select operand; remove
10207 the aarch64_sve_prepare_conditional_op call.
10208 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
10209 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10210 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
10211 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
10212 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
10213 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
10214 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
10215 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
10216 and a splitters to match all of the *_any patterns.
10217 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
10219 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
10220 (SVE_COND_FP_BINARY_REV): Remove.
10221 (sve_int_op_rev, sve_fp_op_rev): New.
10222 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
10223 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
10224 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
10225 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
10226 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
10227 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
10228 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
10229 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
10230 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
10232 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
10233 Remove match_dup 1 from the inner unspec.
10234 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10236 * config/aarch64/aarch64.md (movprfx): New attr.
10237 (length): Default movprfx to 8.
10238 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
10239 (*madd<SVE_I>, *msub<SVE_I): Likewise.
10240 (*<su>mul<SVE_I>3_highpart): Likewise.
10241 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
10242 (*v<ASHIFT><SVE_I>3): Likewise.
10243 (*<su><MAXMIN><SVE_I>3): Likewise.
10244 (*<su><MAXMIN><SVE_F>3): Likewise.
10245 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
10246 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
10247 (*div<SVE_F>4): Likewise.
10249 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
10251 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
10254 2018-07-02 Richard Biener <rguenther@suse.de>
10256 PR tree-optimization/86363
10257 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
10258 memset argument refers to a non-variable address.
10260 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
10262 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
10263 {PLUS,MINUS}_EXPR code to...
10264 (adjust_symbolic_bound): ...here,
10265 (combine_bound): ...here,
10266 (set_value_range_with_overflow): ...and here.
10268 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
10270 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
10272 (extract_range_from_abs_expr): ...here.
10274 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
10276 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
10277 -fno-omit-frame-pointer when not optimizing.
10279 2018-07-02 Martin Liska <mliska@suse.cz>
10282 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
10283 (propagate_nothrow): Likewise.
10285 2018-07-02 Martin Liska <mliska@suse.cz>
10288 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
10290 2018-07-02 David Malcolm <dmalcolm@redhat.com>
10292 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
10293 function in r262149, changing "loc" param from source_location to
10294 const dump_location_t &.
10295 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
10296 declaration, as above.
10298 2018-07-01 Paul Koning <ni1d@arrl.net>
10300 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10301 -munit-asm, -mgnu-asm, -mdec-asm.
10302 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
10303 (pdp11_output_labelref): New.
10304 (pdp11_output_def): New.
10305 (pdp11_output_addr_vec_elt): New.
10306 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
10307 %# and %@ format codes.
10308 (pdp11_option_override): New.
10309 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
10310 (pdp11_output_ident): New.
10311 (pdp11_asm_named_section): New.
10312 (pdp11_asm_init_sections): New.
10313 (pdp11_file_start): New.
10314 (pdp11_file_end): New.
10315 (output_ascii): Use .ascii/.asciz for -mdec-asm.
10316 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
10317 %o, like %c but octal.
10318 (pdp11_option_override): New.
10319 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
10321 (DATA_SECTION_ASM_OP): Ditto.
10322 (READONLY_DATA_SECTION_ASM_OP): New.
10323 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
10324 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
10325 (ASM_OUTPUT_LABELREF): Ditto.
10326 (ASM_OUTPUT_DEF): Ditto.
10327 (ASM_OUTPUT_EXTERNAL): New.
10328 (ASM_OUTPUT_SOURCE_FILENAME): New.
10329 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
10330 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
10331 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
10332 %# and %@ format codes.
10333 * config/pdp11/pdp11.opt (mgnu-asm): New.
10334 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
10335 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
10336 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
10338 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
10340 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
10341 dereferencing path[] beyond its length.
10343 (debug_all_paths): New.
10344 (rewire_first_differing_edge): New.
10345 (adjust_paths_after_duplication): New.
10346 (duplicate_thread_path): Call adjust_paths_after_duplication.
10348 (thread_through_all_blocks): Add new argument to
10349 duplicate_thread_path.
10351 2018-06-30 Jim Wilson <jimw@sifive.com>
10353 * config/riscv/predicates.md (p2m1_shift_operand): New.
10354 (high_mask_shift_operand): New.
10355 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
10356 pattern using p2m1_shift_operand.
10357 (lshsi3_zero_extend_3+2): New combiner pattern using
10358 high_mask_shift_operand.
10360 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10362 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
10364 (vect_recog_rotate_pattern): ...here.
10365 (vect_convert_input): Try to insert casts of invariants in the
10367 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
10368 preheader to be empty.
10370 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10372 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
10373 vector type. If given, install it in the new statement's
10374 STMT_VINFO_VECTYPE.
10375 (vect_element_precision): New function.
10376 (vect_unpromoted_value): New struct.
10377 (vect_unpromoted_value::vect_unpromoted_value): New function.
10378 (vect_unpromoted_value::set_op): Likewise.
10379 (vect_look_through_possible_promotion): Likewise.
10380 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
10381 (vect_widened_op_tree, vect_convert_input): Likewise.
10382 (vect_convert_inputs, vect_convert_output): Likewise.
10383 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
10384 to handle the optional cast of the multiplication result and
10385 vect_widened_op_tree to detect the widened multiplication itself.
10386 Do not require the input and output of promotion casts to have
10387 the same sign, but base the signedness of the operation on the
10388 input rather than the result. If the pattern includes two
10389 promotions, check that those promotions have the same sign.
10390 Do not restrict the MULT_EXPR handling to a double-width result;
10391 handle quadruple-width results and wider. Use vect_convert_inputs
10392 to convert the inputs to the common type.
10393 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
10394 to handle the optional cast of the ABS result. Also allow a sign
10395 change or a sign extension between the ABS and MINUS.
10396 Use vect_widened_op_tree to detect the widened subtraction and use
10397 vect_convert_inputs to convert the inputs to the common type.
10398 (vect_handle_widen_op_by_const): Delete.
10399 (vect_recog_widen_op_pattern): New function.
10400 (vect_recog_widen_mult_pattern): Use it.
10401 (vect_recog_widen_shift_pattern): Likewise.
10402 (vect_recog_widen_sum_pattern): Use
10403 vect_look_through_possible_promotion to handle the promoted
10406 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10408 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
10409 the containing gimple_seq *.
10410 * gimple-iterator.h (gsi_for_stmt): Declare it.
10411 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
10412 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
10413 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10414 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10415 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
10417 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
10419 (vect_mark_pattern_stmts): ...here. Handle cases in which the
10420 statement being replaced is part of an existing pattern
10421 definition sequence, inserting the new pattern statements before
10423 (vect_pattern_recog_1): Don't return a bool. If the statement
10424 is already part of a pattern, instead apply pattern matching
10425 to the pattern definition statements. Don't clear the
10426 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
10427 (vect_pattern_recog): Don't break after the first match;
10428 continue processing the pattern definition statements instead.
10429 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
10431 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10433 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10434 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10435 (vect_recog_widen_sum_pattern): Use it.
10437 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10439 * tree-vect-loop.c (vectorizable_reduction): Assert that the
10440 phi is not a pattern statement and has not been replaced by
10441 a pattern statement.
10442 * tree-vect-patterns.c (type_conversion_p): Don't check
10443 STMT_VINFO_IN_PATTERN_P.
10444 (vect_recog_vector_vector_shift_pattern): Likewise.
10445 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
10446 the pattern statement rather than the original statement; check
10447 directly for a WIDEN_MULT_EXPR here.
10448 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
10449 vect_is_simple_use to return the pattern statement rather
10450 than the original statement; use is_pattern_stmt_p to check
10451 for such a pattern statement.
10452 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
10453 to return the pattern statement rather than the original statement;
10454 don't do the same transformation here.
10455 (vect_is_simple_use): If the defining statement has been replaced
10456 by a pattern statement, return the pattern statement instead.
10457 Remove the corresponding (local) transformation from the vectype
10460 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
10462 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
10463 end and default to null.
10464 * tree-vect-loop.c (vect_create_epilog_for_reduction)
10465 (vectorizable_reduction): Update calls accordingly, dropping the
10466 gimple ** argument if the passed-back statement isn't needed.
10467 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
10468 (vect_recog_rotate_pattern): Likewise.
10469 (vect_recog_mask_conversion_pattern): Likewise.
10470 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10471 (vect_mask_constant_operand_p): Likewise.
10472 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
10473 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
10474 (get_group_load_store_type, get_load_store_type): Likewise.
10475 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
10476 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
10477 (vectorizable_conversion, vectorizable_assignment): Likewise.
10478 (vectorizable_shift, vectorizable_operation): Likewise.
10479 (vectorizable_store, vect_is_simple_cond): Likewise.
10480 (vectorizable_condition, vectorizable_comparison): Likewise.
10481 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
10482 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
10483 and move it to the end. Cope with null def_stmt_outs.
10485 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
10487 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
10489 2018-06-29 Jeff Law <law@redhat.com>
10491 * config/v850/v850.c (v850_legitimate_address_p): Handle large
10492 displacements for TARGET_V850E2V3 and newer.
10493 (TARGET_LRA_P): Remove. Defaults to LRA now.
10494 * config/v850/v850.md (sign23byte_load): Remove.
10495 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
10496 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
10498 2018-06-29 Martin Liska <mliska@suse.cz>
10501 * coverage.c (coverage_init): Mangle full path name.
10502 * doc/invoke.texi: Document the change.
10503 * gcov-io.c (mangle_path): New.
10504 * gcov-io.h (mangle_path): Likewise.
10505 * gcov.c (mangle_name): Use mangle_path for path mangling.
10507 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10509 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
10510 if starting source register is not even.
10512 2018-06-29 Martin Liska <mliska@suse.cz>
10514 PR tree-optimization/86263
10515 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
10516 Make edge redirection.
10518 2018-06-29 David Malcolm <dmalcolm@redhat.com>
10520 * dumpfile.c (dump_loc): Add indentation based on scope depth.
10521 (dump_scope_depth): New variable.
10522 (get_dump_scope_depth): New function.
10523 (dump_begin_scope): New function.
10524 (dump_end_scope): New function.
10525 * dumpfile.h (get_dump_scope_depth): New declaration.
10526 (dump_begin_scope): New declaration.
10527 (dump_end_scope): New declaration.
10528 (class auto_dump_scope): New class.
10529 (AUTO_DUMP_SCOPE): New macro.
10530 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
10533 2018-06-29 Richard Biener <rguenther@suse.de>
10535 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
10536 compute_all_dependences succeeds.
10537 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
10538 exceed --param loop-max-datarefs-for-datadeps.
10540 2018-06-29 Jakub Jelinek <jakub@redhat.com>
10542 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
10544 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
10547 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
10548 alternative 0 in preferred_for_speed attribute.
10550 2018-06-28 Paul Koning <ni1d@arrl.net>
10552 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
10553 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
10554 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
10555 * config/pdp11/pdp11.md: Correct "length" attribute calculation
10556 for shift insn patterns.
10558 2018-06-28 David Malcolm <dmalcolm@redhat.com>
10560 * cgraph.c (cgraph_node::get_body): Replace assignments to
10561 "dump_file" with calls to set_dump_file.
10562 * dumpfile.c (alt_dump_file): Make static, and group with...
10563 (alt_flags): ...this definition.
10564 (dumps_are_enabled): New variable.
10565 (refresh_dumps_are_enabled): New function.
10566 (set_dump_file): New function.
10567 (set_alt_dump_file): New function.
10568 (gcc::dump_manager::dump_start): Replace assignments to
10569 "dump_file" and "alt_dump_file" with calls to set_dump_file and
10571 (gcc::dump_manager::dump_finish): Likewise.
10572 * dumpfile.h (alt_dump_file): Delete decl.
10573 (dumps_are_enabled): New variable decl.
10574 (set_dump_file): New function decl.
10575 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
10577 * tree-nested.c (lower_nested_functions): Replace assignments to
10578 "dump_file" with calls to set_dump_file.
10580 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
10582 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
10583 goto_locus of each outgoing edge of each basic block.
10585 2018-06-28 Richard Biener <rguenther@suse.de>
10587 * dwarf2out.c (decl_scope_table): Remove.
10588 (push_decl_scope): Likewise.
10589 (pop_decl_scope): Likewise.
10590 (gen_type_die_for_member): Do not call push/pop_decl_scope.
10591 (gen_struct_or_union_type_die): Likewise.
10592 (gen_tagged_type_die): Likewise.
10593 (dwarf2out_init): Do not initialize decl_scope_table.
10594 (dwarf2out_c_finalize): Do not free it.
10596 2018-06-28 Richard Biener <rguenther@suse.de>
10598 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
10599 deciding whether to not re-use a DIE.
10601 2018-06-28 Richard Biener <rguenther@suse.de>
10603 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
10604 DW_AT_abstract_origin attribute.
10606 2018-06-28 Martin Liska <mliska@suse.cz>
10608 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10609 Use newly introduced constants.
10610 * tree-switch-conversion.h (struct jump_table_cluster):
10611 Define max_ratio_for_size and max_ratio_for_speed.
10613 2018-06-28 Martin Liska <mliska@suse.cz>
10615 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10616 Add new checking assert to catch invalid state.
10617 (jump_table_cluster::can_be_handled): Handle single case
10619 (jump_table_cluster::is_beneficial): Bail out for such case.
10620 (bit_test_cluster::find_bit_tests):
10621 Add new checking assert to catch invalid state.
10622 (bit_test_cluster::can_be_handled): Handle single case
10624 (bit_test_cluster::is_beneficial): Bail out for such case.
10625 (switch_decision_tree::analyze_switch_statement):
10628 2018-06-28 Martin Liska <mliska@suse.cz>
10630 * common.opt: Introduce -completion option.
10631 * gcc.c (driver_handle_option): Handle it.
10632 (driver::main): Print completions if completion
10634 * opt-suggestions.c (option_proposer::get_completions):
10636 (option_proposer::suggest_completion): Likewise.
10637 (option_proposer::find_param_completions): Likewise.
10638 (verify_autocompletions): Likewise.
10639 (test_completion_valid_options): Likewise.
10640 (test_completion_valid_params): Likewise.
10641 (in_completion_p): Likewise.
10642 (empty_completion_p): Likewise.
10643 (test_completion_partial_match): Likewise.
10644 (test_completion_garbage): Likewise.
10645 (opt_proposer_c_tests): Likewise.
10646 * opt-suggestions.h: Declare new functions.
10647 * opts.c (common_handle_option): Handle OPT__completion_.
10648 * selftest-run-tests.c (selftest::run_tests): Add
10649 opt_proposer_c_tests.
10650 * selftest.c (assert_str_startswith): New.
10651 * selftest.h (assert_str_startswith): Likewise.
10652 (opt_proposer_c_tests): New.
10653 (ASSERT_STR_STARTSWITH): Likewise.
10655 2018-06-28 Martin Liska <mliska@suse.cz>
10657 * Makefile.in: Add opt-suggestions.o.
10658 * gcc-main.c: Include opt-suggestions.h.
10659 * gcc.c (driver::driver): Likewise.
10660 (driver::~driver): Remove m_option_suggestions.
10661 (driver::build_option_suggestions): Moved to option_proposer.
10662 (driver::suggest_option): Likewise.
10663 (driver::handle_unrecognized_options): Use option_proposer.
10664 * gcc.h (class driver): Add new memver m_option_proposer.
10665 * opt-suggestions.c: New file.
10666 * opt-suggestions.h: New file.
10668 2018-06-28 Martin Liska <mliska@suse.cz>
10670 * vec.h (class auto_string_vec): New (moved from auto_argvec).
10671 (auto_string_vec::~auto_string_vec): Likewise.
10673 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
10675 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
10676 prevent_decl_creation_for_types fields up and add reset_location field.
10677 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
10678 statement if id->reset_location is true.
10679 (copy_edges_for_bb): Do not set goto_locus on the new edges if
10680 id->reset_location is true.
10681 (copy_phis_for_bb): Force input_location on the arguments if
10682 id->reset_location is true.
10683 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10684 is set on the function to be inlined.
10686 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
10688 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
10690 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
10692 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
10693 registers for Pmode.
10694 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
10695 hard registers for the clobbered pseudo.
10697 2018-06-27 Paul Koning <ni1d@arrl.net>
10699 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10700 mutually exclusive options.
10701 * config/pdp11/constraints.md (h): New constraint.
10702 (O): Update definition to match shift code generation.
10703 (D): New constraint.
10704 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
10706 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
10708 (output_jump): Change arguments.
10709 (pdp11_fixed_cc_regs): New function.
10710 (pdp11_cc_mode): Ditto.
10711 (pdp11_expand_shift): Ditto.
10712 (pdp11_assemble_shift): Ditto.
10713 (pdp11_small_shift): Ditto.
10714 (pdp11_branch_cost): Remove.
10715 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
10717 (pdp11_register_move_cost): Update for CC registers.
10718 (pdp11_rtx_costs): Add case for LSHIFTRT.
10719 (pdp11_output_jump): Add CCNZ mode conditional branches.
10720 (notice_update_cc_on_set): Remove.
10721 (pdp11_cc_mode): New function.
10722 (simple_memory_operand): Correct pre/post decrement case.
10723 (no_side_effect_operand): New function.
10724 (pdp11_regno_reg_class): Add CC_REGS class.
10725 (pdp11_fixed_cc_regs): New function.
10726 (pdp11_small_shift): New function.
10727 (pdp11_expand_shift): New function to expand shift insns.
10728 (pdp11_assemble_shift): New function to output shifts.
10729 (pdp11_branch_cost): Remove.
10730 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
10731 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
10732 (WCHAR_TYPE): Ditto.
10733 (PTRDIFF_TYPE): Ditto.
10734 (ADJUST_INSN_LENGTH): New macro.
10735 (FIXED_REGISTERS): Add CC registers.
10736 (CALL_USED_REGISTERS): Ditto.
10737 (reg_class): Ditto.
10738 (REG_CLASS_NAMES): Ditto.
10739 (REG_CLASS_CONTENTS): Ditto.
10740 (SELECT_CC_MODE): Use new function.
10741 (TARGET_FLAGS_REGNUM): New macro.
10742 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
10743 (cc0_reg_rtx): Remove.
10744 (CC_STATUS_MDEP): Remove.
10745 (CC_STATUS_MDEFP_INIT): Remove.
10746 (CC_IN_FPU): Remove.
10747 (NOTICE_UPDATE_CC): Remove.
10748 (REGISTER_NAMES): Add CC registers.
10749 (BRANCH_COST): Change to constant 1.
10750 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10752 * config/pdp11/pdp11.opt (mbcopy): Remove.
10753 (mbcopy-builtin): Remove.
10754 (mbranch-cheap): Remove.
10755 (mbranch-expensive): Remove.
10756 * config/pdp11/predicates.md (expand_shift_operand): Update to
10757 match shift code generation.
10758 (ccnz_operator): New predicate.
10759 * doc/invoke.texi (PDP-11 Options): Remove deleted options
10760 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10761 Remove non-existent option -mabshi, -mno-abshi. Document mutually
10763 * doc/md.texi (PDP-11): Document new D and h constraints. Update
10764 description of O constraint.
10766 2018-06-27 Jeff Law <law@redhat.com>
10767 Austin Law <austinklaw@gmail.com>
10769 * config/v850/v850.md (addsi3_set_flags): New pattern.
10770 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10771 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10772 (zero_extendhisi2_v850_set_flags): Likewise.
10773 (zero_extendqisi2_v850_set_flags): Likewise.
10774 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10775 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10776 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10778 * config/v850/v850-protos.h (notice_update_cc): Remove.
10779 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10780 (v850_print_operand): Handle 'D' and "d".
10781 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10782 Add handling of arithmetic/logical operations compared against zero.
10783 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10784 Do not look at v850_compare_op, instead get mode from last argument.
10785 (v850_gen_compare): Remove
10786 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10787 after reload for prologue insns.
10788 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10790 (construct_save_jarl): Likewise.
10791 (TARGET_FLAGS_REGNUM): Define.
10792 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10793 (NOTICE_UPDATE_CC): Remove.
10794 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10795 than cc0. Conditionalize on reload_completed.
10796 (cmpsi_insn, setfcc_insn): Likewise.
10797 (tst1 splitter): Turn into define_and_split which sets the flags
10799 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10800 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10801 (cstoresf4, cstoredf4): Clobber the flags.
10802 (cmpsi, cmpsf, cmpdf): Remove expanders.
10803 (setf_insn): Remove pattern.
10804 (addsi3): Turn into define_and_split which clobbers the flags after
10805 reload and a suitable pattern (addsi3_clobber_flags) for use after
10807 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10808 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10809 (ashrsi3, ashrsi3_v850e2): Likewise.
10810 (bins): Clobber the flags.
10811 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10812 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10813 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10814 (call_value_internal_short, call_value_internal_long): Likewise.
10815 (callt_save_interrupt, callt_return_interrupt): Likewise.
10816 (save_interrupt, return_interrupt): Likewise.
10817 (callt_save_all_interrupt, save_all_interrupt): Likewise.
10818 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10819 (restore_all_interrupt, _restore_all_interrupt): Likewise.
10820 (All FP comparisons): Only allow after reload has completed.
10822 (divh, divhu): Tweak output template.
10823 (branch_z_normal, branch_z_invert): Remove
10824 (branch_nz_normal, branch_nz_invert): Likewise.
10825 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10827 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10828 * config/v850/v850.c (notice_update_cc): Remove.
10829 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
10830 (CC_NO_CARRY): Likewise.
10831 (NOTICE_UPDATE_CC): Define to nothing.
10832 * config/v850/v850.md: Remove block comment on cc0 handling
10833 Remove "cc" attribute from all patterns. Remove cc_status handling
10834 from all patterns. Minor formatting fixes.
10836 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10838 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
10839 (cortex-a76.cortex-a55): Likewise.
10840 * config/aarch64/aarch64-tune.md: Regenerate.
10841 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
10842 cortex-a76.cortex-a55.
10844 2018-06-27 Jeff Law <law@redhat.com>
10846 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
10847 (MULTILIB_DIRNAMES): Similarly.
10849 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
10851 * gimple.h (gimple_return_retbnd): Delete.
10852 (gimple_return_set_retbnd): Likewise.
10853 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
10854 gimple_return_set_retbnd.
10855 * gimple-pretty-print.c (dump_gimple_return): Remove call to
10856 gimple_return_retbnd and adjust.
10857 * tree-inline.h (struct copy_body_data): Remove retbnd field.
10858 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
10859 Explicitly return NULL in a couple more cases. Move assertion
10860 on debug statements and remove unreachable code.
10861 (reset_debug_binding): Do not test id->retbnd.
10862 (expand_call_inline): Do not set it.
10864 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
10866 * configure.ac: Add --disable-gcov option.
10867 * configure: Regenerate.
10868 * Makefile.in: Honour @enable_gcov@.
10869 * doc/install.texi: Document --disable-gcov.
10871 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10873 * config/arm/arm-cpus.in (cortex-a76): New entry.
10874 (cortex-a76.cortex-a55): Likewise.
10875 * config/arm/arm-tables.opt: Regenerate.
10876 * config/arm/arm-tune.md: Likewise.
10877 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10878 * doc/invoke.texi (ARM Options): Document cortex-a76 and
10879 cortex-a76.cortex-a55.
10881 2018-06-27 Tamar Christina <tamar.christina@arm.com>
10884 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10886 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
10888 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10890 (EPILOGUE_USES): Likewise.
10892 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
10894 * tree-inline.c (remap_location): New function extracted from...
10895 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
10896 (copy_phis_for_bb): ...here. Call remap_location.
10897 (copy_cfg_body): Adjust call to copy_edges_for_bb.
10899 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
10901 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10902 unaligned vsx for 16B memset.
10904 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
10907 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10908 ieee128_float_type_node to long_double_type_node unless
10909 TARGET_LONG_DOUBLE_128 is set.
10911 2018-06-26 David Malcolm <dmalcolm@redhat.com>
10913 * cfgloop.c (get_loop_location): Convert return type from
10914 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10915 by implicit construction from rtx_insn *, and using
10916 dump_user_location_t::from_function_decl for the fallback case.
10917 * cfgloop.h (get_loop_location): Convert return type from
10918 location_t to dump_user_location_t.
10919 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10920 dump_printf_loc to pass in a dump_location_t rather than a
10921 location_t, via the gimple stmt.
10922 * coverage.c (get_coverage_counts): Update calls to
10923 dump_printf_loc to pass in dump_location_t rather than a
10925 * doc/optinfo.texi (Dump types): Convert example of
10926 dump_printf_loc from taking "locus" to taking "insn". Update
10927 description of the "_loc" calls to cover dump_location_t.
10928 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10930 (dump_user_location_t::dump_user_location_t): New constructors,
10931 from gimple *stmt and rtx_insn *.
10932 (dump_user_location_t::from_function_decl): New function.
10933 (dump_loc): Make static.
10934 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10935 const dump_location_t &.
10936 (dump_generic_expr_loc): Delete.
10937 (dump_printf_loc): Convert param "loc" from location_t to
10938 const dump_location_t &.
10939 (selftest::test_impl_location): New function.
10940 (selftest::dumpfile_c_tests): New function.
10941 * dumpfile.h: Include "profile-count.h".
10942 (class dump_user_location_t): New class.
10943 (struct dump_impl_location_t): New struct.
10944 (class dump_location_t): New class.
10945 (dump_printf_loc): Convert 2nd param from source_location to
10946 const dump_location_t &.
10947 (dump_generic_expr_loc): Delete.
10948 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10949 const dump_location_t &.
10950 * gimple-fold.c (fold_gimple_assign): Update call to
10951 dump_printf_loc to pass in a dump_location_t rather than a
10952 location_t, via the gimple stmt.
10953 (gimple_fold_call): Likewise.
10954 * gimple-loop-interchange.cc
10955 (loop_cand::analyze_iloop_reduction_var): Update for change to
10956 check_reduction_path.
10957 (tree_loop_interchange::interchange): Update for change to
10958 find_loop_location.
10959 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10960 change in return-type of find_loop_location.
10961 (graphite_regenerate_ast_isl): Likewise.
10962 * graphite-optimize-isl.c (optimize_isl): Likewise.
10963 * graphite.c (graphite_transform_loops): Likewise.
10964 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10965 pass in a dump_location_t rather than a location_t, via the
10967 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10968 * ipa.c (walk_polymorphic_call_targets): Likewise.
10969 * loop-unroll.c (report_unroll): Convert "locus" param from
10970 location_t to dump_location_t.
10971 (decide_unrolling): Update for change to get_loop_location's
10973 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10974 location_t to dump_user_location_t.
10975 (grid_find_single_omp_among_assignments_1): Updates calls to
10976 dump_printf_loc to pass in a dump_location_t rather than a
10977 location_t, via the gimple stmt.
10978 (grid_parallel_clauses_gridifiable): Convert "tloc" from
10979 location_t to dump_location_t. Updates calls to dump_printf_loc
10980 to pass in a dump_location_t rather than a location_t, via the
10982 (grid_inner_loop_gridifiable_p): Likewise.
10983 (grid_dist_follows_simple_pattern): Likewise.
10984 (grid_gfor_follows_tiling_pattern): Likewise.
10985 (grid_target_follows_gridifiable_pattern): Likewise.
10986 (grid_attempt_target_gridification): Convert initialization
10987 of local "grid" from memset to zero-initialization; FIXME: does
10988 this require C++11? Update call to dump_printf_loc to pass in a
10989 optinfo_location rather than a location_t, via the gimple stmt.
10990 * profile.c (read_profile_edge_counts): Updates call to
10991 dump_printf_loc to pass in a dump_location_t rather than a
10993 (compute_branch_probabilities): Likewise.
10994 * selftest-run-tests.c (selftest::run_tests): Call
10996 * selftest.h (dumpfile_c_tests): New decl.
10997 * tree-loop-distribution.c (pass_loop_distribution::execute):
10998 Update for change in return type of find_loop_location.
10999 * tree-parloops.c (parallelize_loops): Likewise.
11000 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11001 "locus" from location_t to dump_user_location_t.
11002 (canonicalize_loop_induction_variables): Likewise.
11003 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11004 for change in return type of find_loop_location.
11005 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11006 to dump_printf_loc to pass in a dump_location_t rather than a
11007 location_t, via the stmt.
11008 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11010 * tree-vect-loop-manip.c (find_loop_location): Convert return
11011 type from source_location to dump_user_location_t.
11012 (vect_do_peeling): Update for above change.
11013 (vect_loop_versioning): Update for change in type of
11015 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11016 from location_t to dump_user_location_t.
11017 (vect_estimate_min_profitable_iters): Update for change in type
11019 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11020 location_t to dump_location_t.
11021 (vect_slp_bb): Update for change in type of vect_location.
11022 * tree-vectorizer.c (vect_location): Convert from source_location
11023 to dump_user_location_t.
11024 (try_vectorize_loop_1): Update for change in vect_location's type.
11025 (vectorize_loops): Likewise.
11026 (increase_alignment): Likewise.
11027 * tree-vectorizer.h (vect_location): Convert from source_location
11028 to dump_user_location_t.
11029 (find_loop_location): Convert return type from source_location to
11030 dump_user_location_t.
11031 (check_reduction_path): Convert 1st param from location_t to
11032 dump_user_location_t.
11033 * value-prof.c (check_counter): Update call to dump_printf_loc to
11034 pass in a dump_user_location_t rather than a location_t; update
11035 call to error_at for change in type of "locus".
11036 (check_ic_target): Update call to dump_printf_loc to
11037 pass in a dump_user_location_t rather than a location_t, via the
11040 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
11042 * config/s390/s390.h (enum processor_flags): Do not use
11043 default tune parameter when -march was specified.
11045 2018-06-26 Jakub Jelinek <jakub@redhat.com>
11048 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
11049 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
11052 2018-06-26 Richard Biener <rguenther@suse.de>
11054 PR tree-optimization/86287
11056 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
11057 (vect_analyze_loop): Initialize n_stmts.
11059 2018-06-26 Richard Biener <rguenther@suse.de>
11061 PR middle-end/86271
11062 * fold-const.c (fold_convertible_p): Pointer extension
11065 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
11068 * dwarf2out.c (loc_list_has_views): Adjust comments.
11069 (dw_loc_list): Split single cross-partition range with
11072 2018-06-25 Jeff Law <law@redhat.com>
11074 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
11075 on -mbig-switch by default.
11077 * config/v850/predicates.md (const_float_1_operand): Fix match_code
11079 (const_float_0_operand): Remove unused predicate.
11080 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
11081 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
11082 (recipsf2): New expander. Original pattern now called
11084 (recipdf2, recipdf2_insn): Similarly.
11085 (rsqrtsf2, rsqrtsf2_insn): Similarly
11086 (rsqrtdf2, rsqrtdf2_insn): Similarly
11088 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
11090 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
11091 Simplify logic for FreeBSD (twice).
11093 2018-06-25 Martin Sebor <msebor@redhat.com>
11095 PR tree-optimization/86204
11096 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
11097 a strnlen result if it's less than the length of the string.
11099 2018-06-25 Martin Sebor <msebor@redhat.com>
11101 PR tree-optimization/85700
11102 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
11103 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
11104 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
11106 2018-06-25 Martin Sebor <msebor@redhat.com>
11108 * doc/extend.texi (Zero-length arrays): Update and clarify.
11110 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
11112 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
11113 added IEEE/IBM long double multilib support on PowerPC little
11114 endian Linux systems.
11115 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11116 (MULTILIB_DEFAULTS): Likewise.
11117 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11119 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
11120 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
11121 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
11123 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
11125 PR middle-end/86311
11126 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
11127 (REORDER_45): Likewise.
11129 2018-06-25 Jeff Law <law@redhat.com>
11131 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
11132 dividend to 32 bits. Adjust length.
11133 (udivmodhi4): Cleanup output template. Fix length.
11135 2018-06-25 Carl Love <cel@us.ibm.com>
11137 * config/rs6000/vsx.md: Change word selector to prefered location.
11139 2018-06-25 Richard Biener <rguenther@suse.de>
11141 PR tree-optimization/86304
11142 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
11143 epilogue-if-converted loops as well.
11145 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
11147 * lto-section-out.c (lto_begin_section): Do not print section
11148 name for noaddr and unnumbered dumps.
11150 2018-06-25 Richard Biener <rguenther@suse.de>
11152 * tree-vectorizer.h (struct vec_info_shared): New structure
11153 with parts split out from struct vec_info and loop_nest from
11154 struct _loop_vec_info.
11155 (struct vec_info): Adjust accordingly.
11156 (struct _loop_vec_info): Likewise.
11157 (LOOP_VINFO_LOOP_NEST): Adjust.
11158 (LOOP_VINFO_DATAREFS): Likewise.
11159 (LOOP_VINFO_DDRS): Likewise.
11160 (struct _bb_vec_info): Likewise.
11161 (BB_VINFO_DATAREFS): Likewise.
11162 (BB_VINFO_DDRS): Likewise.
11163 (struct _stmt_vec_info): Add dr_aux member.
11164 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
11165 (DR_MISALIGNMENT_UNINITIALIZED): New.
11166 (set_dr_misalignment): Adjust.
11167 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
11168 (vect_analyze_loop): Adjust prototype.
11169 (vect_analyze_loop_form): Likewise.
11170 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11171 Compute dependences lazily.
11172 (vect_record_base_alignments): Use shared datarefs/ddrs.
11173 (vect_verify_datarefs_alignment): Likewise.
11174 (vect_analyze_data_refs_alignment): Likewise.
11175 (vect_analyze_data_ref_accesses): Likewise.
11176 (vect_analyze_data_refs): Likewise.
11177 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
11178 constructor parameter for shared part.
11179 (vect_analyze_loop_form): Pass in shared part and adjust.
11180 (vect_analyze_loop_2): Pass in storage for the number of
11181 stmts. Move loop nest finding to the caller. Compute
11183 (vect_analyze_loop): Pass in shared part.
11184 (vect_transform_loop): Verify shared datarefs are unchanged.
11185 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
11186 constructor parameter for shared part.
11187 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
11188 (vect_slp_bb): Verify shared datarefs are unchanged before
11190 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
11192 (new_stmt_vec_info): Initialize DR_AUX misalignment to
11193 DR_MISALIGNMENT_UNINITIALIZED.
11194 * tree-vectorizer.c (vec_info::vec_info): Add constructor
11195 parameter for shared part.
11196 (vec_info::~vec_info): Adjust.
11197 (vec_info_shared::vec_info_shared): New.
11198 (vec_info_shared::~vec_info_shared): Likewise.
11199 (vec_info_shared::save_datarefs): Likewise.
11200 (vec_info_shared::check_datarefs): Likewise.
11201 (try_vectorize_loop_1): Construct shared part live for analyses
11202 of a single loop for multiple vector sizes.
11203 * tree-parloops.c (gather_scalar_reductions): Adjust.
11205 2018-06-25 Richard Biener <rguenther@suse.de>
11207 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
11208 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
11209 (vect_analyze_data_refs): Remove similar code from here and
11210 simplify accordingly.
11212 2018-06-25 Richard Biener <rguenther@suse.de>
11214 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
11215 for reverse storage order accesses rather than asserting
11216 they cannot happen here.
11218 2018-06-25 Tom de Vries <tdevries@suse.de>
11221 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
11222 Use data16 instead of .byte for insn prefix.
11224 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
11227 * parser.c (make_char_string_pack): Pass this literal chars
11228 through cpp_interpret_string.
11229 (cp_parser_userdef_numeric_literal): Check the result of
11230 make_char_string_pack.
11232 2018-06-24 Maya Rashish <coypu@sdf.org>
11234 * ginclude/stddef.h: Simplify conditions around avoiding
11235 re-definition of __size_t.
11237 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
11239 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
11240 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
11242 2018-06-22 Maya Rashish <coypu@sdf.org>
11244 * doc/invoke.texi (mno-fancy-math-387): Update for changes
11245 made to OpenBSD and NetBSD through the years.
11247 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11249 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11250 behavior of vec_pack (vector double, vector double) to match
11251 behavior of vec_float2 (vector double, vector double).
11253 2018-06-22 Olivier Hainque <hainque@adacore.com>
11255 * gimplify.c (gimplify_function_tree): Prevent creation
11256 of a trampoline for the address of the current function
11257 passed to entry/exit instrumentation hooks.
11259 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
11262 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
11265 2018-06-22 Martin Liska <mliska@suse.cz>
11267 PR tree-optimization/86263
11268 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11269 Bail out if is_enabled is false.
11270 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
11272 (jump_table_cluster::is_enabled): New function.
11274 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
11276 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11277 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
11278 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
11279 (lto_input_ts_binfo_tree_pointers): Likewise.
11280 * tree-streamer-out.c (streamer_write_tree_bitfields,
11281 write_ts_binfo_tree_pointers): Likewise.
11282 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
11284 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
11286 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
11288 2018-06-22 Martin Liska <mliska@suse.cz>
11290 * symbol-summary.h (get): Make it pure and inline move
11291 functionality from ::get function.
11292 (get): Remove and inline into ::get and ::get_create.
11293 (get_create): Move code from ::get function.
11295 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11298 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
11299 -x assembler-with-cpp.
11301 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11303 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
11304 _FILE_OFFSET_BITS=64 for C++.
11306 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
11308 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
11309 conversion insn that shows up when pr85657-3.c is compiled using
11310 IEEE 128-bit long double.
11311 (neg<mode>2_internal): Use the correct mode to check whether the
11312 mode is IBM extended.
11313 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
11314 multiply and divide external functions from being created more
11317 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11319 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
11321 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
11322 the edge can be forwarded.
11323 (cfg_layout_merge_blocks): Likewise.
11325 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11327 * except.c (finish_eh_generation): Commit edge insertions only after
11328 the EH edges have been redirected from post-landing to landing pads.
11330 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11332 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
11333 create_tmp_var_for to create the FRAME decl.
11334 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
11336 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
11338 * tree-inline.c (copy_edges_for_bb): Minor tweak.
11339 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
11340 debug statement when resetting its value.
11341 (expand_call_inline): Copy the locus of the call onto the assignment
11342 of the return value, if any. Use local variable in more cases.
11344 2018-06-21 Martin Liska <mliska@suse.cz>
11346 * ipa-pure-const.c (propagate_nothrow): Use
11347 funct_state_summaries->get.
11348 (dump_malloc_lattice): Likewise.
11349 (propagate_malloc): Likewise.
11351 2018-06-21 Richard Biener <rguenther@suse.de>
11353 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
11354 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
11355 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
11357 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
11358 BLOCK_ABSTRACT_ORIGIN unconditionally.
11360 2018-06-21 David Malcolm <dmalcolm@redhat.com>
11362 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
11364 * ipa-reference.c (ipa_reference_c_finalize): Delete
11365 ipa_ref_opt_sum_summaries and set it to NULL.
11367 2018-06-21 Tom de Vries <tdevries@suse.de>
11369 PR tree-optimization/85859
11370 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
11371 test with comment from bb_no_side_effects_p.
11373 2018-06-21 Richard Biener <rguenther@suse.de>
11375 PR tree-optimization/86232
11376 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
11377 max for constant niter.
11379 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
11381 * config/aarch64/aarch64-simd.md
11382 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
11384 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
11386 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
11387 Make opernads of the unspec commutative.
11389 2018-06-21 Richard Biener <rguenther@suse.de>
11391 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
11392 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
11393 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
11394 (vect_analyze_data_ref_dependence): Re-order checks to deal with
11396 (vect_record_base_alignments): Do not record base alignment
11397 for gathers or scatters.
11398 (vect_compute_data_ref_alignment): Drop return value that is always
11399 true. Bail out early for gathers or scatters.
11400 (vect_enhance_data_refs_alignment): Bail out early for gathers
11402 (vect_find_same_alignment_drs): Likewise.
11403 (vect_analyze_data_refs_alignment): Remove dead code.
11404 (vect_slp_analyze_and_verify_node_alignment): Likewise.
11405 (vect_analyze_data_refs): For possible gathers or scatters do
11406 not create an alternate DR, just check their possible validity
11407 and mark them. Adjust DECL_NONALIASED handling to not rely
11408 on DR_BASE_ADDRESS.
11409 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
11410 update inits of gathers or scatters.
11411 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
11412 Also copy gather/scatter flag to pattern vinfo.
11414 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
11416 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
11417 behavior of vec_packsu (vector unsigned long long, vector unsigned
11418 long long) to match behavior of vec_packs with same signature.
11420 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
11421 Thomas Schwinge <thomas@codesourcery.com>
11422 Cesar Philippidis <cesar@codesourcery.com>
11424 * gimplify.c (gimplify_scan_omp_clauses): Add support for
11425 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11426 (gimplify_adjust_omp_clauses): Likewise.
11427 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
11428 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
11429 (gimplify_omp_target_update): Update handling of acc update and
11431 * omp-low.c (install_var_field): Remove unused parameter
11432 base_pointers_restrict.
11433 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
11434 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
11436 (omp_target_base_pointers_restrict_p): Delete.
11437 (scan_omp_target): Update call to scan_sharing_clauses.
11438 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
11440 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
11441 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
11442 (convert_local_omp_clauses): Likewise.
11443 * tree-pretty-print.c (dump_omp_clause): Likewise.
11444 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
11446 (omp_clause_code_name): Likewise.
11448 2018-06-20 Jakub Jelinek <jakub@redhat.com>
11451 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
11454 PR tree-optimization/86231
11455 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
11456 anti-range don't overwrite *vr0min before using it to compute *vr0max.
11458 2018-06-20 Tom de Vries <tdevries@suse.de>
11460 PR tree-optimization/86097
11461 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
11462 iv type if signedness of iv type is not the same as that of *nit.
11464 2018-06-20 Jakub Jelinek <jakub@redhat.com>
11466 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
11467 EDGE_EH edges, verify they are all EDGE_EH.
11469 2018-06-20 Maya Rashish <coypu@sdf.org>
11471 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
11473 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11475 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
11476 * config/aarch64/aarch64.c (xgene1_tunings): Add
11477 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
11478 (aarch64_mode_valid_for_sched_fusion_p):
11479 Allow 16-byte modes.
11480 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
11481 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
11483 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
11485 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
11486 * config/aarch64/iterators.md (VQ2): New mode iterator.
11488 2018-06-20 Martin Liska <mliska@suse.cz>
11490 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11491 Change default ratio from 10 to 8.
11493 2018-06-20 Martin Liska <mliska@suse.cz>
11495 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11497 (bit_test_cluster::find_bit_tests): Likewise.
11498 (switch_decision_tree::analyze_switch_statement): Find clusters.
11499 * tree-switch-conversion.h (struct jump_table_cluster): Document
11502 2018-06-20 Martin Liska <mliska@suse.cz>
11504 * tree-switch-conversion.c (switch_conversion::collect):
11505 Record m_uniq property.
11506 (switch_conversion::expand): Bail out for special conditions.
11507 (group_cluster::~group_cluster): New.
11508 (group_cluster::group_cluster): Likewise.
11509 (group_cluster::dump): Likewise.
11510 (jump_table_cluster::emit): New.
11511 (switch_decision_tree::fix_phi_operands_for_edges): New.
11512 (struct case_node): Remove struct.
11513 (jump_table_cluster::can_be_handled): New.
11514 (case_values_threshold): Moved to header.
11515 (reset_out_edges_aux): Likewise.
11516 (jump_table_cluster::is_beneficial): New.
11517 (bit_test_cluster::can_be_handled): Likewise.
11518 (add_case_node): Remove.
11519 (bit_test_cluster::is_beneficial): New.
11520 (case_bit_test::cmp): New.
11521 (bit_test_cluster::emit): New.
11522 (expand_switch_as_decision_tree_p): Remove.
11523 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
11524 (fix_phi_operands_for_edge): Likewise.
11525 (switch_decision_tree::analyze_switch_statement): New.
11526 (compute_cases_per_edge): Move ...
11527 (switch_decision_tree::compute_cases_per_edge): ... here.
11528 (try_switch_expansion): Likewise.
11529 (switch_decision_tree::try_switch_expansion): Likewise.
11530 (record_phi_operand_mapping): Likewise.
11531 (switch_decision_tree::record_phi_operand_mapping): Likewise.
11532 (emit_case_decision_tree): Likewise.
11533 (switch_decision_tree::emit): Likewise.
11534 (balance_case_nodes): Likewise.
11535 (switch_decision_tree::balance_case_nodes): Likewise.
11536 (dump_case_nodes): Likewise.
11537 (switch_decision_tree::dump_case_nodes): Likewise.
11538 (emit_jump): Likewise.
11539 (switch_decision_tree::emit_jump): Likewise.
11540 (emit_cmp_and_jump_insns): Likewise.
11541 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
11542 (emit_case_nodes): Likewise.
11543 (switch_decision_tree::emit_case_nodes): Likewise.
11544 (conditional_probability): Remove.
11545 * tree-switch-conversion.h (enum cluster_type): New.
11547 (struct cluster): Likewise.
11548 (cluster::cluster): Likewise.
11549 (struct simple_cluster): Likewise.
11550 (simple_cluster::simple_cluster): Likewise.
11551 (struct group_cluster): Likewise.
11552 (struct jump_table_cluster): Likewise.
11553 (struct bit_test_cluster): Likewise.
11554 (struct min_cluster_item): Likewise.
11555 (struct case_tree_node): Likewise.
11556 (case_tree_node::case_tree_node): Likewise.
11557 (jump_table_cluster::case_values_threshold): Likewise.
11558 (struct case_bit_test): Likewise.
11559 (struct switch_decision_tree): Likewise.
11560 (struct switch_conversion): Likewise.
11561 (switch_decision_tree::reset_out_edges_aux): Likewise.
11563 2018-06-20 Martin Liska <mliska@suse.cz>
11565 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
11566 (hoist_edge_and_branch_if_true): Likewise.
11567 (expand_switch_using_bit_tests_p): Likewise.
11568 (struct case_bit_test): Likewise.
11569 (case_bit_test_cmp): Likewise.
11570 (emit_case_bit_tests): Likewise.
11571 (switch_conversion::switch_conversion): New class.
11572 (struct switch_conv_info): Remove old struct.
11573 (collect_switch_conv_info): More to ...
11574 (switch_conversion::collect): ... this.
11575 (check_range): Likewise.
11576 (switch_conversion::check_range): Likewise.
11577 (check_all_empty_except_final): Likewise.
11578 (switch_conversion::check_all_empty_except_final): Likewise.
11579 (check_final_bb): Likewise.
11580 (switch_conversion::check_final_bb): Likewise.
11581 (create_temp_arrays): Likewise.
11582 (switch_conversion::create_temp_arrays): Likewise.
11583 (free_temp_arrays): Likewise.
11584 (gather_default_values): Likewise.
11585 (switch_conversion::gather_default_values): Likewise.
11586 (build_constructors): Likewise.
11587 (switch_conversion::build_constructors): Likewise.
11588 (constructor_contains_same_values_p): Likewise.
11589 (switch_conversion::contains_same_values_p): Likewise.
11590 (array_value_type): Likewise.
11591 (switch_conversion::array_value_type): Likewise.
11592 (build_one_array): Likewise.
11593 (switch_conversion::build_one_array): Likewise.
11594 (build_arrays): Likewise.
11595 (switch_conversion::build_arrays): Likewise.
11596 (gen_def_assigns): Likewise.
11597 (switch_conversion::gen_def_assigns): Likewise.
11598 (prune_bbs): Likewise.
11599 (switch_conversion::prune_bbs): Likewise.
11600 (fix_phi_nodes): Likewise.
11601 (switch_conversion::fix_phi_nodes): Likewise.
11602 (gen_inbound_check): Likewise.
11603 (switch_conversion::gen_inbound_check): Likewise.
11604 (process_switch): Use the newly created class.
11605 (switch_conversion::expand): New.
11606 (switch_conversion::~switch_conversion): New.
11607 * tree-switch-conversion.h: New file.
11609 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11611 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
11612 tree-vect-patterns.c.
11613 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
11614 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
11615 (vect_recog_sad_pattern): Likewise.
11616 (vect_recog_widen_sum_pattern): Likewise.
11617 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
11618 (vect_recog_widen_shift_pattern): Remove the type_in argument.
11619 (vect_recog_rotate_pattern): Likewise.
11620 (vect_recog_mult_pattern): Likewise.
11621 (vect_recog_vector_vector_shift_pattern): Likewise.
11622 (vect_recog_divmod_pattern): Likewise.
11623 (vect_recog_mixed_size_cond_pattern): Likewise.
11624 (vect_recog_bool_pattern): Likewise.
11625 (vect_recog_mask_conversion_pattern): Likewise.
11626 (vect_try_gather_scatter_pattern): Likewise.
11627 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
11628 (vect_recog_over_widening_pattern): Likewise.
11629 (vect_recog_gather_scatter_pattern): Likewise.
11630 (vect_recog_func_ptr): Move from tree-vectorizer.h
11631 (vect_vect_recog_func_ptrs): Move further down the file.
11632 (vect_recog_func): Likewise. Remove the third argument.
11633 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
11634 (vect_pattern_recog_1): Expect the pattern function to do any
11635 necessary target tests. Also expect it to provide a vector type.
11636 Remove the type_in handling.
11638 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11640 * tree-vect-patterns.c (vect_pattern_detected): New function.
11641 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
11642 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
11643 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
11644 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
11645 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11646 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11647 (vect_recog_mask_conversion_pattern)
11648 (vect_try_gather_scatter_pattern): Likewise.
11650 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11652 * tree-vect-patterns.c (vect_get_internal_def): New function.
11653 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11654 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
11655 (search_type_for_mask_1): Use it.
11657 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11659 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11660 redundant WIDEN_SUM_EXPR handling.
11661 (vect_recog_sad_pattern): Likewise.
11663 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11665 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
11666 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
11667 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
11668 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
11669 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
11671 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11673 * tree-vect-stmts.c (vectorizable_call): Make sure that we
11674 use the stmt_vec_info of the original bb statement for the
11675 new zero assignment, even if the call is part of a pattern.
11677 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11679 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
11680 that the sequence is attached to the original statement rather
11681 than the pattern statement.
11682 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
11683 PATTERN_DEF_SEQ from the original statement rather than
11684 the main pattern statement.
11685 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
11686 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
11687 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
11689 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11691 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
11692 definition statements before the early exit for statements that aren't
11694 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
11696 (vect_transform_loop): ...here. Process pattern definition
11697 statements without first checking whether the main pattern
11698 statement is live or relevant.
11700 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11702 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
11703 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
11705 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
11707 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
11708 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
11709 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
11710 (expand_block_compare): Change select_block_compare_mode call.
11711 (expand_strncmp_align_check): Use new functions, fix comment.
11712 (emit_final_str_compare_gpr): New function.
11713 (expand_strn_compare): Refactor and clean up code.
11714 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
11716 2018-06-19 Tony Reix <tony.reix@atos.com>
11717 Damien Bergamini <damien.bergamini@atos.com>
11718 David Edelsohn <dje.gcc@gmail.com>
11720 * collect2.c (static_obj): New variable.
11721 (static_libs): New variable.
11722 (is_in_list): Uncomment declaration.
11723 (main): Track AIX libraries linked statically.
11724 (is_in_list): Uncomment definition.
11725 (scan_prog_file): Don't add AIX shared libraries initializer
11726 to constructor list if linking statically.
11728 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
11730 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
11732 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
11734 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
11736 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
11739 2018-06-19 Martin Liska <mliska@suse.cz>
11741 * config/i386/i386.c (ix86_can_inline_p): Do not use
11742 ipa_fn_summaries::get_create.
11743 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
11745 (devirtualization_time_bonus): Likewise.
11746 (ipcp_propagate_stage): Likewise.
11747 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11748 (edge_set_predicate): Likewise.
11749 (evaluate_conditions_for_known_args): Likewise.
11750 (evaluate_properties_for_edge): Likewise.
11751 (ipa_call_summary::reset): Tranform to ...
11752 (ipa_call_summary::~ipa_call_summary): ... this.
11753 (ipa_fn_summary::reset): Transform to ...
11754 (ipa_fn_summary::~ipa_fn_summary): ... this.
11755 (ipa_fn_summary_t::remove): Rename to ...
11756 (ipa_fn_summary_t::remove_callees): ... this.
11757 (ipa_fn_summary_t::duplicate): Use placement new
11758 instead of memory copy.
11759 (ipa_call_summary_t::duplicate): Likewise.
11760 (ipa_call_summary_t::remove): Remove.
11761 (dump_ipa_call_summary): Change get_create to get.
11762 (ipa_dump_fn_summary): Dump only when summary exists.
11763 (analyze_function_body): Use symbol_summary::get instead
11765 (compute_fn_summary): Likewise.
11766 (estimate_edge_devirt_benefit): Likewise.
11767 (estimate_edge_size_and_time): Likewise.
11768 (inline_update_callee_summaries): Likewise.
11769 (remap_edge_change_prob): Likewise.
11770 (remap_edge_summaries): Likewise.
11771 (ipa_merge_fn_summary_after_inlining): Likewise.
11772 (write_ipa_call_summary): Likewise.
11773 (ipa_fn_summary_write): Likewise.
11774 (ipa_free_fn_summary): Likewise.
11775 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11776 (struct ipa_call_summary): Likewise.
11777 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11779 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11780 (estimate_size_after_inlining): Likewise.
11781 (estimate_growth): Likewise.
11782 (growth_likely_positive): Likewise.
11783 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11784 (inline_call): Likewise.
11785 * ipa-inline.c (caller_growth_limits): Likewise.
11786 (can_inline_edge_p): Likewise.
11787 (can_inline_edge_by_limits_p): Likewise.
11788 (compute_uninlined_call_time): Likewise.
11789 (compute_inlined_call_time): Likewise.
11790 (want_inline_small_function_p): Likewise.
11791 (edge_badness): Likewise.
11792 (update_caller_keys): Likewise.
11793 (update_callee_keys): Likewise.
11794 (inline_small_functions): Likewise.
11795 (inline_to_all_callers_1): Likewise.
11796 (dump_overall_stats): Likewise.
11797 (early_inline_small_functions): Likewise.
11798 (early_inliner): Likewise.
11799 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11800 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11801 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11802 * ipa-split.c (execute_split_functions): Likewise.
11803 * symbol-summary.h: Likewise.
11804 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11806 2018-06-19 Richard Biener <rguenther@suse.de>
11808 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11809 (vectorize_loops): ... here. Fix dbgcnt handling.
11810 (try_vectorize_loop): Wrap try_vectorize_loop_1.
11812 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
11815 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11816 ieee128 argument takes up only one (vector) register, not two (floating
11819 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11821 * gimplify.c (gimplify_init_constructor): Really never clear for an
11822 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11824 2018-06-19 Richard Biener <rguenther@suse.de>
11826 PR tree-optimization/86179
11827 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11828 after failed recognition.
11830 2018-06-18 Martin Sebor <msebor@redhat.com>
11832 PR middle-end/85602
11833 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
11834 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
11835 Handle integer subtraction.
11836 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
11837 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
11839 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11841 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
11842 param from rtx to rtx_insn *.
11843 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
11845 (frv_ifcvt_modify_insn): Likwise.
11846 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
11847 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
11848 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
11849 as_a <rtx_insn *> cast to local "unprotected_region" once
11850 it's been established that it's not NULL or pc_rtx.
11851 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
11852 param "sethi" from rtx to rtx_insn *.
11853 (nds32_group_float_insns): Likewise for param "insn".
11854 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
11856 (vax_output_int_subtract): Likewise.
11857 * config/vax/vax.c (vax_output_int_add): Likewise for param
11859 (vax_output_int_subtract): Likewise.
11860 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
11861 (emit_pattern_after): Likewise for param "after".
11862 (emit_insn_after): Likewise.
11863 (emit_jump_insn_after): Likewise.
11864 (emit_call_insn_after): Likewise.
11865 (emit_debug_insn_after): Likewise.
11866 (emit_pattern_before): Likewise for param "before".
11867 (emit_insn_before): Likewise.
11868 (emit_jump_insn_before): Likewise.
11869 * final.c (get_insn_template): Likewise for param "insn", removing
11871 * output.h (get_insn_template): Likewise for 2nd param.
11872 * rtl.h (emit_insn_before): Likewise.
11873 (emit_jump_insn_before): Likewise.
11874 (emit_debug_insn_before_noloc): Likewise.
11875 (emit_insn_after): Likewise.
11876 (emit_jump_insn_after): Likewise.
11877 (emit_call_insn_after): Likewise.
11878 (emit_debug_insn_after): Likewise.
11879 (set_insn_deleted): Likewise for param.
11881 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
11884 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11885 floating point modes, so that IFmode is numerically greater than
11886 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11887 to declare the ordering. This prevents IFmode from being
11888 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11889 machine. Include rs6000-modes.h to share the fractional values
11890 between genmodes* and the rest of the compiler.
11891 (IFmode): Likewise.
11892 (KFmode): Likewise.
11893 (TFmode): Likewise.
11894 * config/rs6000/rs6000-modes.h: New file.
11895 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11896 meaning of rs6000_long_double_size so that 126..128 selects an
11897 appropriate 128-bit floating point type.
11898 (rs6000_option_override_internal): Likewise.
11899 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11900 (TARGET_LONG_DOUBLE_128): Change the meaning of
11901 rs6000_long_double_size so that 126..128 selects an appropriate
11902 128-bit floating point type.
11903 (LONG_DOUBLE_TYPE_SIZE): Update comment.
11904 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11905 source and destination to match the standard usage.
11906 (truncifkf2): Likewise.
11907 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11908 ISA 2.07 to use an explicit clobber, instead of passing in a
11910 (copysign<mode>3_soft): Likewise.
11912 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11914 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11915 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11916 (vect_slp_analyze_instance_dependence): Likewise.
11917 (vect_enhance_data_refs_alignment): Likewise.
11918 (vect_analyze_data_refs_alignment): Likewise.
11919 (vect_slp_analyze_and_verify_instance_alignment
11920 (vect_analyze_data_ref_accesses): Likewise.
11921 (vect_prune_runtime_alias_test_list): Likewise.
11922 (vect_analyze_data_refs): Likewise.
11923 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11924 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11925 (vect_analyze_scalar_cycles_1): Likewise.
11926 (vect_get_loop_niters): Likewise.
11927 (vect_analyze_loop_form_1): Likewise.
11928 (vect_update_vf_for_slp): Likewise.
11929 (vect_analyze_loop_operations): Likewise.
11930 (vect_analyze_loop): Likewise.
11931 (vectorizable_induction): Likewise.
11932 (vect_transform_loop): Likewise.
11933 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11934 * tree-vect-slp.c (vect_analyze_slp): Likewise.
11935 (vect_make_slp_decision): Likewise.
11936 (vect_detect_hybrid_slp): Likewise.
11937 (vect_slp_analyze_operations): Likewise.
11938 (vect_slp_bb): Likewise.
11939 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11940 (vectorizable_bswap): Likewise.
11941 (vectorizable_call): Likewise.
11942 (vectorizable_simd_clone_call): Likewise.
11943 (vectorizable_conversion): Likewise.
11944 (vectorizable_assignment): Likewise.
11945 (vectorizable_shift): Likewise.
11946 (vectorizable_operation): Likewise.
11947 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11949 2018-06-18 Martin Sebor <msebor@redhat.com>
11951 PR tree-optimization/81384
11952 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11953 * builtins.c (expand_builtin_strnlen): New function.
11954 (expand_builtin): Call it.
11955 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11956 * builtins.def (BUILT_IN_STRNLEN): New.
11957 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11958 Warn for bounds in excess of maximum object size.
11959 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11960 single-value ranges. Handle strnlen.
11961 (handle_builtin_strlen): Handle strnlen.
11962 (strlen_check_and_optimize_stmt): Same.
11963 * doc/extend.texi (Other Builtins): Document strnlen.
11965 2018-06-18 Maya Rashish <coypu@sdf.org>
11967 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11968 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11969 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11971 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11973 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11975 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11977 * tree.c (escaped_string::escape): Replace cast to char * by
11978 const_cast<char *> (unescaped).
11980 2018-06-18 Nick Clifton <nickc@redhat.com>
11983 * tree.c (escaped_string): New class. Converts an unescaped
11984 string into its escaped equivalent.
11985 (warn_deprecated_use): Use the new class to convert the
11986 deprecation message, if present.
11987 (test_escaped_strings): New self test.
11988 (test_c_tests): Add test_escaped_strings.
11989 * doc/extend.texi (deprecated): Add a note that the
11990 deprecation message is affected by the -fmessage-length
11991 option, and that control characters will be escaped.
11992 (#pragma GCC error): Document this pragma.
11993 (#pragma GCC warning): Likewise.
11994 * doc/invoke.texi (-fmessage-length): Document this option's
11995 effect on the #warning and #error preprocessor directives and
11996 the deprecated attribute.
11998 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
12000 * tree.c (decl_value_expr_lookup): Revert latest change.
12001 (decl_value_expr_insert): Likewise.
12003 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
12005 * gimplify.c (nonlocal_vlas): Delete.
12006 (nonlocal_vla_vars): Likewise.
12007 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12009 (gimplify_body): Do not create and destroy nonlocal_vlas.
12010 * tree-nested.c: Include diagnostic.h.
12011 (use_pointer_in_frame): Tweak.
12012 (lookup_field_for_decl): Add assertion and declare the transformation.
12013 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12014 internal error when the reference is in a wrong context. Do not
12015 create a debug decl by default.
12016 (note_nonlocal_block_vlas): Delete.
12017 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12018 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
12019 create a debug decl by default.
12020 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12021 call is in a wrong context.
12022 (fixup_vla_decls): New function.
12023 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
12024 debug variables were created.
12025 * tree.c (decl_value_expr_lookup): Add checking assertion.
12026 (decl_value_expr_insert): Likewise.
12028 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
12030 PR middle-end/82479
12031 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12032 * tree-scalar-evolution.c (interpret_expr): Likewise.
12033 (expression_expensive_p): Likewise.
12034 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12035 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12036 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12037 (ssa_defined_by_minus_one_stmt_p): New.
12039 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
12041 PR middle-end/64946
12042 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12043 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12044 * dojump.c (do_jump): Likewise.
12045 * expr.c (expand_expr_real_2): Check operand type's sign.
12046 * fold-const.c (const_unop): Handle ABSU_EXPR.
12047 (fold_abs_const): Likewise.
12048 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
12049 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12050 (strip_sign_op_1): Likesise.
12051 * match.pd: Add new pattern to generate ABSU_EXPR.
12052 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
12053 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12054 * tree-eh.c (operation_could_trap_helper_p): Likewise.
12055 * tree-inline.c (estimate_operator_cost): Likewise.
12056 * tree-pretty-print.c (dump_generic_node): Likewise.
12057 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
12058 * tree.def (ABSU_EXPR): New.
12060 2018-06-16 Jakub Jelinek <jakub@redhat.com>
12062 PR middle-end/86095
12063 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
12064 documented as preserved for backward compatibility only.
12065 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
12067 PR rtl-optimization/86108
12068 * bb-reorder.c (create_forwarder_block): Renamed to ...
12069 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
12070 jump from new landing pad to the second part.
12071 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
12074 2018-06-15 Jakub Jelinek <jakub@redhat.com>
12076 PR middle-end/85878
12077 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
12078 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
12079 Only call store_expr for halves if the mode is the same.
12081 PR middle-end/86123
12082 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
12083 Fix up comment formatting.
12085 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
12087 * typed-splay-tree.h (typed_splay_tree::remove): New function.
12088 (typed_splay_tree::closure,
12089 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
12090 (typed_splay_tree::typed_splay_tree,
12091 typed_splay_tree::operator =): Declared private.
12092 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
12093 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
12094 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
12095 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
12096 typed_splay_tree::splay_tree_splay,
12097 typed_splay_tree::splay_tree_foreach_helper,
12098 typed_splay_tree::splay_tree_insert,
12099 typed_splay_tree::splay_tree_remove,
12100 typed_splay_tree::splay_tree_lookup,
12101 typed_splay_tree::splay_tree_predecessor,
12102 typed_splay_tree::splay_tree_successor,
12103 typed_splay_tree::splay_tree_min,
12104 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
12105 (typed_splay_tree::root, typed_splay_tree::comp,
12106 typed_splay_tree::delete_key,
12107 typed_splay_tree::delete_value): New data members.
12108 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
12109 typed_splay_tree::remove.
12111 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
12113 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
12114 -mginv and -mno-ginv to the assembler.
12115 * config/mips/mips.opt (-mcrc): New option.
12116 (-mginv): Likewise.
12117 * doc/invoke.text (-mcrc): Document.
12118 (-mginv): Likewise.
12120 2018-06-15 Nick Clifton <nickc@redhat.com>
12123 * tree.c (escaped_string): New class. Converts an unescaped
12124 string into its escaped equivalent.
12125 (warn_deprecated_use): Use the new class to convert the
12126 deprecation message, if present.
12127 (test_escaped_strings): New self test.
12128 (test_c_tests): Add test_escaped_strings.
12129 * doc/extend.texi (deprecated): Add a note that the
12130 deprecation message is affected by the -fmessage-length
12131 option, and that control characters will be escaped.
12132 (#pragma GCC error): Document this pragma.
12133 (#pragma GCC warning): Likewise.
12134 * doc/invoke.texi (-fmessage-length): Document this option's
12135 effect on the #warning and #error preprocessor directives and
12136 the deprecated attribute.
12138 2018-06-15 Richard Biener <rguenther@suse.de>
12140 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
12141 here, also noting vector size used.
12142 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
12143 size used in MSG_OPTIMIZED_LOCATIONS dump.
12144 (pass_slp_vectorize::execute): Adjust.
12146 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
12149 * config/arc/arc.c (arc_return_address_register): Fix
12152 2018-06-15 Richard Biener <rguenther@suse.de>
12154 PR middle-end/86159
12155 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
12156 leave useless conversion stripping to force_gimple_operand_gsi.
12157 (gimplify_build2): Likewise.
12158 (gimplify_build1): Likewise.
12160 2018-06-15 Richard Biener <rguenther@suse.de>
12162 PR middle-end/86076
12163 * tree-cfg.c (move_stmt_op): unshare invariant addresses
12164 before adjusting their block.
12166 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
12168 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
12169 multilibs for *-*-rtems*.
12170 * config/riscv/t-rtems: New file.
12172 2018-06-14 Jakub Jelinek <jakub@redhat.com>
12174 PR middle-end/86122
12175 * match.pd ((A +- CST1) +- CST2): Punt if last resort
12176 unsigned_type_for returns NULL.
12179 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
12180 subregs of multi-word pseudos unless the float mode has word size.
12182 2018-06-14 Richard Biener <rguenther@suse.de>
12184 PR middle-end/86139
12185 * tree-vect-generic.c (build_word_mode_vector_type): Remove
12186 duplicate and harmful type_hash_canon.
12187 * tree.c (type_hash_canon): Assert we didn't find ourselves.
12189 2018-06-14 Richard Biener <rguenther@suse.de>
12192 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
12195 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
12197 * config/rtems.h (STDINT_LONG32): Define.
12199 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
12200 Prachi Godbole <prachi.godbole@imgtec.com>
12202 * config/mips/mips-cpus.def: Define P6600.
12203 * config/mips/mips-tables.opt: Regenerate.
12204 * config/mips/mips.c (mips_ucbranch_type): New enum.
12205 (mips_rtx_cost_data): Add support for P6600.
12206 (mips_issue_rate): Likewise.
12207 (mips_multipass_dfa_lookahead): Likewise.
12208 (mips_avoid_hazard): Likewise.
12209 (mips_reorg_process_insns): Likewise.
12210 (mips_classify_branch_p6600): New function.
12211 * config/mips/mips.h (TUNE_P6600): New define.
12212 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
12213 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
12214 * config/mips/mips.md: Include p6600.md.
12215 (processor): Add p6600.
12216 * config/mips/p6600.md: New file.
12217 * doc/invoke.texi: Add p6600 to supported architectures.
12219 2018-06-13 Martin Sebor <msebor@redhat.com>
12221 PR tree-optimization/86114
12222 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
12224 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
12226 2018-06-13 Richard Biener <rguenther@suse.de>
12228 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
12229 Properly set vector type of the intermediate stmt.
12230 * tree-vect-stmts.c (vectorizable_operation): The destination
12231 var always has vectype_out type.
12233 2018-06-13 Jeff Law <law@redhat.com>
12235 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
12236 integer 0 for argument to print_rtl_with_bb.
12237 (rl78_reorg): Likewise.
12239 2018-06-13 David Malcolm <dmalcolm@redhat.com>
12241 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
12242 from rtx to rtx_insn *.
12243 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
12245 (add_sched_insns_for_speculation): Likewise for local "target",
12246 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12247 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
12248 from rtx_insn ** to rtx_code_label **.
12249 (reorg_emit_nops): Likewise.
12250 (c6x_reorg): Likewise for local "call_labels".
12251 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
12253 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
12254 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
12255 the loops over LABEL_REFS.
12256 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
12258 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
12259 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
12260 (split_branches): Strengthen local "olabel" from rtx to
12261 rtx_insn *, adding a safe_as_a cast.
12262 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
12264 (add_insn_after): Likewise for first two params.
12265 (add_insn_before): Likewise.
12266 (remove_insn): Likewise for param.
12267 (emit_pattern_before_noloc): Likewise for second and third params.
12268 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
12269 (emit_call_insn_before_noloc): Likewise.
12270 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
12272 (emit_barrier_before): Likewise.
12273 (emit_label_before): Strengthen "label" param from "rtx" to
12274 "rtx_code_label *". Strengthen "before" param from "rtx" to
12276 (emit_insn_after_1): Strengthen "after" param from "rtx" to
12278 (emit_pattern_after_noloc): Likewise.
12279 (emit_insn_after_noloc): Likewise.
12280 (emit_jump_insn_after_noloc): Likewise.
12281 (emit_call_insn_after_noloc): Likewise.
12282 (emit_debug_insn_after_noloc): Likewise.
12283 (emit_barrier_after): Likewise.
12284 (emit_label_after): Likewise for both params.
12285 (emit_pattern_after_setloc): Likewise for "after" param. Convert
12286 "loc" param from "int" to "location_t".
12287 (emit_insn_after_setloc): Likewise.
12288 (emit_jump_insn_after_setloc): Likewise.
12289 (emit_call_insn_after_setloc): Likewise.
12290 (emit_debug_insn_after_setloc): Likewise.
12291 (emit_pattern_before_setloc): Likewise for "before" param. Convert
12292 "loc" param from "int" to "location_t".
12293 (emit_pattern_before): Convert NULL_RTX to NULL.
12294 (emit_insn_before_setloc): Convert "loc" param from "int" to
12296 (emit_jump_insn_before_setloc): Likewise.
12297 (emit_call_insn_before_setloc): Likewise.
12298 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
12299 rtx_insn *. Convert "loc" param from "int" to "location_t".
12300 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
12301 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
12302 Convert 3rd param from "int" to "location_t".
12303 (emit_barrier_before, emit_barrier_after, next_real_insn):
12304 Strengthen param from rtx to rtx_insn *.
12305 (emit_label_before): Strengthen 1st param from "rtx" to
12306 "rtx_code_label *". Strengthen 2nd param from "rtx" to
12308 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
12309 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
12310 Strengthen 2nd param from "rtx" to "rtx_insn *".
12311 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
12312 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
12313 Likewise. Convert 3rd param from "int" to "location_t".
12314 (emit_label_after): Strengthen 1st param from "rtx" to
12315 "rtx_code_label *".
12316 (next_real_insn, remove_insn): Strengthen param from "rtx" to
12318 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
12319 from "rtx" to "rtx_insn *".
12321 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
12323 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
12324 bodies streamed in with -Q.
12325 * dumpfile.c (dump_files): Add lto-stream-out dump file.
12326 * dumpfile.h (tree_dump_index): Add lto_stream_out.
12327 * gimple-streamer-out.c: Include gimple-pretty-print.h
12328 (output_bb): Dump stmts streamed.
12329 * lto-section-out.c: Include print-tree.h
12330 (lto_begin_section): Dump sections created.
12331 (lto_output_decl_index): Dump decl encoded.
12332 * lto-streamer-out.c: Include print-tree.h
12333 (create_output_block): Dump output block created.
12334 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
12335 (output_function): Dump function output.
12336 (output_constructor): Dump constructor streamed.
12337 (write_global_stream): Output indexes encoded.
12338 (produce_asm_for_decls): Dump streams encoded.
12339 * lto-streamer.c (streamer_dump_file): New global var.
12340 * lto-streamer.h (streamer_dump_file): Declare.
12341 * passes.c (ipa_write_summaries): Initialize streamer dump.
12342 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
12345 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
12348 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
12349 offsets for register save directives. Emit a second batch of save
12350 directives, if need be, when the function accesses prior frames.
12352 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12354 * config/arc/fpu.md (fmasf4): Force operand to register.
12355 (fnmasf4): Likewise.
12357 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12359 * config/arc/arc-protos.h (arc_pad_return): Remove.
12360 * config/arc/arc.c (machine_function): Remove force_short_suffix
12362 (arc_print_operand): Adjust printing of '&'.
12363 (arc_verify_short): Remove conditional printing of short suffix.
12364 (arc_final_prescan_insn): Remove reference to size_reason.
12365 (pad_return): New function.
12366 (arc_reorg): Call pad_return.
12367 (arc_pad_return): Remove.
12368 (arc_init_machine_status): Remove reference to force_short_suffix.
12369 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
12370 (attr length): When attribute iscompact is true force to 2
12371 regardless; in the case of maybe check if we want to force the
12372 instruction to have 4 bytes length.
12373 (nopv): Change it to generate 4 byte long nop as well.
12374 (blockage): New pattern.
12375 (simple_return): Remove call to arc_pad_return.
12376 (p_return_i): Likewise.
12378 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12380 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
12382 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12384 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
12387 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12389 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
12390 for ARC700 and ARCv2.
12392 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
12395 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
12396 operands[2] instead of operands[1].
12399 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
12401 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
12402 case, check whether the outer register overlaps an unallocatable
12403 register, not just whether it fits the required class.
12405 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
12407 * poly-int.h (can_div_trunc_p): Add new overload in which all values
12409 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
12410 (memrefs_conflict_p): Likewise.
12411 (init_alias_analysis): Likewise.
12412 * cfgexpand.c (expand_debug_expr): Likewise.
12413 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
12414 * cse.c (fold_rtx): Likewise.
12415 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
12416 * expr.c (emit_block_move_hints): Likewise.
12417 (clear_storage_hints, push_block, emit_push_insn): Likewise.
12418 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
12419 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
12420 (emit_group_store): Likewise.
12421 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
12422 to read the PRE/POST_MODIFY increment.
12423 * calls.c (store_one_arg): Use strip_offset.
12424 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
12426 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
12428 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
12429 (simplify_binary_operation_1): Extend CONST_INT handling to
12431 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
12432 than a HOST_WIDE_INT.
12433 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
12435 (adjust_mems, add_stores): Update accodingly.
12436 (vt_canonicalize_addr): Track polynomial offsets.
12437 (emit_note_insn_var_location): Likewise.
12438 (vt_add_function_parameter): Likewise.
12439 (vt_initialize): Likewise.
12441 2018-06-12 Jeff Law <law@redhat.com>
12443 * config.gcc (alpha*-*-freebsd*): Remove.
12444 * config/alpha/freebsd.h: Remove.
12446 2018-06-12 David Malcolm <dmalcolm@redhat.com>
12449 * spellcheck-tree.c (levenshtein_distance): Rename to...
12450 (get_edit_distance): ...this, and update for underlying renaming.
12451 * spellcheck-tree.h (levenshtein_distance): Rename to...
12452 (get_edit_distance): ...this.
12453 * spellcheck.c (levenshtein_distance): Rename to...
12454 (get_edit_distance): ...this. Convert from Levenshtein distance
12455 to Damerau-Levenshtein distance by supporting transpositions of
12456 adjacent characters. Rename "v1" to "v_next" and "v0" to
12458 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
12459 (selftest::test_edit_distance_unit_test_oneway): ...this, and
12460 update for underlying renaming.
12461 (selftest::levenshtein_distance_unit_test): Rename to...
12462 (selftest::test_get_edit_distance_unit): ...this, and update for
12463 underlying renaming.
12464 (selftest::test_find_closest_string): Add example from PR 69968
12465 where transposition helps
12466 (selftest::test_metric_conditions): Update for renaming.
12467 (selftest::test_metric_conditions): Likewise.
12468 (selftest::spellcheck_c_tests): Likewise.
12469 * spellcheck.h (levenshtein_distance): Rename both overloads to...
12470 (get_edit_distance): ...this.
12471 (best_match::consider): Update for renaming.
12473 2018-06-12 Martin Sebor <msebor@redhat.com>
12475 PR tree-optimization/85259
12476 * builtins.c (compute_objsize): Handle constant offsets.
12477 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
12478 true iff a warning has been issued.
12479 * gimple.h (gimple_nonartificial_location): New function.
12480 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
12481 gimple_nonartificial_location and handle -Wno-system-headers.
12482 (handle_builtin_stxncpy): Same.
12484 2018-06-12 Martin Sebor <msebor@redhat.com>
12487 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
12489 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
12491 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
12492 BUILTIN_VEC_XST entries for pointer to double and long long.
12494 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
12497 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12499 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
12502 2018-06-12 Martin Liska <mliska@suse.cz>
12504 * doc/options.texi: Document IntegerRange.
12506 2018-06-12 Martin Liska <mliska@suse.cz>
12508 * config/i386/i386.opt: Make MPX-related options as Deprecated.
12509 * opt-functions.awk: Handle Deprecated flag.
12510 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
12512 (read_cmdline_option): Report warning for a deprecated option.
12513 * opts.h (struct cl_option): Add new field cl_deprecated.
12514 (CL_ERR_DEPRECATED): New.
12516 2018-06-12 Martin Liska <mliska@suse.cz>
12518 * doc/options.texi: Document Deprecated option flag.
12520 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12522 * config/arc/arc-arch.h (arc_extras): New enum.
12523 (arc_cpu_t):Add field extra.
12524 (arc_cpu_types): Consider the extras.
12525 * config/arc/arc-cpus.def: Add extras info.
12526 * config/arc/arc-opts.h (processor_type): Consider extra field.
12527 * config/arc/arc.c (arc_override_options): Handle extra field.
12529 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
12531 * config/arc/arc-arch.h: Update ARC_OPTX macro.
12532 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
12534 * config/arc/arc.c (arc_init): Update pic warning.
12535 (irq_range): Update irq range parsing warnings.
12536 (arc_override_options): Update various warning messages.
12537 (arc_handle_aux_attribute): Likewise.
12539 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
12541 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
12543 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12545 * doc/sourcebuild.texi: Document usage of line number 0 in verify
12546 compiler messages directives.
12548 2018-06-12 Matthew Fortune <mfortune@gmail.com>
12550 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
12551 * config/mips/mips-tables.opt: Regenerate.
12552 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
12554 * doc/invoke.texi: Document -march=i6500.
12556 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
12558 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
12559 (i6400_gpmul): Add cpu_unit.
12560 (i6400_gpdiv): Likewise.
12561 (i6400_msa_add_d): Update reservations.
12562 (i6400_msa_int_add) Likewise.
12563 (i6400_msa_short_logic3) Likewise.
12564 (i6400_msa_short_logic2) Likewise.
12565 (i6400_msa_short_logic) Likewise.
12566 (i6400_msa_move) Likewise.
12567 (i6400_msa_cmp) Likewise.
12568 (i6400_msa_short_float2) Likewise.
12569 (i6400_msa_div_d) Likewise.
12570 (i6400_msa_long_logic1) Likewise.
12571 (i6400_msa_long_logic2) Likewise.
12572 (i6400_msa_mult) Likewise.
12573 (i6400_msa_long_float2) Likewise.
12574 (i6400_msa_long_float4) Likewise.
12575 (i6400_msa_long_float5) Likewise.
12576 (i6400_msa_long_float8) Likewise.
12577 (i6400_fpu_fadd): Include frint type.
12578 (i6400_fpu_store): New define_insn_reservation.
12579 (i6400_fpu_load): Likewise.
12580 (i6400_fpu_move): Likewise.
12581 (i6400_fpu_fcmp): Likewise.
12582 (i6400_fpu_fmadd): Likewise.
12583 (i6400_int_mult): Include imul3nc type and update reservation.
12584 (i6400_int_div): Include idiv3 type and update reservation.
12585 (i6400_int_load): Update to check type not move_type.
12586 (i6400_int_store): Likewise.
12587 (i6400_int_prefetch): Set zero latency.
12589 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
12591 * gcc.c: Document new %@{...} sequence.
12592 (LINK_COMMAND_SPEC): Use it for the -L switches.
12593 (cpp_unique_options): Use it for the -I switches.
12594 (at_file_argbuf): New global variable.
12595 (in_at_file): Likewise.
12596 (alloc_args): Create at_file_argbuf.
12597 (clear_args): Truncate at_file_argbuf.
12598 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
12599 (open_at_file): New function.
12600 (close_at_file): Likewise.
12601 (create_at_file): Delete.
12602 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
12605 (validate_switches_from_spec): Deal with %@{...} sequence.
12606 (validate_switches): Likewise.
12607 (driver::finalize): Call clear_args.
12609 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
12611 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
12613 2018-06-11 Martin Sebor <msebor@redhat.com>
12615 * doc/invoke.texi (-Wall): List -Wc++17-compat.
12616 (Wno-class-memaccess): Add @opindex.
12617 (Wno-templates, Wno-multiple-inheritance): Same.
12618 (Wno-virtual-inheritance, Wno-namespaces): Same.
12619 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
12620 (Wno-format-overflow, Wno-format-truncation): Same.
12621 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
12622 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
12623 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
12624 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
12625 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
12626 (Wno-misspelled-isr): Same.
12628 2018-06-11 Martin Sebor <msebor@redhat.com>
12630 * PR tree-optimization/86083
12631 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
12633 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
12635 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
12637 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
12640 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
12641 on the correct operand.
12642 (*movdi_internal64): Ditto.
12644 2018-06-11 Martin Liska <mliska@suse.cz>
12646 PR tree-optimization/86089
12647 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
12649 2018-06-11 Julia Koval <julia.koval@intel.com>
12651 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
12652 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
12653 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
12655 2018-06-11 Olivier Hainque <hainque@adacore.com>
12657 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
12658 for Ada with strict dwarf2.
12660 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
12663 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
12666 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
12668 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
12670 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
12672 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
12675 2018-06-08 Martin Liska <mliska@suse.cz>
12677 * tree-cfg.h (debug_function): Fix argument type to match
12680 2018-06-08 Martin Liska <mliska@suse.cz>
12682 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
12683 Remove usage of MPX-related (and removed) fields.
12684 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
12686 2018-06-08 David Malcolm <dmalcolm@redhat.com>
12688 * cfg.c (debug): Use TDF_NONE rather than 0.
12689 * cfghooks.c (debug): Likewise.
12690 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
12691 (struct dump_option_value_info): Convert to...
12692 (struct kv_pair): ...this template type.
12693 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
12695 (optinfo_verbosity_options): Likewise.
12696 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
12698 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
12699 than int for "optgroup_flags" param.
12700 (dump_generic_expr_loc): Use dump_flags_t rather than int for
12702 (dump_dec): Likewise.
12703 (dump_finish): Use TDF_NONE rather than 0.
12704 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
12705 rather than int for "optgroup_flags" param. Use TDF_NONE rather
12706 than 0. Update for change to option_ptr.
12707 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
12708 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
12709 0. Update for changes to optinfo_verbosity_options and
12711 (opt_info_switch_p): Convert optgroup_flags from int to
12713 (dump_basic_block): Use dump_flags_t rather than int
12714 for "dump_kind" param.
12715 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
12716 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
12717 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
12718 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
12719 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
12720 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
12721 TDF_NONE): Convert from macros to...
12722 (enum dump_flag): ...this new enum.
12723 (dump_flags_t): Update to use enum.
12724 (operator|, operator&, operator~, operator|=, operator&=):
12725 Implement for dump_flags_t.
12726 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
12727 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
12728 Convert from macros to...
12729 (enum optgroup_flag): ...this new enum.
12730 (optgroup_flags_t): New typedef.
12731 (operator|, operator|=): Implement for optgroup_flags_t.
12732 (struct dump_file_info): Convert field "alt_flags" to
12733 dump_flags_t. Convert field "optgroup_flags" to
12735 (dump_basic_block): Use dump_flags_t rather than int for param.
12736 (dump_generic_expr_loc): Likewise.
12737 (dump_dec): Likewise.
12738 (dump_register): Convert param "optgroup_flags" to
12740 (opt_info_enable_passes): Likewise.
12741 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
12743 * gimple-pretty-print.c (debug): Likewise.
12744 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
12745 (merged_store_group::apply_stores): Likewise.
12746 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12747 * gimple.c (verify_gimple_pp): Likewise.
12748 * graphite-poly.c (print_pbb_body): Likewise.
12749 * passes.c (pass_manager::register_one_dump_file): Convert
12750 local "optgroup_flags" to optgroup_flags_t.
12751 * print-tree.c (print_node): Use TDF_NONE rather than 0.
12753 (debug_body): Likewise.
12754 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12755 to optgroup_flags_t.
12756 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12758 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12759 (convert_mult_to_fma): Likewise.
12760 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12761 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12762 * tree-vect-data-refs.c (dump_lower_bound): Convert param
12763 "dump_kind" to dump_flags_t.
12765 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12767 * config/rs6000/rs6000.c (min, max): Delete.
12769 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12771 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12772 -mabi=spe and -mabi=no-spe.
12774 2018-06-08 Martin Liska <mliska@suse.cz>
12776 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12777 where we expect an existing summary.
12779 2018-06-08 Martin Liska <mliska@suse.cz>
12781 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12782 * ipa-inline.h (estimate_edge_growth): Likewise.
12784 2018-06-08 Martin Liska <mliska@suse.cz>
12786 * cgraph.c (function_version_hasher::hash): Use
12787 cgraph_node::get_uid ().
12788 (function_version_hasher::equal):
12789 * cgraph.h (cgraph_node::get_uid): New method.
12790 * ipa-inline.c (update_caller_keys): Use
12791 cgraph_node::get_uid ().
12792 (update_callee_keys): Likewise.
12793 * ipa-utils.c (searchc): Likewise.
12794 (ipa_reduced_postorder): Likewise.
12795 * lto-cgraph.c (input_node): Likewise.
12796 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12797 * symbol-summary.h (symtab_insertion): Likewise.
12798 (symtab_removal): Likewise.
12799 (symtab_duplication): Likewise.
12800 * tree-pretty-print.c (dump_function_header): Likewise.
12801 * tree-sra.c (convert_callers_for_node): Likewise.
12803 2018-06-08 Martin Liska <mliska@suse.cz>
12805 * cgraph.c (symbol_table::create_edge): Always assign a new
12807 (symbol_table::free_edge): Do not recycle numbers.
12808 * cgraph.h (cgraph_edge::get): New method.
12809 * symbol-summary.h (symtab_removal): Use it.
12810 (symtab_duplication): Likewise.
12811 (call_summary::hashable_uid): Remove.
12813 2018-06-08 Martin Liska <mliska@suse.cz>
12815 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12816 (initialize_growth_caches): Remove.
12817 (free_growth_caches): Likewise.
12818 (do_estimate_edge_time): Use edge_growth_cache.
12819 (do_estimate_edge_size): Likewise.
12820 (do_estimate_edge_hints): Likewise.
12821 * ipa-inline.c (reset_edge_caches): Likewise.
12822 (recursive_inlining): Likewise.
12823 (inline_small_functions): Likewise.
12824 * ipa-inline.h (initialize_growth_caches): Remove.
12825 (estimate_edge_size): Likewise.
12826 (estimate_edge_time): Likewise.
12827 (estimate_edge_hints): Likewise.
12828 (reset_edge_growth_cache): Likewise.
12829 * symbol-summary.h (call_summary::remove): New method.
12831 2018-06-08 Martin Liska <mliska@suse.cz>
12833 * ipa-cp.c (class edge_clone_summary): New summary.
12834 (grow_edge_clone_vectors): Remove.
12835 (ipcp_edge_duplication_hook): Remove.
12836 (class edge_clone_summary_t): New call_summary class.
12837 (ipcp_edge_removal_hook): Remove.
12838 (edge_clone_summary_t::duplicate): New function.
12839 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
12840 (create_specialized_node): Likewise.
12841 (ipcp_driver): Initialize edge_clone_summaries and do not
12844 2018-06-08 Martin Liska <mliska@suse.cz>
12846 * symbol-summary.h (get): New function.
12847 (call_summary::m_initialize_when_cloning): New class member.
12849 2018-06-08 Martin Liska <mliska@suse.cz>
12851 * cgraph.c (cgraph_node::remove): Do not recycle uid.
12852 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
12853 (symbol_table::allocate_cgraph_symbol): Do not set uid.
12854 * passes.c (uid_hash_t): Record removed_nodes by their uids.
12855 (remove_cgraph_node_from_order): Use the removed_nodes set.
12856 (do_per_function_toporder): Likwise.
12857 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
12858 instead of summary_uid.
12859 (symtab_removal): Likewise.
12860 (symtab_duplication): Likewise.
12862 2018-06-08 Martin Liska <mliska@suse.cz>
12864 * ipa-cp.c (ipcp_store_bits_results): Use
12865 ipcp_transformation_sum.
12866 (ipcp_store_vr_results): Likewise.
12867 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12869 (ipcp_transformation_initialize): ... this.
12870 (ipa_set_node_agg_value_chain):
12871 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12872 (write_ipcp_transformation_info): Likewise.
12873 (read_ipcp_transformation_info): Likewise.
12874 (ipcp_update_bits): Likewise.
12875 (ipcp_update_vr): Likewise.
12876 (ipcp_transform_function): Likewise.
12877 * ipa-prop.h: Rename ipcp_transformation_summary to
12878 ipcp_transformation.
12879 (class ipcp_transformation_t): New function summary.
12880 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12881 (ipa_get_agg_replacements_for_node): Likewise.
12883 2018-06-08 Martin Liska <mliska@suse.cz>
12885 * ipa-pure-const.c (struct funct_state_d): Do it class instead
12887 (class funct_state_summary_t): New function_summary class.
12888 (has_function_state): Remove.
12889 (get_function_state): Likewise.
12890 (set_function_state): Likewise.
12891 (add_new_function): Likewise.
12892 (funct_state_summary_t::insert): New function.
12893 (duplicate_node_data): Remove.
12894 (remove_node_data): Remove.
12895 (funct_state_summary_t::duplicate): New function.
12896 (register_hooks): Create new funct_state_summaries.
12897 (pure_const_generate_summary): Use it.
12898 (pure_const_write_summary): Likewise.
12899 (pure_const_read_summary): Likewise.
12900 (propagate_pure_const): Likewise.
12901 (propagate_nothrow): Likewise.
12902 (dump_malloc_lattice): Likewise.
12903 (propagate_malloc): Likewise.
12904 (execute): Do not register hooks, just remove summary
12906 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12909 2018-06-08 Martin Liska <mliska@suse.cz>
12911 * ipa-reference.c (remove_node_data): Remove.
12912 (duplicate_node_data): Likewise.
12913 (class ipa_ref_var_info_summary_t): New class.
12914 (class ipa_ref_opt_summary_t): Likewise.
12915 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12916 (get_reference_optimization_summary): Use
12917 ipa_ref_opt_sum_summaries.
12918 (set_reference_vars_info): Remove.
12919 (set_reference_optimization_summary): Likewise.
12920 (ipa_init): Create summaries.
12921 (init_function_info): Use function summary.
12922 (ipa_ref_opt_summary_t::duplicate): New function.
12923 (ipa_ref_opt_summary_t::remove): New function.
12924 (get_read_write_all_from_node): Fix GNU coding style.
12925 (propagate): Use function summary.
12926 (write_node_summary_p): Fix GNU coding style.
12927 (stream_out_bitmap): Likewise.
12928 (ipa_reference_read_optimization_summary): Use function summary.
12929 (ipa_reference_c_finalize): Do not release hooks.
12931 2018-06-08 Martin Liska <mliska@suse.cz>
12933 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12934 (analyze_function_body): Extract multiple calls of get_create.
12935 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12936 * ipa-inline.c (recursive_inlining): Use ::get method.
12937 * ipa-inline.h (estimate_edge_growth): Likewise.
12939 2018-06-08 Martin Liska <mliska@suse.cz>
12941 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12943 (hsa_function_summary::hsa_function_summary): Use the new enum
12945 (hsa_gpu_implementation_p): Use hsa_summaries::get.
12946 * hsa-gen.c (hsa_get_host_function): Likewise.
12947 (get_brig_function_name): Likewise.
12948 * ipa-hsa.c (process_hsa_functions): Likewise.
12949 (ipa_hsa_write_summary): Likewise.
12950 * symbol-summary.h (symtab_duplication): Use ::get function/
12951 (get): New function.
12953 2018-06-08 Martin Liska <mliska@suse.cz>
12955 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12957 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12958 (hsa_register_kernel): Likewise.
12959 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12960 * hsa-gen.c (hsa_get_host_function): Likewise.
12961 (get_brig_function_name): Likewise.
12962 (generate_hsa): Likewise.
12963 (pass_gen_hsail::execute): Likewise.
12964 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12965 (devirtualization_time_bonus): Likewise.
12966 (ipcp_propagate_stage): Likewise.
12967 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12968 (edge_set_predicate): Likewise.
12969 (evaluate_conditions_for_known_args): Likewise.
12970 (evaluate_properties_for_edge): Likewise.
12971 (ipa_fn_summary::reset): Likewise.
12972 (ipa_fn_summary_t::duplicate): Likewise.
12973 (dump_ipa_call_summary): Likewise.
12974 (ipa_dump_fn_summary): Likewise.
12975 (analyze_function_body): Likewise.
12976 (compute_fn_summary): Likewise.
12977 (estimate_edge_devirt_benefit): Likewise.
12978 (estimate_edge_size_and_time): Likewise.
12979 (estimate_calls_size_and_time): Likewise.
12980 (estimate_node_size_and_time): Likewise.
12981 (inline_update_callee_summaries): Likewise.
12982 (remap_edge_change_prob): Likewise.
12983 (remap_edge_summaries): Likewise.
12984 (ipa_merge_fn_summary_after_inlining): Likewise.
12985 (ipa_update_overall_fn_summary): Likewise.
12986 (read_ipa_call_summary): Likewise.
12987 (inline_read_section): Likewise.
12988 (write_ipa_call_summary): Likewise.
12989 (ipa_fn_summary_write): Likewise.
12990 (ipa_free_fn_summary): Likewise.
12991 * ipa-hsa.c (process_hsa_functions): Likewise.
12992 (ipa_hsa_write_summary): Likewise.
12993 (ipa_hsa_read_section): Likewise.
12994 * ipa-icf.c (sem_function::merge): Likewise.
12995 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12996 (do_estimate_edge_time): Likewise.
12997 (estimate_size_after_inlining): Likewise.
12998 (estimate_growth): Likewise.
12999 (growth_likely_positive): Likewise.
13000 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13001 (inline_call): Likewise.
13002 * ipa-inline.c (caller_growth_limits): Likewise.
13003 (can_inline_edge_p): Likewise.
13004 (can_inline_edge_by_limits_p): Likewise.
13005 (compute_uninlined_call_time): Likewise.
13006 (compute_inlined_call_time): Likewise.
13007 (want_inline_small_function_p): Likewise.
13008 (edge_badness): Likewise.
13009 (update_caller_keys): Likewise.
13010 (update_callee_keys): Likewise.
13011 (recursive_inlining): Likewise.
13012 (inline_small_functions): Likewise.
13013 (inline_to_all_callers_1): Likewise.
13014 (dump_overall_stats): Likewise.
13015 (early_inline_small_functions): Likewise.
13016 (early_inliner): Likewise.
13017 * ipa-inline.h (estimate_edge_growth): Likewise.
13018 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13019 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13020 * ipa-prop.h (IPA_NODE_REF): Likewise.
13021 (IPA_EDGE_REF): Likewise.
13022 * ipa-pure-const.c (malloc_candidate_p): Likewise.
13023 (propagate_malloc): Likewise.
13024 * ipa-split.c (execute_split_functions): Likewise.
13025 * symbol-summary.h: Rename get to get_create.
13027 (get_create): Likewise.
13028 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13030 2018-06-08 Martin Liska <mliska@suse.cz>
13032 * symbol-summary.h (release): Move definition out of class
13034 (symtab_removal): Likewise.
13035 (symtab_duplication): Likewise.
13037 2018-06-08 Martin Liska <mliska@suse.cz>
13039 * symbol-summary.h (function_summary): Move constructor
13040 implementation out of class declaration.
13041 (release): Likewise.
13042 (symtab_insertion): Likewise.
13043 (symtab_removal): Likewise.
13044 (symtab_duplication): Likewise.
13047 2018-06-08 Martin Liska <mliska@suse.cz>
13049 * Makefile.in: Remove support for MPX (macros, related functions,
13050 fields in cgraph_node, ...).
13051 * builtin-types.def (BT_BND): Likewise.
13052 (BT_FN_BND_CONST_PTR): Likewise.
13053 (BT_FN_CONST_PTR_BND): Likewise.
13054 (BT_FN_VOID_PTR_BND): Likewise.
13055 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
13056 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
13057 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
13058 (expand_builtin_mempcpy_with_bounds): Likewise.
13059 (expand_builtin_memset_with_bounds): Likewise.
13060 (expand_builtin_memset_args): Likewise.
13061 (std_expand_builtin_va_start): Likewise.
13062 (expand_builtin): Likewise.
13063 (expand_builtin_with_bounds): Likewise.
13064 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
13065 (DEF_LIB_BUILTIN_CHKP): Likewise.
13066 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
13067 (DEF_CHKP_BUILTIN): Likewise.
13068 (BUILT_IN_MEMCPY): Likewise.
13069 (BUILT_IN_MEMMOVE): Likewise.
13070 (BUILT_IN_MEMPCPY): Likewise.
13071 (BUILT_IN_MEMSET): Likewise.
13072 (BUILT_IN_STPCPY): Likewise.
13073 (BUILT_IN_STRCAT): Likewise.
13074 (BUILT_IN_STRCHR): Likewise.
13075 (BUILT_IN_STRCPY): Likewise.
13076 (BUILT_IN_STRLEN): Likewise.
13077 (BUILT_IN_MEMCPY_CHK): Likewise.
13078 (BUILT_IN_MEMMOVE_CHK): Likewise.
13079 (BUILT_IN_MEMPCPY_CHK): Likewise.
13080 (BUILT_IN_MEMSET_CHK): Likewise.
13081 (BUILT_IN_STPCPY_CHK): Likewise.
13082 (BUILT_IN_STRCAT_CHK): Likewise.
13083 (BUILT_IN_STRCPY_CHK): Likewise.
13084 * calls.c (store_bounds): Likewise.
13085 (emit_call_1): Likewise.
13086 (special_function_p): Likewise.
13087 (maybe_warn_nonstring_arg): Likewise.
13088 (initialize_argument_information): Likewise.
13089 (finalize_must_preallocate): Likewise.
13090 (compute_argument_addresses): Likewise.
13091 (expand_call): Likewise.
13092 * cfgexpand.c (expand_call_stmt): Likewise.
13093 (expand_return): Likewise.
13094 (expand_gimple_stmt_1): Likewise.
13095 (pass_expand::execute): Likewise.
13096 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
13097 (cgraph_node::remove): Likewise.
13098 (cgraph_node::dump): Likewise.
13099 (cgraph_node::verify_node): Likewise.
13100 * cgraph.h (chkp_function_instrumented_p): Likewise.
13101 (symtab_node::get_alias_target): Likewise.
13102 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
13103 (cgraph_local_p): Likewise.
13104 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
13105 (cgraph_edge::rebuild_references): Likewise.
13106 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
13107 (walk_polymorphic_call_targets): Likewise.
13108 (cgraph_node::expand_thunk): Likewise.
13109 (symbol_table::output_weakrefs): Likewise.
13110 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
13111 (ix86_handle_option): Likewise.
13112 * config/i386/constraints.md: Likewise.
13113 * config/i386/i386-builtin-types.def (BND): Likewise.
13117 * config/i386/i386-builtin.def (BDESC_END): Likewise.
13118 (BDESC_FIRST): Likewise.
13120 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
13121 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
13122 * config/i386/i386.c (enum reg_class): Likewise.
13123 (ix86_target_string): Likewise.
13124 (ix86_option_override_internal): Likewise.
13125 (ix86_conditional_register_usage): Likewise.
13126 (ix86_valid_target_attribute_inner_p): Likewise.
13127 (ix86_set_indirect_branch_type): Likewise.
13128 (ix86_set_current_function): Likewise.
13129 (ix86_function_arg_regno_p): Likewise.
13130 (init_cumulative_args): Likewise.
13131 (ix86_function_arg_advance): Likewise.
13132 (ix86_function_arg): Likewise.
13133 (ix86_pass_by_reference): Likewise.
13134 (ix86_function_value_regno_p): Likewise.
13135 (ix86_function_value_1): Likewise.
13136 (ix86_function_value_bounds): Likewise.
13137 (ix86_return_in_memory): Likewise.
13138 (ix86_setup_incoming_vararg_bounds): Likewise.
13139 (ix86_va_start): Likewise.
13140 (indirect_thunk_need_prefix): Likewise.
13141 (print_reg): Likewise.
13142 (ix86_print_operand): Likewise.
13143 (ix86_expand_call): Likewise.
13144 (ix86_output_function_return): Likewise.
13145 (reg_encoded_number): Likewise.
13146 (BDESC_VERIFYS): Likewise.
13147 (ix86_init_mpx_builtins): Likewise.
13148 (ix86_init_builtins): Likewise.
13149 (ix86_emit_cmove): Likewise.
13150 (ix86_emit_move_max): Likewise.
13151 (ix86_expand_builtin): Likewise.
13152 (ix86_builtin_mpx_function): Likewise.
13153 (ix86_get_arg_address_for_bt): Likewise.
13154 (ix86_load_bounds): Likewise.
13155 (ix86_store_bounds): Likewise.
13156 (ix86_load_returned_bounds): Likewise.
13157 (ix86_store_returned_bounds): Likewise.
13158 (ix86_class_likely_spilled_p): Likewise.
13159 (ix86_hard_regno_mode_ok): Likewise.
13160 (x86_order_regs_for_local_alloc): Likewise.
13161 (ix86_mitigate_rop): Likewise.
13162 (ix86_bnd_prefixed_insn_p): Likewise.
13163 (ix86_mpx_bound_mode): Likewise.
13164 (ix86_make_bounds_constant): Likewise.
13165 (ix86_initialize_bounds): Likewise.
13166 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
13167 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
13168 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
13169 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
13170 (TARGET_CHKP_BOUND_MODE): Likewise.
13171 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
13172 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
13173 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
13174 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
13175 * config/i386/i386.h (TARGET_MPX): Likewise.
13176 (TARGET_MPX_P): Likewise.
13177 (VALID_BND_REG_MODE): Likewise.
13178 (FIRST_BND_REG): Likewise.
13179 (LAST_BND_REG): Likewise.
13180 (enum reg_class): Likewise.
13181 (BND_REG_P): Likewise.
13182 (BND_REGNO_P): Likewise.
13183 (BNDmode): Likewise.
13184 (ADJUST_INSN_LENGTH): Likewise.
13185 * config/i386/i386.md: Likewise.
13186 * config/i386/i386.opt: Likewise.
13187 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
13188 (defined): Likewise.
13189 (LINK_MPX): Likewise.
13190 (MPX_SPEC): Likewise.
13191 (LIBMPX_SPEC): Likewise.
13192 (LIBMPXWRAPPERS_SPEC): Likewise.
13193 (CHKP_SPEC): Likewise.
13194 * config/i386/predicates.md: Likewise.
13195 * dbxout.c (dbxout_type): Likewise.
13196 * doc/extend.texi: Likewise.
13197 * doc/invoke.texi: Likewise.
13198 * doc/md.texi: Likewise.
13199 * doc/tm.texi: Likewise.
13200 * doc/tm.texi.in: Likewise.
13201 * dwarf2out.c (is_base_type): Likewise.
13202 (gen_formal_types_die): Likewise.
13203 (gen_subprogram_die): Likewise.
13204 (gen_type_die_with_usage): Likewise.
13205 (gen_decl_die): Likewise.
13206 (dwarf2out_late_global_decl): Likewise.
13207 * expr.c (expand_assignment): Likewise.
13208 (emit_storent_insn): Likewise.
13209 (store_expr_with_bounds): Likewise.
13210 (store_expr): Likewise.
13211 (expand_expr_real_1): Likewise.
13212 * expr.h (store_expr_with_bounds): Likewise.
13213 * function.c (use_register_for_decl): Likewise.
13214 (struct bounds_parm_data): Likewise.
13215 (assign_parms_augmented_arg_list): Likewise.
13216 (assign_parm_find_entry_rtl): Likewise.
13217 (assign_parm_is_stack_parm): Likewise.
13218 (assign_parm_load_bounds): Likewise.
13219 (assign_bounds): Likewise.
13220 (assign_parms): Likewise.
13221 (expand_function_start): Likewise.
13222 * gcc.c (CHKP_SPEC): Likewise.
13223 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
13224 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
13225 (wrestrict_dom_walker::check_call): Likewise.
13226 * gimple.c (gimple_build_call_from_tree): Likewise.
13227 * gimple.h (enum gf_mask): Likewise.
13228 (gimple_call_with_bounds_p): Likewise.
13229 (gimple_call_set_with_bounds): Likewise.
13230 * gimplify.c (gimplify_init_constructor): Likewise.
13231 * ipa-cp.c (initialize_node_lattices): Likewise.
13232 (propagate_constants_across_call): Likewise.
13233 (find_more_scalar_values_for_callers_subset): Likewise.
13234 * ipa-hsa.c (process_hsa_functions): Likewise.
13235 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
13236 * ipa-icf.c (sem_function::merge): Likewise.
13237 * ipa-inline.c (early_inliner): Likewise.
13238 * ipa-pure-const.c (warn_function_noreturn): Likewise.
13239 (warn_function_cold): Likewise.
13240 (propagate_pure_const): Likewise.
13241 * ipa-ref.h (enum GTY): Likewise.
13242 * ipa-split.c (find_retbnd): Likewise.
13243 (consider_split): Likewise.
13244 (split_function): Likewise.
13245 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
13246 * ipa.c (walk_polymorphic_call_targets): Likewise.
13247 (symbol_table::remove_unreachable_nodes): Likewise.
13248 (process_references): Likewise.
13249 (cgraph_build_static_cdtor_1): Likewise.
13250 * lto-cgraph.c (lto_output_node): Likewise.
13251 (output_refs): Likewise.
13252 (compute_ltrans_boundary): Likewise.
13253 (input_overwrite_node): Likewise.
13254 (input_node): Likewise.
13255 (input_cgraph_1): Likewise.
13256 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
13257 * passes.c (pass_manager::execute_early_local_passes): Likewise.
13258 (class pass_chkp_instrumentation_passes): Likewise.
13259 (make_pass_chkp_instrumentation_passes): Likewise.
13260 * passes.def: Likewise.
13261 * rtl.h (struct GTY): Likewise.
13262 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
13263 * stor-layout.c (layout_type): Likewise.
13264 * symtab.c: Likewise.
13265 * target.def: Likewise.
13266 * targhooks.c (default_chkp_bound_type): Likewise.
13267 (default_chkp_bound_mode): Likewise.
13268 (default_builtin_chkp_function): Likewise.
13269 (default_chkp_function_value_bounds): Likewise.
13270 (default_chkp_make_bounds_constant): Likewise.
13271 (default_chkp_initialize_bounds): Likewise.
13272 * targhooks.h (default_chkp_bound_type): Likewise.
13273 (default_chkp_bound_mode): Likewise.
13274 (default_builtin_chkp_function): Likewise.
13275 (default_chkp_function_value_bounds): Likewise.
13276 (default_chkp_make_bounds_constant): Likewise.
13277 (default_chkp_initialize_bounds): Likewise.
13278 * toplev.c (compile_file): Likewise.
13279 (process_options): Likewise.
13280 * tree-core.h (DEF_BUILTIN): Likewise.
13281 (DEF_BUILTIN_CHKP): Likewise.
13282 * tree-inline.c (declare_return_variable): Likewise.
13283 (remap_gimple_stmt): Likewise.
13284 (copy_bb): Likewise.
13285 (initialize_inlined_parameters): Likewise.
13286 (expand_call_inline): Likewise.
13287 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
13288 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
13289 (make_pass_ipa_chkp_produce_thunks): Likewise.
13290 (make_pass_chkp): Likewise.
13291 (make_pass_chkp_opt): Likewise.
13292 (make_pass_chkp_instrumentation_passes): Likewise.
13293 * tree-pretty-print.c (dump_generic_node): Likewise.
13294 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
13295 * tree-ssa-dce.c (propagate_necessity): Likewise.
13296 (eliminate_unnecessary_stmts): Likewise.
13297 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
13298 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
13299 * tree-ssa-sccvn.h: Likewise.
13300 * tree-ssa-strlen.c (get_string_length): Likewise.
13301 (valid_builtin_call): Likewise.
13302 (adjust_last_stmt): Likewise.
13303 (handle_builtin_strchr): Likewise.
13304 (handle_builtin_strcpy): Likewise.
13305 (handle_builtin_stxncpy): Likewise.
13306 (handle_builtin_memcpy): Likewise.
13307 (handle_builtin_strcat): Likewise.
13308 (strlen_check_and_optimize_stmt): Likewise.
13309 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
13310 * tree-streamer-in.c: Likewise.
13311 * tree-streamer.c (record_common_node): Likewise.
13312 * tree.c (tree_code_size): Likewise.
13313 (wide_int_to_tree_1): Likewise.
13314 (type_contains_placeholder_1): Likewise.
13315 (build_common_tree_nodes): Likewise.
13316 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
13317 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
13318 (POINTER_BOUNDS_P): Likewise.
13319 (BOUNDED_TYPE_P): Likewise.
13320 (BOUNDED_P): Likewise.
13321 (CALL_WITH_BOUNDS_P): Likewise.
13322 (pointer_bounds_type_node): Likewise.
13323 * value-prof.c (gimple_ic): Likewise.
13324 * var-tracking.c (vt_add_function_parameters): Likewise.
13325 * varasm.c (make_decl_rtl): Likewise.
13326 (assemble_start_function): Likewise.
13327 (output_constant): Likewise.
13328 (maybe_assemble_visibility): Likewise.
13329 * varpool.c (ctor_for_folding): Likewise.
13330 * chkp-builtins.def: Remove.
13331 * ipa-chkp.c: Remove.
13332 * ipa-chkp.h: Remove.
13333 * rtl-chkp.c: Remove.
13334 * rtl-chkp.h: Remove.
13335 * tree-chkp-opt.c: Remove.
13336 * tree-chkp.c: Remove.
13337 * tree-chkp.h: Remove.
13339 2018-06-07 Carl Love <cel@us.ibm.com>
13341 * config/rs6000/vsx.md (vextract_fp_from_shorth,
13342 vextract_fp_from_shortl): Add BE support.
13344 2018-06-07 Paul Koning <ni1d@arrl.net>
13346 * compare-elim.c (try_merge_compare): Don't merge compare if
13347 address contains a side effect.
13348 (try_eliminate_compare): Likewise.
13350 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
13352 * config.gcc: Support "tremont".
13353 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
13354 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13356 * config/i386/i386.c (m_TREMONT): Define.
13357 (processor_target_table): Add "tremont".
13358 (PTA_TREMONT): Define.
13359 (ix86_lea_outperforms): Add TARGET_TREMONT.
13360 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
13361 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
13362 and M_INTEL_GOLDMONT_PLUS.
13363 (fold_builtin_cpu): Add "tremont".
13364 (ix86_add_stmt_cost): Add TARGET_TREMONT.
13365 (ix86_option_override_internal): Add "tremont".
13366 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
13367 (processor_type): Add PROCESSOR_TREMONT.
13368 * config/i386/x86-tune.def: Add m_TREMONT.
13369 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
13371 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13373 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
13374 symbol defined for msp430i* devices to be lower case.
13376 2018-06-07 Richard Biener <rguenther@suse.de>
13378 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
13379 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
13380 Properly wrap signed arithmetic if overflow wraps.
13382 2018-06-07 Jakub Jelinek <jakub@redhat.com>
13384 PR tree-optimization/69615
13385 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
13386 of a cast from a same precision integral SSA_NAME in a bb dominated
13387 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
13388 cast to utype if rhs2 has already a compatible type.
13390 2018-06-07 Richard Biener <rguenther@suse.de>
13392 PR tree-optimization/85935
13393 * graphite-scop-detection.c (find_params_in_bb): Analyze
13394 condition operands with respect to the correct loop. Assert
13395 the analysis doesn't fail.
13397 2018-06-04 Carl Love <cel@us.ibm.com>
13399 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
13400 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
13401 as it is slightly cheaper.
13402 (first_match_or_eos_index_<mode>):
13403 Calculate index using natural element order.
13404 (first_match_index_<mode>):
13405 Calculate index using natural element order.
13406 (first_match_or_eos_index_<mode>):
13407 Calculate index using natural order.
13408 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
13409 for BE and LE modes.
13410 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
13411 P9V_BUILTIN_VCLZLSBB_V16QI.
13412 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
13415 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
13417 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
13418 indentation and line wrap for many prototypes. Add missing
13419 @smallexample directives around block of prototypes for vec_xl and
13422 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
13424 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
13425 track if we pass or return IEEE 128-bit floating point.
13426 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
13427 C++ mangling that is compatible with GCC 8.1.
13428 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
13429 (init_cumulative_args): Note if we pass or return IEEE 128-bit
13430 floating point types.
13431 (rs6000_function_arg_advance_1): Likewise.
13432 (rs6000_mangle_type): Optionally generate mangled names that match
13433 what GCC 8.1 generated for IEEE 128-bit floating point types.
13434 (rs6000_globalize_decl_name): If we have an external function that
13435 passes or returns IEEE 128-bit types, generate a weak reference
13436 from the mangled name used in GCC 8.1 to the current mangled
13438 (rs6000_init_builtins): Make __ibm128 use the long double type if
13439 long double is IBM extended double. Make __float128 use the long
13440 double type if long double is IEEE 128-bit.
13443 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
13444 macro for __ibm128 built-in functions.
13445 (PACK_IF): Add __ibm128 pack/unpack functions.
13446 (UNPACK_IF): Likewise.
13447 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
13448 enable long double built-in functions if long double is IEEE
13449 128-bit floating point.
13450 (rs6000_invalid_builtin): Update long double built-in function
13452 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
13453 functions, adjust the built-in function to use the long double
13454 built-in function if __ibm128 and long double are the same type.
13455 * doc/extend.texi (PowerPC builtins): Update documention for
13456 __builtin_{,un}pack_longdouble. Add documentation for
13457 __builtin_{,un}pack_ibm128.
13459 2018-06-06 Jim Wilson <jimw@sifive.com>
13461 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
13462 (struct machine_function): New field interrupt_mode.
13463 (riscv_handle_type_attribute): New function. Add forward declaration.
13464 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
13465 (riscv_expand_epilogue): Check interrupt_mode field.
13466 (riscv_set_current_function): Check interrupt attribute args and
13467 set interrupt_mode field.
13468 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
13469 (riscv_sret, riscv_uret): New.
13470 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
13471 new arguments to interrupt attribute.
13473 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
13476 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
13477 Don't handle -mcpu=power8 if -mpower9-vector is also used.
13479 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
13481 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
13482 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
13483 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13484 several redundant entries.
13486 2018-06-06 David Malcolm <dmalcolm@redhat.com>
13488 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
13489 type from "rtx" to "rtx_insn *".
13490 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
13491 for local "call_insn", removing cast.
13492 (ix86_expand_call): Likewise, introducing a "call_insn" local.
13494 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
13496 PR tree-optimization/86066
13497 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
13498 for BIT_INSERT_EXPR stores.
13500 2018-06-06 Richard Biener <rguenther@suse.de>
13502 PR tree-optimization/86062
13503 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13504 component refs ontop
13505 of to be offsetted base.
13507 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13509 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
13510 to be static and remove check on interrupt attribute name.
13512 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
13514 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13515 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
13517 2018-06-05 Steve Ellcey <sellcey@cavium.com>
13520 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
13522 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
13523 Remove second argument, change how error is called.
13524 (aarch64_layout_arg): Remove second argument from
13525 aarch64_err_no_fpadvsimd call.
13526 (aarch64_init_cumulative_args): Ditto.
13527 (aarch64_gimplify_va_arg_expr): Ditto.
13528 * config/aarch64/aarch64.md (mov<mode>): Ditto.
13530 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
13532 * config/i386/i386.md (simple_return_indirect_internal): New expander.
13533 (*simple_return_indirect_internal<mode>): Rename from
13534 simple_return_indirect_internal. Use W mode iterator.
13535 (rstorssp): New expander.
13536 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
13537 (clrssbsy): New expander.
13538 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
13540 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
13542 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
13544 (cmse_check_pointed_object): Likewise.
13546 2018-06-05 Martin Liska <mliska@suse.cz>
13548 PR gcov-profile/47618
13549 * doc/invoke.texi: Document how -fprofile-dir format
13552 2018-06-05 Richard Biener <rguenther@suse.de>
13554 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
13555 removal pretend DOM info isn't available so we do not update
13556 it and only remove edges, not dominated blocks. Actually free
13557 DOM info in case we removed something. Remove unreachable blocks.
13558 (mfb_keep_latches): Work with either DOM info or marked backedges.
13559 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
13560 first. Mark backedges if DOM info isn't available.
13561 (Re-)compute DOM info after cleanup_control_flow_pre.
13563 2018-06-05 Richard Biener <rguenther@suse.de>
13565 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
13566 (locus_discrim_hasher::hash): Adjust.
13567 (locus_discrim_hasher::equal): Likewise.
13568 (next_discriminator_for_locus): Work on line directly.
13569 (same_line_p): Pass in expanded locus1 as well.
13570 (assign_discriminators): Avoid redundant location expansions.
13572 2018-06-05 Richard Biener <rguenther@suse.de>
13574 PR tree-optimization/86046
13575 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
13576 if required after clearing TREE_ADDRESSABLE.
13578 2018-06-05 Richard Biener <rguenther@suse.de>
13580 PR tree-optimization/86047
13581 * tree-ssa-loop.c (for_each_index): Glob handling of all
13582 decls and constants and really handle all of them.
13584 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13587 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
13588 qualifier_void_pointer and qualifier_const_void_pointer.
13589 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
13590 (arm_init_builtins): Handle the above.
13591 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
13592 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
13593 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
13596 2018-06-05 Martin Liska <mliska@suse.cz>
13598 * auto-profile.c (read_autofdo_file): Do not use
13599 gcov_ctr_summary struct.
13600 (afdo_callsite_hot_enough_for_early_inline): Likewise.
13601 * coverage.c (struct counts_entry): Likewise.
13602 (read_counts_file): Read just single summary entry.
13603 (get_coverage_counts): Use gcov_summary struct.
13604 * coverage.h (get_coverage_counts): Likewise.
13605 * gcov-dump.c (dump_working_sets): Likewise.
13606 (tag_summary): Dump just single summary.
13607 * gcov-io.c (gcov_write_summary): Write just histogram
13609 (gcov_read_summary): Read just single summary.
13610 (compute_working_sets): Use gcov_summary struct.
13611 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
13612 of GCOV_COUNTERS_SUMMABLE.
13613 (GCOV_COUNTERS_SUMMABLE): Remove.
13614 (GCOV_FIRST_VALUE_COUNTER): Replace with
13615 GCOV_COUNTER_V_INTERVAL.
13616 (struct gcov_ctr_summary): Remove.
13617 (struct gcov_summary): Directly use fields of former
13619 (compute_working_sets): Use gcov_summary struct.
13620 * gcov.c (read_count_file): Do not use ctrs fields.
13621 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
13623 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
13625 * profile.c: Likewise.
13626 * profile.h: Likewise.
13628 2018-06-05 Martin Liska <mliska@suse.cz>
13630 PR gcov-profile/84846
13631 * gcov.c (output_lines): Print working directory only
13632 in intermediate format.
13634 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
13636 * config/s390/s390-builtin-types.def: Add void function type.
13637 * config/s390/s390-builtins.def: Use the function type for the
13640 2018-06-04 Jim Wilson <jimw@sifive.com>
13642 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
13644 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
13645 and maybe_eh_return. Change regno to unsigned int. Use new args to
13646 handle EH_RETURN_DATA_REGNO registers properly.
13647 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
13648 (riscv_expand_epilogue): Update comment. Change argument name and
13649 type. Update code to use new name and type. Pass new args to
13650 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
13652 * config/riscv/riscv.md (NORMAL_RETURN): New.
13653 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
13654 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
13655 (eh_return): Call gen_eh_return_internal and emit barrier.
13656 (eh_return_internal): Call riscv_expand_epilogue.
13658 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
13660 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
13661 bit_insertion field and declare can_be_merged_into method.
13662 (merged_store_group::can_be_merged_into): New method.
13663 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
13664 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
13665 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
13667 2018-06-04 Richard Biener <rguenther@suse.de>
13669 PR tree-optimization/85955
13670 * builtins.c (fold_builtin_sincos): Convert pointers to
13671 destination to appropriate type before dereferencing.
13673 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
13675 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
13677 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
13679 * expr.c (expand_expr_real_1): Force the operand into memory if
13680 its TYPE_MODE is BLKmode and if there is no integer mode for
13681 the number of bits being extracted.
13683 2018-06-04 Jakub Jelinek <jakub@redhat.com>
13687 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13688 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
13690 2018-06-04 Richard Biener <rguenther@suse.de>
13692 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
13693 (cleanup_tree_cfg_noloop): ... single caller. Do
13694 start_recording_case_labels later.
13696 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
13698 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
13699 to _IMMINTRIN_H_INCLUDED.
13700 * config/i386/pconfigintrin.h: Ditto.
13701 * config/i386/waitpkgintrin.h: Ditto.
13702 * config/i386/immintrin.h: Add includes for sgxintrin.h,
13703 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
13704 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
13705 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
13706 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
13707 waitpkgintrin.h and cldemoteintrin.h.
13709 2018-06-04 Richard Biener <rguenther@suse.de>
13711 PR tree-optimization/86038
13712 * tracer.c (find_best_successor): Check probability for
13713 being initialized, bail out if not.
13715 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
13718 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
13719 of bits to ignore when comparing architectures.
13721 2018-06-04 Jakub Jelinek <jakub@redhat.com>
13723 PR tree-optimization/69615
13724 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
13725 maximum or minimum of the type, try to merge it also as if
13726 range1 is + [-, x - 1] or + [x + 1, -].
13729 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
13731 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
13733 PR tree-optimization/86034
13734 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
13735 the unsigned bitfield type in a bit insertion sequence if it does not
13736 have a larger precision than the bitfield size.
13737 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
13739 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
13741 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
13743 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13745 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
13746 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13747 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13748 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13750 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13752 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13753 Disable -fdelete-null-pointer-checks for ELF toolchain.
13755 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13756 Kito Cheng <kito.cheng@gmail.com>
13758 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13759 (nds32le-*-*, nds32be-*-*): Integrate checking process.
13760 (nds32*-*-*): Add glibc and uclibc conditions.
13761 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13762 (TARGET_EXCEPT_UNWIND_INFO): Define.
13763 * config/nds32/elf.h: New file.
13764 * config/nds32/linux.h: New file.
13765 * config/nds32/nds32-elf.opt: New file.
13766 * config/nds32/nds32-linux.opt: New file.
13767 * config/nds32/nds32-fp-as-gp.c
13768 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13769 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13771 (nds32_asm_file_end): Ditto.
13772 (nds32_print_operand): Ditto.
13773 (nds32_insert_attributes): Ditto.
13774 (nds32_init_libfuncs): New function.
13775 (TARGET_HAVE_TLS): Define.
13776 (TARGET_INIT_LIBFUNCS): Define.
13777 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13779 (TARGET_ELF): Apply different mcmodel setting.
13780 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13781 been migrated into elf.h and linux.h files.
13782 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13783 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13784 (mcmodel): The content has been migrated into nds32-elf.opt and
13785 nds32-linux.opt files.
13786 * config/nds32/t-elf: New file.
13787 * config/nds32/t-linux: New file.
13789 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13790 Shiva Chen <shiva0217@gmail.com>
13792 * config/nds32/constants.md (unspec_volatile_element): Add
13793 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13794 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13796 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13797 (make_pass_nds32_fp_as_gp): Declare.
13798 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13800 (nds32_asm_function_end_prologue): Remove unused asm output.
13801 (nds32_asm_function_begin_epilogue): Remove unused asm output.
13802 (nds32_asm_file_start): Output necessary fp_as_gp information.
13803 (nds32_option_override): Adjust register usage.
13804 (nds32_expand_prologue): Consider fp_as_gp situation.
13805 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13806 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13809 (simple_return): Ditto.
13810 (omit_fp_begin): Output special directive for fp_as_gp.
13811 (omit_fp_end): Output special directive for fp_as_gp.
13812 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13813 mforbid-fp-as-gp): New options.
13815 2018-06-01 Mark Wielaard <mark@klomp.org>
13817 * dwarf2out.c (dwarf2out_finish): Remove generation of
13818 DW_AT_loclists_base.
13820 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
13822 * gimple-ssa-store-merging.c: Include gimple-fold.h.
13823 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13824 (struct merged_store_group): Add bit_insertion field.
13825 (dump_char_array): Use standard hexadecimal format.
13826 (merged_store_group::merged_store_group): Set bit_insertion to false.
13827 (merged_store_group::apply_stores): Use optimal buffer size. Deal
13828 with BIT_INSERT_EXPR stores. Move up code updating the mask and
13829 also print the mask in the dump file.
13830 (pass_store_merging::gate): Minor tweak.
13831 (imm_store_chain_info::coalesce_immediate): Fix wrong association
13832 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
13833 stores with INTEGER_CST stores.
13834 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
13835 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
13836 and use it throughout. Generate bit insertion sequences if need be.
13837 (pass_store_merging::process_store): Remove redundant condition.
13838 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
13840 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
13842 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
13843 the 128-bit floating point types. Fix function comment.
13845 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13847 * config/aarch64/aarch64-simd.md
13848 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
13850 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13853 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13855 PR tree-optimization/85989
13856 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
13858 (backprop::intersect_uses): Check it when deciding whether this
13859 is a backedge reference.
13860 (backprop::process_block): Add each phi to m_visited_phis
13861 after visiting it, then clear it at the end.
13863 2018-06-01 Richard Biener <rguenther@suse.de>
13865 * tree-vectorizer.h (vect_dr_stmt): New function.
13866 (vect_get_load_cost): Adjust.
13867 (vect_get_store_cost): Likewise.
13868 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13869 Use vect_dr_stmt instead of DR_SMTT.
13870 (vect_record_base_alignments): Likewise.
13871 (vect_calculate_target_alignment): Likewise.
13872 (vect_compute_data_ref_alignment): Likewise and make static.
13873 (vect_update_misalignment_for_peel): Likewise.
13874 (vect_verify_datarefs_alignment): Likewise.
13875 (vector_alignment_reachable_p): Likewise.
13876 (vect_get_data_access_cost): Likewise. Pass down
13877 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13878 (vect_get_peeling_costs_all_drs): Likewise.
13879 (vect_peeling_hash_get_lowest_cost): Likewise.
13880 (vect_enhance_data_refs_alignment): Likewise.
13881 (vect_find_same_alignment_drs): Likewise.
13882 (vect_analyze_data_refs_alignment): Likewise.
13883 (vect_analyze_group_access_1): Likewise.
13884 (vect_analyze_group_access): Likewise.
13885 (vect_analyze_data_ref_access): Likewise.
13886 (vect_analyze_data_ref_accesses): Likewise.
13887 (vect_vfa_segment_size): Likewise.
13888 (vect_small_gap_p): Likewise.
13889 (vectorizable_with_step_bound_p): Likewise.
13890 (vect_prune_runtime_alias_test_list): Likewise.
13891 (vect_analyze_data_refs): Likewise.
13892 (vect_supportable_dr_alignment): Likewise.
13893 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13894 (vect_gen_prolog_loop_niters): Likewise.
13895 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13896 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13898 (vect_recog_mask_conversion_pattern): Likewise.
13899 (vect_try_gather_scatter_pattern): Likewise.
13900 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13901 to vect_get_store_cost.
13902 (vect_get_store_cost): Get stmt_info instead of DR.
13903 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13904 (vect_get_load_cost): Get stmt_info instead of DR.
13906 2018-06-01 Richard Biener <rguenther@suse.de>
13908 PR middle-end/86017
13909 * gimple-fold.c (var_decl_component_p): Also allow offsetted
13910 vars wrapped in MEM_REFs.
13912 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13914 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13915 Fix subreg tests so that we only return a choice between
13916 GENERAL_REGS and FP_REGS if the original classes included both.
13918 2018-06-01 Richard Biener <rguenther@suse.de>
13921 * tree-ssa-structalias.c (get_function_part_constraint):
13922 Handle NULL fi->decl.
13923 (find_func_aliases_for_call): Properly handle indirect
13924 fi from direct call.
13925 (find_func_clobbers): Likewise.
13926 (ipa_pta_execute): Likewise.
13927 (create_variable_info_for): For functions that are ifunc_resolver
13928 resolve to a varinfo that contains the result of the resolver call.
13929 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13932 2018-05-31 Michael Collison <michael.collison@arm.com>
13934 * config/aarch64/aarch64.md:
13935 (*fix_to_zero_extenddfdi2): New pattern.
13936 * gcc.target/aarch64/fix_extend1.c: New testcase.
13938 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
13940 PR middle-end/78809
13941 PR middle-end/83026
13942 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13943 and BUILT_IN_STRNCMP_EQ.
13944 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13945 BUILT_IN_STRNCMP_EQ.
13946 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13947 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13948 (gimple_fold_builtin): Likewise.
13949 * tree-ssa-strlen.c (compute_string_length): New function.
13950 (determine_min_obsize): New function.
13951 (handle_builtin_string_cmp): New function to handle calls to
13952 string compare functions.
13953 (strlen_optimize_stmt): Add handling to builtin string compare
13955 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13956 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13957 * tree.c (build_common_builtin_nodes): Add new defines of
13958 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13960 2018-05-31 Jakub Jelinek <jakub@redhat.com>
13963 * bb-reorder.c (pass_partition_blocks::gate): Return false for
13964 functions with naked attribute.
13966 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
13968 * config/i386/sse.md (avx_vec_concat<mode>):
13969 Substitute concat_tg_mode mode attribute with xtg_mode.
13970 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13971 (concat_tg_mode): Remove mode attribute.
13973 2018-05-31 Martin Sebor <msebor@redhat.com>
13976 * calls.c (alloc_max_size): Correct a logic error/typo.
13977 Treat excessive arguments as infinite. Warn for invalid arguments.
13978 * doc/invoke.texi (-Walloc-size-larger-than): Update.
13980 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
13983 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13984 and movx for Haswell.
13986 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
13987 Cesar Philippidis <cesar@codesourcery.com>
13989 PR middle-end/85879
13990 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13991 when emitting error on private/firstprivate reductions.
13992 * omp-low.c (lower_omp_target): Avoid reference-type processing
13993 on pointers for firstprivate clause.
13995 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13997 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14000 * config/aarch64/aarch64-simd.md
14001 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14002 (aarch64_ld1_x3_<mode>): Likewise
14003 (aarch64_st1x2<VALLDIF:mode>): Likewise
14004 (aarch64_st1_x2_<mode>): Likewise
14005 (aarch64_st1x3<VALLDIF:mode>): Likewise
14006 (aarch64_st1_x3_<mode>): Likewise
14007 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14008 (vld1_s8_x3): Likewise.
14009 (vld1_u16_x3): Likewise.
14010 (vld1_s16_x3): Likewise.
14011 (vld1_u32_x3): Likewise.
14012 (vld1_s32_x3): Likewise.
14013 (vld1_u64_x3): Likewise.
14014 (vld1_s64_x3): Likewise.
14015 (vld1_f16_x3): Likewise.
14016 (vld1_f32_x3): Likewise.
14017 (vld1_f64_x3): Likewise.
14018 (vld1_p8_x3): Likewise.
14019 (vld1_p16_x3): Likewise.
14020 (vld1_p64_x3): Likewise.
14021 (vld1q_u8_x3): Likewise.
14022 (vld1q_s8_x3): Likewise.
14023 (vld1q_u16_x3): Likewise.
14024 (vld1q_s16_x3): Likewise.
14025 (vld1q_u32_x3): Likewise.
14026 (vld1q_s32_x3): Likewise.
14027 (vld1q_u64_x3): Likewise.
14028 (vld1q_s64_x3): Likewise.
14029 (vld1q_f16_x3): Likewise.
14030 (vld1q_f32_x3): Likewise.
14031 (vld1q_f64_x3): Likewise.
14032 (vld1q_p8_x3): Likewise.
14033 (vld1q_p16_x3): Likewise.
14034 (vld1q_p64_x3): Likewise.
14035 (vst1_s64_x2): Likewise.
14036 (vst1_u64_x2): Likewise.
14037 (vst1_f64_x2): Likewise.
14038 (vst1_s8_x2): Likewise.
14039 (vst1_p8_x2): Likewise.
14040 (vst1_s16_x2): Likewise.
14041 (vst1_p16_x2): Likewise.
14042 (vst1_s32_x2): Likewise.
14043 (vst1_u8_x2): Likewise.
14044 (vst1_u16_x2): Likewise.
14045 (vst1_u32_x2): Likewise.
14046 (vst1_f16_x2): Likewise.
14047 (vst1_f32_x2): Likewise.
14048 (vst1_p64_x2): Likewise.
14049 (vst1q_s8_x2): Likewise.
14050 (vst1q_p8_x2): Likewise.
14051 (vst1q_s16_x2): Likewise.
14052 (vst1q_p16_x2): Likewise.
14053 (vst1q_s32_x2): Likewise.
14054 (vst1q_s64_x2): Likewise.
14055 (vst1q_u8_x2): Likewise.
14056 (vst1q_u16_x2): Likewise.
14057 (vst1q_u32_x2): Likewise.
14058 (vst1q_u64_x2): Likewise.
14059 (vst1q_f16_x2): Likewise.
14060 (vst1q_f32_x2): Likewise.
14061 (vst1q_f64_x2): Likewise.
14062 (vst1q_p64_x2): Likewise.
14063 (vst1_s64_x3): Likewise.
14064 (vst1_u64_x3): Likewise.
14065 (vst1_f64_x3): Likewise.
14066 (vst1_s8_x3): Likewise.
14067 (vst1_p8_x3): Likewise.
14068 (vst1_s16_x3): Likewise.
14069 (vst1_p16_x3): Likewise.
14070 (vst1_s32_x3): Likewise.
14071 (vst1_u8_x3): Likewise.
14072 (vst1_u16_x3): Likewise.
14073 (vst1_u32_x3): Likewise.
14074 (vst1_f16_x3): Likewise.
14075 (vst1_f32_x3): Likewise.
14076 (vst1_p64_x3): Likewise.
14077 (vst1q_s8_x3): Likewise.
14078 (vst1q_p8_x3): Likewise.
14079 (vst1q_s16_x3): Likewise.
14080 (vst1q_p16_x3): Likewise.
14081 (vst1q_s32_x3): Likewise.
14082 (vst1q_s64_x3): Likewise.
14083 (vst1q_u8_x3): Likewise.
14084 (vst1q_u16_x3): Likewise.
14085 (vst1q_u32_x3): Likewise.
14086 (vst1q_u64_x3): Likewise.
14087 (vst1q_f16_x3): Likewise.
14088 (vst1q_f32_x3): Likewise.
14089 (vst1q_f64_x3): Likewise.
14090 (vst1q_p64_x3): Likewise.
14092 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14094 * config/msp430/msp430.c (msp430_output_labelref): Prepend
14095 user_label_prefix to name.
14097 * tree-core.h: Update comment about the format of NAME string
14098 passed to handler in attribute_spec.
14100 * config/msp430/msp430.md: Remove erroneous subreg expression from
14101 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
14102 zero_extend{q,h}isi2.
14104 2018-05-30 Borislav Petkov <bp@suse.de>
14106 * doc/extend.texi: Document some architecture specific
14107 constraints and sort entries.
14109 2018-05-30 Martin Sebor <msebor@redhat.com>
14111 PR middle-end/85369
14112 * builtins.c (expand_builtin_stpcpy_1): New function.
14113 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
14114 only if the former succeeds.
14116 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
14118 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
14121 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14123 * doc/invoke.texi (-flinker-output): Document
14125 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14127 * passes.c (ipa_write_summaries): Only modify statements if body
14129 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
14130 incrementally linking.
14131 (ipa_passes): Likewise.
14132 * lto-cgraph.c (lto_output_node): When incrementally linking do not
14133 pass down resolution info.
14134 * common.opt (flag_incremental_link): Update info.
14135 * gcc.c (plugin specs): Turn flinker-output=* to
14136 -plugin-opt=-linker-output-known
14137 * toplev.c (compile_file): Also cut compilation when doing incremental
14139 * flag-types. (enum lto_partition_model): Add
14140 LTO_LINKER_OUTPUT_NOLTOREL.
14141 (invoke.texi): Add -flinker-output docs.
14142 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
14143 link same way as WPA; do not stream in dead initializers.
14145 * dwarf2out.c (dwarf2out_die_ref_for_decl,
14146 darf2out_register_external_decl): Support incremental link.
14148 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14150 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
14152 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
14154 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
14155 it down to simple_object_copy_lto_debug_sections.
14156 (run_gcc): Determine incremental LTO link time and configure
14157 lto1 into non-wpa mode, disable renaming of debug sections.
14159 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
14161 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14162 descriptions of various incorrectly documented functions.
14164 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
14167 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14168 address check not strict.
14170 2018-05-30 Richard Biener <rguenther@suse.de>
14172 PR tree-optimization/85964
14173 * tracer.c (better_p): Drop initialized count check, we only
14174 call the function with initialized counts now.
14175 (find_best_successor): Do find a best edge if one
14176 has uninitialized count.
14177 (find_best_predecessor): Likewise. Do BB frequency check only
14178 if count is initialized.
14180 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
14182 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14183 (aarch64_ldrstr_offset_compare): New.
14184 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
14185 load/store orderings.
14186 (aarch64_gen_adjusted_ldpstp): Likewise.
14188 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
14190 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14191 Check for subset of GENERAL_REGS and FP_REGS.
14192 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14195 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
14197 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
14198 and wi::to_poly_offset. Add the current offset and then check
14199 whether the sum fits, rather than using an unchecked addition of
14200 a checked term. Check for a shwi rather than a uhwi.
14201 * expr.c (get_bit_range): Use tree_to_poly_uint64.
14202 (store_constructor): Use poly_int_tree_p.
14203 (expand_expr_real_1): Likewise.
14204 * function.c (assign_temp): Likewise.
14205 * fold-const.c (const_binop): Use poly_int_tree_p and
14206 wi::to_poly_offset.
14207 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
14209 * ipa-icf-gimple.c (func_checker::compare_operand): Use
14210 to_poly_offset for MEM offsets.
14211 * ipa-icf.c (sem_variable::equals): Likewise.
14212 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
14213 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
14214 wi::to_poly_offset for BIT_FIELD_REF offsets.
14215 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
14216 wi::to_poly_offset.
14217 * var-tracking.c (emit_note_insn_var_location): Use
14218 tree_to_poly_uint64.
14220 2018-05-29 Jim Wilson <jimw@sifive.com>
14222 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
14224 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
14227 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
14228 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
14230 (sse4_1_round<mode>2): Use nonimmediate_operand
14231 for operand 1 predicate.
14233 2018-05-29 Martin Sebor <msebor@redhat.com>
14234 Richard Biener <rguenther@suse.de>
14237 * calls.c (get_size_range): Call determine_value_range instead
14238 of get_value_range..
14239 * tree-vrp.h (determine_value_range): Declared new function.
14240 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
14242 2018-05-29 Richard Biener <rguenther@suse.de>
14244 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
14245 sure to use non-pattern stmts for get_earlier_stmt arguments.
14246 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
14247 called on pattern stmts.
14248 (get_later_stmt): Likewise.
14250 2018-05-29 Martin Liska <mliska@suse.cz>
14252 PR gcov-profile/85759
14253 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
14256 2018-05-29 Jakub Jelinek <jakub@redhat.com>
14258 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
14260 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
14264 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14265 VEC_PACK_FLOAT_EXPR): New tree codes.
14266 * tree-pretty-print.c (op_code_prio): Handle
14267 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14268 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14269 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14270 * tree-inline.c (estimate_operator_cost): Likewise.
14271 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
14272 * fold-const.c (const_binop): Likewise.
14273 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
14274 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
14275 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14276 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
14277 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
14278 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
14279 * expr.c (expand_expr_real_2): Likewise.
14280 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
14281 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
14282 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
14284 * optabs.c (expand_widen_pattern_expr): For
14285 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
14286 sign from result type rather than operand's type.
14287 (expand_binop_directly): For vec_packu_float_optab and
14288 vec_packs_float_optab allow result type to be different from operand's
14290 * optabs-tree.c (optab_for_tree_code): Handle
14291 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14292 VEC_PACK_FLOAT_EXPR. Formatting fixes.
14293 * tree-vect-generic.c (expand_vector_operations_1): Handle
14294 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
14295 VEC_PACK_FLOAT_EXPR.
14296 * tree-vect-stmts.c (supportable_widening_operation): Handle
14298 (supportable_narrowing_operation): Handle FLOAT_EXPR.
14299 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
14300 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
14301 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
14302 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
14304 (vec_pack<floatprefix>_float_<mode>): New expander.
14305 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
14307 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
14308 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
14309 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
14310 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
14311 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
14313 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
14314 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
14315 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
14316 VEC_PACK_FLOAT_EXPR): Document.
14318 2018-05-29 Richard Biener <rguenther@suse.de>
14320 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
14322 (stmt_vec_info_vec): Make pointer.
14323 (init_stmt_vec_info_vec): Remove.
14324 (free_stmt_vec_info_vec): Likewise.
14325 (set_stmt_vec_info_vec): New function.
14326 (free_stmt_vec_infos): Likewise.
14327 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
14328 (set_vinfo_for_stmt): Likewise.
14329 (get_earlier_stmt): Likewise.
14330 (get_later_stmt): Likewise.
14331 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
14332 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
14333 (vec_info::~vec_info): Free stmt_vec_infos.
14334 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
14335 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
14336 (pass_slp_vectorize::execute): Likewise.
14337 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
14338 (free_stmt_vec_info_vec): Likewise.
14339 (set_stmt_vec_info_vec): New function.
14340 (free_stmt_vec_infos): Likewise.
14341 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
14342 the global stmt_vec_info_vec.
14343 * tree-parloops.c (gather_scalar_reductions): Use
14344 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
14347 2018-05-29 Richard Biener <rguenther@suse.de>
14349 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
14351 2018-05-29 Martin Liska <mliska@suse.cz>
14352 David Malcolm <dmalcolm@redhat.com>
14354 * vec.c (test_reverse): New.
14355 (vec_c_tests): Add new test.
14356 * vec.h (vl_ptr>::reverse): New function.
14358 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
14360 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
14362 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
14365 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14367 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
14369 2018-05-28 Richard Biener <rguenther@suse.de>
14371 PR tree-optimization/85933
14372 * tree-vect-data-refs.c (vect_record_base_alignments): Only
14373 look at stmts marked as vectorizable.
14375 2018-05-28 Richard Biener <rguenther@suse.de>
14377 PR tree-optimization/85934
14378 * tree-vect-generic.c (expand_vector_operations_1): Hoist
14379 vector boolean check before scalar optimization.
14381 2018-05-28 Jakub Jelinek <jakub@redhat.com>
14383 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
14386 2018-05-28 Mark Wielaard <mark@klomp.org>
14388 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
14389 if it is an expression containing a minus sign.
14391 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
14393 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
14395 2018-05-27 Paul Koning <ni1d@arrl.net>
14397 * config/pdp11/pdp11.md (truncsihi2): Remove.
14399 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
14400 Chung-Ju Wu <jasonwucj@gmail.com>
14402 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
14404 (unaligned_store_dw): Ditto.
14405 * config/nds32/nds32-memory-manipulation.c
14406 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
14407 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
14408 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
14409 (emit_setmem_word_loop): Rename to ...
14410 (emit_setmem_doubleword_loop): ... this.
14411 (nds32_gen_dup_4_byte_to_word_value): New function.
14412 (nds32_gen_dup_8_byte_to_double_word_value): New function.
14413 (nds32_expand_setmem_loop): Refine implementation.
14414 (nds32_expand_setmem_loop_v3m): Ditto.
14415 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
14418 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
14420 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
14422 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
14424 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
14425 (nds32_init_machine_status): Initialize machine->attr_naked_p and
14426 machine->attr_no_prologue_p.
14427 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
14428 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
14429 (nds32_expand_epilogue): Consider attr_naked_p.
14430 (nds32_expand_epilogue_v3pop): Likewise.
14431 (nds32_can_use_return_insn): Likewise.
14432 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
14433 attr_no_prologue_p fields.
14434 * config/nds32/nds32.opt (mret-in-naked-func): New option.
14436 2018-05-27 Jakub Jelinek <jakub@redhat.com>
14439 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
14441 * config/i386/sse.md
14442 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
14444 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
14446 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
14448 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
14450 (*<floatsuffix>floatv2div2sf2): Rename to ...
14451 (*float<floatunssuffix>v2div2sf2): ... this.
14452 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
14453 (float<floatunssuffix>v2div2sf2_mask): ... this.
14454 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
14455 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
14456 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
14458 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
14460 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14462 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
14464 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14466 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
14468 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
14469 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
14470 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
14471 gen_ufix_truncv8dfv8si2.
14472 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
14473 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
14474 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
14475 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
14476 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
14477 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
14478 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
14479 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
14481 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
14485 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
14487 2018-05-25 Jim Wilson <jimw@sifive.com>
14489 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
14490 * config/riscv/riscv.c (struct machine_function): Add
14491 interrupt_handler_p and attribute_checked_p fields.
14492 (riscv_attribute_table): Add interrupt.
14493 (riscv_interrupt_type_p): New.
14494 (riscv_save_reg_p): Save extra regs for interrupt handler.
14495 (riscv_use_save_libcall): Return false for interrupt handler.
14496 (riscv_first_stack_step): Add forward declaration.
14497 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
14498 for interrupt handler with large frame. Use it for saved reg list.
14499 (riscv_expand_prologue): Move flag_stack_usage_info support to
14500 eliminate duplication.
14501 (riscv_expand_epilogue): Generate mret for interrupt handler.
14502 (riscv_epilogue_uses): New.
14503 (riscv_can_use_return_insn): Return false for interrupt handler.
14504 (riscv_function_ok_for_sibcall): Likewise.
14505 (riscv_set_current_function): Add interrupt handler support.
14506 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
14507 * config/riscv/riscv.md (UNSPECV_MRET): New.
14509 (riscv_frflags, riscv_fsflags): Use tab after opcode.
14511 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
14513 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
14515 PR tree-optimization/85712
14516 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
14517 this candidate has already been replaced in-situ by a copy.
14519 2018-05-25 Jason Merrill <jason@redhat.com>
14521 PR c++/80485 - inline function non-zero address.
14522 * symtab.c (nonzero_address): Check DECL_COMDAT.
14524 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
14527 * config/alpha/alpha.md (ashlsi3): New insn pattern.
14528 (*ashlsi_se): Rename from *ashldi_se. Define as sign
14529 extension of SImode operation. Use const123_operand predicate.
14530 (*saddsi_1): Remove.
14531 (*saddl_se_1): Ditto.
14532 (*ssubsi_1): Ditto.
14533 (*ssubl_se_1): Ditto.
14534 * config/alpha/predicates.md (const123_operand): New predicate.
14535 * config/alpha/constraints.md (P): Use IN_RANGE.
14537 2018-05-25 Richard Biener <rguenther@suse.de>
14539 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
14541 (ref_maybe_used_by_stmt_p): Likewise.
14542 (stmt_may_clobber_ref_p): Likewise.
14543 (stmt_may_clobber_ref_p_1): Likewise.
14544 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
14546 (ref_maybe_used_by_stmt_p): Likewise.
14547 (stmt_may_clobber_ref_p): Likewise.
14548 (stmt_may_clobber_ref_p_1): Likewise.
14549 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
14550 the alias oracle to disambiguate DRs with stmts DR analysis
14552 (vect_analyze_data_refs): Do not give up on not analyzable
14553 DRs for BB vectorization. Remove code truncating the dataref
14556 2018-05-25 Jakub Jelinek <jakub@redhat.com>
14559 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14560 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
14561 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
14563 2018-05-25 Richard Biener <rguenther@suse.de>
14565 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
14566 function, combining stmt data ref gathering and fatal analysis
14568 (vect_analyze_data_refs): Remove now redudnant code and simplify.
14569 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
14570 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
14571 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
14572 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
14574 2018-05-25 Bin Cheng <bin.cheng@arm.com>
14576 PR tree-optimization/85720
14577 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
14578 SCC if all partitions are builtins.
14579 (version_loop_by_alias_check): New parameter. Generate cancelable
14580 runtime alias check if all partitions are builtins.
14581 (distribute_loop): Update call to above function.
14583 2018-05-25 Bin Cheng <bin.cheng@arm.com>
14585 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
14586 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
14587 (parm_default_def_partition_arg): Ditto.
14588 (set_parm_default_def_partition): Ditto.
14589 (get_parm_default_def_partitions): Ditto and make it static.
14590 (get_undefined_value_partitions): Ditto and make it static.
14591 (remove_ssa_form): Refactor call to init_var_map here.
14592 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
14593 computation for loop region.
14594 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
14595 (register_default_def): Delete.
14596 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
14597 (parm_default_def_partition_arg): Ditto.
14598 (set_parm_default_def_partition): Ditto.
14599 (get_parm_default_def_partitions): Ditto and make it static.
14600 (get_undefined_value_partitions): Ditto and make it static.
14601 (coalesce_with_default, coalesce_with_default): Update comment.
14602 (create_coalesce_list_for_region): New func factored out from
14603 create_outofssa_var_map.
14604 (populate_coalesce_list_for_outofssa): New func factored out from
14605 create_outofssa_var_map and coalesce_ssa_name.
14606 (create_outofssa_var_map): Delete.
14607 (coalesce_ssa_name): Refactor to support live range computation.
14608 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
14609 (get_parm_default_def_partitions): Delete.
14610 (get_undefined_value_partitions): Ditto.
14611 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
14612 computation for loop region.
14613 (new_tree_live_info, loe_visit_block): Ditto.
14614 (live_worklist, set_var_live_on_entry): Ditto.
14615 (calculate_live_on_exit, verify_live_on_entry): Ditto.
14616 * tree-ssa-live.h (struct _var_map): New fields.
14617 (init_var_map): Change decl.
14618 (region_contains_p): New.
14620 2018-05-25 Bin Cheng <bin.cheng@arm.com>
14622 * tree-ssa-live.h (live_merge_and_clear): Delete.
14624 2018-05-25 Richard Biener <rguenther@suse.de>
14627 * tree-dump.c (dequeue_and_dump): Remove access to removed
14628 operand 2 of a SWITCH_EXPR.
14630 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14632 * doc/sourcebuild.texi (vect_double_cond_arith): Include
14633 multiplication and division.
14634 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
14635 (cond_udiv@var{m}, cond_umod@var{m}): Document.
14636 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
14637 (cond_udiv_optab, cond_umod_optab): New optabs.
14638 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
14639 (IFN_COND_RDIV): New internal functions.
14640 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
14641 TRUNC_MOD_EXPR and RDIV_EXPR.
14642 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
14643 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
14645 (SVE_INT_BINARY): Include mult.
14646 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
14647 (optab, sve_int_op): Handle mult.
14648 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
14650 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
14651 for SVE_INT_BINARY_SD.
14653 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14655 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
14656 (optab, sve_int_op): Handle div and udiv.
14657 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
14658 for SVE_INT_BINARY_SD.
14659 (*<optab><mode>3): New insn for the same.
14661 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14663 * tree-vect-patterns.c: Include predict.h.
14664 (vect_recog_divmod_pattern): Restrict check for division support
14665 to when optimizing for size.
14667 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14669 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
14670 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
14671 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
14672 (gimple_match_op::set_op): Likewise.
14673 (gimple_resimplify4): Declare.
14674 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
14675 (expr::gen_transform): Likewise.
14676 (decision_tree::gen): Generate a simplification routine for 4 operands.
14677 * gimple-match-head.c (gimple_simplify): Add an overload for
14678 4 operands. In the top-level function, handle up to 4 call
14679 arguments and call gimple_resimplify4.
14680 (gimple_resimplify4): New function.
14681 (build_call_internal): Pass a fourth operand.
14682 (maybe_push_to_seq): Likewise.
14683 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
14684 Fold VEC_COND_EXPRs of an operation and a default value into
14685 an IFN_COND_* function if possible.
14686 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
14688 (SVE_COND_FP_BINARY): Include them.
14689 (optab, sve_fp_op): Handle them.
14690 (SVE_INT_BINARY_REV): New code iterator.
14691 (SVE_COND_FP_BINARY_REV): New int iterator.
14692 (commutative): New int attribute.
14693 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
14695 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
14697 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
14698 (*cond_<optab><mode>): New patterns for reversed operands.
14700 2018-05-25 Richard Biener <rguenther@suse.de>
14702 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
14703 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
14704 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
14705 (STMT_VINFO_GROUPED_ACCESS): Adjust.
14706 * tree-vect-data-refs.c (everywhere): Adjust users.
14707 * tree-vect-loop.c (everywhere): Likewise.
14708 * tree-vect-slp.c (everywhere): Likewise.
14709 * tree-vect-stmts.c (everywhere): Likewise.
14710 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14712 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14714 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
14716 (gcc_cv_as_section_exclude): ... this.
14717 Try Solaris as #exclude syntax.
14718 * configure: Regenerate.
14719 * config.in: Regenerate.
14720 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
14722 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
14723 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
14725 * varasm.c (default_elf_asm_named_section): Don't check if
14726 HAVE_GAS_SECTION_EXCLUDE is defined.
14728 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14730 * doc/md.texi: Update the documentation of the cond_* optabs
14731 to mention the new final operand. Fix GET_MODE_NUNITS call.
14732 Describe the scalar case too.
14733 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
14734 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
14736 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
14737 (get_conditional_internal_fn): Update comment.
14738 * tree-vect-loop.c (vectorizable_reduction): Pass the original
14739 accumulator value as a final argument to conditional functions.
14740 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
14741 a define_expand and add an "else" operand. Assert for now that
14742 the else operand is equal to operand 2. Use SVE_INT_BINARY and
14743 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
14744 (*cond_<optab><mode>): New patterns.
14745 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
14746 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14747 (UNSPEC_COND_EOR): Delete.
14748 (optab): Remove associated mappings.
14749 (SVE_INT_BINARY): New code iterator.
14750 (sve_int_op): Remove int attribute and add "minus" to the code
14752 (SVE_COND_INT_OP): Delete.
14753 (SVE_COND_FP_OP): Rename to...
14754 (SVE_COND_FP_BINARY): ...this.
14756 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14758 * optabs.c (can_reuse_operands_p): New function.
14759 (maybe_legitimize_operands): Try to reuse the results for
14762 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14764 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14765 Add {q} suffix to insn mnemonic.
14767 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14769 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14770 (msp430_warn_func_return): New.
14772 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
14774 * fold-const.c (tree_nonzero_bits): New function.
14775 * fold-const.h (tree_nonzero_bits): Likewise.
14776 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14777 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14779 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
14783 * varasm.c (assemble_alias): Check ifunc_resolver only on
14786 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14789 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14790 when memory input operand is handled.
14792 2018-05-24 Luis Machado <luis.machado@linaro.org>
14794 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14796 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14798 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14800 * match.pd: Delay FMA folds until after vectorization.
14802 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
14805 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14806 address check not strict.
14808 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14810 * gimple-match.h (gimple_match_op): New class.
14811 (mprts_hook): Replace parameters with a gimple_match_op *.
14812 (maybe_build_generic_op): Likewise.
14813 (gimple_simplified_result_is_gimple_val): Replace parameters with
14814 a const gimple_match_op *.
14815 (gimple_simplify): Replace code_helper * and tree * parameters with
14816 a gimple_match_op * parameter.
14817 (gimple_resimplify1): Replace code_helper *, tree and tree *
14818 parameters with a gimple_match_op * parameter.
14819 (gimple_resimplify2): Likewise.
14820 (gimple_resimplify3): Likewise.
14821 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14822 parameters with a gimple_match_op * parameter.
14823 * gimple-match-head.c (gimple_simplify): Change prototypes of
14824 auto-generated functions to take a gimple_match_op * instead of
14825 separate code_helper * and tree * parameters. Make the same
14826 change in the top-level overload and update calls to the
14827 gimple_resimplify routines. Update calls to the auto-generated
14828 functions and to maybe_push_res_to_seq in the publicly-facing
14829 operation-specific gimple_simplify overloads.
14830 (gimple_match_op::MAX_NUM_OPS): Define.
14831 (gimple_resimplify1): Replace rcode and ops with a single res_op
14832 parameter. Update call to gimple_simplify.
14833 (gimple_resimplify2): Likewise.
14834 (gimple_resimplify3): Likewise.
14835 (mprts_hook): Replace parameters with a gimple_match_op *.
14836 (maybe_build_generic_op): Likewise.
14837 (build_call_internal): Replace type, nargs and ops with
14838 a gimple_match_op *.
14839 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
14840 with a single gimple_match_op *. Update calls to mprts_hook,
14841 build_call_internal and gimple_simplified_result_is_gimple_val.
14842 Factor out code that is common to the tree_code and combined_fn cases.
14843 * genmatch.c (expr::gen_transform): Replace tem_code and
14844 tem_ops with a gimple_match_op called tem_op. Update calls
14845 to the gimple_resimplify functions and maybe_push_res_to_seq.
14846 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
14847 res_ops. Update call to the gimple_resimplify functions.
14848 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
14849 (decision_tree::gen): Make the functions take a gimple_match_op *
14850 called res_op instead of separate res_code and res_ops parameters.
14851 Update call accordingly.
14852 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
14853 and ops with a single res_op parameter. Update calls to
14854 maybe_build_generic_op and maybe_push_res_to_seq.
14855 (fold_stmt_1): Update calls to gimple_simplify and
14856 replace_stmt_with_simplification.
14857 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
14858 and gimple_simplified_result_is_gimple_val.
14859 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
14861 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
14862 with a gimple_match_op *.
14863 (vn_nary_build_or_lookup): Likewise. Update call to
14864 vn_nary_build_or_lookup_1.
14865 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
14866 gimple_match_op *. Update calls to the gimple_resimplify routines
14867 and to gimple_simplified_result_is_gimple_val.
14868 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14869 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14870 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14871 (visit_nary_op): Likewise.
14872 (visit_reference_op_load): Likewise.
14874 2018-05-23 Luis Machado <luis.machado@linaro.org>
14876 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14877 modifier for printing the step amount.
14879 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
14882 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14885 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
14887 * doc/sourcebuild.texi (Endianness): New subsubsection.
14889 2018-05-23 Luis Machado <luis.machado@linaro.org>
14891 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14892 <prefetch_dynamic_strides>: New const bool field.
14893 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14894 prefetch_dynamic_strides.
14895 (exynosm1_prefetch_tune): Likewise.
14896 (thunderxt88_prefetch_tune): Likewise.
14897 (thunderx_prefetch_tune): Likewise.
14898 (thunderx2t99_prefetch_tune): Likewise.
14899 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14901 (aarch64_override_options_internal): Update to set
14902 PARAM_PREFETCH_DYNAMIC_STRIDES.
14903 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14904 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14905 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14906 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14907 prefetch-dynamic-strides setting.
14909 2018-05-23 Luis Machado <luis.machado@linaro.org>
14911 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14912 <minimum_stride>: New const int field.
14913 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14914 minimum_stride field defaulting to -1.
14915 (exynosm1_prefetch_tune): Likewise.
14916 (thunderxt88_prefetch_tune): Likewise.
14917 (thunderx_prefetch_tune): Likewise.
14918 (thunderx2t99_prefetch_tune): Likewise.
14919 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14920 <default_opt_level>: Set to 3.
14921 (aarch64_override_options_internal): Update to set
14922 PARAM_PREFETCH_MINIMUM_STRIDE.
14923 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14924 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14925 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14926 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14927 stride is constant and is below the minimum stride threshold.
14929 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14931 * config/arm/arm-cpus.in (mode26): Delete.
14932 (armv4): Delete mode26 reference.
14933 * config/arm/arm.c (arm_configure_build_target): Delete use of
14936 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14938 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14940 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14941 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
14943 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
14944 (floatunsdidf2): Ditto.
14946 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14948 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14949 (fixuns_trunc<mode>si2_avx512f): Ditto.
14950 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14951 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14952 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14954 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14956 PR rtl-optimization/79985
14957 * df-scan.c (df_insn_refs_collect): Remove special case for
14958 global registers and asm statements.
14960 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14962 * extend.texi (Global Register Variables): Rewrite the bullet list.
14963 Note that the register is available for allocation. Note that access
14964 via inline asm must use constraints. Add note about async-signal
14965 handlers. Remove paragraph about automagic register selection.
14967 2018-05-23 Richard Biener <rguenther@suse.de>
14969 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14970 of fixed offset from memset VN.
14972 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
14974 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14975 first_interp field.
14976 (alloc_cand_and_find_basis): Initialize first_interp field.
14977 (slsr_process_mul): Modify first_interp field.
14978 (slsr_process_add): Likewise.
14979 (slsr_process_cast): Modify first_interp field for each new
14981 (slsr_process_copy): Likewise.
14982 (dump_candidate): Dump first_interp field.
14983 (replace_mult_candidate): Process all interpretations, not just
14985 (replace_rhs_if_not_dup): Likewise.
14986 (replace_one_candidate): Likewise.
14988 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
14990 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14992 (aarch64_needs_frame_chain): New function.
14993 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14995 2018-05-23 Sudakshina Das <sudi.das@arm.com>
14998 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14999 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15000 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15001 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15002 as true for strict-align.
15003 (aarch64_can_inline_p): Perform checks even when callee has no
15004 attributes to check for strict alignment.
15005 * doc/extend.texi (AArch64 Function Attributes): Document
15007 * doc/invoke.texi: (AArch64 Options): Likewise.
15009 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
15011 PR tree-optimization/85853
15012 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15013 the handling of the root of the node to...
15014 (vect_slp_analyze_node_operations_1): ...this new function,
15015 and run the whole thing with the child nodes' def types
15016 set according to their SLP node's def type.
15018 2018-05-23 Richard Biener <rguenther@suse.de>
15020 PR middle-end/85874
15021 * tree-data-ref.c (create_runtime_alias_checks): Defer
15022 and ignore overflow warnings.
15024 2018-05-23 Yury Gribov <tetra2005@gmail.com>
15026 PR tree-optimization/85822
15027 * tree-vrp.c (is_masked_range_test): Fix handling of negative
15030 2018-05-23 Richard Biener <rguenther@suse.de>
15032 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15033 memset constants via native_interpret_expr.
15035 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
15038 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15040 (cgraph_node::create_alias): Likewise.
15041 (cgraph_node::get_availability): Check ifunc_resolver instead
15042 of looking up ifunc attribute.
15043 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15044 * varasm.c (do_assemble_alias): Likewise.
15045 (assemble_alias): Likewise.
15046 (default_binds_local_p_3): Likewise.
15047 * cgraph.h (cgraph_node): Add ifunc_resolver.
15048 (cgraph_node::only_called_directly_or_aliased_p): Return false
15049 for IFUNC resolver.
15050 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
15052 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
15053 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
15054 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
15055 instead of looking up ifunc attribute.
15057 2018-05-22 Luis Machado <luis.machado@linaro.org>
15059 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
15061 2018-05-22 Martin Sebor <msebor@redhat.com>
15063 PR middle-end/85359
15064 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
15065 only when expasion succeeds.
15066 (expand_builtin_strcmp): Same.
15067 (expand_builtin_strncmp): Same.
15069 2018-05-22 Martin Sebor <msebor@redhat.com>
15071 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
15073 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
15074 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15076 * config/aarch64/aarch64-ldpstp.md: Replace uses of
15077 aarch64_mem_pair_operand with memory_operand and delete operand swapping
15079 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15080 Add check for legitimate_address.
15081 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
15082 (aarch64_swap_ldrstr_operands): New.
15083 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
15086 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
15087 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15089 * config/aarch64/aarch64.md: New patterns to generate stp
15091 (store_pair_sw, store_pair_dw): New patterns to generate stp for
15092 single words and double words.
15093 (load_pair_sw, load_pair_dw): Likewise.
15094 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
15096 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
15098 * config/aarch64/aarch64-ldpstp.md: Modify peephole
15099 for different mode ldpstp and add peephole for merged zero stores.
15100 Likewise for loads.
15101 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15103 (aarch64_gen_store_pair): Rename calls to match new patterns.
15104 (aarch64_gen_load_pair): Rename calls to match new patterns.
15105 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
15106 (load_pair<DREG:mode><DREG2:mode>): ... This.
15107 (store_pair<mode>): Rename to...
15108 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
15109 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
15110 New mode iterators.
15111 (V_INT_EQUIV): Handle SImode.
15112 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
15115 2018-05-22 Martin Sebor <msebor@redhat.com>
15118 * calls.c (maybe_warn_nonstring_arg): Use string length to set
15119 or ajust the presumed bound on an operation to avoid unnecessary
15122 2018-05-22 Martin Sebor <msebor@redhat.com>
15124 PR tree-optimization/85826
15125 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
15126 assuming that a DECL necesarily has a constant size.
15128 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
15130 PR middle-end/85862
15131 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
15133 2018-05-22 Richard Biener <rguenther@suse.de>
15135 PR tree-optimization/85834
15136 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
15137 non-constant and non-zero memset arguments.
15139 2018-05-22 Martin Liska <mliska@suse.cz>
15142 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
15144 2018-05-22 Richard Biener <rguenther@suse.de>
15146 PR tree-optimization/85863
15147 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
15148 comparisons when vectype is specified.
15149 (vectorizable_condition): Do not specify vectype for
15150 vect_is_simple_cond when SLP vectorizing.
15152 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
15155 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
15156 define __ibm128 as long double.
15157 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
15158 as a distinct type when IEEE 128-bit support is enabled.
15159 (init_float128_ieee): Fix up conversions between IFmode and IEEE
15160 128-bit types to use the correct functions.
15161 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
15162 convert between 128-bit floating point types that have different
15163 modes but the same representation, instead of using gen_lowpart to
15165 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
15167 (IFKF_reg): New attributes to give the register constraints for
15169 (extend<mode>tf2_internal): New insns to mark an explicit
15170 conversion between 128-bit floating point types that have a
15171 different mode but share the same representation.
15173 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
15175 PR tree-optimization/85814
15176 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
15177 a null return from get_strinfo when unsharing the next
15178 strinfo in the chain.
15180 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
15183 * varasm.c (weak_finish): Clean up weak_decls.
15185 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15187 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
15188 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
15189 UNSPEC_UADALP values.
15190 * config/aarch64/iterators.md (ABAL): New int iterator.
15193 (sur): Add mappings for the above.
15194 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
15196 (aarch64_<sur>abal<mode>_4): Likewise.
15197 (aarch64_<sur>adalp<mode>_3): Likewise.
15198 (<sur>sadv16qi): New define_expand.
15200 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
15202 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
15203 (*movdf_internal): Ditto.
15204 (*rcpsf2_sse): Ditto.
15205 (*rsqrtsf2_sse): Ditto.
15206 (*sqrt<mode>2_sse): Ditto.
15208 2018-05-21 Tamar Christina <tamar.christina@arm.com>
15210 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
15212 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
15213 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15214 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
15215 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
15217 * config/aarch64/arm_neon.h: Likewise.
15218 * config/aarch64/iterators.md (VQ_I): New.
15220 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
15222 * config.gcc: Add arc/t-multilib-linux to tmake_file for
15224 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
15227 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
15229 * config/nds32/constraints.md (S): New constraint.
15230 * config/nds32/nds32.md (call_internal): Use constraint S.
15231 (call_value_internal): Likewise.
15232 (sibcall_internal): Likewise.
15233 (sibcall_value_internal): Likewise.
15235 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
15236 Chung-Ju Wu <jasonwucj@gmail.com>
15238 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15239 into consideration.
15241 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
15242 Chung-Ju Wu <jasonwucj@gmail.com>
15244 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
15245 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
15246 (nds32_rtx_costs_impl): Simplify.
15247 (nds32_address_cost_impl): Simplify.
15248 (nds32_init_rtx_costs): New function.
15249 (nds32_rtx_costs_speed_prefer): Likewise.
15250 (nds32_rtx_costs_size_prefer): Likewise.
15251 (nds32_address_cost_speed_prefer): Likewise.
15252 (nds32_address_cost_speed_fwprop): Likewise.
15253 (nds32_address_cost_size_prefer): Likewise.
15254 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
15255 * config/nds32/nds32.c (nds32_option_override): Use
15256 nds32_init_rtx_costs function.
15258 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
15260 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
15261 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
15262 (TARGET_PIPELINE_N8): Likewise.
15263 (TARGET_PIPELINE_N10): Likewise.
15264 (TARGET_PIPELINE_N13): Likewise.
15265 (TARGET_PIPELINE_GRAYWOLF): Likewise.
15267 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
15269 * config/nds32/nds32-fpu.md: Update copyright year.
15271 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15273 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
15275 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15277 * config/nds32/nds32.c
15278 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
15279 * config/nds32/nds32.opt (minline-asm-r15): New option.
15281 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15283 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
15285 * config/nds32/nds32.md (abssi2): New pattern.
15287 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
15289 * config/i386/i386.md (rex64namesuffix): New mode attribute.
15290 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
15291 Merge insn pattern from sse_cvtsi2ss<round_name> and
15292 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
15293 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
15294 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
15295 using SWI48 mode iterator.
15296 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
15297 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
15298 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
15299 pattern from sse_cvttss2si<round_saeonly_name>
15300 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
15301 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
15302 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
15303 using SWI48 mode iterator.
15304 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
15305 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
15306 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
15307 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
15308 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
15309 using SWI48 mode iterator.
15310 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
15311 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
15312 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
15313 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
15314 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
15315 SWI48 mode iterator.
15316 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
15317 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
15318 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
15319 pattern from sse_cvttsd2si<round_saeonly_name>
15320 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
15322 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15324 * config/nds32/nds32-md-auxiliary.c
15325 (nds32_valid_smw_lwm_base_p): Refine.
15326 (nds32_output_smw_single_word): Refine.
15327 (nds32_output_smw_double_word): New.
15328 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
15330 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15332 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
15333 (nds32_output_stack_pop): Refine.
15334 (nds32_expand_unaligned_load): Refine.
15335 (nds32_expand_unaligned_store): Refine.
15337 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15338 Chung-Ju Wu <jasonwucj@gmail.com>
15340 * config/nds32/constants.md: Add TP_REGNUM constant.
15341 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
15342 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
15344 * config/nds32/nds32-doubleword.md: Consider flag_pic.
15345 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
15346 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
15347 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
15348 and PIC code generation.
15349 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
15351 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
15353 * config/nds32/nds32.md: Support TLS and PIC.
15354 * config/nds32/nds32.c: Support TLS and PIC.
15355 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
15356 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
15359 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
15361 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
15362 mode with E_ prefix.
15364 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15365 Chung-Ju Wu <jasonwucj@gmail.com>
15367 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
15368 * config/nds32/nds32-md-auxiliary.c
15369 (symbolic_reference_mentioned_p): New.
15370 (nds32_legitimize_ict_address): New.
15371 (nds32_expand_ict_move): New.
15372 (nds32_indirect_call_referenced_p): New.
15373 (nds32_symbol_binds_local_p): Delete.
15374 (nds32_long_call_p): Modify.
15375 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
15376 * config/nds32/nds32-protos.h
15377 (symbolic_reference_mentioned_p): Declare.
15378 (nds32_legitimize_ict_address): Declare.
15379 (nds32_expand_ict_move): Declare.
15380 (nds32_indirect_call_referenced_p): Declare.
15381 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
15382 (nds32_relax_group): Use nds32_ict_const_p as condition.
15383 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
15384 (nds32_asm_file_start): Output ict_model directive in asm code.
15385 (nds32_legitimate_address_p): Consider indirect call.
15386 (nds32_print_operand): Consider indirect call.
15387 (nds32_print_operand_address): Consider indirect call.
15388 (nds32_insert_attributes): Handle "indirect_call" attribute.
15389 (TARGET_LEGITIMATE_ADDRESS_P): Define.
15390 (TARGET_LEGITIMATE_CONSTANT_P): Define.
15391 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
15392 (TARGET_DELEGITIMIZE_ADDRESS): Define.
15393 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
15394 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
15395 (TARGET_ICT_MODEL_SMALL): Define.
15396 (TARGET_ICT_MODEL_LARGE): Define.
15397 * config/nds32/nds32.md (movsi): Consider ict model.
15398 (call, call_value): Consider ict model.
15399 (sibcall, sibcall_value): Consider ict model.
15400 * config/nds32/nds32.opt (mict-model): New option.
15401 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
15404 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
15405 Monk Chiang <sh.chiang04@gmail.com>
15406 Jim Wilson <jimw@sifive.com>
15408 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
15409 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
15410 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
15411 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
15412 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
15413 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
15414 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
15415 compute save_libcall_adjustment properly.
15416 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
15417 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
15418 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
15419 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
15420 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
15421 (ABI_SPEC): Handle mabi=ilp32e.
15422 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
15423 (RVE): Add RVE mask.
15424 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
15425 <-march>: Add rv32e as an example.
15427 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
15430 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
15431 (intra_create_variable_infos): Handle C++ constructors.
15433 2018-05-18 Martin Liska <mliska@suse.cz>
15435 * passes.def: Remove a redundant pass.
15437 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
15440 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
15442 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15444 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
15446 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
15448 (armv2, armv2a, armv3, armv3m): Delete architectures.
15449 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
15450 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
15451 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
15453 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
15454 (*mulsidi3adddi): Likewise.
15455 (mulsidi3): Likewise.
15456 (*mulsidi3_nov6): Likewise.
15457 (umulsidi3): Likewise.
15458 (umulsidi3_nov6): Likewise.
15459 (umaddsidi4): Likewise.
15460 (*umulsidi3adddi): Likewise.
15461 (smulsi3_highpart): Likewise.
15462 (*smulsi3_highpart_nov6): Likewise.
15463 (umulsi3_highpart): Likewise.
15464 (*umulsi3_highpart_nov6): Likewise.
15465 * config/arm/arm.h (arm_arch3m): Delete.
15466 * config/arm/arm.c (arm_arch3m): Delete.
15467 (arm_option_override_internal): Update armv3-related comment.
15468 (arm_configure_build_target): Delete use of isa_bit_mode32.
15469 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
15470 (arm_rtx_costs_internal): Delete check of arm_arch3m.
15471 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
15472 (mulsa3): Likewise.
15473 (mulusa3): Likewise.
15474 * config/arm/arm-protos.h (arm_arch3m): Delete.
15475 * config/arm/arm-tables.opt: Regenerate.
15476 * config/arm/arm-tune.md: Likewise.
15477 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
15478 deleted architectures.
15480 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15482 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
15483 (armv5t, armv5te): New features.
15484 (ARMv5, ARMv5e): Delete fgroups.
15485 (ARMv5t, ARMv5te): Adjust for above changes.
15486 (ARMv6m): Likewise.
15487 (armv5, armv5e): Delete arches.
15488 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
15490 (*call_reg_arm): Likewise.
15491 (*call_value_reg_armv5): Likewise.
15492 (*call_value_reg_arm): Likewise.
15493 (*call_symbol): Likewise.
15494 (*call_value_symbol): Likewise.
15495 (*sibcall_insn): Likewise.
15496 (*sibcall_value_insn): Likewise.
15497 (clzsi2): Likewise.
15498 (prefetch): Likewise.
15499 (define_split and define_peephole2 dependent on arm_arch5):
15501 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
15503 (TARGET_ARM_QBIT): Likewise.
15504 (TARGET_DSP_MULTIPLY): Likewise.
15505 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
15506 (arm_arch5, arm_arch5e): Delete.
15507 (arm_arch5t, arm_arch5te): Declare.
15508 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
15509 (arm_arch5t): Declare.
15510 (arm_option_reconfigure_globals): Update for the above.
15511 (arm_options_perform_arch_sanity_checks): Update comment, replace
15512 use of arm_arch5 with arm_arch5t.
15513 (use_return_insn): Likewise.
15514 (arm_emit_call_insn): Likewise.
15515 (output_return_instruction): Likewise.
15516 (arm_final_prescan_insn): Likewise.
15517 (arm_coproc_builtin_available): Likewise.
15518 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
15519 arm_arch5e with arm_arch5t and arm_arch5te.
15520 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
15521 (arm_arch5t, arm_arch5te): Declare.
15522 * config/arm/arm-tables.opt: Regenerate.
15523 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
15524 * config/arm/t-multilib: Likewise.
15525 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
15526 instead of arm_arch5.
15527 (*call_reg_thumb1): Likewise.
15528 (*call_value_reg_thumb1_v5): Likewise.
15529 (*call_value_reg_thumb1): Likewise.
15530 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
15532 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
15534 2018-05-18 Martin Liska <mliska@suse.cz>
15536 PR gcov-profile/84846
15537 * doc/gcov.texi: Document -t option of gcov tool.
15539 2018-05-18 Martin Liska <mliska@suse.cz>
15541 PR gcov-profile/84846
15542 * gcov.c (print_usage): Add new -t option.
15543 (process_args): Handle the option.
15544 (generate_results): Use stdout as output when requested by
15547 2018-05-18 Martin Liska <mliska@suse.cz>
15549 PR gcov-profile/84846
15550 * coverage.c (coverage_init): Write PWD to .gcno file.
15551 * doc/gcov.texi: Document how working directory is printed.
15552 * gcov-dump.c (dump_gcov_file): Print PWD.
15553 * gcov.c (output_intermediate_file): Likewise.
15554 (read_graph_file): Read PWD string.
15555 (output_lines): Print PWD.
15557 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15559 PR middle-end/85817
15560 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
15561 for retval and return false if all args to phi are zero.
15563 2018-05-18 Richard Biener <rguenther@suse.de>
15565 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
15567 (evrp_dom_walker::before_dom_children): Call it.
15569 2018-05-18 Richard Biener <rguenther@suse.de>
15571 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
15572 results when processing array refs with variable index.
15574 2018-05-18 Toon Moene <toon@moene.org>
15576 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
15577 directly after that of -floop-interchange. Indicate that both
15578 options are enabled by default when specifying -O3.
15580 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15582 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
15583 iterator. Delete separate integer-mode vec_set<mode> expander.
15584 (aarch64_simd_vec_setv2di): Delete.
15585 (vec_setv2di): Delete.
15586 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
15587 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
15588 the "w, r" alternative.
15590 2018-05-18 Martin Liska <mliska@suse.cz>
15592 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
15593 * tree-pass.h (make_pass_lower_switch_O0): New function.
15594 * tree-switch-conversion.c (node_has_low_bound): Remove.
15595 (node_has_high_bound): Likewise.
15596 (node_is_bounded): Likewise.
15597 (class pass_lower_switch): Make it a template type and create
15599 (pass_lower_switch::execute): Add template argument.
15600 (make_pass_lower_switch): New function.
15601 (make_pass_lower_switch_O0): New function.
15602 (do_jump_if_equal): Remove.
15603 (emit_case_nodes): Simplify to just handle all 3 cases and leave
15604 all the hard work to tree optimization passes.
15606 2018-05-18 Martin Liska <mliska@suse.cz>
15608 * dbgcnt.c (limit_low): Renamed from limit.
15609 (limit_high): New variable.
15610 (dbg_cnt_is_enabled): Check for upper limit.
15611 (dbg_cnt): Adjust dumping.
15612 (dbg_cnt_set_limit_by_index): Add new argument for high
15614 (dbg_cnt_set_limit_by_name): Likewise.
15615 (dbg_cnt_process_single_pair): Parse new format.
15616 (dbg_cnt_process_opt): Use strtok.
15617 (dbg_cnt_list_all_counters): Remove 'value' and add
15619 * doc/invoke.texi: Document changes.
15621 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
15623 * doc/sourcebuild.texi (scalar_all_fma): Document.
15624 * tree.def (FMA_EXPR): Delete.
15625 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
15626 * internal-fn.c (ternary_direct): New macro.
15627 (expand_ternary_optab_fn): Likewise.
15628 (direct_ternary_optab_supported_p): Likewise.
15629 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
15630 * builtins.c (fold_builtin_fma): Delete.
15631 (fold_builtin_3): Don't call it.
15632 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
15633 * expr.c (expand_expr_real_2): Likewise.
15634 * fold-const.c (operand_equal_p): Likewise.
15635 (fold_ternary_loc): Likewise.
15636 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
15637 * gimple.c (DEFTREECODE): Likewise.
15638 * gimplify.c (gimplify_expr): Likewise.
15639 * optabs-tree.c (optab_for_tree_code): Likewise.
15640 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
15641 * tree-eh.c (operation_could_trap_p): Likewise.
15642 (stmt_could_throw_1_p): Likewise.
15643 * tree-inline.c (estimate_operator_cost): Likewise.
15644 * tree-pretty-print.c (dump_generic_node): Likewise.
15645 (op_code_prio): Likewise.
15646 * tree-ssa-loop-im.c (stmt_cost): Likewise.
15647 * tree-ssa-operands.c (get_expr_operands): Likewise.
15648 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
15649 * fold-const-call.h (fold_fma): Delete.
15650 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
15651 CFN_FNMA and CFN_FNMS.
15652 (fold_fma): Delete.
15653 * genmatch.c (combined_fn): New enum.
15654 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
15655 (commutative_op): New function.
15656 (commutate): Use it. Handle more than 2 operands.
15657 (dt_operand::gen_gimple_expr): Use commutative_op.
15658 (parser::parse_expr): Allow :c to be used with non-binary
15659 operators if the commutative operand is known.
15660 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
15661 CFN_FMS, CFN_FNMA and CFN_FNMS.
15662 (backprop::process_assign_use): Remove FMA_EXPR handling.
15663 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
15664 (gen_hsa_fma): New function.
15665 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
15666 IFN_FNMA and IFN_FNMS.
15667 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
15668 * gimple-fold.h (follow_all_ssa_edges): Declare.
15669 * gimple-fold.c (follow_all_ssa_edges): New function.
15670 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
15671 gimple_build interface and use follow_all_ssa_edges to fold the result.
15672 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
15673 instead of checking for optabs directly.
15674 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
15675 rather than FMA_EXPRs.
15676 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
15677 call to IFN_FMA instead of an FMA_EXPR.
15679 2018-05-17 Jim Wilson <jimw@sifive.com>
15681 * expr.c (do_tablejump): When converting index to Pmode, if we have a
15682 sign extended promoted subreg, and the range does not have the sign bit
15683 set, then do a sign extend.
15685 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
15686 test, check for sign extended subreg and/or constant operands, and
15687 do a sign extend in that case.
15689 2018-05-17 Steve Ellcey <sellcey@cavium.com>
15691 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
15692 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
15694 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
15695 Change logics_shift_reg to logics_shift_imm.
15696 (thunderx2t99_fp_loadpair_basic): Delete.
15697 (thunderx2t99_fp_storepair_basic): Delete.
15698 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
15699 (thunderx2t99_asimd_polynomial): Delete.
15700 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
15701 and neon_fp_mul_d_scalar_q.
15702 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15703 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
15704 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
15705 (thunderx2t99_asimd_lut): Add missing tbl types.
15706 (thunderx2t99_asimd_ext): Delete.
15707 (thunderx2t99_asimd_load1_1_mult): Delete.
15708 (thunderx2t99_asimd_load1_2_mult): Delete.
15709 (thunderx2t99_asimd_load1_ldp): New.
15710 (thunderx2t99_asimd_load1): New.
15711 (thunderx2t99_asimd_load2): Add missing *load2* types.
15712 (thunderx2t99_asimd_load3): New.
15713 (thunderx2t99_asimd_load4): New.
15714 (thunderx2t99_asimd_store1_1_mult): Delete.
15715 (thunderx2t99_asimd_store1_2_mult): Delete.
15716 (thunderx2t99_asimd_store2_mult): Delete.
15717 (thunderx2t99_asimd_store2_onelane): Delete.
15718 (thunderx2t99_asimd_store_stp): New.
15719 (thunderx2t99_asimd_store1): New.
15720 (thunderx2t99_asimd_store2): New.
15721 (thunderx2t99_asimd_store3): New.
15722 (thunderx2t99_asimd_store4): New.
15724 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
15726 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
15727 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
15729 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
15730 Segher Boessenkool <segher@kernel.crashing.org>
15733 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
15736 2018-05-17 Richard Biener <rguenther@suse.de>
15738 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
15739 for pruning loop and prune defs feeding only already visited PHIs.
15741 2018-05-17 Richard Biener <rguenther@suse.de>
15743 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
15745 2018-05-17 Bin Cheng <bin.cheng@arm.com>
15746 Richard Biener <rguenther@suse.de>
15748 PR tree-optimization/85793
15749 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15750 for VMAT_ELEMENTWISE.
15752 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15754 * internal-fn.h (lookup_internal_fn): Declare
15755 * internal-fn.c (lookup_internal_fn): New function.
15756 * gimple.c (gimple_build_call_from_tree): Handle calls to
15757 internal functions.
15758 * gimple-pretty-print.c (dump_gimple_call): Print "." before
15759 internal function names.
15760 * tree-pretty-print.c (dump_generic_node): Likewise.
15761 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15763 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15765 * gimple-fold.h (gimple_build): Make the function forms take
15766 combined_fn rather than built_in_function.
15767 (gimple_simplify): Likewise.
15768 * gimple-match-head.c (gimple_simplify): Likewise.
15769 * gimple-fold.c (gimple_build): Likewise.
15770 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15771 rather than gimple_build_call_internal.
15772 (get_initial_defs_for_reduction): Likewise.
15773 (vect_create_epilog_for_reduction): Likewise.
15774 (vectorizable_live_operation): Likewise.
15776 2018-05-17 Martin Liska <mliska@suse.cz>
15778 * gimple-ssa-sprintf.c (format_directive): Do not use
15779 space in between 'G_' and '('.
15781 2018-05-17 Jakub Jelinek <jakub@redhat.com>
15784 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15785 even if the mask is not all ones.
15788 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15790 (ix86_gimple_fold_builtin): Likewise.
15793 * config/i386/i386.c: Include tree-vector-builder.h.
15794 (ix86_vector_shift_count): New function.
15795 (ix86_fold_builtin): Fold shift builtins by scalar count.
15796 (ix86_gimple_fold_builtin): Likewise.
15798 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15799 _mm512_setzero): New intrinsics.
15801 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
15802 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15804 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15805 code generation for cases where splatting a value is not useful.
15806 * simplify-rtx.c (simplify_ternary_operation): Simplify
15807 vec_merge across a vec_duplicate and a paradoxical subreg forming
15808 a vector mode to a vec_concat.
15810 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
15812 * config.gcc: Support "goldmont-plus".
15813 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15815 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15816 PROCESSOR_GOLDMONT_PLUS.
15817 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15818 (processor_target_table): Add "goldmont-plus".
15819 (PTA_GOLDMONT_PLUS): Define.
15820 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15821 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15822 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15823 (fold_builtin_cpu): Add "goldmont-plus".
15824 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15825 (ix86_option_override_internal): Add "goldmont-plus".
15826 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15827 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15828 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15829 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15831 2018-05-17 Richard Biener <rguenther@suse.de>
15833 PR tree-optimization/85757
15834 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
15835 remove defs that only feed that PHI from further processing.
15837 2018-05-16 Jim Wilson <jimw@sifive.com>
15839 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
15841 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
15843 2018-05-16 Mark Wielaard <mark@klomp.org>
15845 * dwarf2out.c (count_index_strings): New function.
15846 (output_indirect_strings): Call count_index_strings and generate
15847 header for dwarf_version >= 5.
15849 2018-05-16 Mark Wielaard <mark@klomp.org>
15851 * dwarf2out.c (dwarf_FORM): New function.
15852 (set_indirect_string): Use dwarf_FORM.
15853 (reset_indirect_string): Likewise.
15854 (size_of_die): Likewise.
15855 (value_format): Likewise.
15856 (output_die): Likewise.
15857 (add_skeleton_AT_string): Likewise.
15858 (output_macinfo_op): Likewise.
15859 (index_string): Likewise.
15860 (output_index_string_offset): Likewise.
15861 (output_index_string): Likewise.
15862 (count_index_strings): Likewise.
15864 2018-05-16 Carl Love <cel@us.ibm.com>
15866 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15867 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15869 2018-05-16 Martin Jambor <mjambor@suse.cz>
15871 * ipa-prop.c (ipa_free_all_edge_args): Remove.
15872 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15874 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
15876 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15877 (fnma<mode>4): Likewise.
15878 (fms<mode>4): Likewise.
15879 (fnms<mode>4): Likewise.
15880 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15881 (aarch64_fnma<mode>4): Likewise.
15882 (aarch64_fms<mode>4): Likewise.
15883 (aarch64_fnms<mode>4): Likewise.
15884 (aarch64_fnmadd<mode>4): Likewise.
15886 2018-05-16 Jason Merrill <jason@redhat.com>
15888 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
15890 2018-05-16 Richard Biener <rguenther@suse.de>
15892 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15893 (dump_stmt_cost): Declare.
15894 (add_stmt_cost): Dump cost we add.
15895 (add_stmt_costs): New function.
15896 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15897 No longer exported.
15898 (vect_analyze_stmt): Adjust prototype.
15899 (vectorizable_condition): Likewise.
15900 (vectorizable_live_operation): Likewise.
15901 (vectorizable_reduction): Likewise.
15902 (vectorizable_induction): Likewise.
15903 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15904 cost vector to pass to vectorizable_ and record afterwards.
15905 (vect_model_reduction_cost): Take cost vector argument and adjust.
15906 (vect_model_induction_cost): Likewise.
15907 (vectorizable_reduction): Likewise.
15908 (vectorizable_induction): Likewise.
15909 (vectorizable_live_operation): Likewise.
15910 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15911 SLP_TREE_NUMBER_OF_VEC_STMTS.
15912 (vect_analyze_slp_cost_1): Remove.
15913 (vect_analyze_slp_cost): Likewise.
15914 (vect_slp_analyze_node_operations): Take visited args and
15915 a target cost vector. Avoid processing already visited stmt sets.
15916 (vect_slp_analyze_operations): Use a local cost vector to gather
15917 costs and register those of non-discarded instances.
15918 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15919 (vect_schedule_slp_instance): Remove copying of
15920 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
15922 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15923 adding cost. Record cost entry location.
15924 (vect_prologue_cost_for_slp_op): Function to compute cost of
15925 a constant or invariant generated for SLP vect in the prologue,
15926 split out from vect_analyze_slp_cost_1.
15927 (vect_model_simple_cost): Make static. Adjust for SLP costing.
15928 (vect_model_promotion_demotion_cost): Likewise.
15929 (vect_model_store_cost): Likewise, make static.
15930 (vect_model_load_cost): Likewise.
15931 (vectorizable_bswap): Add cost vector arg and adjust.
15932 (vectorizable_call): Likewise.
15933 (vectorizable_simd_clone_call): Likewise.
15934 (vectorizable_conversion): Likewise.
15935 (vectorizable_assignment): Likewise.
15936 (vectorizable_shift): Likewise.
15937 (vectorizable_operation): Likewise.
15938 (vectorizable_store): Likewise.
15939 (vectorizable_load): Likewise.
15940 (vectorizable_condition): Likewise.
15941 (vectorizable_comparison): Likewise.
15942 (can_vectorize_live_stmts): Likewise.
15943 (vect_analyze_stmt): Likewise.
15944 (vect_transform_stmt): Adjust calls to vectorizable_*.
15945 * tree-vectorizer.c: Include gimple-pretty-print.h.
15946 (dump_stmt_cost): New function.
15948 2018-05-16 Richard Biener <rguenther@suse.de>
15950 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15951 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15952 * tree-ssa-dse.c: Include tree-ssa-loop.h.
15953 (check_name): New callback.
15954 (dse_classify_store): Track cycles via a visited bitmap of PHI
15955 defs and simplify handling of in-loop and across loop dead stores
15956 and properly fail for loop-variant refs. Handle byte-tracking with
15957 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15960 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
15962 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15963 (vect_get_mask_type_for_stmt): Likewise.
15964 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15966 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
15967 to determine the statement's vector type and the vector type that
15968 should be used for calculating nunits. Deal with cases in which
15969 the type has to be deferred.
15970 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15971 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15972 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15973 (vect_determine_vf_for_stmt): New functions, split out from...
15974 (vect_determine_vectorization_factor): ...here.
15975 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15976 (vect_get_mask_type_for_stmt): New functions, split out from
15977 vect_determine_vectorization_factor.
15979 2018-05-16 Richard Biener <rguenther@suse.de>
15981 * tree-cfg.c (verify_gimple_assign_ternary): Properly
15982 verify the [VEC_]COND_EXPR embedded comparison.
15984 2018-05-15 Martin Sebor <msebor@redhat.com>
15986 PR tree-optimization/85753
15987 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15988 RECORD_TYPE in addition to ARRAY_TYPE.
15990 2018-05-15 Martin Sebor <msebor@redhat.com>
15992 PR middle-end/85643
15993 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15995 2018-05-15 Richard Biener <rguenther@suse.de>
15997 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15998 add by_clobber_p one. Change algorithm to collect all defs
15999 representing uses we need to walk and try reducing them to
16000 a single one before failing.
16001 (dse_dom_walker::dse_optimize_stmt): Adjust.
16003 2018-05-13 Mark Wielaard <mark@klomp.org>
16005 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16006 (size_of_loc_descr): Likewise.
16007 (output_loc_operands): Likewise.
16008 (output_loc_operands_raw): Likewise.
16009 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16010 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16011 (hash_loc_operands): Likewise.
16012 (compare_loc_operands): Likewise.
16014 2018-05-14 Mark Wielaard <mark@klomp.org>
16016 * dwarf2out.c (count_index_addrs): New function.
16017 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16019 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16021 PR tree-optimization/83648
16022 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16023 return value as malloc candidate.
16025 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16028 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16029 param as true in call to suggest_attribute.
16031 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
16033 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16034 -mreadonly-in-sdata.
16036 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16038 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16040 (aarch64_crypto_aesd_fused): Likewise.
16042 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
16044 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
16045 (movsi_aarch64): Likewise.
16046 (load_pairsi): Likewise.
16047 (load_pairdi): Likewise.
16048 (store_pairsi): Likewise.
16049 (store_pairdi): Likewise.
16050 (load_pairsf): Likewise.
16051 (load_pairdf): Likewise.
16052 (store_pairsf): Likewise.
16053 (store_pairdf): Likewise.
16054 (zero_extend): Likewise.
16055 (trunc): Swap alternatives.
16056 (fcvt_target): Add '?' to prefer w over r.
16058 2018-05-14 Jakub Jelinek <jakub@redhat.com>
16061 * config/i386/i386.md: Disallow non-commutative arithmetics in
16062 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
16063 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
16064 in the peephole2 before it.
16066 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
16068 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
16069 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
16070 (ix86_handle_option): Handle -mcldemote.
16071 * config.gcc: New header.
16072 * config/i386/cldemoteintrin.h: New file.
16073 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
16074 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16076 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16077 OPTION_MASK_ISA_CLDEMOTE.
16078 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
16079 (ix86_valid_target_attribute_inner_p): Ditto.
16080 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
16081 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
16082 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
16083 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
16084 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
16086 * config/i386/i386.opt: Add -mcldemote.
16087 * config/i386/x86intrin.h: New header.
16088 * doc/invoke.texi: Add -mcldemote.
16090 2018-05-14 Richard Biener <rguenther@suse.de>
16092 * doc/match-and-simplify.texi: Adjust :s documentation.
16094 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
16096 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
16097 intended memcpy size.
16098 (REORDER_45): Likewise.
16100 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
16102 * sort.cc: New file.
16103 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
16104 * vec.c (qsort_chk): Use gcc_qsort.
16105 * Makefile.in (OBJS-libcommon): Add sort.o.
16106 (build/sort.o): New target. Use it...
16107 (BUILD_RTL): ... here, and...
16108 (build/gencfn-macros): ... here, and...
16109 (build/genmatch): ... here.
16111 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16112 Chung-Ju Wu <jasonwucj@gmail.com>
16114 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
16115 * config/nds32/nds32-graywolf.md: New file.
16116 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
16117 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
16119 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
16120 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
16121 * config/nds32/nds32.md (pipeline_model): Add graywolf.
16122 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
16123 * config/nds32/pipelines.md: Include n15 settings.
16125 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16126 Chung-Ju Wu <jasonwucj@gmail.com>
16128 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
16129 * config/nds32/nds32-n13.md: New file.
16130 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
16131 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
16133 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
16134 * config/nds32/nds32.md (pipeline_model): Add n13.
16135 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
16136 * config/nds32/pipelines.md: Include n13 settings.
16138 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
16139 Chung-Ju Wu <jasonwucj@gmail.com>
16141 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
16142 * config/nds32/nds32-n10.md: New file.
16143 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
16144 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
16146 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
16147 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
16148 * config/nds32/nds32.md (pipeline_model): Add n10.
16149 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
16150 * config/nds32/pipelines.md: Include n10 settings.
16152 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
16153 Kito Cheng <kito.cheng@gmail.com>
16154 Chung-Ju Wu <jasonwucj@gmail.com>
16156 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
16157 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16158 Add enum values for DSP extension instructions.
16159 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
16161 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
16162 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
16163 New code iterators.
16164 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
16165 * config/nds32/nds32-dspext.md: New file for DSP implementation.
16166 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
16167 * config/nds32/nds32-intrinsic.md: Likewise.
16168 * config/nds32/nds32_intrinsic.h: Likewise.
16169 * config/nds32/nds32-md-auxiliary.c: Likewise.
16170 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
16171 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
16172 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
16173 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
16174 * config/nds32/nds32-protos.h: New declarations for DSP extension.
16175 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
16176 TYPE_DMAC in switch statement.
16177 * config/nds32/nds32.c: New checking and implementation for DSP
16178 extension instructions.
16179 * config/nds32/nds32.h: Likewise.
16180 * config/nds32/nds32.md: Likewise.
16181 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
16182 * config/nds32/predicates.md: Implement new predicates for DSP
16185 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16187 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
16188 Reformat alternatives and attributes so it is easier to identify
16189 which constraints/attributes go with which instruction.
16190 (mov<mode>_hardfloat32, FMOVE64): Likewise.
16191 (mov<mode>_softfloat32, FMOVE64): Likewise.
16192 (mov<mode>_hardfloat64, FMOVE64): Likewise.
16193 (mov<mode>_softfloat64, FMOVE64): Likewise.
16195 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
16197 * doc/extend.texi (PowerPC Built-in Functions): Rename this
16199 (Basic PowerPC Built-in Functions): The new name of the
16200 subsection previously known as "PowerPC Built-in Functions".
16201 (Basic PowerPC Built-in Functions Available on all Configurations):
16203 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
16204 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
16205 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
16206 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
16208 2018-05-11 Martin Jambor <mjambor@suse.cz>
16211 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
16214 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
16217 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
16219 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
16221 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
16222 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
16223 (ix86_handle_option): Handle -mwaitpkg.
16224 * config.gcc: New header.
16225 * config/i386/cpuid.h (bit_WAITPKG): New bit.
16226 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
16227 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
16229 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16230 OPTION_MASK_ISA_WAITPKG.
16231 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
16232 (ix86_option_override_internal): Add PTA_WAITPKG.
16233 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
16234 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
16235 IX86_BUILTIN_TPAUSE.
16236 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
16237 __builtin_ia32_umwait and __builtin_ia32_tpause.
16238 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
16239 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
16240 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
16241 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
16242 UNSPECV_TPAUSE): New.
16243 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
16244 * config/i386/i386.opt: Add -mwaitpkg.
16245 * config/i386/waitpkgintrin.h: New file.
16246 * config/i386/x86intrin.h: New header.
16247 * doc/invoke.texi: Add -mwaitpkg.
16249 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
16252 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
16254 (cortex-m0): Use armv6s-m isa.
16255 (cortex-m0plus): Likewise.
16256 (cortex-m1): Likewise.
16257 (cortex-m0.small-multiply): Likewise.
16258 (cortex-m0plus.small-multiply): Likewise.
16259 (cortex-m1.small-multiply): Likewise.
16261 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
16262 Jakub Jelinek <jakub@redhat.com>
16264 PR tree-optimization/85692
16265 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
16266 source permute as well.
16268 2018-05-11 Martin Liska <mliska@suse.cz>
16271 * doc/extend.texi: Document LLVM style format for no_sanitize
16274 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16276 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
16277 mode_supports_vsx_dform_quad to mode_supports_dq_form.
16278 (mode_supports_vsx_dform_quad): Likewise.
16279 (mode_supports_vmx_dform): Move these functions to be next to the
16280 other mode_supports functions.
16281 (mode_supports_dq_form): Likewise.
16282 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
16283 mode_supports_dq_form.
16284 (reg_offset_addressing_ok_p): Likewise.
16285 (offsettable_ok_by_alignment): Likewise.
16286 (rs6000_legitimate_offset_address_p): Likewise.
16287 (legitimate_lo_sum_address_p): Likewise.
16288 (rs6000_legitimize_address): Likewise.
16289 (rs6000_legitimize_reload_address): Likewise.
16290 (rs6000_secondary_reload_inner): Likewise.
16291 (rs6000_preferred_reload_class): Likewise.
16292 (rs6000_output_move_128bit): Likewise.
16294 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
16296 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
16297 Generate SImode target register for null target.
16298 <case IX86_BUILTIN_XGETBV>: Ditto.
16299 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
16300 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
16302 2018-05-10 Carl Love <cel@us.ibm.com>
16304 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16305 dcbtt and dcbtstt if operands[2] is 0.
16307 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
16310 * config/i386/sse.md (usadv64qi): New expander.
16312 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
16314 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
16315 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
16316 -maltivec=be support.
16317 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
16318 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
16319 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
16320 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
16321 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
16322 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
16323 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
16324 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
16325 altivec_vsumsws): Adjust.
16326 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
16327 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
16328 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
16330 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
16331 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
16332 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
16333 (altivec_lve<VI_char>x): Delete expand.
16334 (*altivec_lve<VI_char>x_internal): Rename to...
16335 (altivec_lve<VI_char>x): ... this.
16336 (altivec_lvxl_<mode>): Delete expand.
16337 (*altivec_lvxl_<mode>_internal): Rename to ...
16338 (altivec_lvxl_<mode>): ... this.
16339 (altivec_stvxl_<mode>): Delete expand.
16340 (*altivec_stvxl_<mode>_internal): Rename to ...
16341 (altivec_stvxl_<mode>): ... this.
16342 (altivec_stve<VI_char>x): Delete expand.
16343 (*altivec_stve<VI_char>x_internal): Rename to ...
16344 (altivec_stve<VI_char>x): ... this.
16345 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
16346 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
16347 reduc_plus_scal_<mode>): Adjust.
16348 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
16350 (rs6000_cpu_cpp_builtins): Adjust.
16351 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
16352 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
16353 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
16354 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
16355 -maltivec=be support.
16356 (rs6000_split_vec_extract_var): Adjust.
16357 (rs6000_split_v4si_init): Adjust.
16358 (swap_selector_for_mode): Delete.
16359 (altivec_expand_lvx_be, altivec_expand_stvx_be,
16360 altivec_expand_stvex_be): Delete.
16361 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
16362 -maltivec=be support.
16363 (rs6000_gimple_fold_builtin): Ditto.
16364 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
16366 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
16367 (TARGET_DIRECT_MOVE_64BIT): Adjust.
16368 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
16369 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
16370 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
16371 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
16372 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
16373 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
16374 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
16375 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
16376 anonymous split): Adjust.
16377 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
16378 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
16380 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
16382 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
16383 when --with-gxx-include-dir is also specified.
16384 * configure: Regenerate.
16386 2018-05-09 Jim Wilson <jimw@sifive.com>
16389 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
16390 * config/riscv/t-withmultilib: New.
16391 * config/riscv/withmultilib.h: New.
16392 * doc/install.texi: Document RISC-V --with-multilib-list support.
16394 2018-05-09 Richard Biener <rguenther@suse.de>
16396 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
16398 (vect_bb_vectorization_profitable_p): Adjust. Compute
16399 actual scalar cost using the cost vector and the add_stmt_cost
16402 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16404 PR rtl-optimization/85645
16405 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
16406 in the REG_CFA_REGISTER note for LR, don't leave it empty.
16408 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16410 PR rtl-optimization/85645
16411 * shrink-wrap.c (spread_components): Return a boolean saying if
16412 anything was changed.
16413 (try_shrink_wrapping_separate): Iterate spread_components until
16414 nothing changes anymore.
16416 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16418 PR rtl-optimization/85645
16419 * regrename.c (build_def_use): Also kill the chains that include the
16420 destination of a REG_CFA_REGISTER note.
16422 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
16424 PR rtl-optimization/85645
16425 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
16426 insn that has a REG_CFA_REGISTER note.
16428 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
16430 * cfgexpand.c (expand_clobber): New function.
16431 (expand_gimple_stmt_1): Use it.
16432 * tree-vect-stmts.c (vect_clobber_variable): New function,
16434 (vectorizable_simd_clone_call): ...here.
16435 (vectorizable_store): Emit a clobber either side of an
16436 IFN_STORE_LANES sequence.
16437 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
16439 2018-05-09 Tom de Vries <tom@codesourcery.com>
16442 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
16443 (define_insn "trap_if_false"): Add exit after trap.
16445 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
16447 PR rtl-optimization/85638
16448 * bb-reorder.c: Include common/common-target.h.
16449 (create_forwarder_block): New function extracted from...
16450 (fix_up_crossing_landing_pad): ...here. Rename into...
16451 (dw2_fix_up_crossing_landing_pad): ...this.
16452 (sjlj_fix_up_crossing_landing_pad): New function.
16453 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
16454 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
16455 from both partitions and exit the loop after one iteration.
16457 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
16460 * doc/extend.texi (PowerPC Built-in Functions): Rename this
16462 (Basic PowerPC Built-in Functions): The new name of the
16463 subsection previously known as "PowerPC Built-in Functions".
16464 (Basic PowerPC Built-in Functions Available on all Configurations):
16466 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16468 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16470 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16472 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16475 2018-05-08 Jim Wilson <jimw@sifive.com>
16477 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
16478 (LD_EMUL_SUFFIX): New.
16479 (LINK_SPEC): Use it.
16481 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
16483 * doc/extend.texi (PowerPC Built-in Functions): Rename this
16485 (Basic PowerPC Built-in Functions): The new name of the
16486 subsection previously known as "PowerPC Built-in Functions".
16487 (Basic PowerPC Built-in Functions Available on all Configurations):
16489 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
16491 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
16493 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
16495 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
16498 2018-05-08 Jakub Jelinek <jakub@redhat.com>
16501 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
16502 after cmpelim optimization.
16504 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
16506 * config.gcc: Support "goldmont".
16507 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
16508 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16509 PROCESSOR_GOLDMONT.
16510 * config/i386/i386.c (m_GOLDMONT): Define.
16511 (processor_target_table): Add "goldmont".
16512 (PTA_GOLDMONT): Define.
16513 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
16514 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
16515 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
16516 (fold_builtin_cpu): Add "goldmont".
16517 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
16518 (ix86_option_override_internal): Add "goldmont".
16519 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
16520 (processor_type): Add PROCESSOR_GOLDMONT.
16521 * config/i386/i386.md: Add CPU "glm".
16522 * config/i386/glm.md: New file.
16523 * config/i386/x86-tune.def: Add m_GOLDMONT.
16524 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
16526 2018-05-08 Jakub Jelinek <jakub@redhat.com>
16529 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
16531 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
16532 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
16533 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
16536 * config/i386/i386.c (ix86_fold_builtin): Handle
16537 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
16540 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
16541 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
16543 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
16546 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
16547 (check_arch): Likewise.
16548 (check_fpu): Return the result rather than printing it.
16549 (end arch): Fix operator precedence.
16550 (end cpu): Likewise.
16551 (END): Print the result from check_fpu.
16553 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
16554 Alan Hayward <alan.hayward@arm.com>
16555 David Sherwood <david.sherwood@arm.com>
16557 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
16558 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
16559 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
16560 (*fcmuo<mode>_and): New patterns.
16562 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
16564 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
16565 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
16566 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
16567 (cmp_op, sve_imm_con): New code attributes.
16568 (SVE_COND_INT_CMP, imm_con): Delete.
16569 (cmp_op): Remove above unspecs from int attribute.
16570 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
16572 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
16573 comparison-specific unspecs.
16574 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
16575 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
16576 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
16577 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
16578 (*vec_fcm<cmp_op><mode>): Rename to...
16579 (*fcm<cmp_op><mode>): ...this and adjust likewise.
16580 (*vec_fcmuo<mode>): Rename to...
16581 (*fcmuo<mode>): ...this and adjust likewise.
16582 (*pred_fcm<cmp_op><mode>): New pattern.
16583 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
16584 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
16586 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
16588 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
16589 (aarch64_emit_sve_predicated_cond): New function.
16590 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
16591 (aarch64_emit_unspec_cond_or): Replace with...
16592 (aarch64_emit_sve_or_conds): ...this new function. Use
16593 aarch64_emit_sve_ptrue_op for the individual comparisons and
16594 aarch64_emit_binop to OR them together.
16595 (aarch64_emit_inverted_unspec_cond): Replace with...
16596 (aarch64_emit_sve_inverted_cond): ...this new function. Use
16597 aarch64_emit_sve_ptrue_op for the comparison and
16598 aarch64_emit_unop to invert the result.
16599 (aarch64_expand_sve_vec_cmp_float): Update after the above
16600 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
16602 2018-05-07 Nathan Sidwell <nathan@acm.org>
16604 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
16605 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
16606 (Backwards Compatibility): Likewise.
16608 2018-05-07 Luis Machado <luis.machado@linaro.org>
16612 2018-05-07 Luis Machado <luis.machado@linaro.org>
16614 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16615 <prefetch_dynamic_strides>: New const bool field.
16616 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16617 prefetch_dynamic_strides.
16618 (exynosm1_prefetch_tune): Likewise.
16619 (thunderxt88_prefetch_tune): Likewise.
16620 (thunderx_prefetch_tune): Likewise.
16621 (thunderx2t99_prefetch_tune): Likewise.
16622 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16624 (aarch64_override_options_internal): Update to set
16625 PARAM_PREFETCH_DYNAMIC_STRIDES.
16626 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16627 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16628 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16629 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16630 prefetch-dynamic-strides setting.
16632 2018-05-07 Luis Machado <luis.machado@linaro.org>
16634 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16635 <minimum_stride>: New const int field.
16636 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16637 minimum_stride field.
16638 (exynosm1_prefetch_tune): Likewise.
16639 (thunderxt88_prefetch_tune): Likewise.
16640 (thunderx_prefetch_tune): Likewise.
16641 (thunderx2t99_prefetch_tune): Likewise.
16642 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16643 (aarch64_override_options_internal): Update to set
16644 PARAM_PREFETCH_MINIMUM_STRIDE.
16645 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16646 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16647 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16648 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16649 stride is constant and is below the minimum stride threshold.
16651 2018-05-07 Luis Machado <luis.machado@linaro.org>
16653 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
16656 2018-05-07 Luis Machado <luis.machado@linaro.org>
16658 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16659 <prefetch_dynamic_strides>: New const bool field.
16660 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16661 prefetch_dynamic_strides.
16662 (exynosm1_prefetch_tune): Likewise.
16663 (thunderxt88_prefetch_tune): Likewise.
16664 (thunderx_prefetch_tune): Likewise.
16665 (thunderx2t99_prefetch_tune): Likewise.
16666 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
16668 (aarch64_override_options_internal): Update to set
16669 PARAM_PREFETCH_DYNAMIC_STRIDES.
16670 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16671 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16672 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16673 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16674 prefetch-dynamic-strides setting.
16676 2018-05-07 Luis Machado <luis.machado@linaro.org>
16678 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16679 <minimum_stride>: New const int field.
16680 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16681 minimum_stride field.
16682 (exynosm1_prefetch_tune): Likewise.
16683 (thunderxt88_prefetch_tune): Likewise.
16684 (thunderx_prefetch_tune): Likewise.
16685 (thunderx2t99_prefetch_tune): Likewise.
16686 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16687 (aarch64_override_options_internal): Update to set
16688 PARAM_PREFETCH_MINIMUM_STRIDE.
16689 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16690 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16691 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16692 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16693 stride is constant and is below the minimum stride threshold.
16695 2018-05-06 Jakub Jelinek <jakub@redhat.com>
16698 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
16699 the type is addressable. Don't force op into register if it has
16702 2018-05-05 Roland McGrath <mcgrathr@google.com>
16705 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
16706 any section for which we don't know a specific type it should have,
16707 regardless of name. Previously this was done only for the exact
16708 names ".init_array", ".fini_array", and ".preinit_array".
16709 (default_elf_asm_named_section): Add comment about
16710 relationship with default_section_type_flags and SECTION_NOTYPE.
16711 (get_section): Don't consider it a type conflict if one side has
16712 SECTION_NOTYPE and the other doesn't, as long as neither has the
16713 SECTION_BSS et al used in the default_section_type_flags logic.
16715 2018-05-05 Tom de Vries <tom@codesourcery.com>
16718 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
16719 (workaround_barsyncs): New function.
16720 (nvptx_reorg): Use workaround_barsyncs.
16721 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16722 (define_expand "nvptx_membar_cta"): New define_expand.
16723 (define_insn "*nvptx_membar_cta"): New insn.
16725 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
16727 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
16728 To improve optimization opportunities.
16729 * builtin-types.def: The new needed builtin types for the above.
16731 2018-05-04 Richard Biener <rguenther@suse.de>
16733 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
16734 * gimple-ssa-store-merging.c
16735 (imm_store_chain_info::output_merged_store): Remove redundant create,
16736 release split_store vector contents on failure.
16737 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
16738 scalar stmt vector on cache hit.
16740 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16742 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
16744 * config.gcc (powerpc-xilinx-eabi*): Remove.
16745 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16747 (fusion_addis_mem_combo_load): Ditto.
16748 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16750 (rs6000_cpu_cpp_builtins): Ditto.
16751 * config/rs6000/rs6000-linux.c
16752 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16753 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16754 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16756 (rs6000_setup_reg_addr_masks): Ditto.
16757 (rs6000_init_hard_regno_mode_ok): Ditto.
16758 (rs6000_option_override_internal): Ditto.
16759 (legitimate_lo_sum_address_p): Ditto.
16760 (rs6000_legitimize_address): Ditto.
16761 (rs6000_legitimize_reload_address): Ditto.
16762 (rs6000_legitimate_address_p): Ditto.
16763 (abi_v4_pass_in_fpr): Ditto.
16764 (setup_incoming_varargs): Ditto.
16765 (rs6000_gimplify_va_arg): Ditto.
16766 (rs6000_split_multireg_move): Ditto.
16767 (rs6000_savres_strategy): Ditto.
16768 (rs6000_emit_prologue_components): Ditto.
16769 (rs6000_emit_epilogue_components): Ditto.
16770 (rs6000_emit_prologue): Ditto.
16771 (rs6000_emit_epilogue): Ditto.
16772 (rs6000_elf_file_end): Ditto.
16773 (rs6000_function_value): Ditto.
16774 (rs6000_libcall_value): Ditto.
16775 * config/rs6000/rs6000.h: Ditto.
16776 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16777 (TARGET_MINMAX): ... this. New.
16778 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16779 * config/rs6000/rs6000.md: Remove Xilinx FP support.
16780 (*movsi_internal1_single): Delete.
16781 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16782 mfpu=, mxilinx-fpu): Delete.
16783 * config/rs6000/singlefp.h: Delete.
16784 * config/rs6000/sysv4.h: Remove Xilinx FP support.
16785 * config/rs6000/t-rs6000: Ditto.
16786 * config/rs6000/t-xilinx: Delete.
16787 * config/rs6000/titan.md: Adjust for fp_type removal.
16788 * config/rs6000/vsx.md: Remove Xilinx FP support.
16789 (VStype_simple): Delete.
16790 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16791 * config/rs6000/xfpu.h: Delete.
16792 * config/rs6000/xfpu.md: Delete.
16793 * config/rs6000/xilinx.h: Delete.
16794 * config/rs6000/xilinx.opt: Delete.
16795 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16796 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16798 2018-05-04 Tom de Vries <tom@codesourcery.com>
16801 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16804 2018-05-04 Richard Biener <rguenther@suse.de>
16806 PR middle-end/85627
16807 * tree-complex.c (update_complex_assignment): We are always in SSA form.
16808 (expand_complex_div_wide): Likewise.
16809 (expand_complex_operations_1): Likewise.
16810 (expand_complex_libcall): Preserve EH info of the original stmt.
16811 (tree_lower_complex): Handle removed blocks.
16812 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16813 on complex multiplication and division libcall builtins.
16815 2018-05-04 Richard Biener <rguenther@suse.de>
16817 PR middle-end/85574
16818 * fold-const.c (negate_expr_p): Restrict negation of operand
16819 zero of a division to when we know that can happen without
16821 (fold_negate_expr_1): Likewise.
16823 2018-05-04 Jakub Jelinek <jakub@redhat.com>
16826 * real.h (real_nextafter): Declare.
16827 * real.c (real_nextafter): New function.
16828 * fold-const-call.c (fold_const_nextafter): New function.
16829 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16830 CASE_CFN_NEXTTOWARD.
16831 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16832 even when arg1_mode is different from arg0_mode.
16834 2018-05-03 Nathan Sidwell <nathan@acm.org>
16836 * doc/extend.texi (Deprecated Features): Remove
16837 -ffriend-injection.
16838 (Backwards Compatibility): Likewise.
16839 * doc/invoke.texi (C++ Language Options): Likewise.
16840 (C++ Dialect Options): Likewise.
16842 2018-05-03 Jakub Jelinek <jakub@redhat.com>
16845 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
16846 _mm512_mask_mullox_epi64): New intrinsics.
16848 2018-05-03 Tom de Vries <tom@codesourcery.com>
16851 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16852 dump files): Add offload-tree.
16854 2018-05-03 Richard Biener <rguenther@suse.de>
16856 PR tree-optimization/85615
16857 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
16858 to loops not nested in BBs loop father to avoid creating multi-entry
16861 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16863 PR tree-optimization/70291
16864 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
16865 arguments. Change return type to tree. Emit libcall as a new
16866 statement rather than replacing existing one when inplace_p is true.
16867 (expand_complex_multiplication_components): New function.
16868 (expand_complex_multiplication): Expand floating-point complex
16869 multiplication using the above.
16870 (expand_complex_division): Rename inner_type parameter to type.
16871 Update expand_complex_libcall call-site.
16872 (expand_complex_operations_1): Update expand_complex_multiplication
16873 and expand_complex_division call-sites.
16875 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16878 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16879 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16880 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16881 the highest significant bit of the shift count mask is clear. In
16882 check whether and[sq]i3 is needed verify that all significant bits
16883 of the shift count other than the highest are set.
16885 2018-05-02 Tom de Vries <tom@codesourcery.com>
16888 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16889 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16890 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16891 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16892 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16893 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16894 __builtin_goacc_parlevel_size.
16896 2018-05-02 Richard Biener <rguenther@suse.de>
16898 PR tree-optimization/85597
16899 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16900 do not use split vect_get_vec_defs call but call vect_get_slp_defs
16903 2018-05-02 Tom de Vries <tom@codesourcery.com>
16906 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16907 dump files): Add ltrans-tree.
16909 2018-05-02 Tom de Vries <tom@codesourcery.com>
16912 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16913 dump files): Add wpa-ipa.
16915 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
16917 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
16918 powerpc*-*-linux*paired* target.
16919 * config/rs6000/750cl.h: Delete.
16920 * config/rs6000/paired.h: Delete.
16921 * config/rs6000/paired.md: Delete.
16922 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16924 * config/rs6000/rs6000-builtin.def: Remove paired float support.
16925 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16926 comment. Remove paired float support.
16927 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16928 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16930 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16931 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16933 * config/rs6000/rs6000.c: Remove paired float support.
16934 (paired_expand_vector_init, paired_expand_vector_move,
16935 paired_emit_vector_compare, paired_emit_vector_cond_expr,
16936 (paired_expand_lv_builtin, paired_expand_stv_builtin,
16937 paired_expand_builtin, paired_expand_predicate_builtin,
16938 paired_init_builtins): Delete.
16939 * config/rs6000/rs6000.h: Remove paired float support.
16940 * config/rs6000/rs6000.md: Remove paired float support.
16941 (move_from_CR_ov_bit): Delete.
16942 * config/rs6000/rs6000.opt (mpaired): Delete.
16943 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16944 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16946 2018-05-02 Richard Biener <rguenther@suse.de>
16948 PR middle-end/85567
16949 * gimplify.c (gimplify_save_expr): When in SSA form allow
16950 SAVE_EXPRs to compute to SSA vars.
16952 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16955 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16956 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16957 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16958 clobber operands[2], instead use a new pseudo. Formatting fixes.
16960 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
16962 PR tree-optimization/85586
16963 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16964 exit early for statements in the same group if the accesses are
16967 2018-05-02 Tom de Vries <tom@codesourcery.com>
16970 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16973 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
16975 PR tree-optimization/85143
16976 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16978 2018-05-01 Tom de Vries <tom@codesourcery.com>
16981 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16982 not found" error message.
16984 2018-05-01 Tom de Vries <tom@codesourcery.com>
16987 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16988 * vec.c (test_ordered_remove_if): New function.
16989 (vec_c_tests): Call test_ordered_remove_if.
16990 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16991 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16992 * tree-vect-patterns.c (vect_pattern_recog_1): Use
16993 VEC_ORDERED_REMOVE_IF.
16995 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16997 PR tree-optimization/82665
16998 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16999 pointer subtraction where arguments come from a memchr call.
17001 2018-05-01 Jakub Jelinek <jakub@redhat.com>
17003 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17004 --push-state --as-needed and --pop-state instead of --as-needed and
17005 --no-as-needed if ld supports it.
17006 * configure: Regenerated.
17009 * doc/install.texi2html: Replace _002d with - and _002a with * in
17010 generated html files using sed.
17012 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17015 * gcc-rich-location.c (blank_line_before_p): New function.
17016 (use_new_line): New function.
17017 (gcc_rich_location::add_fixit_insert_formatted): New function.
17018 * gcc-rich-location.h
17019 (gcc_rich_location::add_fixit_insert_formatted): New function.
17021 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17023 * selftest.c (assert_streq): Rename "expected" and "actual" to
17024 "val1" and "val2". Extend NULL-handling to cover both inputs
17025 symmetrically, while still requiring both to be non-NULL for a pass.
17026 * selftest.h (assert_streq): Rename "expected" and "actual" to
17028 (ASSERT_EQ): Likewise.
17029 (ASSERT_EQ_AT): Likewise.
17030 (ASSERT_KNOWN_EQ): Likewise.
17031 (ASSERT_KNOWN_EQ_AT): Likewise.
17032 (ASSERT_NE): Likewise.
17033 (ASSERT_MAYBE_NE): Likewise.
17034 (ASSERT_MAYBE_NE_AT): Likewise.
17035 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
17036 the assertion to pass.
17037 (ASSERT_STREQ_AT): Likewise.
17039 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
17041 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17042 interaction with -pie.
17044 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17046 * selftest.h: Fix alphabetization of per-source-file selftest
17049 2018-04-30 Jason Merrill <jason@redhat.com>
17051 PR c++/61982 - dead stores to destroyed objects.
17052 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
17055 2018-04-30 Jason Merrill <jason@redhat.com>
17057 * tree.c (build_clobber): New.
17058 * tree.h: Declare it.
17059 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
17061 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17063 * diagnostic-show-locus.c (layout::layout): Update for
17064 location_get_source_line returning a char_span.
17065 (struct char_span): Move to input.h.
17066 (struct correction): Update for fields in char_span becoming
17068 (struct source_line): Update for location_get_source_line
17069 returning a char_span.
17070 (layout::print_line): Likewise.
17071 * edit-context.c (edited_file::print_content): Likewise.
17072 (edited_file::print_diff_hunk): Likewise.
17073 (edited_file::print_run_of_changed_lines): Likewise.
17074 (edited_file::get_num_lines): Likewise.
17075 (edited_line::edited_line): Likewise.
17076 * final.c (asm_show_source): Likewise.
17077 * input.c (location_get_source_line): Convert return type
17078 from const char * to char_span, losing the final "line_len"
17080 (dump_location_info): Update for the above.
17081 (get_substring_ranges_for_loc): Likewise. Use a char_span
17082 when handling the literal within the line.
17083 (test_reading_source_line): Update for location_get_source_line
17084 returning a char_span.
17085 * input.h (class char_span): Move here from
17086 diagnostic-show-locus.c, converting from a struct to a class.
17087 Make data members private.
17088 (char_span::operator bool): New.
17089 (char_span::length): New.
17090 (char_span::get_buffer): New.
17091 (char_span::operator[]): New.
17092 (char_span::subspan): Make const.
17093 (char_span::xstrdup): New.
17094 (location_get_source_line): Convert return type from const char *
17095 to char_span, losing the final "line_size" param.
17097 2018-04-30 Jan Hubicka <jh@suse.cz>
17099 * lto-wrapper.c (ltrans_priorities): New static var.
17100 (cmp_priority): New.
17101 (run_gcc): Read priorities and if doing parallel build order
17102 the Makefile by them.
17104 2018-04-30 David Malcolm <dmalcolm@redhat.com>
17106 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
17108 2018-04-30 Richard Biener <rguenther@suse.de>
17110 * tree-cfg.c (verify_address): Remove base argument, add
17111 flag whether to check TREE_ADDRESSABLE and do that.
17112 (verify_expr): Remove.
17113 (verify_types_in_gimple_reference): Add pieces from verify_expr.
17114 (verify_gimple_assign_single): Likewise.
17115 (verify_gimple_switch): Likewise.
17116 (verify_expr_location_1): Dereference tp once. Add (disabled)
17117 piece from verify_expr.
17118 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
17120 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
17122 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17124 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
17126 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
17127 (small_data_pattern): Likewise.
17128 (arc_rewrite_small_data): Likewise.
17129 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
17130 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
17131 (get_symbol_alignment): New function.
17132 (legitimate_small_data_address_p): Likewise.
17133 (legitimate_scaled_address): Update, call
17134 legitimate_small_data_address_p.
17135 (output_sdata): New static variable.
17136 (arc_print_operand): Update how we handle small data operands.
17137 (arc_print_operand_address): Likewise.
17138 (arc_legitimate_address_p): Update, use
17139 legitimate_small_data_address_p.
17140 (arc_rewrite_small_data_p): Remove.
17141 (arc_rewrite_small_data_1): Likewise.
17142 (arc_rewrite_small_data): Likewise.
17143 (small_data_pattern): Likewise.
17144 (compact_sda_memory_operand): Update to use
17145 legitimate_small_data_address_p and get_symbol_alignment.
17146 (prepare_move_operands): Don't rewite sdata pattern.
17147 (prepare_extend_operands): Remove.
17148 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
17150 (zero_extendqisi2): Likewise.
17151 (zero_extendhisi2): Likewise.
17152 (extendqihi2): Likewise.
17153 (extendqisi2): Likewise.
17154 (extendhisi2): Likewise.
17155 (addsi3): Likewise.
17156 (subsi3): Likewise.
17157 (andsi3): Likewise.
17158 * config/arc/constraints.md (Usd): Change it to memory constraint.
17160 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
17162 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
17163 as source of std instructions.
17164 * config/arc/arc.md (movsi_insn): Update pattern predicate to
17165 allow 6-bit constants as source for store instructions.
17166 (movdi_insn): Update instruction pattern to allow 6-bit constants
17167 as source for store instructions.
17169 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
17171 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
17173 2018-04-30 Nathan Sidwell <nathan@acm.org>
17174 Sandra Loosemore <sandra@codesourcery.com>
17176 * dumpfile.c (dump_open): Allow '-' for stdout.
17177 * doc/invoke.texi (Developer Options): Document dump filename
17178 determination early. Document stdin/stdout selection.
17180 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
17182 Microblaze Target: PIC data text relative
17184 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
17185 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
17187 * config/microblaze/microblaze.h (microblaze_constant_address_p):
17188 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
17189 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
17190 New addressing mode for data-text relative position indepenedent code.
17191 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
17192 'ADDRESS_SYMBOLIC_TXT_REL'.
17193 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
17194 (microblaze_legitimate_pic_operand): Exclude function calls from
17195 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
17196 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
17198 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17199 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17200 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
17201 for 'address + offset'.
17202 (microblaze_expand_prologue): Add new function prologue call for
17204 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
17205 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
17206 table in case of TARGET_PIC_DATA_TEXT_REL.
17207 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
17208 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
17209 Add new macros 'UNSPEC_TEXT',
17210 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
17211 + exclude function calls from 'UNSPEC_PLT' in case of data text
17213 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17214 new target hook for generating address diff vector tables in case of
17216 * doc/tm.texi : Regenerate.
17217 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
17218 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
17219 of addr diff vector generation.
17220 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17221 target hook definition.
17222 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
17223 Add default function for generate_pic_addr_diff_vec -> flag_pic.
17224 * doc/invoke.texi (Add new pic option): Add new microblaze pic
17225 option for data text relative.
17227 2018-04-30 Richard Biener <rguenther@suse.de>
17229 * tree-chrec.h (evolution_function_is_constant_p): Remove
17231 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
17233 2018-04-30 Richard Biener <rguenther@suse.de>
17236 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
17238 2018-04-30 Richard Biener <rguenther@suse.de>
17240 PR tree-optimization/28364
17241 PR tree-optimization/85275
17242 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17243 copying first exit test.
17245 2018-04-28 Mark Wielaard <mark@klomp.org>
17247 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
17248 dwarf_version >= 5.
17249 (dwarf_AT): Handle DW_AT_addr_base.
17250 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
17252 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
17255 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
17256 (*ashl<dwi>3_doubleword_mask_1): Ditto.
17257 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
17258 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
17260 2018-04-28 Richard Biener <rguenther@suse.de>
17262 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
17263 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
17264 to reflect use. Only add interesting stmts.
17266 2018-04-27 Martin Jambor <mjambor@suse.cz>
17269 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
17270 the jump function allows for passing through aggregate values.
17272 2018-04-27 David Malcolm <dmalcolm@redhat.com>
17274 * input.h (in_system_header_at): Convert from macro to inline
17276 (from_macro_expansion_at): Likewise.
17277 (from_macro_definition_at): Likewise.
17279 2018-04-27 Jeff Law <law@redhat.com>
17281 * config.gcc: Mark tile* targets as deprecated/obsolete.
17283 2018-04-27 Richard Biener <rguenther@suse.de>
17285 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
17288 2018-04-27 Richard Biener <rguenther@suse.de>
17290 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
17292 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
17294 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
17295 with Yd constraint. Set "preferred_for_speed" attribute from
17296 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
17297 with Yd constraint.
17298 (*movdi_internal): Ditto.
17299 (movti_interunit splitters): Remove
17300 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
17301 (movdi_interunit splitters): Ditto.
17302 * config/i386/constraints.md (Ye): Remove.
17303 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
17305 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17308 * config/aarch64/constraints.md (Usg): Limit to 31.
17309 (Usj): Limit to 63.
17311 2018-04-27 Jakub Jelinek <jakub@redhat.com>
17313 PR tree-optimization/85529
17314 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
17315 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
17316 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
17317 zero extension or masking of the MSB bit.
17318 (optimize_range_tests): Add FIRST_BB argument, pass it through
17319 to optimize_range_tests_var_bound.
17320 (maybe_optimize_range_tests, reassociate_bb): Adjust
17321 optimize_range_tests callers.
17323 2018-04-26 Richard Biener <rguenther@suse.de>
17324 Jakub Jelinek <jakub@redhat.com>
17326 * cgraph.h (symbol_table): Just declare debug method here.
17327 * symtab.c (symbol_table::debug): Define.
17329 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
17331 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
17333 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
17335 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
17336 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
17337 (*movdi_internal): Substitute Yi and Yj constraint with x
17338 and Ym and Yn constraint with y constraint. Update "isa"
17339 attribute and set "preferred_for_speed" attribute from
17340 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
17341 (*movsi_internal): Ditto.
17342 (*movdf_internal): Ditto.
17343 (*movsf_internal): Ditto.
17344 (*zero_extendsidi2): Ditto.
17345 * config/i386/sse.md (vec_set<mode>_0): Ditto.
17346 (sse2_loadld): Ditto.
17347 (*vec_extract<ssevecmodelower>_0): Ditto.
17348 (*vec_extractv4si_0_zext_sse4): Ditto.
17349 (vec_concatv2di): Ditto.
17350 (*vec_dup<mode>): Ditto.
17351 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
17352 * config/i386/constraints.md (Yi): Remove.
17357 2018-04-26 Nathan Sidwell <nathan@acm.org>
17359 * dumpfile.c (dump_open): New.
17360 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
17361 (dump_finish): Detect stdio/stderr by value not name.
17363 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
17365 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
17367 2018-04-26 Tom de Vries <tom@codesourcery.com>
17370 * config/nvptx/nvptx.c (verify_neutering_jumps)
17371 (verify_neutering_labels): New function
17372 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
17374 2018-04-26 Tom de Vries <tom@codesourcery.com>
17377 * config/nvptx/nvptx.c (needs_neutering_p): New function.
17378 (nvptx_single): Use needs_neutering_p to skip over insns that do not
17381 2018-04-26 Richard Biener <rguenther@suse.de>
17382 Tom de Vries <tom@codesourcery.com>
17385 * lto-streamer-out.c (output_function): Fixup loops if required to match
17386 discovery done in the reader.
17388 2018-04-26 Richard Biener <rguenther@suse.de>
17390 PR tree-optimization/85116
17391 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
17392 have a loop exit from the single latch predecessor. Remove
17393 case of header with just condition.
17394 (ch_base::copy_headers): Exclude infinite loops from any
17396 (pass_ch::execute): Record exits.
17398 2018-04-26 Richard Biener <rguenther@suse.de>
17400 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
17401 prologue cost vector and pass it to vect_get_load_cost.
17402 (vect_get_peeling_costs_all_drs): Likewise.
17403 (vect_peeling_hash_get_lowest_cost): Likewise.
17404 (vect_enhance_data_refs_alignment): Likewise.
17406 2018-04-26 Richard Biener <rguenther@suse.de>
17408 PR middle-end/85450
17409 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
17410 checking of integer<->pointer conversions.
17411 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
17412 sign-/zero-extending pointer types.
17413 (expand_omp_for_static_chunk): Likewise.
17415 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
17416 Jean Lee <xiaoyur347@gmail.com>
17418 * config/mips/mips.c (mips_asan_shadow_offset): New function.
17419 (TARGET_ASAN_SHADOW_OFFSET): Define.
17420 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
17421 true for -fsanitize=address.
17423 2018-04-25 Mark Wielaard <mark@klomp.org>
17425 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
17428 2018-04-25 Jakub Jelinek <jakub@redhat.com>
17430 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
17431 than "alu", remove explicit "memory" and "imm_disp" attributes.
17432 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
17434 PR middle-end/85414
17435 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
17436 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
17437 gen_lowpart_no_emit.
17439 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
17442 * config/i386/i386.c (ix86_expand_builtin): Change memory
17443 operand to XI, extend p0 to Pmode.
17444 * config/i386/i386.md: Change unspec volatile and operand
17445 1 mode to XI, change operand 0 mode to P.
17447 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17449 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
17450 GET_MODE_MASK before any checking.
17451 (nds32_can_use_bset_p): Likewise.
17452 (nds32_can_use_btgl_p): Likewise.
17454 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17456 * config/nds32/nds32-doubleword.md: New define_split pattern for
17457 illegal register number.
17459 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17461 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
17463 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
17465 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
17467 2018-04-25 Richard Biener <rguenther@suse.de>
17469 * lto-streamer.h (LTO_major_version): Bump to 8.
17471 2018-04-25 Jakub Jelinek <jakub@redhat.com>
17473 * BASE-VER: Set to 9.0.0.
17475 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
17477 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
17478 in __abskf2 and __powikf2.
17480 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17483 * config/aarch64/constraints.md (Usg, Usj): New constraints.
17484 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
17485 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
17486 Use the above on operand 2. Reindent.
17487 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
17489 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
17492 * common/config/i386/i386-common.c (ix86_handle_option): Don't
17494 * config/i386/i386.opt (mcet): Removed.
17495 * doc/install.texi: Remove -mcet documentation.
17496 * doc/invoke.texi: Likewise.
17498 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
17501 * doc/install.texi: Remove -mcet from bootstrap-cet.
17503 2018-04-24 Jakub Jelinek <jakub@redhat.com>
17506 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
17507 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
17511 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
17512 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
17513 containing a CONST_VECTOR.
17515 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
17517 * doc/install.texi: Update newlib dependency for nvptx.
17519 2018-04-24 Jakub Jelinek <jakub@redhat.com>
17522 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
17523 instead of INTVAL when shifting x left.
17525 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
17527 PR tree-optimization/85478
17528 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
17529 vect_grouped_store_supported for single element vectors.
17531 2018-04-24 Richard Biener <rguenther@suse.de>
17534 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
17535 load cost increase to the case of non-constant step.
17537 2018-04-24 Jakub Jelinek <jakub@redhat.com>
17540 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
17541 destination if any_malformed_asm.
17543 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
17545 PR middle-end/85496
17546 * expr.c (store_field): In the bitfield case, if the value comes from
17547 a function call and is returned in registers by means of a PARALLEL,
17548 do not change the mode of the temporary unless BLKmode and VOIDmode.
17550 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
17552 PR rtl-optimization/85423
17553 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
17554 dependencies to debug insns when the previous insn is non-debug.
17556 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
17558 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
17559 enums into a single definition.
17560 (fls): Fix predicates and printing.
17563 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
17565 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
17566 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
17567 and short u6 immediate.
17568 (check_if_valid_sleep_operand): Remove.
17569 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
17571 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17573 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
17574 flag_always_save_lp condition.
17575 * config/nds32/nds32.opt (malways-save-lp): New option.
17577 2018-04-22 Shiva Chen <shiva0217@gmail.com>
17579 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
17580 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
17581 * config/nds32/nds32.h
17582 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
17583 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
17585 2018-04-22 Shiva Chen <shiva0217@gmail.com>
17587 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
17588 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
17590 2018-04-22 Shiva Chen <shiva0217@gmail.com>
17591 Chung-Ju Wu <jasonwucj@gmail.com>
17593 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
17595 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
17596 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
17598 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17600 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
17602 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17604 * config/nds32/nds32-protos.h (nds32_data_alignment,
17605 nds32_local_alignment): Declare.
17606 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
17607 nds32_local_alignment): New functions.
17608 (TARGET_CONSTANT_ALIGNMENT): Define.
17609 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
17611 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17613 * config/nds32/nds32.c
17614 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
17615 (TARGET_MODES_TIEABLE_P): Likewise.
17617 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
17619 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
17620 level Ofast and Og.
17622 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
17623 Chung-Ju Wu <jasonwucj@gmail.com>
17625 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17626 for unaligned access.
17627 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17629 * config/nds32/nds32-intrinsic.md: Likewise.
17630 * config/nds32/nds32_intrinsic.h: Likewise.
17631 * config/nds32/nds32.h (nds32_builtins): Likewise.
17632 * config/nds32/nds32.opt (munaligned-access): New option.
17633 * config/nds32/nds32.c (nds32_asm_file_start): Display
17634 flag_unaligned_access status.
17636 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
17638 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
17639 -mno-relax is present.
17640 * config/riscv/linux.h (LINK_SPEC): Ditto.
17642 2018-04-20 Martin Sebor <msebor@redhat.com>
17645 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
17647 (gimple_fold_builtin_stxcpy_chk): Same.
17648 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
17650 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
17653 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
17654 __powikf2 when long double is IEEE 128-bit.
17656 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
17658 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
17659 step to make sure stack always aligned.
17661 2018-04-20 Carl Love <cel@us.ibm.com>
17664 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
17665 size check for arg0.
17667 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
17668 Tom de Vries <tom@codesourcery.com>
17671 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17672 Emit insns for calls too.
17673 (nvptx_find_par): Always look for worker-level predecessor insn.
17674 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
17676 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
17677 (nvptx_process_pars): Propagate frames for calls.
17679 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
17682 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
17684 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
17685 (ix86_handle_option): Don't handle OPT_mibt.
17686 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
17688 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
17690 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
17691 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
17692 (ix86_target_macros): Define __CET__ with flag_cf_protection
17693 for -fcf-protection.
17694 * config/i386/i386.c (isa2_opts): Remove -mibt.
17695 * config/i386/i386.h (TARGET_IBT): Removed.
17696 (TARGET_IBT_P): Likewise.
17697 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
17698 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
17699 * config/i386/i386.opt (mcet): Update help message.
17700 (mshstk): Likewise.
17702 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
17703 -mcet as an alias for -mshstk.
17705 2018-04-20 Richard Biener <rguenther@suse.de>
17707 PR middle-end/85475
17708 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
17709 complexity by forcing a single use of the multiply operand.
17711 2018-04-20 Martin Jambor <mjambor@suse.cz>
17714 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
17715 recursion dependency to only apply to non-clones.
17717 2018-04-20 Martin Jambor <mjambor@suse.cz>
17720 * ipa-cp.c (create_specialized_node): Check that clones of
17721 self-recursive edges exist during IPA-CP.
17723 2018-04-19 Toon Moene <toon@moene.org>
17725 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
17728 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17730 PR tree-optimization/85467
17731 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
17732 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
17733 VECTOR_CST element to type.
17735 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17738 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
17739 * config/i386/i386.md (builtin_setjmp_setup): Removed.
17740 (builtin_longjmp): Likewise.
17741 (save_stack_nonlocal): New pattern.
17742 (restore_stack_nonlocal): Likewise.
17744 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17747 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17748 Replace ASM_OUTPUT_LABEL with fprintf.
17750 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17753 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17754 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17755 * config/i386/i386-c.c (ix86_target_macros_internal): Also
17756 define __IBT__ and __SHSTK__ for -fcf-protection.
17757 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17759 (ix86_trampoline_init): Likewise.
17760 (x86_output_mi_thunk): Likewise.
17761 (ix86_notrack_prefixed_insn_p): Likewise.
17762 (ix86_option_override_internal): Don't disallow -fcf-protection.
17763 * config/i386/i386.md (rdssp<mode>): Also enable for
17765 (incssp<mode>): Likewise.
17766 (nop_endbr): Likewise.
17767 * config/i386/i386.opt (mcet): Change help message to built-in
17770 (mshstk): Likewise.
17771 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17772 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
17773 enable CET built-in functions.
17775 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
17777 * common/config/i386/i386-common.c
17778 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17779 OPTION_MASK_ISA_MOVDIRI_UNSET,
17780 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17781 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17782 * config.gcc (movdirintrin.h): New header.
17783 * config/i386/cpuid.h (bit_MOVDIRI,
17784 bit_MOVDIR64B): New bits.
17785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17787 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17788 (VOID, PVOID, PCVOID)): New function types.
17789 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17790 __builtin_ia32_directstoreu_u64,
17791 __builtin_ia32_movdir64b): New builtins.
17792 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17793 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17795 (ix86_valid_target_attribute_inner_p): Ditto.
17796 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17797 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17798 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17799 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17800 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17801 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17802 (movdiri<mode>, movdir64b_<mode>): New.
17803 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17804 * config/i386/immintrin.h: Include movdirintrin.h.
17805 * config/i386/movdirintrin.h: New file.
17806 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17808 2018-04-19 Richard Biener <rguenther@suse.de>
17810 PR middle-end/85455
17811 * cfg.c (clear_bb_flags): When loop state says we have
17812 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17814 2018-04-19 Richard Biener <rguenther@suse.de>
17816 PR tree-optimization/84737
17817 * tree-vect-data-refs.c (vect_copy_ref_info): New function
17818 copying restrict info.
17819 (vect_setup_realignment): Use it.
17820 * tree-vectorizer.h (vect_copy_ref_info): Declare.
17821 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17822 the first DR to all generated stores.
17823 (vectorizable_load): Likewise for loads.
17825 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17827 PR tree-optimization/85446
17828 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17829 the integral and pointer types to have the same precision.
17831 * doc/install.texi: Document --disable-cet being the default and
17834 2018-04-18 Martin Liska <mliska@suse.cz>
17836 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17839 2018-04-18 Martin Liska <mliska@suse.cz>
17842 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17845 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17846 arguments if they are comparable.
17848 2018-04-18 Martin Liska <mliska@suse.cz>
17851 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17854 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17857 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
17860 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
17861 ENDBR after calling __morestack.
17863 2018-04-18 David Malcolm <dmalcolm@redhat.com>
17866 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17867 by using gcc_base_ver to generate a gcc_driver_version, and use
17868 it when generating GCC_DRIVER_NAME.
17869 * configure: Regenerate.
17871 2018-04-18 Jakub Jelinek <jakub@redhat.com>
17874 * config.gcc: Obsolete powerpc*-*-*spe*.
17876 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17879 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17880 (stabstr_D): Change type of unum from unsigned int to
17881 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
17884 2018-04-17 Jim Wilson <jimw@sifive.com>
17887 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17888 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17889 Set arg_pointer_offset after using pretend_args_size.
17891 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17893 PR rtl-optimization/85431
17894 * dse.c (record_store): Ignore zero width stores.
17897 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
17898 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17899 __builtin_stack_restore rather than after it.
17900 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17901 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17902 argument instead of virtual_dynamic_stack_rtx.
17904 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
17906 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17908 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17909 Add note to error message to explain internal mapping of overloaded
17910 built-in function name to non-overloaded built-in function name.
17911 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17914 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17917 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17918 where the inputs overlap with the output.
17920 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17922 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17923 (=v, v) alternative and explicit "memory" attribute.
17924 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
17925 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17927 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17928 "sselog1" type instead of "sselog".
17929 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17930 "sselog". Remove explicit "memory" attribute.
17931 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17932 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17934 (vec_extract_hi_v32hi): Merge all alternatives into one, use
17935 "sselog1" type instead of "sselog". Remove explicit "memory"
17937 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17938 use "sselog1" type instead of "sselog". Remove explicit "memory"
17940 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17941 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17943 (vec_extract_hi_v64qi): Merge all alternatives into one, use
17944 "sselog1" type instead of "sselog". Remove explicit "memory"
17946 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17947 use "sselog1" type instead of "sselog". Remove explicit "memory"
17951 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17953 PR middle-end/85414
17954 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17957 2018-04-17 Martin Jambor <mjambor@suse.cz>
17960 * ipa-cp.c (create_specialized_node): Call
17961 expand_all_artificial_thunks if necessary.
17963 2018-04-17 Martin Liska <mliska@suse.cz>
17966 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17967 in message, remote space in between '_G' and '('.
17969 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17972 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17973 avx512f_vmcmp<mode>3<round_saeonly_name>,
17974 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17975 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17976 avx512f_rndscale<mode><round_saeonly_name>,
17977 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17978 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17979 Use %<iptr>2 instead of %2 for -masm=intel.
17980 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17981 avx512f_vcvttss2usi<round_saeonly_name>,
17982 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17984 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17985 avx512f_vcvttsd2usi<round_saeonly_name>,
17986 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17987 Use %q1 instead of %1 for -masm=intel.
17988 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17989 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17990 of %3 for -masm=intel.
17991 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17993 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17995 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17997 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17998 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18000 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18002 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18003 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18004 %g1 and one with %0 and %1.
18005 (avx512er_vmrcp28<mode><round_saeonly_name>,
18006 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18007 %1 for -masm=intel.
18008 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18009 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18010 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18011 of %0 and %{%4%} for -masm=intel.
18012 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18013 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18014 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18015 order of %0 and %{%5%}%{z%} for -masm=intel.
18017 2018-04-17 Jan Hubicka <jh@suse.cz>
18020 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18022 2018-04-17 Martin Liska <mliska@suse.cz>
18025 * multiple_target.c (create_dispatcher_calls): Set apostrophes
18026 for target_clone error message. Make default implementation
18027 clone to be a local declaration.
18028 (separate_attrs): Add new argument and check for an empty
18030 (expand_target_clones): Handle it.
18031 (ipa_target_clone): Make redirection just for target_clones
18034 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
18035 Tom de Vries <tom@codesourcery.com>
18037 PR middle-end/84955
18038 * omp-expand.c (expand_oacc_for): Add dummy false branch for
18039 tiled basic blocks without omp continue statements.
18041 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
18044 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
18045 vec_extract expression as having side effects to make sure it gets
18048 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
18051 * config/i386/i386.c (get_builtin_code_for_version): Check
18054 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
18057 * config.gcc: Support "skylake".
18058 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18060 * config/i386/i386.c (m_SKYLAKE): Define.
18061 (processor_target_table): Add "skylake".
18062 (ix86_option_override_internal): Add "skylake".
18063 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
18064 PROCESSOR_CANNONLAKE.
18065 (get_builtin_code_for_version): Fix priority for
18066 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
18067 PROCESSOR_SKYLAKE-AVX512.
18068 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
18069 (processor_type): Add PROCESSOR_SKYLAKE.
18071 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
18072 Jason Merrill <jason@redhat.com>
18075 * convert.c (convert_to_integer_1): Use direct recursion for
18076 enumeral types and types with a precision less than the number
18077 of bits in their mode.
18079 2018-04-16 Julia Koval <julia.koval@intel.com>
18082 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
18083 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
18085 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
18088 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
18089 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
18090 and -mno-direct-move.
18092 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
18095 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
18096 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
18097 Ensure negative shifts result in {0}.
18099 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
18101 PR rtl-optimization/79916
18102 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
18103 regs (if any) to define how to gnerate SD moves when LRA is in
18106 2018-04-13 Jakub Jelinek <jakub@redhat.com>
18108 PR rtl-optimization/85393
18109 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
18110 * except.c (expand_dw2_landing_pad_for_region): Make static.
18111 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
18112 a label and unconditional jump to old_bb, rather than
18113 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
18116 PR rtl-optimization/85376
18117 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
18118 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
18119 instead of a specific value.
18121 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
18122 Bin Cheng <bin.cheng@arm.com>
18124 PR tree-optimization/82965
18125 PR tree-optimization/83991
18126 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
18127 by_profile_only parameter.
18128 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
18129 information if the loop was predicted to iterate too many times.
18130 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
18132 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
18135 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
18138 2018-04-13 Martin Liska <mliska@suse.cz>
18139 Jakub Jelinek <jakub@redhat.com>
18141 PR middle-end/81657
18142 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
18143 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
18144 * builtins.c (expand_builtin_memory_copy_args): Use
18145 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
18146 handle dest_addr == pc_rtx.
18148 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
18151 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
18152 asked to not generate direct moves.
18153 (fix_trunc<mode>si2_stfiwx): Similar.
18154 (fix_trunc<mode>si2_internal): Similar.
18156 2018-04-12 Jakub Jelinek <jakub@redhat.com>
18159 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
18160 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
18161 lookup if dest in some wider mode is known to be const0_rtx and
18162 if so, record permanent equivalence for it to be ZERO_EXTEND of
18163 the narrower mode destination.
18165 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
18167 * lto-streamer-out.c (output_function): Revert 259346.
18168 * omp-expand.c (expand_oacc_for): Likewise.
18170 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
18172 PR rtl-optimization/85354
18173 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
18174 * sel-sched.c (sel_global_init): ... here.
18176 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
18179 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
18180 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
18181 mode for PE-COFF targets.
18182 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
18183 (i386_pe_asm_lto_end): Likewise.
18184 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
18185 (TARGET_ASM_LTO_END): Likewise.
18186 * config/i386/winnt.c (saved_debug_info_level): New static variable.
18187 (i386_pe_asm_lto_start): New function.
18188 (i386_pe_asm_lto_end): Likewise.
18190 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
18191 Richard Biener <rguenther@suse.de>
18193 PR middle-end/84955
18194 * lto-streamer-out.c (output_function): Fix CFG loop state before
18196 * omp-expand.c (expand_oacc_for): Handle calls to internal
18197 functions like regular functions.
18199 2018-04-12 Richard Biener <rguenther@suse.de>
18202 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
18203 for the early LTO debug to properly generate references to it
18204 during DIE emission. Do not re-use that for the skeleton for
18206 (dwarf2out_early_finish): Likewise.
18208 2018-04-12 Jakub Jelinek <jakub@redhat.com>
18211 * config/i386/sse.md
18212 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
18213 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
18214 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
18215 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
18216 and output is a reg, avoid creating invalid lowpart subreg, but
18217 instead split into a 512-bit move. Don't split if not AVX512VL,
18218 input is xmm16+ reg and output is a mem.
18219 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
18220 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
18221 xmm16+ reg and output is a mem.
18223 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18225 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
18226 also for flag_dwarf2_cfi_asm.
18228 2018-04-12 Jakub Jelinek <jakub@redhat.com>
18230 PR rtl-optimization/85342
18231 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
18232 a bool scalar var inside of the loop instead. Don't try to update
18233 recog_data.operand after failed apply_change_group.
18235 2018-04-12 Tom de Vries <tom@codesourcery.com>
18238 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
18239 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
18240 array with flexible array member as array without given dimension.
18241 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
18242 argument for undefined param to true.
18244 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
18247 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18248 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
18249 from PowerPC section.
18250 * config/rs6000/sysv4.opt (mcall-): Improve help text.
18251 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
18252 help text that is too long.
18253 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
18254 help text that is too long.
18255 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
18256 help text that is too long.
18258 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
18260 * config/alpha/alpha.md (stack_probe_internal): Rename
18261 from "probe_stack". Update all callers.
18263 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
18265 PR rtl-optimization/84566
18266 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
18267 sched_macro_fuse_insns.
18269 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
18272 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
18273 (compute_block_dependences): ... from here.
18275 2018-04-11 Jakub Jelinek <jakub@redhat.com>
18277 PR tree-optimization/85331
18278 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
18279 from int to HOST_WIDE_INT.
18281 2018-04-11 Martin Jambor <mjambor@suse.cz>
18284 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
18285 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
18286 not the same as the source val.
18287 (cgraph_edge_brings_value_p): New parameter.
18288 (gather_edges_for_value): Pass destination value to
18289 cgraph_edge_brings_value_p.
18290 (perhaps_add_new_callers): Likewise.
18291 (get_info_about_necessary_edges): Likewise and exclude values brought
18292 only by self-recursive edges.
18293 (create_specialized_node): Redirect only clones of self-calling edges.
18294 (+self_recursive_pass_through_p): New function.
18295 (find_more_scalar_values_for_callers_subset): Use it.
18296 (find_aggregate_values_for_callers_subset): Likewise.
18297 (known_aggs_to_agg_replacement_list): Removed.
18298 (decide_whether_version_node): Re-calculate known constants for all
18299 remaining context clones.
18301 2018-04-11 Richard Biener <rguenther@suse.de>
18304 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
18305 from early DWARF output.
18306 (dwarf2out_early_finish): Output line info unconditionally into
18307 early DWARF and add reference to it.
18309 2018-04-11 Jakub Jelinek <jakub@redhat.com>
18312 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
18313 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
18314 other than V2DFmode using iptr mode attribute.
18315 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
18317 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
18319 PR rtl-optimization/84659
18320 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
18322 2018-04-11 Jakub Jelinek <jakub@redhat.com>
18325 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
18327 (output_loc_list): Pass address of a dummy size variable even in the
18328 locview handling loop.
18329 (index_location_lists): Add comment on why skip_loc_list_entry can't
18332 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
18335 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
18338 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
18341 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
18342 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
18343 and -mstring-compare-inline-limit.
18345 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
18348 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
18349 for stack clash protection in a register whenever we need it to be in
18352 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
18354 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
18355 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
18357 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
18360 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
18362 (mlong-double-): Ditto.
18363 * config/rs6000/sysv4.opt (msdata=): Ditto.
18364 (mtls-size=): Ditto.
18366 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
18368 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18369 erroneous entries for
18370 "vector int vec_ldl (int, long int *)", and
18371 "vector unsigned int vec_ldl (int, unsigned long int *)".
18372 Add comments and entries for
18373 "vector bool char vec_ldl (int, bool char *)",
18374 "vector bool short vec_ldl (int, bool short *)",
18375 "vector bool int vec_ldl (int, bool int *)",
18376 "vector bool long long vec_ldl (int, bool long long *)",
18377 "vector pixel vec_ldl (int, pixel *)",
18378 "vector long long vec_ldl (int, long long *)",
18379 "vector unsigned long long vec_ldl (int, unsigned long long *)".
18380 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
18381 type tree bool_long_long_type_node and correct definition of
18382 bool_V2DI_type_node to make reference to this new type tree.
18383 (rs6000_mangle_type): Replace erroneous reference to
18384 bool_long_type_node with bool_long_long_type_node.
18385 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
18386 comments to emphasize sign distinctions for char and int types and
18387 replace RS6000_BTI_bool_long constant with
18388 RS6000_BTI_bool_long_long constant. Also add comment to restrict
18389 use of RS6000_BTI_pixel.
18390 (bool_long_type_node): Remove this macro definition.
18391 (bool_long_long_type_node): New macro definition
18393 2018-04-10 Jakub Jelinek <jakub@redhat.com>
18395 PR rtl-optimization/85300
18396 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
18397 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
18398 simplify_unary_operation fails.
18400 2018-04-10 Martin Liska <mliska@suse.cz>
18402 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
18403 cgraph_edge and ipa_ref.
18405 2018-04-10 Jakub Jelinek <jakub@redhat.com>
18409 * config/i386/sse.md
18410 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
18411 computation of the VEC_MERGE selector from mask.
18412 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
18413 Fix decoding of the VEC_MERGE selector into mask.
18415 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
18417 PR tree-optimization/85286
18418 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
18420 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
18422 * final.c (final_1): Set insn_last_address as well as
18423 insn_current_address.
18425 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18428 * explow.c (emit_stack_probe): Call validize_mem on memory location
18429 before passing it to gen_probe_stack. Create address operand and
18430 legitimize it for the probe_stack_address case.
18432 2018-04-09 Jan Hubicka <jh@suse.cz>
18435 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
18436 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
18437 * tree.c (free_lang_data_in_type): Fix handling of binfos;
18439 (free_lang_data): Rebuild type inheritance graph.
18441 2018-04-09 Martin Sebor <msebor@redhat.com>
18443 * invoke.texi (-finline-small-functions): Mention other optimization
18445 (-findirect-inlining, -fpartial-inlining): Same.
18446 (-finline-functions-called-once): Same.
18447 (-freorder-blocks-and-partition): Same.
18449 2018-04-09 Jan Hubicka <jh@suse.cz>
18452 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
18453 jumps; choose last target that matches the criteria (i.e.
18454 no partition changes for non-crossing jumps).
18455 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
18456 support for redirecting crossing jumps to non-crossing.
18458 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
18460 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
18461 also for naked functions.
18463 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
18465 * config/arc/arc.md (add_shift): New pattern.
18466 (add_shift2): Likewise.
18467 (sub_shift): Likewise.
18468 (sub_shift_cmp0_noout): Likewise.
18469 (compare_si_ashiftsi): Likewise.
18470 (xbfu_cmp0_noout): New combine pattern.
18471 (xbfu_cmp0"): Likewise.
18472 (movsi_set_cc_insn): Place the predicable variant first.
18473 (commutative_binary_cmp0_noout): Remove clobber.
18474 (commutative_binary_cmp0): New pattern.
18475 (noncommutative_binary_cmp0): Likewise.
18476 (noncommutative_binary_cmp0_noout): Likewise.
18477 (noncommutative_binary_comparison_result_used): Removed.
18478 (rsub_cmp0): New pattern.
18479 (rsub_cmp0_noout): Likewise.
18480 (extzvsi): Changed, keep only meaningful variants.
18481 (SQH, SEZ): New iterators.
18482 (SQH_postfix): New mode attribute.
18483 (SEZ_prefix): New code attribute.
18484 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
18485 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
18486 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
18487 of numerical value.
18488 (noncommutative_operator): Check the availability of barrel
18491 2018-04-09 Richard Biener <rguenther@suse.de>
18493 PR tree-optimization/85284
18494 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
18495 Only use the niter constraining form of simple_iv when the exit
18496 is always executed.
18498 2018-04-09 Tom de Vries <tom@codesourcery.com>
18501 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18502 (define_expand "*memory_barrier"): New define_expand.
18503 (define_insn "memory_barrier"): New insn.
18505 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18507 PR rtl-optimization/80463
18508 PR rtl-optimization/83972
18509 PR rtl-optimization/83480
18511 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
18512 correct producer for the insn.
18513 (tidy_control_flow): Fixup seqnos in case of debug insns.
18515 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18517 PR rtl-optimization/83913
18519 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
18520 different sched-times when merging exprs.
18522 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18524 PR rtl-optimization/83962
18526 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
18527 tidy_fallthru_edge and tidy_control_flow.
18529 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
18531 PR rtl-optimization/83530
18533 * sel-sched.c (force_next_insn): New global variable.
18534 (remove_insn_for_debug): When force_next_insn is true, also leave only
18535 next insn in the ready list.
18536 (sel_sched_region): When the region wasn't scheduled, make another pass
18537 over it with force_next_insn set to 1.
18539 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
18541 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
18543 * config/nds32/constants.md (unspec_volatile_element): Add enum values
18544 for interrupt control.
18545 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18546 functions for interrupt control.
18547 * config/nds32/nds32-intrinsic.md: Likewise.
18548 * config/nds32/nds32_intrinsic.h: Likewise.
18549 * config/nds32/nds32.h (nds32_builtins): Likewise.
18551 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
18553 * config/nds32/nds32.c (nds32_init_machine_status,
18554 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
18555 strict_aligned_p field.
18556 (nds32_expand_to_rtl_hook): New function.
18557 (TARGET_EXPAND_TO_RTL_HOOK): Define.
18558 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
18560 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18561 Chung-Ju Wu <jasonwucj@gmail.com>
18563 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
18564 * config/nds32/nds32-n7.md: New file.
18565 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
18566 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
18568 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
18569 * config/nds32/nds32.md (pipeline_model): Add n7.
18570 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
18571 * config/nds32/pipelines.md: Include n7 settings.
18573 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18574 Chung-Ju Wu <jasonwucj@gmail.com>
18576 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
18577 * config/nds32/nds32-e8.md: New file.
18578 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
18579 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
18581 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
18582 * config/nds32/nds32.md (pipeline_model): Add e8.
18583 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
18584 * config/nds32/pipelines.md: Include e8 settings.
18586 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18587 Chung-Ju Wu <jasonwucj@gmail.com>
18589 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
18590 * config/nds32/nds32-n8.md: New file.
18591 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
18592 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
18594 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
18595 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
18596 * config/nds32/nds32.md (pipeline_model): Add n8.
18597 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
18598 * config/nds32/pipelines.md: Include n8 settings.
18600 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
18601 Chung-Ju Wu <jasonwucj@gmail.com>
18603 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
18604 * config/nds32/nds32-n9-2r1w.md: New file.
18605 * config/nds32/nds32-n9-3r2w.md: New file.
18606 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
18607 nds32_register_ports): New or modify for cpu n9.
18608 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
18610 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
18611 * config/nds32/nds32-utils.c: New file.
18612 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
18613 TARGET_MUL_SLOW): Define.
18614 * config/nds32/nds32.md (pipeline_model): New attribute.
18615 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
18616 New options that support cpu n9.
18617 * config/nds32/pipelines.md: Include n9 settings.
18618 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
18620 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
18622 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
18623 information if necessary.
18624 (output_cond_branch_compare_zero): Likewise.
18625 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
18626 (nds32_target_alignment): Refine for alignment.
18627 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
18628 (FUNCTION_BOUNDARY): Modify.
18629 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
18631 * config/nds32/nds32.opt (malways-align, malign-functions): New.
18633 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
18635 * config/nds32/constants.md (unspec_volatile_element): Add values for
18636 TLB operation and data prefetch.
18637 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
18638 functions for TLB operation and data prefetch.
18639 * config/nds32/nds32-intrinsic.md: Likewise.
18640 * config/nds32/nds32_intrinsic.h: Likewise.
18641 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
18642 (nds32_print_operand): Likewise.
18643 * config/nds32/nds32.h (nds32_builtins): Likewise.
18645 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
18646 Andrew Pinski <pinsika@gcc.gnu.org>
18648 PR middle-end/82976
18649 * match.pd: Use constant_boolean_node of correct type instead of
18650 boolean_true_node or boolean_false_node for simplifying
18651 pointer comparisons to zero.
18653 2018-04-07 Jakub Jelinek <jakub@redhat.com>
18655 PR tree-optimization/80021
18656 * tree.c (verify_type_variant): Make error call in verify_variant_match
18657 translatable and remove final full stop.
18659 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18661 * config/nds32/constants.md (unspec_volatile_element): Add
18662 UNSPEC_VOLATILE_EH_RETURN.
18663 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
18664 nds32_output_stack_pop): Support dwarf exception handling process.
18665 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
18666 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
18667 exception handling process.
18668 (nds32_compute_stack_frame): Likewise.
18669 (nds32_return_addr_rtx): Likewise.
18670 (nds32_initial_elimination_offset): Likewise.
18671 (nds32_expand_prologue): Likewise.
18672 (nds32_expand_epilogue): Likewise.
18673 (nds32_dynamic_chain_address): New function.
18674 * config/nds32/nds32.h (machine_function): Add fields for dwarf
18675 exception handling.
18676 (DYNAMIC_CHAIN_ADDRESS): Define.
18677 (EH_RETURN_DATA_REGNO): Define.
18678 (EH_RETURN_STACKADJ_RTX): Define.
18679 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
18680 patterns for dwarf exception handling.
18682 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18684 * config/nds32/nds32.h: Clean up obsolete macros.
18686 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18688 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18689 Add enum values for particular instructions.
18690 * config/nds32/nds32-intrinsic.c: Implementation of expanding
18691 particular intrinsic functions.
18692 * config/nds32/nds32-intrinsic.md: Likewise.
18693 * config/nds32/nds32_intrinsic.h: Likewise.
18694 * config/nds32/nds32.h (nds32_builtins): Likewise.
18695 * config/nds32/nds32.md (type): Add pbsad and pbsada.
18696 (btst, ave): New patterns for particular instructions.
18698 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18700 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18701 Add enum values for atomic load/store and memory sync.
18702 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
18704 * config/nds32/nds32-intrinsic.md: Likewise.
18705 * config/nds32/nds32_intrinsic.h: Likewise.
18706 * config/nds32/nds32.h (nds32_builtins): Likewise.
18708 2018-04-07 Jakub Jelinek <jakub@redhat.com>
18710 PR tree-optimization/85257
18711 * fold-const.c (native_encode_vector): If not all elts could fit
18712 and off is -1, return 0 rather than offset.
18713 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
18714 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
18715 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
18716 adjust buffer in native_interpret_expr call.
18718 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18720 * config/nds32/constants.md (unspec_volatile_element): Add cache
18721 control enum values.
18722 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
18723 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
18724 * config/nds32/nds32.c (nds32_cctl_names): New.
18725 (nds32_print_operand): Handle cache control register names.
18726 * config/nds32/nds32.h (nds32_builtins): New enum values.
18727 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
18729 * config/nds32/nds32.md (type): Add mmu.
18730 * config/nds32/pipelines.md (simple_insn): Add mmu.
18732 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18734 * config/nds32/nds32.md (type): Remove call.
18735 * config/nds32/pipelines.md (simple_insn): Likewise.
18737 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18739 * config/nds32/constants.md (unspec_volatile_element): Add
18740 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
18741 UNSPEC_VOLATILE_FMFCFG.
18742 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
18743 description for fmfcfg and fmfcsr.
18744 (bdesc_1arg): Add fmtcsr.
18745 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
18746 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18747 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18748 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18749 unspec_fmfcfg): New patterns.
18750 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18751 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18752 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18753 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18754 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18755 __nds32__fmfcfg): Define.
18757 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18759 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18760 intrinsic register names.
18761 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18762 intrinsic register enum values and macros.
18764 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18766 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18767 for load/store addressing form.
18768 (nds32_print_operand_address): Likewise.
18770 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
18773 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18774 based on LABEL_REF. Remove useless assertion.
18775 (pic_address_needs_scratch): Fix formatting.
18776 (sparc_legitimize_pic_address): Minor tweaks.
18777 (sparc_delegitimize_address): Adjust assertion accordingly.
18778 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18779 into symbolic_operand.
18780 (movsi_high_pic_label_ref): Likewise.
18781 (movsi_lo_sum_pic_label_ref): Likewise.
18782 (movdi_pic_label_ref): Likewise.
18783 (movdi_high_pic_label_ref): Likewise.
18784 (movdi_lo_sum_pic_label_ref): Likewise.
18786 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
18788 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18789 custom LIB_SPEC setup.
18791 2018-04-06 Ruslan Bukin <br@bsdpad.com>
18792 Kito Cheng <kito.cheng@gmail.com>
18794 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18795 * config/riscv/freebsd.h: New.
18797 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18799 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18800 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18803 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18804 Kito Cheng <kito.cheng@gmail.com>
18806 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18807 nds32_output_call, nds32_symbol_binds_local_p): New functions.
18808 * config/nds32/nds32-protos.h (nds32_output_call,
18809 nds32_output_return): Declare.
18810 * config/nds32/nds32.md: Refine all the call and return patterns.
18812 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18815 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18816 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18818 PR rtl-optimization/84872
18819 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18820 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18821 EDGE_CROSSING edge.
18823 2018-04-06 Tamar Christina <tamar.christina@arm.com>
18825 * expr.c (copy_blkmode_to_reg): Revert 254862.
18826 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18828 2018-04-06 Richard Biener <rguenther@suse.de>
18830 PR middle-end/85244
18831 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18832 after seeing a component reference with an adjacent field. Treat
18833 refs to arrays at struct end of external decls similar to
18834 refs to unconstrained commons.
18836 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18839 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
18840 look through SAVE_EXPRs with non-side-effects argument. Adjust
18842 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18845 2018-04-06 Richard Biener <rguenther@suse.de>
18847 PR middle-end/85180
18848 * alias.c (find_base_term): New wrapper around find_base_term
18849 unwinding CSELIB_VAL_PTR changes.
18850 (find_base_term): Do not restore CSELIB_VAL_PTR during the
18853 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18855 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18857 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
18858 constant definitions.
18859 ("nop"): lr 0,0 -> nopr r0
18860 ("nop_lr0", "nop_lr1"): New insn definitions.
18862 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18864 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
18865 NDS32_V3PUSH_AVAILABLE_P macro.
18867 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
18868 Chung-Ju Wu <jasonwucj@gmail.com>
18870 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18871 (nds32*-*-*): Add float and fpu_config into supported_defaults.
18872 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18873 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18874 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18875 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18876 * config/nds32/constraints.md: New constraints and checking for hard
18877 float configuration.
18878 * config/nds32/iterators.md: New mode iterator and attribute for hard
18879 float configuration.
18880 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18882 * config/nds32/nds32-fpu.md: New file.
18883 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18884 deal with hard float code generation.
18885 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18887 (abi_type, float_reg_number): New enum type.
18888 * config/nds32/nds32-predicates.c: New predicates for hard float.
18889 * config/nds32/nds32-protos.h: Declare functions for hard float.
18890 * config/nds32/nds32.c: Implementation for hard float configuration.
18891 * config/nds32/nds32.h: Definitions for hard float configuration.
18892 * config/nds32/nds32.md: Include hard float machine description and
18893 modify patterns for hard float configuration.
18894 * config/nds32/nds32.opt: New options for hard float configuration.
18895 * config/nds32/predicates.md: New predicates for hard float
18898 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18900 * common/config/nds32/nds32-common.c
18901 (nds32_option_optimization_table): Enable -mreleax-hint by default.
18903 2018-04-05 Jakub Jelinek <jakub@redhat.com>
18905 PR middle-end/85195
18906 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18907 CONSTRUCTOR_ELT (ctor, ...)->value.
18909 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
18912 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18914 2018-04-05 Tom de Vries <tom@codesourcery.com>
18917 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18920 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18921 Kito Cheng <kito.cheng@gmail.com>
18923 * config/nds32/constraints.md (U33): Fine-tune checking condition.
18924 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18925 * config/nds32/nds32.h (nds32_16bit_address_type): Add
18926 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18928 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18929 Kito Cheng <kito.cheng@gmail.com>
18931 * config/nds32/constraints.md (Ufe): New memory constraint.
18932 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18933 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18934 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18936 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18937 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18939 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18941 * config/nds32/nds32.md: Use optimize_size in the condition for
18942 alu-shift instructions.
18944 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18946 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18948 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18950 * config/nds32/nds32.md (negsi2): Refine pattern.
18952 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18953 Chung-Ju Wu <jasonwucj@gmail.com>
18955 * config/nds32/iterators.md (shift_rotate): New code iterator.
18956 (shift): New code attribute.
18957 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18958 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18959 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18960 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18961 bit-wise operations.
18962 (andsi3, *andsi3): Ditto.
18963 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18964 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18965 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18966 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18967 nds32_ior_operand, nds32_xor_operand): New predicates.
18969 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18971 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18972 (addsi3, subsi3): ... this.
18974 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18976 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18978 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18980 * config/nds32/nds32.md: Adjust indention.
18982 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18984 * config/nds32/nds32.md (feature): New attribute.
18986 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18988 * config/nds32/nds32.md (subtype): New attribute.
18990 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
18993 * config/arm/arm-builtins.c (arm_expand_builtin): Change
18994 expansion to perform a bitwise AND of the argument followed by a
18995 boolean negation of the result.
18997 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
18999 PR rtl-optimization/84878
19000 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19001 the basic block. Assert the use reference is not artificial and that
19002 it has an associated insn.
19004 2018-04-04 Michael Matz <matz@suse.de>
19006 * builtins.c (compute_objsize): Pass correct operand
19007 to array_at_struct_end_p.
19009 2018-04-04 Richard Biener <rguenther@suse.de>
19012 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19013 from contexts for DINFO_LEVEL_TERSE and below.
19015 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
19017 * config/nds32/nds32-doubleword.md (move_<mode>): Require
19018 resiter_operand condition.
19019 * config/nds32/nds32.md (*move<mode>): Ditto.
19021 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
19022 Monk Chiang <sh.chiang04@gmail.com>
19024 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19026 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19028 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19030 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19031 Kito Cheng <kito.cheng@gmail.com>
19033 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19034 nds32_cond_code_str, output_cond_branch,
19035 output_cond_branch_compare_zero, nds32_expand_cbranch,
19036 nds32_expand_cstore, nds32_expand_movcc,
19037 nds32_output_cbranchsi4_equality_zero,
19038 nds32_output_cbranchsi4_equality_reg,
19039 nds32_output_cbranchsi4_equality_reg_or_const_int,
19040 nds32_output_cbranchsi4_greater_less_zero: New functions.
19041 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19042 nds32_expand_cstore, nds32_expand_movcc,
19043 nds32_output_cbranchsi4_equality_zero,
19044 nds32_output_cbranchsi4_equality_reg,
19045 nds32_output_cbranchsi4_equality_reg_or_const_int,
19046 nds32_output_cbranchsi4_greater_less_zero): Declare.
19047 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19048 nds32_rimm11s_operand): New predicates.
19049 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
19050 * config/nds32/nds32.md: Rewrite all the branch and conditional move
19053 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
19055 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
19056 * config/nds32/nds32.md: Ditto.
19057 * config/nds32/pipelines.md: Ditto.
19059 2018-04-04 Richard Biener <rguenther@suse.de>
19061 PR tree-optimization/85168
19062 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
19063 propagating abnormals.
19065 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19067 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
19069 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
19070 Kito Cheng <kito.cheng@gmail.com>
19072 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
19073 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
19074 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
19075 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
19076 * config/nds32/nds32.md (sibcall_internal): New.
19077 (sibcall_register): Remove.
19078 (sibcall_immediate): Remove.
19079 (sibcall_value_internal): New.
19080 (sibcall_value_register): Remove.
19081 (sibcall_value_immediate): Remove.
19082 * config/nds32/predicates.md (nds32_general_register_operand): New.
19083 (nds32_call_address_operand): New.
19085 2018-04-03 Jakub Jelinek <jakub@redhat.com>
19087 PR rtl-optimization/85167
19088 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
19089 bb_defs if *split_p, instead preinitialize it to NULL.
19091 PR tree-optimization/85156
19092 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
19093 evaluating the argument multiple times.
19095 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
19097 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
19099 (_mm_cvtpd_ps): Likewise.
19100 (_mm_cvttpd_epi32): Likewise.
19101 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
19102 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
19103 vector, pixel, and bool following altivec.h include.
19105 2018-04-03 Martin Sebor <msebor@redhat.com>
19107 * doc/extend.texi (Common Function Attributes): Clarify.
19108 (const attribute): Likewise.
19109 (pure attribute): Likewise.
19111 2018-04-03 Jakub Jelinek <jakub@redhat.com>
19114 * config/i386/i386.c (ix86_expand_vector_set): Use
19115 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
19117 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
19119 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
19120 instructions when changing rounding bits to preserve precision bits
19121 in the x87 control word.
19123 2018-04-03 Martin Liska <mliska@suse.cz>
19125 PR tree-optimization/82491
19126 * rtl.h (strip_offset_and_add): Replace += suboffset with
19127 poly_uint64 () + suboffset.
19129 2018-03-29 Martin Liska <mliska@suse.cz>
19130 Martin Jambor <mjambor@suse.cz>
19133 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
19134 param_type is not an integral or pointer type.
19136 2018-04-03 Richard Biener <rguenther@suse.de>
19138 * sese.h (recompute_all_dominators): Remove.
19140 2018-04-02 Martin Sebor <msebor@redhat.com>
19142 * doc/invoke.texi (-Wrestrict): Fix typos.
19144 2018-04-02 Jim Wilson <jimw@sifive.com>
19146 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
19147 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
19148 (<optab>di3, <optab>si3_extend): Likewise.
19149 (<optab>si3_mask, <optab>si3_mask_1): New.
19150 (<optab>di3_mask, <optab>di3_mask_1): New.
19151 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
19152 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
19153 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
19155 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
19157 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
19160 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
19162 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
19163 (nds32_canonicalize_comparison): New function.
19165 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
19166 Kito Cheng <kito.cheng@gmail.com>
19167 Kuan-Lin Chen <kuanlinchentw@gmail.com>
19169 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
19170 * config/nds32/constants.md (unspec_volatile_element): Add
19171 UNSPEC_VOLATILE_RELAX_GROUP.
19172 * config/nds32/nds32-relax-opt.c: New file.
19173 * config/nds32/nds32-predicates.c
19174 (nds32_symbol_load_store_p): New function.
19175 * config/nds32/nds32-protos.h
19176 (nds32_symbol_load_store_p): Declare function.
19177 (make_pass_nds32_relax_opt): Declare new rtl pass function.
19178 * config/nds32/nds32.c
19179 (nds32_register_pass): New function to register pass.
19180 (nds32_register_passes): New function to register passes.
19181 * config/nds32/nds32.md (relax_group): New pattern.
19182 * config/nds32/nds32.opt (mrelax-hint): New option.
19183 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
19185 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
19187 * config/nds32/t-nds32: Modify files dependency.
19189 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
19191 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
19192 (PROFILE_HOOK): Define its implementation.
19194 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
19196 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
19197 type and 32-bit size.
19199 2018-04-01 Jakub Jelinek <jakub@redhat.com>
19201 PR middle-end/85090
19202 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
19203 (V_128_256): New mode iterator.
19204 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
19205 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
19206 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
19208 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
19209 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
19211 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
19214 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
19215 NaN inputs correctly.
19217 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
19220 * config/rs6000/vsx.md (??r): New mode attribute.
19221 (*vsx_mov<mode>_64bit): Use it.
19222 (*vsx_mov<mode>_32bit): Likewise.
19224 2018-03-30 Martin Sebor <msebor@redhat.com>
19226 PR tree-optimization/84818
19227 * builtins.c (check_access): Use warning_n.
19229 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19232 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
19234 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
19237 2018-03-30 Julia Koval <julia.koval@intel.com>
19240 * x86-tune.def (movx, partial_reg_dependency): Enable for
19243 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
19245 PR inline-asm/84985
19246 * lra-constraints.c (process_alt_operands): Move setting
19247 this_alternative_matches below.
19249 2018-03-29 Martin Liska <mliska@suse.cz>
19252 * doc/invoke.texi: Document how LTO works with debug info.
19253 Describe auto-load support of binutils. Mention 'x86-64'
19254 as valid option value of -march option.
19256 2018-03-29 Jakub Jelinek <jakub@redhat.com>
19258 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
19261 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
19262 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
19263 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
19266 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
19269 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
19270 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
19271 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
19272 for RS6000_BTM_POWERPC64.
19273 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
19274 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
19275 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
19278 (DIVDEU): Likewise.
19280 2018-03-28 Carl Love <cel@us.ibm.com>
19283 2017-09-27 Carl Love <cel@us.ibm.com>
19285 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
19286 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
19287 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
19290 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19292 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
19293 instead of __vector bool.
19294 (_mm_max_pu8): Likewise.
19295 (_mm_min_pi16): Likewise.
19297 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
19300 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
19301 (DIVWEUO): Likewise.
19302 (DIVDEO): Likewise.
19303 (DIVDEUO): Likewise.
19304 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
19305 DIVWEUO and DIVDEUO.
19306 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
19307 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
19308 (div_extend): Likewise.
19309 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
19311 (__builtin_divweuo): Likewise.
19312 (__builtin_divdeo): Likewise.
19313 (__builtin_divdeuo): Likewise.
19315 2018-03-28 Jakub Jelinek <jakub@redhat.com>
19318 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
19319 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
19321 PR tree-optimization/82004
19322 * gimple-match-head.c (optimize_pow_to_exp): New function.
19323 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
19324 Don't fold to exp if optimize_pow_to_exp is false.
19326 2018-03-28 Martin Liska <mliska@suse.cz>
19329 * calls.c (initialize_argument_information): Fix trailing space.
19330 * common.opt: Fix typo and provide better explanation for
19331 -fsanitize-coverage option.
19332 * config/i386/i386.opt: Fix typo.
19334 2018-03-28 Jakub Jelinek <jakub@redhat.com>
19335 Martin Liska <mliska@suse.cz>
19338 * gimplify.c (asan_poison_variable): Don't do the check for
19339 gimplify_omp_ctxp here.
19340 (gimplify_decl_expr): Do it here.
19341 (gimplify_target_expr): Likewise.
19343 2018-03-28 Martin Liska <mliska@suse.cz>
19346 * config/i386/i386.c (ix86_function_arg_advance): Do not call
19347 chkp_type_bounds_count if MPX is not enabled.
19349 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
19351 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
19353 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
19356 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
19357 function to create the function decl for complex long double
19358 multiply and divide for -mabi=ieeelongdouble.
19359 (init_float128_ieee): Call it.
19361 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
19364 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
19365 -fcf-protection=branch -mibt.
19366 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
19368 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19371 * config/arm/arm.c (arm_valid_symbolic_address): Handle
19372 arm_word_relocations.
19374 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
19377 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
19378 extern array declarations.
19380 2018-03-27 Richard Biener <rguenther@suse.de>
19382 PR middle-end/84067
19383 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
19384 explicit single_use checks.
19386 2018-03-27 Richard Biener <rguenther@suse.de>
19388 PR tree-optimization/85082
19389 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
19392 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19394 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
19395 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
19396 Turn on fasynchronous-unwind-tables and funwind-tables.
19398 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
19401 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
19402 (*bmi_blsr_<mode>_ccz): Ditto.
19404 2018-03-26 Tom de Vries <tom@codesourcery.com>
19406 PR tree-optimization/85063
19407 * omp-general.c (offloading_function_p): New function. Factor out
19409 * omp-offload.c (pass_omp_target_link::gate): ... here.
19410 * omp-general.h (offloading_function_p): Declare.
19411 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
19412 with attribute omp declare target for offloading functions.
19414 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
19416 PR tree-optimization/84005
19417 * tree-data-ref.h (get_base_for_alignment): Declare.
19418 * tree-data-ref.c (get_base_for_alignment_1): New function.
19419 (get_base_for_alignment): Likewise.
19420 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
19421 get_base_for_alignment to find a suitable base object, instead
19422 of always using drb->base_address.
19424 2018-03-23 Jakub Jelinek <jakub@redhat.com>
19426 PR inline-asm/85022
19427 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
19428 known size by default.
19430 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
19432 PR inline-asm/85030
19433 * lra-constraints.c (process_alt_operands): Don't match BLKmode
19434 and non BLKmode operands.
19436 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19439 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
19440 Clean up attributes.
19442 2018-03-23 Richard Biener <rguenther@suse.de>
19445 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
19446 we are going to emit early debug for LTO.
19448 2018-03-23 Jakub Jelinek <jakub@redhat.com>
19450 PR inline-asm/85034
19451 * function.c (match_asm_constraints_1): Don't optimize if input
19452 doesn't satisfy general_operand predicate for output's mode.
19454 PR inline-asm/85022
19455 * alias.c (write_dependence_p): Don't require for x_canonicalized
19456 non-VOIDmode if x has VOIDmode.
19459 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
19460 just don't try to optimize it rather than assert it never happens.
19462 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
19464 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
19465 macro expansions for definition of ST_INTERNAL_<mode> and
19466 LD_INTERNAL_<mode> builtins.
19467 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
19469 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
19471 (altivec_expand_st_builtin): Likewise.
19472 (altivec_expand_builtin): Remove calls to deleted functions.
19473 (rs6000_address_for_altivec): Delete this function.
19474 * config/rs6000/vector.md: Remove expands for
19475 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
19477 2018-03-22 Sudakshina Das <sudi.das@arm.com>
19480 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
19481 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
19482 re-computing once computed.
19483 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
19484 (arm_init_machine_status): Initialize
19485 machine->static_chain_stack_bytes.
19487 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
19490 * doc/extend.texi: Add four new prototypes for vec_ld.
19491 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
19492 definitions for more logical presentation.
19493 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
19494 entries for V1TI variants of __builtin_altivec_ld builtin.
19495 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
19496 handling of V1TI variant of LVX icode pattern.
19497 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
19498 (rs6000_gimple_fold_builtin): Likewise.
19499 (altivec_init_builtins): Add code to define
19500 __builtin_altivec_lvx_v1ti function.
19502 2018-03-22 Jakub Jelinek <jakub@redhat.com>
19504 PR inline-asm/84941
19505 * function.c (match_asm_constraints_1): Don't do the optimization
19506 if input isn't a REG, SUBREG, MEM or constant.
19508 2018-03-22 Tom de Vries <tom@codesourcery.com>
19510 PR tree-optimization/84956
19511 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
19512 bb_has_abnormal_pred.
19514 2018-03-22 Jakub Jelinek <jakub@redhat.com>
19517 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
19518 DECL_INITIAL (decl) to decl at the end.
19519 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
19520 adjust the comment.
19522 2018-03-21 Joseph Myers <joseph@codesourcery.com>
19524 * doc/extend.texi (__builtin_tgmath): Document when complex
19525 integer types are treated as _Complex _Float64.
19527 2018-03-21 Tom de Vries <tom@codesourcery.com>
19529 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
19531 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19533 PR tree-optimization/84960
19534 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
19535 if it is ENTRY block, move them into single succ of ENTRY in that case.
19537 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
19539 PR tree-optimization/84811
19540 * poly-int.h (poly_span_traits): Remove the T3 parameter and
19541 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
19542 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
19543 (known_subrange_p): Update accordingly. Cast each value involved
19544 in the size comparison, rather than casting the result of the
19547 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19549 PR tree-optimization/84982
19550 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
19551 by flipping the least significant bit rather than all bits from
19552 bitpos to bitpos + bitsize - 1.
19554 2018-03-21 Nathan Sidwell <nathan@acm.org>
19556 * doc/extend.texi (Deprecated Features): Remove mention of
19557 long-deleted deprecations.
19559 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19562 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
19563 * configure: Regenerate.
19565 2018-03-21 Tom de Vries <tom@codesourcery.com>
19567 PR tree-optimization/83126
19568 * tree-parloops.c (num_phis): New function.
19569 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
19571 2018-03-21 Nathan Sidwell <nathan@acm.org>
19573 * doc/extend.texi (Deprecated Features): Update deprecated flags,
19574 mention anon-struct/union members and trailing attributes.
19576 2018-03-21 Bin Cheng <bin.cheng@arm.com>
19578 PR tree-optimization/84969
19579 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
19580 builtin memset partitions if they set different rhs values.
19582 2018-03-21 Jakub Jelinek <jakub@redhat.com>
19584 PR rtl-optimization/84989
19585 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
19586 VEC_DUPLICATE with scalar result mode.
19588 2018-03-21 Martin Liska <mliska@suse.cz>
19591 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
19592 not intended return statement.
19594 2018-03-21 Martin Liska <mliska@suse.cz>
19597 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
19598 (chkp_find_bound_slots_1): Limit number of iterations.
19600 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
19603 * Minor grammar fixes for x86 options.
19605 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19608 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
19609 holding REG_CFA_RESTORE notes, instead turn them into a USE.
19611 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
19614 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
19615 (altivec_lvx_<mode>_1op): Likewise.
19616 (altivec_stvx_<mode>_2op): Likewise.
19617 (altivec_stvx_<mode>_1op): Likewise.
19618 (altivec_lvx_<VM2:mode>): New define_expand.
19619 (altivec_stvx_<VM2:mode>): Likewise.
19620 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
19621 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19622 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
19623 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
19624 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
19625 (rs6000_gen_lvx): Likewise.
19626 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
19627 (altivec_expand_stv_builtin): Likewise.
19628 (altivec_expand_builtin): Likewise.
19629 * config/rs6000/vector.md: Likewise.
19631 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19634 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
19637 2018-03-20 Richard Biener <rguenther@suse.de>
19640 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
19641 sign-conversions as zero, fall back to standard scalar_stmt
19644 2018-03-20 Martin Liska <mliska@suse.cz>
19647 * predict.c (rebuild_frequencies): Handle case when we have
19648 PROFILE_ABSENT, but flag_guess_branch_prob is false.
19650 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19653 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
19654 flag_section_anchors.
19655 * varasm.c (use_blocks_for_decl_p): Remove hack for
19656 dw2_force_const_mem.
19659 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
19661 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
19662 be created, use lowpart_subreg of operands[0] rather than operands[0]
19664 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
19665 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
19666 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
19667 and n constraint instead of aarch64_shift_imm_di and Usd.
19668 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
19669 (*aarch64_<optab>_reg_minus<mode>3): ... this.
19671 2018-03-20 Sudakshina Das <sudi.das@arm.com>
19674 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
19675 to favor GPR over NEON registers.
19676 (<shift>di3_neon): Likewise.
19678 2018-03-20 Tom de Vries <tom@codesourcery.com>
19681 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
19682 (nvptx_process_pars): Emit bar.sync asap and alap.
19684 2018-03-20 Tom de Vries <tom@codesourcery.com>
19687 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
19688 seen_label if seen_label is already set.
19690 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19693 * config/i386/i386.c (fold_builtin_cpu): For features above 31
19694 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
19695 Use 1U instead of 1. Formatting fixes.
19698 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
19699 instead of TREE_TYPE (s1) for the return value.
19701 2018-03-19 Jakub Jelinek <jakub@redhat.com>
19703 PR tree-optimization/84946
19704 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
19705 bitsize + bitsize in poly_uint64 rather than poly_int64.
19708 * dwarf2asm.c: Include fold-const.c.
19709 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
19710 of decl rather than decl itself.
19712 PR rtl-optimization/84643
19713 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
19715 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
19718 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
19719 calling assemble_variable.
19721 2018-03-19 Sudakshina Das <sudi.das@arm.com>
19724 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
19725 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
19727 2018-03-19 Jim Wilson <jimw@sifive.com>
19730 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
19731 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
19732 (riscv_first_stack_step): Likewise.
19733 (riscv_option_override): Use STACK_BOUNDARY instead of
19734 MIN_STACK_BOUNDARY.
19735 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
19736 MIN_STACK_BOUNDARY.
19737 (BIGGEST_ALIGNMENT): Set to 128.
19738 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19739 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
19742 2018-03-19 Richard Biener <rguenther@suse.de>
19744 PR tree-optimization/84933
19745 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
19746 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19748 2018-03-19 Richard Biener <rguenther@suse.de>
19750 PR tree-optimization/84859
19751 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19752 (cond_if_else_store_replacement): Perform sinking operation on
19753 single-store BBs regardless of MAX_STORES_TO_SINK setting.
19754 Generalize what a BB with a single eligible store is.
19756 2018-03-19 Richard Biener <rguenther@suse.de>
19758 PR tree-optimization/84929
19759 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19760 chrec_is_positive against non-chrec arg.
19762 2018-03-19 Tamar Christina <tamar.christina@arm.com>
19765 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19767 2018-03-18 Martin Liska <mliska@suse.cz>
19769 PR rtl-optimization/84635
19770 * regrename.c (build_def_use): Use matches_mode only when
19773 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
19775 PR tree-optimization/84913
19776 * tree-vect-loop.c (vectorizable_reduction): Don't try to
19777 vectorize chains of COND_EXPRs.
19779 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19781 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19783 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19785 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19787 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19789 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19791 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19792 Kito Cheng <kito.cheng@gmail.com>
19794 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19795 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19796 (nds32_adjust_reg_alloc_order): New function.
19797 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19799 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
19801 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19802 nds32_print_operand, nds32_print_operand_address): Use
19803 HOST_WIDE_INT_PRINT_DEC instead.
19805 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19807 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19809 2018-03-17 Jakub Jelinek <jakub@redhat.com>
19812 * config/i386/i386.c (initial_ix86_tune_features,
19813 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19814 unsigned long long.
19815 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19816 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19817 rather than 1u << ix86_tune. Formatting fix.
19818 (ix86_option_override_internal): Change ix86_arch_mask from
19819 unsigned int to unsigned HOST_WIDE_INT, initialize to
19820 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19821 (ix86_function_specific_restore): Likewise.
19823 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19826 * postreload.c (reload_combine_recognize_pattern): Perform
19827 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19828 truncate_int_for_mode the result for the destination's mode.
19831 * hsa-gen.c (mem_type_for_type): Fix comment typo.
19832 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19834 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19837 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
19840 * lra-assigns.c (lra_split_hard_reg_for): Don't use
19841 regno_allocno_class_array and sorted_pseudos.
19842 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
19843 insns where regno is used.
19845 2018-03-16 Martin Liska <mliska@suse.cz>
19848 * multiple_target.c (create_dispatcher_calls): Redirect
19849 reference in the symbol table.
19851 2018-03-16 Martin Liska <mliska@suse.cz>
19854 * multiple_target.c (create_dispatcher_calls): Redirect also
19857 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19861 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
19862 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
19863 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
19865 2018-03-16 Julia Koval <julia.koval@intel.com>
19867 * doc/invoke.texi (Skylake Server): Add CLWB.
19868 Cannonlake): Remove CLWB.
19870 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19872 PR tree-optimization/84841
19873 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19875 (FLOAT_ONE_CONST_TYPE): Define.
19876 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19877 (sort_by_operand_rank): Put entries with higher constant_type last
19878 rather than first to match comments.
19880 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
19882 * config/nios2/nios2.md (movsi_internal): Fix thinko in
19885 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19888 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19889 check and use address of target always.
19891 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
19894 * config/i386/i386.c (indirect_thunk_needed): Update comments.
19895 (indirect_thunk_bnd_needed): Likewise.
19896 (indirect_thunks_used): Likewise.
19897 (indirect_thunks_bnd_used): Likewise.
19898 (indirect_return_needed): New.
19899 (indirect_return_bnd_needed): Likewise.
19900 (output_indirect_thunk_function): Add a bool argument for
19902 (output_indirect_thunk_function): Don't generate alias for
19903 function return thunk.
19904 (ix86_code_end): Call output_indirect_thunk_function to generate
19905 function return thunks.
19906 (ix86_output_function_return): Set indirect_return_bnd_needed
19907 and indirect_return_needed instead of indirect_thunk_bnd_needed
19908 and indirect_thunk_needed.
19910 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
19912 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19913 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19914 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19916 2018-03-15 David Malcolm <dmalcolm@redhat.com>
19917 Paul Hua <paul.hua.gm@gmail.com>
19920 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19922 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
19924 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19925 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19926 resp. SFmode cases.
19928 2018-03-15 Tamar Christina <tamar.christina@arm.com>
19931 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19932 instead of GET_MODE_SIZE when comparing Units.
19934 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19937 * varasm.c (hash_section): Return an unchangeble hash value
19938 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19939 Return !aarch64_can_use_per_function_literal_pools_p ().
19941 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19944 * optabs.c (emit_conditional_move): Pass address of cmode's copy
19945 rather than address of cmode as last argument to prepare_cmp_insn.
19947 2018-03-15 Julia Koval <julia.koval@intel.com>
19949 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19950 F_AVX512VNNI, F_AVX512BITALG): New.
19952 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
19955 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
19956 insn for floating-point loads and stores.
19958 2018-03-14 Carl Love <cel@us.ibm.com>
19960 * config/rs6000/rs6000-c.c: Add macro definitions for
19961 ALTIVEC_BUILTIN_VEC_PERMXOR.
19962 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19963 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19964 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19965 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19967 * config/doc/extend.texi: Add prototypes for vec_permxor.
19969 2018-03-14 David Malcolm <dmalcolm@redhat.com>
19972 * diagnostic-show-locus.c (class layout_point): Convert m_line
19973 from int to linenum_type.
19974 (line_span::comparator): Use linenum "compare" function when
19975 comparing line numbers.
19976 (test_line_span): New function.
19977 (layout_range::contains_point): Convert param "row" from int to
19979 (layout_range::intersects_line_p): Likewise.
19980 (layout::will_show_line_p): Likewise.
19981 (layout::print_source_line): Likewise.
19982 (layout::should_print_annotation_line_p): Likewise.
19983 (layout::print_annotation_line): Likewise.
19984 (layout::print_leading_fixits): Likewise.
19985 (layout::annotation_line_showed_range_p): Likewise.
19986 (struct line_corrections): Likewise for field m_row.
19987 (line_corrections::line_corrections): Likewise for param "row".
19988 (layout::print_trailing_fixits): Likewise.
19989 (layout::get_state_at_point): Likewise.
19990 (layout::get_x_bound_for_row): Likewise.
19991 (layout::print_line): Likewise.
19992 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19993 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19994 * input.c (selftest::test_linenum_comparisons): New function.
19995 (selftest::input_c_tests): Call it.
19996 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19997 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19998 * selftest.h (ASSERT_GT): New macro.
19999 (ASSERT_GT_AT): New macro.
20000 (ASSERT_LT): New macro.
20001 (ASSERT_LT_AT): New macro.
20003 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
20005 PR rtl-optimization/84780
20006 * combine.c (distribute_links): Don't make a link based on pc_rtx.
20008 2018-03-14 Martin Liska <mliska@suse.cz>
20010 * tree.c (record_node_allocation_statistics): Use
20011 get_stats_node_kind.
20012 (get_stats_node_kind): New function extracted from
20013 record_node_allocation_statistics.
20014 (free_node): Use get_stats_node_kind.
20016 2018-03-14 Richard Biener <rguenther@suse.de>
20018 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20019 that the value-set of ANTIC_IN doesn't grow.
20022 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20024 (BB_VISITED_WITH_VISITED_SUCCS): New define.
20025 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20027 2018-03-14 Julia Koval <julia.koval@intel.com>
20029 * config.gcc (icelake-client, icelake-server): New.
20031 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20032 (initial_ix86_arch_features): Ditto.
20033 (PTA_SKYLAKE): Add SGX.
20034 (PTA_ICELAKE): Remove.
20035 (PTA_ICELAKE_CLIENT): New.
20036 (PTA_ICELAKE_SERVER): New.
20037 (ix86_option_override_internal): Split up icelake on icelake client and
20039 (get_builtin_code_for_version): Ditto.
20040 (fold_builtin_cpu): Ditto.
20041 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20042 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20043 * config/i386/i386.h (processor_type): Ditto.
20044 * doc/invoke.texi: Ditto.
20046 2018-03-14 Jakub Jelinek <jakub@redhat.com>
20049 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
20050 INTEGER_CST offset, add it together with bitpos / 8 and
20051 sign extend based on POINTER_SIZE.
20055 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
20058 * config/i386/constraints.md (Yc): New register constraint.
20059 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
20060 Use Yc constraint for alternative 2 of operand 0. Remove
20061 preferred_for_speed attribute.
20063 2018-03-14 Richard Biener <rguenther@suse.de>
20065 PR tree-optimization/84830
20066 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
20067 with the old one to avoid oscillations.
20069 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
20072 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
20074 (assign_by_spills): Return a flag of reload assignment failure.
20075 Do not process the reload assignment failures. Do not spill other
20076 reload pseudos if they has the same reg class. Update n if
20078 (lra_assign): Add a return arg. Set up from the result of
20079 assign_by_spills call.
20080 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20081 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
20082 usage_insns if it is not NULL.
20083 (spill_hard_reg_in_range): New function.
20084 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20085 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20086 function prototypes.
20087 (lra_assign): Change prototype.
20088 * lra.c (lra): Add code to deal with fails by splitting hard reg
20091 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
20093 * config/riscv/riscv.opt (mrelax): New option.
20094 * config/riscv/riscv.c (riscv_file_start): Emit ".option
20095 "norelax" when riscv_mrelax is disabled.
20096 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
20098 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20101 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
20102 reassociation for int modes.
20104 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
20106 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
20107 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
20109 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
20110 * config/aarch64/aarch64-sve.md
20111 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
20112 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
20113 (*extend<mode><Vwide>2): Rename to...
20114 (aarch64_sve_extend<mode><Vwide>2): ...this.
20115 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
20116 renaming the old pattern to...
20117 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
20119 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
20120 define_expand, renaming the old pattern to...
20121 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
20122 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
20123 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
20124 account when deciding which SVE instruction the optab should use.
20125 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
20127 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
20129 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
20130 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
20131 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
20132 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
20133 (tlsdesc_small_<mode>): Turn a define_expand and use
20134 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
20135 (tlsdesc_small_advsimd_<mode>): ...this.
20136 (tlsdesc_small_sve_<mode>): New pattern.
20138 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
20140 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
20141 (UNSPEC_UMUL_HIGHPART): New constants.
20142 (MUL_HIGHPART): New int iteraor.
20143 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
20144 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20146 (*<su>mul<mode>3_highpart): New define_insn.
20148 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
20151 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20154 2018-03-13 Martin Liska <mliska@suse.cz>
20157 * (sem_item_optimizer::sem_item_optimizer): Initialize new
20159 (sem_item_optimizer::~sem_item_optimizer): Release it.
20160 (sem_item_optimizer::merge_classes): Register variable aliases.
20161 (sem_item_optimizer::fixup_pt_set): New function.
20162 (sem_item_optimizer::fixup_points_to_sets): Likewise.
20163 * ipa-icf.h: Declare new variables and functions.
20165 2018-03-13 Jakub Jelinek <jakub@redhat.com>
20167 PR middle-end/84834
20168 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
20169 integer_pow2p@2 and test integer_pow2p in condition.
20170 (A < 0 ? C : 0): Similarly for @1.
20172 PR middle-end/84831
20173 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
20174 characters starting at p contain '\0' character, don't look beyond
20178 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
20179 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
20182 * reg-stack.c (change_stack): Change update_end var from int to
20183 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
20184 also call set_block_for_insn on the newly added insns and rescan.
20187 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
20188 on the last operand.
20191 * tree.c (stabilize_reference_1): Return save_expr (e) for
20192 STATEMENT_LIST even if it doesn't have side-effects.
20194 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
20196 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
20198 2018-03-12 Renlin Li <renlin.li@arm.com>
20200 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
20201 aarch64_output_scalar_simd_mov_immediate.
20203 2018-03-12 Martin Sebor <msebor@redhat.com>
20205 PR tree-optimization/83456
20206 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
20207 for perfectly overlapping calls to memcpy.
20208 (gimple_fold_builtin_memory_chk): Same.
20209 (gimple_fold_builtin_strcpy): Handle no-warning.
20210 (gimple_fold_builtin_stxcpy_chk): Same.
20211 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
20213 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
20215 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
20216 parameter. Use it for SFmode.
20217 (rs6000_function_arg_advance_1): Adjust.
20218 (rs6000_function_arg): Adjust.
20219 (rs6000_gimplify_va_arg): Pass false for that new parameter.
20221 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
20223 PR rtl-optimization/84169
20224 PR rtl-optimization/84780
20225 * combine.c (can_combine_p): Check for a 2-insn combination whether
20226 the destination register is used between the two insns, too.
20228 2018-03-12 Richard Biener <rguenther@suse.de>
20230 PR tree-optimization/84803
20231 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20232 for refs DR analysis didn't process.
20234 2018-03-12 Richard Biener <rguenther@suse.de>
20236 PR tree-optimization/84777
20237 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
20238 force-vectorize loops ignore whether we are optimizing for size.
20240 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
20242 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
20243 (TARGET_MD_ASM_ADJUST): Define.
20245 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
20246 Kito Cheng <kito.cheng@gmail.com>
20247 Chung-Ju Wu <jasonwucj@gmail.com>
20249 * config/nds32/nds32.c (nds32_compute_stack_frame,
20250 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
20251 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
20252 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
20253 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
20254 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
20255 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
20256 * config/nds32/nds32.md (prologue, epilogue): Use macro
20257 NDS32_V3PUSH_AVAILABLE_P to do checking.
20259 2018-03-11 Jakub Jelinek <jakub@redhat.com>
20262 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
20263 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
20264 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
20265 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
20266 addition of most attributes on !orig_type_die or the attribute not
20267 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
20269 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
20270 Chung-Ju Wu <jasonwucj@gmail.com>
20272 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
20273 __NDS32_VH__ macro.
20274 * config/nds32/nds32.opt (mvh): New option.
20276 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
20277 Chung-Ju Wu <jasonwucj@gmail.com>
20279 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
20281 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
20282 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
20285 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
20286 Chung-Ju Wu <jasonwucj@gmail.com>
20288 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
20290 * config/nds32/nds32-multiple.md (strlensi): New pattern.
20291 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
20293 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
20294 Kito Cheng <kito.cheng@gmail.com>
20295 Chung-Ju Wu <jasonwucj@gmail.com>
20297 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
20298 UNSPEC_FFMISM and UNSPEC_FLMISM.
20299 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
20300 for ffb, ffmism and flmism.
20301 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
20302 (unspec_ffmism): Ditto.
20303 (unspec_flmism): Ditto.
20304 (nds32_expand_builtin_impl): Check if string extension is available.
20305 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
20306 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
20308 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
20311 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
20314 * lra-assigns.c (assign_by_spills): Return a flag of reload
20315 assignment failure. Do not process the reload assignment
20316 failures. Do not spill other reload pseudos if they has the same
20318 (lra_assign): Add a return arg. Set up from the result of
20319 assign_by_spills call.
20320 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20321 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
20322 usage_insns if it is not NULL.
20323 (spill_hard_reg_in_range): New function.
20324 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20325 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20326 function prototypes.
20327 (lra_assign): Change prototype.
20328 * lra.c (lra): Add code to deal with fails by splitting hard reg
20331 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
20334 * config/i386/i386.opt: Replace Enforcment with Enforcement.
20336 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
20339 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
20340 (dw_val_node): Add val_symbolic_view.
20341 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
20342 (symview_upper_bound): New.
20343 (new_line_info_table): Initialize symviews_since_reset.
20344 (dwarf2out_source_line): Count symviews_since_reset and set
20345 symview_upper_bound.
20346 (dw_val_equal_p): Handle symview.
20347 (add_AT_symview): New.
20348 (print_dw_val): Handle symview.
20349 (attr_checksum, attr_checksum_ordered): Likewise.
20350 (same_dw_val_p, size_of_die): Likewise.
20351 (value_format, output_die): Likewise.
20352 (add_high_low_attributes): Use add_AT_symview for entry_view.
20353 (dwarf2out_finish): Reset symview_upper_bound, clear
20356 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
20359 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
20360 Add strict argument and use it.
20361 (rs6000_split_multireg_move): Update for new strict argument.
20362 (mem_operand_gpr): Disallow all non-offsettable addresses.
20363 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
20365 2018-03-09 Jakub Jelinek <jakub@redhat.com>
20368 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
20369 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
20370 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
20373 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
20374 decl, use remap_type if we want to use the type.
20376 2018-03-09 Martin Sebor <msebor@redhat.com>
20378 PR tree-optimization/84526
20379 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20381 (builtin_access::generic_overlap): Be prepared to handle non-array
20384 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
20386 PR rtl-optimization/84682
20387 * lra-constraints.c (process_address_1): Check is_address flag
20388 for address constraints.
20389 (process_alt_operands): Likewise.
20390 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
20391 preprocess_constraints.
20392 * recog.h (preprocess_constraints): Add oploc parameter.
20394 * recog.c (preprocess_constraints): Test address_operand for
20395 CT_ADDRESS constraints.
20397 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
20400 * lra-assigns.c (assign_by_spills): Return a flag of reload
20401 assignment failure. Do not process the reload assignment
20402 failures. Do not spill other reload pseudos if they has the same
20404 (lra_assign): Add a return arg. Set up from the result of
20405 assign_by_spills call.
20406 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20407 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
20408 usage_insns if it is not NULL.
20409 (spill_hard_reg_in_range): New function.
20410 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20411 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20412 function prototypes.
20413 (lra_assign): Change prototype.
20414 * lra.c (lra): Add code to deal with fails by splitting hard reg
20417 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20420 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
20421 Accept complain bool parameter. Only emit errors if it is true.
20422 (arm_parse_cpu_option_name): Likewise.
20423 (arm_target_thumb_only): Adjust callers of the above.
20424 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
20425 prototype to take a default true bool parameter.
20426 (arm_parse_arch_option_name): Likewise.
20428 2018-03-09 David Malcolm <dmalcolm@redhat.com>
20429 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
20433 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
20434 * configure: Regenerate.
20435 * configure.ac ("linker --version-script option"): New.
20436 ("linker soname option"): New.
20438 2018-03-09 Richard Biener <rguenther@suse.de>
20440 PR tree-optimization/84775
20441 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
20442 immediate uses of predicate stmts and mark them modified.
20445 PR tree-optimization/84178
20446 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20448 (version_loop_for_if_conversion): Delay update_ssa call.
20449 (tree_if_conversion): Delay update_ssa until after predicate
20452 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
20455 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
20456 when the function accesses prior frames.
20458 2018-03-08 Jakub Jelinek <jakub@redhat.com>
20461 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
20462 gen_llsym, otherwise call maybe_gen_llsym.
20464 PR inline-asm/84742
20465 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
20466 has ',' character inside of it.
20468 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20471 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
20472 as clobbering CC_REGNUM.
20474 2018-03-08 Richard Biener <rguenther@suse.de>
20476 PR middle-end/84552
20477 * tree-scalar-evolution.c: Include tree-into-ssa.h.
20478 (follow_copies_to_constant): Do not follow SSA names registered
20481 2018-03-08 Richard Biener <rguenther@suse.de>
20483 PR tree-optimization/84178
20484 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
20486 (version_loop_for_if_conversion): Delay update_ssa call.
20487 (tree_if_conversion): Delay update_ssa until after predicate
20490 2018-03-08 David Malcolm <dmalcolm@redhat.com>
20492 PR tree-optimization/84178
20493 * tree-if-conv.c (release_bb_predicate): Remove the
20494 the assertion that the stmts have NULL use_ops.
20495 Discard the statements, asserting that they haven't
20496 yet been added to a BB.
20498 2018-03-08 Richard Biener <rguenther@suse.de>
20500 PR tree-optimization/84746
20501 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
20502 (phi_translate): Pass in destination ANTIC_OUT set.
20503 (phi_translate_1): Likewise. For a simplified result lookup
20504 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
20505 (phi_translate_set): Adjust.
20506 (do_pre_regular_insertion): Likewise.
20507 (do_pre_partial_partial_insertion): Likewise.
20509 2018-03-08 Martin Liska <mliska@suse.cz>
20511 PR gcov-profile/84735
20512 * doc/gcov.texi: Document usage of profile files.
20513 * gcov-io.h: Document changes in the format.
20515 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
20519 * dwarf2out.c (struct dw_line_info_table): Update comments for
20521 (FORCE_RESET_NEXT_VIEW): New.
20522 (FORCE_RESETTING_VIEW_P): New.
20523 (RESETTING_VIEW_P): Check for -1 too.
20524 (ZERO_VIEW_P): Likewise.
20525 (new_line_info_table): Force-reset next view.
20526 (dwarf2out_begin_function): Likewise.
20527 (dwarf2out_source_line): Simplify zero_view_p initialization.
20528 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
20529 view directly. Omit view when omitting .loc at line 0.
20531 2018-03-08 Jakub Jelinek <jakub@redhat.com>
20533 PR tree-optimization/84740
20534 * tree-switch-conversion.c (process_switch): Call build_constructors
20535 only if info.phi_count is non-zero.
20537 PR tree-optimization/84739
20538 * tree-tailcall.c (find_tail_calls): Check call arguments against
20539 DECL_ARGUMENTS (current_function_decl) rather than
20540 DECL_ARGUMENTS (func) when checking for tail recursion.
20542 2018-03-07 Jakub Jelinek <jakub@redhat.com>
20544 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
20545 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
20546 Volker Reichelt's entry and add entries for people that perform
20547 GCC fuzzy testing and report numerous bugs.
20549 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
20552 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
20553 readonly data in sdata, if that is disabled.
20554 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
20555 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
20556 -mreadonly-in-sdata option.
20558 2018-03-07 Martin Sebor <msebor@redhat.com>
20560 PR tree-optimization/84468
20561 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
20562 basic block when looking for nul assignment.
20564 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
20567 * except.h (output_function_exception_table): Adjust prototype.
20568 * except.c (output_function_exception_table): Remove FNNAME parameter
20569 and add SECTION parameter. Ouput one part of the table at a time.
20570 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
20571 the first part of the exception table and emit unwind directives.
20572 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
20573 (i386_pe_seh_cold_init): Likewise.
20574 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
20575 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
20576 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
20577 (ix86_output_call_insn): Emit a nop in one more case for SEH.
20578 * config/i386/winnt.c: Include except.h.
20579 (struct seh_frame_state): Add reg_offset, after_prologue and
20580 in_cold_section fields.
20581 (i386_pe_seh_end_prologue): Set seh->after_prologue.
20582 (i386_pe_seh_cold_init): New function.
20583 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
20584 to seh->in_cold_section.
20585 (seh_emit_push): Record the offset of the push.
20586 (seh_emit_save): Record the offet of the save.
20587 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
20588 Test seh->after_prologue to disregard the epilogue.
20589 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
20590 (i386_pe_end_cold_function): New function.
20592 2018-03-07 Jakub Jelinek <jakub@redhat.com>
20595 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
20596 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
20599 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
20601 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
20602 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
20604 PR middle-end/84723
20605 * multiple_target.c: Include tree-inline.h and intl.h.
20606 (expand_target_clones): Diagnose and fail if node->definition and
20607 !tree_versionable_function_p (node->decl).
20609 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
20611 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
20613 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
20614 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20615 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
20617 2018-03-06 Jakub Jelinek <jakub@redhat.com>
20620 * combine.c (try_combine): Use reg_or_subregno instead of handling
20621 just paradoxical SUBREGs and REGs.
20623 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
20625 * config/arc/arc.c (arc_finalize_pic): Remove function.
20626 (arc_must_save_register): We use single base PIC register, remove
20627 checks to save/restore the PIC register.
20628 (arc_expand_prologue): Likewise.
20629 * config/arc/arc-protos.h (arc_set_default_type_attributes):
20631 (arc_verify_short): Likewise.
20632 (arc_attr_type): Likewise.
20633 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
20634 (walk_stores): Likewise.
20635 (arc_address_cost): Make it static.
20636 (arc_verify_short): Likewise.
20637 (branch_dest): Likewise.
20638 (arc_attr_type): Likewise.
20639 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
20640 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
20641 (arc_final_prescan_insn): Remove inserting the nops due to
20642 hardware hazards. It is done in reorg step.
20643 (insn_length_variant_t): Remove.
20644 (insn_length_parameters_t): Likewise.
20645 (arc_insn_length_parameters): Likewise.
20646 (arc_get_insn_variants): Likewise.
20647 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
20649 2018-03-06 Jakub Jelinek <jakub@redhat.com>
20651 PR inline-asm/84683
20652 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
20655 PR tree-optimization/84687
20656 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
20658 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
20660 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20662 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
20663 Rename to ppc_speculation_barrier.
20664 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
20665 __builtin_ppc_speculation_barrier.
20667 2018-03-05 Jakub Jelinek <jakub@redhat.com>
20670 * combine.c (combine_simplify_rtx): Don't try to simplify if
20671 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
20674 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
20676 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
20677 to 32 bytes when compiling for POWER9.
20679 2018-03-05 Jakub Jelinek <jakub@redhat.com>
20682 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
20683 regparm >= 3 with no arg reg available also for calls with
20684 flag_force_indirect_call. Pass decl to ix86_function_regparm.
20687 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
20689 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
20691 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
20694 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
20696 2018-03-05 Richard Biener <rguenther@suse.de>
20698 PR tree-optimization/84486
20699 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
20700 When inserting a __builtin_assume_aligned call set the LHS
20701 SSA name alignment info accordingly.
20703 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
20705 PR tree-optimization/84114
20706 * config/aarch64/aarch64.c (aarch64_reassociation_width)
20707 Avoid reassociation of FLOAT_MODE addition.
20709 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
20711 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
20712 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
20713 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
20714 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
20715 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
20716 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
20717 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
20719 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
20720 __builtin_ia32_wbinvd): New builtins.
20721 (SPECIAL_ARGS2): New.
20722 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
20723 (SPECIAL_ARGS2): New.
20724 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
20725 (ix86_valid_target_attribute_inner_p): Ditto.
20726 (ix86_init_mmx_sse_builtins): Add special_args2.
20727 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
20728 TARGET_WBNOINVD_P): New.
20729 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
20730 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
20731 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
20732 * config/i386/immintrin.h (_wbinvd): New intrinsic.
20733 * config/i386/pconfigintrin.h: New file.
20734 * config/i386/wbnoinvdintrin.h: Ditto.
20735 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
20737 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
20739 2018-03-05 Richard Biener <rguenther@suse.de>
20741 PR tree-optimization/84670
20742 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20744 (BB_VISITED_WITH_VISITED_SUCCS): New define.
20745 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20746 (compute_antic_aux): Only assert the number of values in ANTIC_IN
20747 doesn't grow if all successors (recursively) were visited at least
20750 2018-03-05 Richard Biener <rguenther@suse.de>
20752 PR tree-optimization/84650
20753 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20754 if executed in the loop pipeline.
20756 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
20758 * doc/configfiles.texi (Configuration Files): Move info about
20759 conditionalizing $target-protos.h to...
20760 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
20761 differs from $target-protos.h.
20763 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
20764 Chung-Ju Wu <jasonwucj@gmail.com>
20766 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20767 * config/nds32/nds32-multiple.md (setmemsi): Define.
20768 * config/nds32/nds32-memory-manipulation.c
20769 (nds32_gen_dup_4_byte_to_word_value): New.
20770 (emit_setmem_word_loop): New.
20771 (emit_setmem_byte_loop): New.
20772 (nds32_expand_setmem_loop): New.
20773 (nds32_expand_setmem_loop_v3m): New.
20774 (nds32_expand_setmem_unroll): New.
20775 (nds32_expand_setmem): New.
20777 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20778 Chung-Ju Wu <jasonwucj@gmail.com>
20780 * config/nds32/nds32-memory-manipulation.c
20781 (nds32_emit_load_store): New.
20782 (nds32_emit_post_inc_load_store): New.
20783 (nds32_emit_mem_move): New.
20784 (nds32_emit_mem_move_block): New.
20785 (nds32_expand_movmemsi_loop_unknown_size): New.
20786 (nds32_expand_movmemsi_loop_known_size): New.
20787 (nds32_expand_movmemsi_loop): New.
20788 (nds32_expand_movmemsi_unroll): New.
20789 (nds32_expand_movmemqi): Rename ...
20790 (nds32_expand_movmemsi): ... to this.
20791 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20792 (movmemsi): ... to this.
20793 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20794 (nds32_expand_movmemsi): ... to this.
20796 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20797 Monk Chiang <sh.chiang04@gmail.com>
20798 Chung-Ju Wu <jasonwucj@gmail.com>
20800 * config/nds32/nds32-protos.h
20801 (nds32_expand_load_multiple): New arguments.
20802 (nds32_expand_store_multiple): Ditto.
20803 (nds32_valid_multiple_load_store): Rename ...
20804 (nds32_valid_multiple_load_store_p): ... to this.
20805 * config/nds32/nds32-memory-manipulation.c
20806 (nds32_expand_load_multiple): Refine implementation.
20807 (nds32_expand_store_multiple): Ditto.
20808 * config/nds32/nds32-multiple.md
20809 (load_multiple): Update nds32_expand_load_multiple interface.
20810 (store_multiple): Update nds32_expand_store_multiple interface.
20811 * config/nds32/nds32-predicates.c
20812 (nds32_valid_multiple_load_store): Rename ...
20813 (nds32_valid_multiple_load_store_p): ... to this and refine
20815 * config/nds32/predicates.md
20816 (nds32_load_multiple_and_update_address_operation): New predicate.
20817 (nds32_store_multiple_and_update_address_operation): New predicate.
20819 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20820 Chung-Ju Wu <jasonwucj@gmail.com>
20822 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20823 (combo): New attribute.
20824 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20826 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
20828 * config/nds32/nds32.opt: Change -mcmodel= default value.
20830 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
20831 Monk Chiang <sh.chiang04@gmail.com>
20832 Chung-Ju Wu <jasonwucj@gmail.com>
20834 * config/nds32/constants.md (unspec_element): New enum.
20835 * config/nds32/constraints.md (Umw): New constraint.
20836 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
20837 * config/nds32/nds32-intrinsic.md: Likewise.
20838 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
20839 (nds32_valid_smw_lwm_base_p): New.
20840 (nds32_output_smw_single_word): New.
20841 (nds32_output_lmw_single_word): New.
20842 (nds32_expand_unaligned_load): New.
20843 (nds32_expand_unaligned_store): New.
20844 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20845 (nds32_output_smw_single_word): Declare.
20846 (nds32_output_lmw_single_word): Declare.
20847 (nds32_expand_unaligned_load): Declare.
20848 (nds32_expand_unaligned_store): Declare.
20849 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
20850 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
20851 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
20852 NDS32_BUILTIN_UASTORE_DW.
20853 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
20856 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
20857 Kito Cheng <kito.cheng@gmail.com>
20858 Chung-Ju Wu <jasonwucj@gmail.com>
20860 * config/nds32/nds32-intrinsic.c
20861 (nds32_expand_builtin_null_ftype_reg): Delete.
20862 (nds32_expand_builtin_reg_ftype_imm): Ditto.
20863 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
20864 (nds32_read_argument): New.
20865 (nds32_legitimize_target): Ditto.
20866 (nds32_legitimize_argument): Ditto.
20867 (nds32_check_constant_argument): Ditto.
20868 (nds32_expand_unop_builtin): Ditto.
20869 (nds32_expand_unopimm_builtin): Ditto.
20870 (nds32_expand_binop_builtin): Ditto.
20871 (nds32_builtin_decl_impl): Ditto.
20872 (builtin_description): Ditto.
20873 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20874 (nds32_init_builtins_impl): Ditto.
20875 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20876 (nds32_builtin_decl): New.
20877 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20878 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20880 2018-03-02 Jeff Law <law@redhat.com>
20882 * reorg.c (stop_search_p): Handle DEBUG_INSN.
20883 (redundant_insn, fill_simple_delay_slots): Likewise.
20884 (fill_slots_from_thread): Likewise.
20885 * resource.c (mark_referenced_resources): Likewise.
20886 (mark_set_resources, find_dead_or_set_registers): Likewise.
20888 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20890 * substring-locations.h (format_warning_va): Formatting fix for
20891 ATTRIBUTE_GCC_DIAG.
20892 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20894 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20895 * substring-locations.c: Include intl.h.
20896 (format_warning_va): Turned into small wrapper around
20897 format_warning_n_va, renamed to ...
20898 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20899 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20901 (format_warning_at_substring_n): New function.
20902 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20903 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
20904 format_warning_at_substring with just a shorter name instead of
20905 const function pointer.
20906 (fmtwarn_n): New function.
20907 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20908 appropriate, get rid of all the fmtstr temporaries, move conditionals
20909 with G_() wrapped string literals directly into fmtwarn arguments,
20910 cast dir.len to (int), formatting fixes.
20912 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
20914 * doc/invoke.texi: Remove "Cilk Plus" references.
20916 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20917 Richard Biener <rguenther@suse.de>
20920 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20921 for error or warning attributes if CALL_FROM_THUNK_P is set.
20924 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20927 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20928 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20931 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20932 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20934 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20935 instead of -1U in last predictors element's probability member.
20937 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
20940 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20941 arguments if they are comparable.
20943 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20945 PR tree-optimization/84634
20946 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20947 masks and masked_loop_p with a single loop_masks, making sure it's
20948 null for bb vectorization.
20950 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20952 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20953 (vect_analyze_data_ref_access): Use loop->safe_len rather than
20954 loop->force_vectorize to check whether there is no alias.
20956 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20959 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20961 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20963 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20964 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20965 instead of a loop around prev_real_insn.
20966 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20969 PR inline-asm/84625
20970 * config/i386/i386.c (ix86_print_operand): Use conditional
20971 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20974 2018-03-02 Richard Biener <rguenther@suse.de>
20976 PR tree-optimization/84427
20977 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20978 (bitmap_set_subtract_values): Rewrite to handle multiple
20981 (prune_clobbered_mems): Likewise.
20982 (phi_translate): Take edge instead of pred/phiblock.
20983 (phi_translate_1): Likewise.
20984 (phi_translate_set): Likewise. Insert all translated
20985 exprs for a value into the set, keeping possibly multiple
20986 expressions per value.
20987 (compute_antic_aux): Adjust for phi_translate changes.
20988 When intersecting union the expressions and prune those
20989 not in the final value set, keeping possibly multiple
20990 expressions per value. Do not use value-insertion
20991 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20992 all expressions. Add verification that the value-sets
20993 only shrink during iteration.
20994 (compute_partial_antic_aux): Adjust for the phi_translate changes.
20995 (do_pre_regular_insertion): Likewise.
20996 (do_pre_partial_partial_insertion): Likewise.
20998 2018-03-02 Richard Biener <rguenther@suse.de>
21001 * config/darwin.c (saved_debug_info_level): New static global.
21002 (darwin_asm_lto_start): Disable debug info generation for LTO out.
21003 (darwin_asm_lto_end): Restore debug info generation settings.
21005 2018-03-01 Martin Liska <mliska@suse.cz>
21008 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21009 volatile arguments.
21011 2018-03-01 Richard Biener <rguenther@suse.de>
21014 * dwarf2out.c (gen_variable_die): Properly handle late VLA
21015 type annotation with LTO when debug was disabled at compile-time.
21017 2018-03-01 Matthew Fortune <mfortune@gmail.com>
21019 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21021 (mips_final_postscan_insn): Likewise.
21023 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
21025 PR rtl-optimization/84528
21026 * alias.c (init_alias_target): Add commentary.
21027 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21028 a unique base value if the frame pointer is not eliminated
21029 to the stack pointer.
21031 2018-03-01 Tom de Vries <tom@codesourcery.com>
21033 PR rtl-optimization/83327
21034 * lra-int.h (hard_regs_spilled_into): Declare.
21035 * lra.c (hard_regs_spilled_into): Define.
21036 (init_reg_info): Init hard_regs_spilled_into.
21037 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21038 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21039 (process_bb_lives): Handle hard_regs_spilled_into.
21040 (lra_create_live_ranges_1): Before doing liveness propagation, clear
21041 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21043 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
21045 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
21046 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
21047 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
21048 * config/rs6000/aix72.h: New file.
21050 2018-02-28 Jakub Jelinek <jakub@redhat.com>
21052 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
21053 instead of warning_at with conditional singular and plural messages
21057 * stor-layout.c (update_alignment_for_field): For
21058 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
21059 && !DECL_PACKED (field), do the alignment update, just use
21060 only desired_align instead of MAX (type_align, desired_align)
21062 (place_field): Don't do known_align < desired_align handling
21063 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
21064 is non-NULL, instead do it after rli->prev_field handling and
21065 only if not within a bitfield word. For DECL_PACKED (field)
21066 use type_align of BITS_PER_UNIT.
21068 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
21070 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
21071 superfluous parentheses and trailing spaces.
21073 2018-02-28 Richard Biener <rguenther@suse.de>
21075 PR tree-optimization/84584
21076 * graphite-scop-detection.c (scop_detection::add_scop): Discard
21077 SCoPs with fake exit edge.
21079 2018-02-28 Martin Liska <mliska@suse.cz>
21082 * timevar.c (timer::print): Fix format to properly print 100%
21085 2018-02-28 Richard Biener <rguenther@suse.de>
21087 PR middle-end/84607
21088 * genmatch.c (capture_info::walk_match): Do not mark
21089 captured expressions without operands as expr_p given
21090 they act more like predicates and should be subject to
21091 "lost tail" side-effect preserving.
21093 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
21095 PR rtl-optimization/81611
21096 * auto-inc-dec.c (attempt_change): Move dead note from
21097 mem_insn if it's the next use of regno
21098 (find_address): Take address use of reg holding
21099 non-incremented value. Add parm to limit search to the named
21101 (merge_in_block): Attempt to use a mem insn that is the next
21102 use of the original regno.
21104 2018-02-27 Martin Sebor <msebor@redhat.com>
21107 * doc/invoke.texi (-Wmissing-attributes): New option.
21108 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
21110 2018-02-27 Martin Sebor <msebor@redhat.com>
21112 PR translation/84207
21113 * diagnostic-core.h (warning_n, error_n, inform_n): Change
21114 n argument to unsigned HOST_WIDE_INT.
21115 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
21116 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
21117 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
21118 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
21120 2018-02-27 Richard Biener <rguenther@suse.de>
21122 PR tree-optimization/84512
21123 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21124 Do not use the estimate returned from record_stmt_cost for
21125 the scalar iteration cost but sum properly using add_stmt_cost.
21127 2018-02-27 Richard Biener <rguenther@suse.de>
21129 PR tree-optimization/84466
21130 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21131 Adjust last change to less strictly validate use operands.
21133 2018-02-27 Martin Liska <mliska@suse.cz>
21135 PR gcov-profile/84548
21136 * gcov.c (process_file): Allow partial overlap and consider it
21137 also as group functions.
21138 (output_lines): Properly calculate range of lines for a group.
21140 2018-02-27 Martin Liska <mliska@suse.cz>
21142 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
21143 'ggc' suffixes. Change first column width.
21144 (timer::print): Fix formatting of the column.
21146 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
21148 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
21149 preserve inline entry blocks for the sake of debug inline
21150 entry point markers alone.
21151 (remove_unused_locals): Suggest in comments a better place to
21152 force the preservation of inline entry blocks that are
21153 otherwise unused, but do not preserve them.
21155 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
21157 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
21159 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
21162 * config/i386/constraints.md (Bs): Replace
21163 ix86_indirect_branch_register with
21164 TARGET_INDIRECT_BRANCH_REGISTER.
21166 * config/i386/i386.md (indirect_jump): Likewise.
21167 (tablejump): Likewise.
21168 (*sibcall_memory): Likewise.
21169 (*sibcall_value_memory): Likewise.
21170 Peepholes of indirect call and jump via memory: Likewise.
21171 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
21172 (*sibcall_value_GOT_32): Likewise.
21173 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21174 (GOT_memory_operand): Likewise.
21175 (call_insn_operand): Likewise.
21176 (sibcall_insn_operand): Likewise.
21177 (GOT32_symbol_operand): Likewise.
21178 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
21180 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
21182 PR rtl-optimization/83496
21183 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
21184 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
21185 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
21186 redundant insn, if any.
21187 (relax_delay_slots): Likewise.
21188 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
21190 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
21192 PR tree-optimization/83965
21193 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
21194 that grouped statements are part of a reduction chain. Return
21195 true if the statement is not marked as a reduction itself but
21196 is part of a group.
21197 (vect_recog_dot_prod_pattern): Don't check whether the statement
21198 is part of a group here.
21199 (vect_recog_sad_pattern): Likewise.
21200 (vect_recog_widen_sum_pattern): Likewise.
21202 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
21205 * final.c (rest_of_clean_state): Also look for calls inside sequences.
21207 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
21210 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
21212 (ix86_output_indirect_function_return): New prototype.
21213 (ix86_split_simple_return_pop_internal): Likewise.
21214 * config/i386/i386.c (indirect_return_via_cx): New.
21215 (indirect_return_via_cx_bnd): Likewise.
21216 (indirect_thunk_name): Handle return va CX_REG.
21217 (output_indirect_thunk_function): Create alias for
21218 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
21219 (ix86_output_indirect_jmp): Remove the bool argument.
21220 (ix86_output_indirect_function_return): New function.
21221 (ix86_split_simple_return_pop_internal): Likewise.
21222 * config/i386/i386.md (*indirect_jump): Don't pass false
21223 to ix86_output_indirect_jmp.
21224 (*tablejump_1): Likewise.
21225 (simple_return_pop_internal): Change it to define_insn_and_split.
21226 Call ix86_split_simple_return_pop_internal to split it for
21227 -mfunction-return=.
21228 (simple_return_indirect_internal): Call
21229 ix86_output_indirect_function_return instead of
21230 ix86_output_indirect_jmp.
21232 2018-02-26 Jakub Jelinek <jakub@redhat.com>
21235 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
21236 memset and value initialization afterwards.
21238 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
21240 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
21242 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
21245 * common/config/aarch64/aarch64-common.c
21246 (aarch_option_optimization_table[]): Switch
21247 off fomit-frame-pointer
21249 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
21250 Chung-Ju Wu <jasonwucj@gmail.com>
21252 * config/nds32/nds32-multiple.md (load_multiple): Disallow
21254 (store_multiple): Ditto.
21256 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
21258 * config.gcc: Add --with-cpu support for nds32 target.
21259 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
21260 * config/nds32/nds32.opt: Add -mcpu= option.
21262 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
21264 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
21265 isel=yes): Warn for these deprecated options.
21267 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
21269 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
21270 ISA_2_5_MASKS_EMBEDDED.
21272 2018-02-23 Jakub Jelinek <jakub@redhat.com>
21274 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
21275 p->max as pointers rather than using iterative_hash_expr.
21277 2018-02-23 Carl Love <cel@us.ibm.com>
21279 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
21280 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
21282 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
21283 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
21284 P8V_BUILTIN_VEC_VUNSIGNED2.
21286 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
21289 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
21290 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
21292 * lra-constraints.c (curr_insn_transform): Set up
21293 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
21295 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
21296 * lra-eliminations.c (spill_pseudos): Ditto.
21297 (process_insn_for_elimination): Ditto.
21298 * lra-lives.c (reg_early_clobber_p): Use the new macros.
21299 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
21300 LRA_NON_CLOBBERED_ALT.
21302 2018-02-22 Martin Sebor <msebor@redhat.com>
21304 PR tree-optimization/84480
21305 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
21306 to maybe_diag_stxncpy_trunc. Call it.
21307 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
21308 from gimple_fold_builtin_strcpy. Print inlining stack.
21309 (handle_builtin_stxncpy): Print inlining stack.
21310 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
21312 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
21315 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
21316 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
21317 and -fcheck-pointer-bounds are used together.
21318 (indirect_thunk_prefix): New enum.
21319 (indirect_thunk_need_prefix): New function.
21320 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
21321 "_nt" instead of "_bnd" for NOTRACK prefix.
21322 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
21323 (output_indirect_thunk_function): Likewise.
21325 (ix86_code_end): Update output_indirect_thunk_function calls.
21326 (ix86_output_indirect_branch_via_reg): Replace
21327 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
21328 (ix86_output_indirect_branch_via_push): Likewise.
21329 (ix86_output_function_return): Likewise.
21330 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
21331 incompatible with -fcf-protection=branch and
21332 -fcheck-pointer-bounds.
21334 2018-02-22 Steve Ellcey <sellcey@cavium.com>
21337 * config/aarch64/aarch64.c (aarch64_print_address_internal):
21338 Change gcc_assert call to output_operand_lossage.
21340 2018-02-22 Steve Ellcey <sellcey@cavium.com>
21342 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
21344 2018-02-22 DJ Delorie <dj@redhat.com>
21345 Sebastian Perta <sebastian.perta@renesas.com>
21346 Oleg Endo <olegendo@gcc.gnu.org>
21348 * config/rx/rx.c (rx_rtx_costs): New function.
21349 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
21351 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
21353 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
21355 2018-02-22 Martin Liska <mliska@suse.cz>
21358 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
21359 Add "native" as a possible value.
21361 2018-02-22 Martin Liska <mliska@suse.cz>
21364 * config/i386/i386.c (ix86_option_override_internal):
21365 Add "native" as a possible value for -march and -mtune.
21367 2018-02-22 Jakub Jelinek <jakub@redhat.com>
21370 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
21371 to all type variants.
21373 PR tree-optimization/84503
21374 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
21375 width as info->bitpos + info->bitsize - start.
21376 (merged_store_group::merge_overlapping): Simplify width computation.
21377 (check_no_overlap): New function.
21378 (imm_store_chain_info::try_coalesce_bswap): Compute expected
21379 start + width and last_order of the group, fail if check_no_overlap
21381 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
21382 to group if check_no_overlap fails.
21384 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
21386 * config/rs6000/altivec.md: Delete contraint arguments to
21387 define_expand, define_split, and define_peephole2, and in
21388 define_insn_and_split if always unused.
21389 * config/rs6000/darwin.md: Ditto.
21390 * config/rs6000/dfp.md: Ditto.
21391 * config/rs6000/rs6000.md: Ditto.
21392 * config/rs6000/sync.md: Ditto.
21393 * config/rs6000/vector.md: Ditto.
21394 * config/rs6000/vsx.md: Ditto.
21396 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
21398 * config/rs6000/altivec.md: Write output control strings as braced
21399 blocks instead of double-quoted strings.
21400 * config/rs6000/darwin.md: Ditto.
21401 * config/rs6000/rs6000.md: Ditto.
21402 * config/rs6000/vector.md: Ditto.
21403 * config/rs6000/vsx.md: Ditto.
21405 2018-02-21 Jason Merrill <jason@redhat.com>
21407 PR c++/84314 - ICE with templates and fastcall attribute.
21408 * attribs.c (build_type_attribute_qual_variant): Remove assert.
21410 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
21412 * ipa-cp.c (determine_versionability): Fix comment typos.
21414 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
21417 * ipa-cp.c (determine_versionability): Do not version functions caling
21420 2018-02-21 Martin Liska <mliska@suse.cz>
21423 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
21424 Add "native" as a possible value.
21425 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
21426 the macro when native cpu detection is available.
21428 2018-02-21 Martin Liska <mliska@suse.cz>
21431 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
21432 Add "native" as a possible value.
21433 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
21434 when native cpu detection is available.
21436 2018-02-21 Jakub Jelinek <jakub@redhat.com>
21437 Martin Sebor <msebor@redhat.com>
21439 PR tree-optimization/84478
21440 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
21442 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
21443 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
21444 support which is conservatively correct, for 2 only stay conservative
21445 for maxlen. Formatting and comment capitalization fixes. Add STRICT
21446 argument to the 2 argument get_range_strlen, adjust 6 arg
21447 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
21449 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
21450 (gimple_fold_builtin_strlen): Pass true as last argument to
21453 2018-02-20 Martin Sebor <msebor@redhat.com>
21455 PR middle-end/84095
21456 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
21457 (builtin_memref::set_base_and_offset): Same. Handle inner references.
21458 (builtin_memref::builtin_memref): Factor out parts into
21459 set_base_and_offset and call it.
21461 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
21463 PR middle-end/84406
21464 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
21465 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
21466 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
21467 search at the associated MODE_INT.
21469 2018-02-20 Jeff Law <law@redhat.com>
21471 PR middle-end/82123
21472 PR tree-optimization/81592
21473 PR middle-end/79257
21474 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
21475 for range data rather than using global data.
21476 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
21477 range data rather than using global data.
21478 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
21479 pass it to children as needed.
21480 (struct directive::fmtresult): Similarly.
21481 (struct directive::set_width): Similarly.
21482 (struct directive::set_precision): Similarly.
21483 (format_integer, format_directive, parse_directive): Similarly.
21484 (format_none): Accept unnamed vr_values parameter.
21485 (format_percent, format_floating, format_character): Similarly.
21486 (format_string, format_plain): Similarly.
21487 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
21488 the EVRP range analyzer for range data rather than using global data.
21489 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
21490 gimple-ssa-evrp-analyze.h
21491 (class sprintf_dom_walker): Add after_dom_children member function.
21492 Add evrp_range_analyzer member.
21493 (sprintf_dom_walker::before_dom_children): Call into the EVRP
21494 range analyzer as needed.
21495 (sprintf_dom_walker::after_dom_children): New member function.
21496 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
21498 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
21499 (evrp_range_analyzer::pop_to_marker): Likewise.
21501 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
21503 PR tree-optimization/84419
21504 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
21505 with the required type if its current type is compatible but
21508 2018-02-20 Jakub Jelinek <jakub@redhat.com>
21510 PR middle-end/82004
21511 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
21512 after vectorization.
21514 2018-02-20 Martin Liska <mliska@suse.cz>
21517 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
21518 possible values if we don't have a hint.
21520 2018-02-20 Martin Liska <mliska@suse.cz>
21524 * final.c (shorten_branches): Build align_tab array with one
21526 * opts.c (finish_options): Add alignment option limit check.
21527 (MAX_CODE_ALIGN): Likewise.
21528 (MAX_CODE_ALIGN_VALUE): Likewise.
21529 * doc/invoke.texi: Document maximum allowed option value for
21530 all -falign-* options.
21532 2018-02-19 Jakub Jelinek <jakub@redhat.com>
21535 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
21536 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
21537 * var-tracking.c (emit_note_insn_var_location): Remove all references
21538 to NOTE_INSN_CALL_ARG_LOCATION.
21539 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
21540 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
21541 Use copy_rtx_if_shared.
21542 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
21543 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
21544 (dwarf2out_var_location): Remove handling of
21545 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
21547 * final.c (final_scan_insn): Remove all references to
21548 NOTE_INSN_CALL_ARG_LOCATION.
21549 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
21550 before dumping final insns.
21551 * except.c (emit_note_eh_region_end): Remove all references to
21552 NOTE_INSN_CALL_ARG_LOCATION.
21553 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
21554 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
21555 * config/arc/arc.c (hwloop_optimize): Likewise.
21556 * config/arm/arm.c (create_fix_barrier): Likewise.
21557 * config/s390/s390.c (s390_chunkify_start): Likewise.
21558 * config/sh/sh.c (find_barrier): Likewise.
21559 * config/i386/i386.c (rest_of_insert_endbranch,
21560 ix86_seh_fixup_eh_fallthru): Likewise.
21561 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
21562 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
21563 * config/frv/frv.c (frv_function_prologue): Likewise.
21564 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
21566 (note_outside_basic_block_p): Remove all references to
21567 NOTE_INSN_CALL_ARG_LOCATION.
21568 * gengtype.c (adjust_field_rtx_def): Likewise.
21569 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
21571 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
21572 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
21575 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
21578 PR tree-optimization/84452
21579 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
21580 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
21583 2018-02-19 Martin Liska <mliska@suse.cz>
21586 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
21588 2018-02-19 Martin Liska <mliska@suse.cz>
21589 Richard Sandiford <richard.sandiford@linaro.org>
21591 PR tree-optimization/82491
21592 * gimple-fold.c (get_base_constructor): Make earlier bail out
21595 2018-02-19 Carl Love <cel@us.ibm.com>
21597 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
21598 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
21600 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
21601 P8V_BUILTIN_VEC_NEG.
21603 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
21605 * config/rl78/rl78.md (movdf): New define expand.
21607 2018-02-19 Martin Liska <mliska@suse.cz>
21610 * doc/invoke.texi: Fix typo.
21611 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
21613 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
21615 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
21616 handle rs6000_single_float and rs6000_double_float specially for
21619 2018-02-16 Jeff Law <law@redhat.com>
21621 * config/rx/rx.c (add_pop_cfi_notes): New function.;
21622 (pop_regs): Use it.
21624 2018-02-16 Jakub Jelinek <jakub@redhat.com>
21627 * ipa-inline.c (inline_small_functions): Fix a typo.
21629 2018-02-16 Nathan Sidwell <nathan@acm.org>
21631 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
21633 2018-02-16 Carl Love <cel@us.ibm.com>
21635 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
21636 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
21637 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
21638 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
21639 expansion to P8V_BUILTIN_VEC_FLOAT2.
21641 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
21643 PR rtl-optimization/70023
21644 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
21645 src_regno into account.
21647 2018-02-16 Carl Love <cel@us.ibm.com>
21649 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
21650 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
21651 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
21652 * config/rs6000/rs6000.c: Remove case statements for
21653 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
21654 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
21655 and P9V_BUILTIN_VEC_VINSERT4B.
21656 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
21657 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
21658 * config/rs6000/vsx.md:
21659 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
21662 2018-02-16 Carl Love <cel@us.ibm.com>
21664 * config/rs6000/altivec.h: Add builtin names vec_extract4b
21666 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
21668 * config/rs6000/rs6000-c.c: Add the definitions for
21669 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
21670 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
21671 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
21672 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
21673 definition for insert4b and define insn *insert3b_internal.
21674 * doc/extend.texi: Add documentation for vec_extract4b.
21676 2018-02-16 Nathan Sidwell <nathan@acm.org>
21678 * doc/extend.texi (Backwards Compatibility): Mention friend
21679 injection. Note for-scope is deprecated.
21680 * doc/invoke.texi (-ffriend-injection): Deprecate.
21682 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
21684 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
21685 that moved to I2, also allow destinations that are a paradoxical
21686 subreg (instead of a normal reg).
21688 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
21691 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
21694 2018-02-16 Richard Biener <rguenther@suse.de>
21696 PR tree-optimization/84037
21697 PR tree-optimization/84016
21699 * config/i386/i386.c (ix86_builtin_vectorization_cost):
21700 Adjust vec_construct for the fact we need additional higher latency
21701 128bit inserts for AVX256 and AVX512 vector builds.
21702 (ix86_add_stmt_cost): Scale vector construction cost for
21705 2018-02-16 Richard Biener <rguenther@suse.de>
21707 PR tree-optimization/84417
21708 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
21709 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
21710 (non_rewritable_lvalue_p): Likewise, use poly-ints.
21712 2018-02-16 Martin Liska <mliska@suse.cz>
21715 * internal-fn.def (ASAN_CHECK): Set proper flags.
21716 (ASAN_MARK): Likewise.
21718 2018-02-16 Julia Koval <julia.koval@intel.com>
21720 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
21721 from PTA_CANNONLAKE.
21723 2018-02-16 Jakub Jelinek <jakub@redhat.com>
21726 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
21727 Use ++iter rather than iter++ for std::list iterators.
21728 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
21729 defer deleting them until all nodes in the forest are processed. Do
21730 free even leaf nodes. Change to_process into auto_vec.
21733 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
21734 * vec.h (vec_default_construct): Use memset instead of placement new
21735 if BROKEN_VALUE_INITIALIZATION is defined.
21736 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
21737 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
21740 PR rtl-optimization/83723
21741 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
21742 * lra.c (lra_substitute_pseudo): Likewise. If true, use
21743 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
21745 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21747 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21749 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
21751 PR rtl-optimization/81443
21752 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21753 from inner REGs to paradoxical SUBREGs.
21755 2018-02-16 Richard Biener <rguenther@suse.de>
21757 PR tree-optimization/84399
21758 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21759 For operands we can analyze at their definition make sure we can
21760 analyze them at each use as well.
21762 2018-02-16 Richard Biener <rguenther@suse.de>
21764 PR tree-optimization/84190
21765 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21766 volatile accesses if the decl isn't volatile.
21768 2018-02-15 Jason Merrill <jason@redhat.com>
21770 PR c++/84314 - ICE with templates and fastcall attribute.
21771 * attribs.c (build_type_attribute_qual_variant): Don't clobber
21772 TYPE_CANONICAL on an existing type.
21774 2018-02-15 Jakub Jelinek <jakub@redhat.com>
21776 PR tree-optimization/84383
21777 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21778 dstoff nor call operand_equal_p if dstbase is NULL.
21780 PR tree-optimization/84334
21781 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21782 also a CONSTANT_CLASS_P, punt.
21784 2018-02-14 Jim Wilson <jimw@sifive.com>
21786 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21787 first SMALL_OPERAND check. New local min_second_step. Move assert
21788 to where locals are set. Add TARGET_RVC support.
21789 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21791 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
21793 * doc/invoke.texi: Correct -Wformat-overflow code sample.
21795 2018-02-14 Martin Sebor <msebor@redhat.com>
21797 PR tree-optimization/83698
21798 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21799 arrays constrain the offset range to their bounds.
21800 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21801 (builtin_access::overlap): Avoid setting the size of overlap if it's
21803 (maybe_diag_overlap): Also consider arrays when deciding what values
21804 of offsets to include in diagnostics.
21806 2018-02-14 Martin Sebor <msebor@redhat.com>
21809 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21810 that correspond to the kind of a declaration.
21812 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
21815 * config/pa/pa.md: Load address of PIC label using the linkage table
21816 if the label is nonlocal.
21818 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
21820 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21821 warning message if user requests -maltivec=be.
21822 * doc/invoke.texi: Document deprecation of -maltivec=be.
21824 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
21827 * config/rs6000/rs6000-c.c: Update definitions for
21828 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21829 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21831 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21834 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
21835 add _get_ssp intrinsics. Remove argument from
21836 __builtin_ia32_rdssp[d|q].
21837 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
21838 * config/i386/i386-builtin.def: Remove argument from
21839 __builtin_ia32_rdssp[d|q].
21840 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
21841 ix86_expand_special_args_builtin for _rdssp[d|q].
21842 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
21843 Clear register before usage.
21844 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
21845 Add documentation for new _get_ssp and _inc_ssp intrinsics.
21847 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
21849 PR tree-optimization/84357
21850 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
21851 operand 1 of an ARRAY_REF too.
21853 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
21856 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
21857 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21859 (set_of_reg): New struct.
21860 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
21861 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
21862 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21864 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
21865 Split into bitclr, bitset, bitinvert patterns if appropriate.
21866 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21867 use rx_fuse_in_memory_bitop.
21868 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21869 to named insn, correct maximum insn length.
21871 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
21874 * machmode.def: Define a complex mode for PARTIAL_INT.
21875 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21877 * doc/rtl.texi: Document CSPImode.
21878 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21880 (msp430_hard_regno_nregs_with_padding): Likewise.
21882 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
21885 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21887 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
21889 PR rtl-optimization/84169
21890 * combine.c (try_combine): New variable split_i2i3. Set it to true if
21891 we generated a parallel as new i3 and we split that to new i2 and i3
21892 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
21893 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
21894 those to i2, not i1. Partially rewrite this scan code.
21896 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21899 * stor-layout.c (place_field): For variable length fields, adjust
21900 offset_align afterwards not just based on the field's alignment,
21901 but also on the size.
21903 PR middle-end/84309
21904 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21905 of exps and logs in the use_exp2 case.
21907 2018-02-13 Jeff Law <law@redhat.com>
21909 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21910 entry for "vector".
21912 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21915 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
21919 * common.opt (gas-loc-support, gas-locview-support): New.
21920 (ginline-points, ginternal-reset-location-views): New.
21921 * doc/invoke.texi: Document them. Use @itemx where intended.
21922 (gvariable-location-views): Adjust.
21923 * target.def (reset_location_view): New.
21924 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21925 (TARGET_RESET_LOCATION_VIEW): New.
21926 * doc/tm.texi: Rebuilt.
21927 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21928 (dwarf2out_default_as_locview_support): New.
21929 (output_asm_line_debug_info): Use option variables.
21930 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21931 (output_loc_list): Likewise.
21932 (add_high_low_attributes): Check option variables.
21933 Don't output entry view attribute in strict mode.
21934 (gen_inlined_subroutine_die): Check option variables.
21935 (dwarf2out_inline_entry): Likewise.
21936 (init_sections_and_labels): Likewise.
21937 (dwarf2out_early_finish): Likewise.
21938 (maybe_reset_location_view): New, from...
21939 (dwarf2out_var_location): ... here. Call it.
21940 * debug.h (dwarf2out_default_as_loc_support): Declare.
21941 (dwarf2out_default_as_locview_support): Declare.
21942 * hooks.c (hook_int_rtx_insn_0): New.
21943 * hooks.h (hook_int_rtx_insn_0): Declare.
21944 * toplev.c (process_options): Take -gas-loc-support and
21945 -gas-locview-support from dwarf2out. Enable
21946 -gvariable-location-views by default only with locview
21947 assembler support. Enable -ginternal-reset-location-views by
21948 default only if the target defines the corresponding hook.
21949 Enable -ginline-points by default if location views are
21950 enabled; force it disabled if statement frontiers are
21952 * tree-inline.c (expand_call_inline): Check option variables.
21953 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21955 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
21957 PR tree-optimization/84321
21958 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21959 handling. Also check whether the anti-range contains any values
21960 that satisfy the mask; switch to a VR_RANGE if not.
21962 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
21965 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21967 2018-02-13 Martin Jambor <mjambor@suse.cz>
21970 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21971 of call statements, also set location of a load to a temporary.
21973 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
21975 * config/rl78/rl78.c (add_vector_labels): New function.
21976 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21977 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21978 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21979 which checks that no arguments are passed.
21980 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21981 * doc/extend.texi: Documentation for the new attribute.
21983 2018-02-13 Andreas Schwab <schwab@suse.de>
21985 * config/riscv/linux.h (CPP_SPEC): Define.
21987 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21990 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21991 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21992 OPTION_MASK_ISA_AES as first argument to def_builtin_const
21993 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21994 instead of OPTION_MASK_ISA_PCLMUL as first argument to
21995 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21996 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21997 temporarily for AES and PCLMUL builtins.
21999 PR tree-optimization/84339
22000 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22001 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22004 PR middle-end/84309
22005 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22006 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22007 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22009 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22011 * omp-simd-clone.h: New file.
22012 * omp-simd-clone.c: Include omp-simd-clone.h.
22013 (expand_simd_clones): No longer static.
22014 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22015 cgraph.h and omp-simd-clone.h.
22016 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22017 (vect_recog_widen_shift_pattern): Formatting fix.
22018 (vect_pattern_recog_1): Don't check optab for calls.
22021 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22022 operands[2] into a REG before using gen_lowpart on it.
22024 2018-02-12 Jeff Law <law@redhat.com>
22027 * config/sh/sh.c (find_barrier): Consider a sibling call
22030 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22031 successfully back substituting a reg.
22033 2018-02-12 Richard Biener <rguenther@suse.de>
22035 PR tree-optimization/84037
22036 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22037 parameter, move visited init to caller.
22038 (vect_slp_analyze_operations): Separate cost from validity
22039 check, initialize visited once for all instances.
22040 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22042 * tree-vect-stmts.c (vect_model_simple_cost): Make early
22044 (vect_model_promotion_demotion_cost): Likewise.
22045 (vectorizable_bswap): Guard cost modeling with !slp_node
22046 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
22048 (vectorizable_call): Likewise.
22049 (vectorizable_conversion): Likewise.
22050 (vectorizable_assignment): Likewise.
22051 (vectorizable_shift): Likewise.
22052 (vectorizable_operation): Likewise.
22053 (vectorizable_store): Likewise.
22054 (vectorizable_load): Likewise.
22055 (vectorizable_condition): Likewise.
22056 (vectorizable_comparison): Likewise.
22058 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
22061 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
22062 (ASAN_MARK): Fix fnspec to account for return value, change pointer
22063 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
22065 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
22067 PR middle-end/83665
22068 * params.def (inline-min-speedup): Increase from 8 to 15.
22069 (max-inline-insns-auto): Decrease from 40 to 30.
22070 * ipa-split.c (consider_split): Add some buffer for function to
22071 be considered inlining candidate.
22072 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
22075 2018-02-12 Richard Biener <rguenther@suse.de>
22077 PR tree-optimization/84037
22078 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
22079 matched stmts if we cannot swap the non-matched ones.
22081 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
22083 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
22084 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
22085 _mm_maskz_scalef_round_ss): New intrinsics.
22086 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
22087 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
22088 __builtin_ia32_scalefss_round): Remove.
22089 (__builtin_ia32_scalefsd_mask_round,
22090 __builtin_ia32_scalefss_mask_round): New intrinsics.
22091 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
22092 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
22093 ((match_operand:VF_128 2 "<round_nimm_predicate>"
22094 "<round_constraint>")): Changed to ...
22095 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
22096 "<round_scalar_constraint>")): ... this.
22097 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
22098 %0, %1, %2<round_op3>}"): Changed to ...
22099 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
22100 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
22101 %2<round_scalar_mask_op3>}"): ... this.
22102 * config/i386/subst.md (round_scalar_nimm_predicate): New.
22104 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
22106 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
22107 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
22108 (_mm_maskz_sqrt_round_ss): New intrinsics.
22109 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
22110 (__builtin_ia32_sqrtsd_mask_round)
22111 (__builtin_ia32_sqrtss_mask_round): New builtins.
22112 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
22113 (__builtin_ia32_sqrtss_round): Remove.
22114 (__builtin_ia32_sqrtsd_mask_round)
22115 (__builtin_ia32_sqrtss_mask_round): New builtins.
22116 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
22117 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
22118 ((match_operand:VF_128 1 "vector_operand"
22119 "xBm,<round_constraint>")): Changed to ...
22120 ((match_operand:VF_128 1 "vector_operand"
22121 "xBm,<round_scalar_constraint>")): ... this.
22122 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
22123 %0, %2, %<iptr>1<round_op3>}): Changed to ...
22124 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
22125 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
22126 %<iptr>1<round_scalar_mask_op3>}): ... this.
22127 ((set_attr "prefix" "<round_prefix>")): Changed to ...
22128 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
22130 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
22133 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
22134 Cast vec_cmpeq result to correct type.
22135 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
22136 Cast vec_cmpgt result to correct type.
22138 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
22140 * final.c (final_scan_insn_1): Renamed from...
22141 (final_scan_insn): ... this. New wrapper, to recover
22142 seen from the outermost call in recursive ones.
22143 * config/sparc/sparc.c (output_return): Drop seen from call.
22144 (output_sibcall): Likewise.
22145 * config/visium/visium.c (output_branch): Likewise.
22147 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
22149 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
22152 2018-02-10 Alan Modra <amodra@gmail.com>
22155 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
22156 Specify LR as an input.
22158 2018-02-10 Jakub Jelinek <jakub@redhat.com>
22161 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
22162 remove_member_access_dummy_vars): New functions.
22163 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
22164 lower_omp_1, execute_lower_omp): Use them.
22166 PR rtl-optimization/84308
22167 * shrink-wrap.c (spread_components): Release todo vector.
22169 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
22171 PR rtl-optimization/57193
22172 * ira-color.c (struct allocno_color_data): Add member
22173 conflict_allocno_hard_prefs.
22174 (update_conflict_allocno_hard_prefs): New.
22175 (bucket_allocno_compare_func): Add a preference based on
22176 conflict_allocno_hard_prefs.
22177 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
22178 (color_allocnos): Remove a dead code. Initiate
22179 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
22181 2018-02-09 Jakub Jelinek <jakub@redhat.com>
22184 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
22185 constraint from =wa to wa. Avoid a subreg on the output operand,
22186 instead use a pseudo and subreg it in a move.
22187 (p9_xxbrd_<mode>): Changed to ...
22188 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
22189 (p9_xxbrd_v2df): New expander.
22190 (p9_xxbrw_<mode>): Changed to ...
22191 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
22192 (p9_xxbrw_v4sf): New expander.
22194 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
22196 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
22198 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
22201 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
22202 multiply in 32-bit mode.
22203 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
22204 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
22207 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
22209 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
22210 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
22211 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
22212 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
22214 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
22217 * dwarf2out.c (is_trivial_indirect_ref): New function.
22218 (dwarf2out_late_global_decl): Do not generate a location
22219 attribute for variables that have a non-trivial DECL_VALUE_EXPR
22220 and that are not defined in the current unit.
22222 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
22224 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
22225 instead of a libcall for UNORDERED.
22227 2018-02-09 Tamar Christina <tamar.christina@arm.com>
22230 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
22231 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
22233 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22236 * config/s390/s390.c (s390_set_current_function): Invoke
22237 s390_indirect_branch_settings also if fndecl didn't change.
22239 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
22241 * config/rs6000/rs6000.md (blockage): Set length to zero.
22243 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
22245 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
22247 2018-02-09 Jakub Jelinek <jakub@redhat.com>
22250 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
22251 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
22255 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
22256 PARALLEL incoming that failed vt_get_decl_and_offset check.
22258 PR middle-end/84237
22259 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
22260 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
22262 (get_variable_section): For decls in named .bss* sections pass true as
22263 second argument to bss_initializer_p.
22265 2018-02-09 Marek Polacek <polacek@redhat.com>
22266 Jakub Jelinek <jakub@redhat.com>
22269 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
22270 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
22271 Sync some changes from cxx_fold_indirect_ref.
22273 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
22275 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
22277 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
22278 (BLOCK_INLINE_ENTRY_LABEL): New.
22279 (dwarf2out_var_location): Disregard inline entry markers.
22280 (inline_entry_data): New struct.
22281 (inline_entry_data_hasher): New hashtable type.
22282 (inline_entry_data_hasher::hash): New.
22283 (inline_entry_data_hasher::equal): New.
22284 (inline_entry_data_table): New variable.
22285 (add_high_low_attributes): Add DW_AT_entry_pc and
22286 DW_AT_GNU_entry_view attributes if a pending entry is found
22287 in inline_entry_data_table. Add old entry_pc attribute only
22288 if debug nonbinding markers are disabled.
22289 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
22290 markers are enabled.
22291 (block_within_block_p, dwarf2out_inline_entry): New.
22292 (dwarf2out_finish): Check that no entries remained in
22293 inline_entry_data_table.
22294 * final.c (reemit_insn_block_notes): Handle inline entry notes.
22295 (final_scan_insn, notice_source_line): Likewise.
22296 (rest_of_clean_state): Skip inline entry markers.
22297 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
22299 * gimple.c (gimple_build_debug_inline_entry): New.
22300 * gimple.h (enum gimple_debug_subcode): Add
22301 GIMPLE_DEBUG_INLINE_ENTRY.
22302 (gimple_build_debug_inline_entry): Declare.
22303 (gimple_debug_inline_entry_p): New.
22304 (gimple_debug_nonbind_marker_p): Adjust.
22305 * insn-notes.def (INLINE_ENTRY): New.
22306 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
22307 inline entry marker notes.
22308 (print_insn): Likewise.
22309 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
22310 (INSN_DEBUG_MARKER_KIND): Likewise.
22311 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
22312 * tree-inline.c (expand_call_inline): Build and insert
22313 debug_inline_entry stmt.
22314 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
22315 inline entry blocks early, if nonbind markers are enabled.
22316 (dump_scope_block): Dump fragment info.
22317 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
22318 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
22319 (gimple_build_debug_inline_entry): New.
22320 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
22321 Enable/disable inline entry points too.
22322 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
22323 (DEBUG_INSN): Describe inline entry markers.
22325 * common.opt (gvariable-location-views): New.
22326 (gvariable-location-views=incompat5): New.
22327 * config.in: Rebuilt.
22328 * configure: Rebuilt.
22329 * configure.ac: Test assembler for view support.
22330 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
22331 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
22332 * dwarf2out.c (var_loc_view): New typedef.
22333 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
22334 (dwarf2out_locviews_in_attribute): New.
22335 (dwarf2out_locviews_in_loclist): New.
22336 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
22337 (enum dw_line_info_opcode): Add LI_adv_address.
22338 (struct dw_line_info_table): Add view.
22339 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
22340 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
22341 (zero_view_p): New variable.
22342 (ZERO_VIEW_P): New macro.
22343 (output_asm_line_debug_info): New.
22344 (struct var_loc_node): Add view.
22345 (add_AT_view_list, AT_loc_list): New.
22346 (add_var_loc_to_decl): Add view param. Test it against last.
22347 (new_loc_list): Add view params. Record them.
22348 (AT_loc_list_ptr): Handle loc and view lists.
22349 (view_list_to_loc_list_val_node): New.
22350 (print_dw_val): Handle dw_val_class_view_list.
22351 (size_of_die): Likewise.
22352 (value_format): Likewise.
22353 (loc_list_has_views): New.
22354 (gen_llsym): Set vl_symbol too.
22355 (maybe_gen_llsym, skip_loc_list_entry): New.
22356 (dwarf2out_maybe_output_loclist_view_pair): New.
22357 (output_loc_list): Output view list or entries too.
22358 (output_view_list_offset): New.
22359 (output_die): Handle dw_val_class_view_list.
22360 (output_dwarf_version): New.
22361 (output_compilation_unit_header): Use it.
22362 (output_skeleton_debug_sections): Likewise.
22363 (output_rnglists, output_line_info): Likewise.
22364 (output_pubnames, output_aranges): Update version comments.
22365 (output_one_line_info_table): Output view numbers in asm comments.
22366 (dw_loc_list): Determine current endview, pass it to new_loc_list.
22367 Call maybe_gen_llsym.
22368 (loc_list_from_tree_1): Adjust.
22369 (add_AT_location_description): Create view list attribute if
22370 needed, check it's absent otherwise.
22371 (convert_cfa_to_fb_loc_list): Adjust.
22372 (maybe_emit_file): Call output_asm_line_debug_info for test.
22373 (dwarf2out_var_location): Reset views as needed. Precompute
22374 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
22375 attribute. Set view.
22376 (new_line_info_table): Reset next view.
22377 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
22378 (dwarf2out_source_line): Likewise. Output view resets and labels to
22379 the assembler, or select appropriate line info opcodes.
22380 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
22381 (optimize_string_length): Catch it. Adjust.
22382 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
22383 dw_val_class_view_list, and remove it if no longer needed.
22384 (hash_loc_list): Hash view numbers.
22385 (loc_list_hasher::equal): Compare them.
22386 (optimize_location_lists): Check whether a view list symbol is
22387 needed, and whether the locview attribute is present, and
22388 whether they match. Remove the locview attribute if no longer
22390 (index_location_lists): Call skip_loc_list_entry for test.
22391 (dwarf2out_finish): Call output_asm_line_debug_info for test.
22392 Use output_dwarf_version.
22393 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
22394 (struct dw_val_node): Add val_view_list.
22395 * final.c (SEEN_NEXT_VIEW): New.
22396 (set_next_view_needed): New.
22397 (clear_next_view_needed): New.
22398 (maybe_output_next_view): New.
22399 (final_start_function): Rename to...
22400 (final_start_function_1): ... this. Take pointer to FIRST,
22401 add SEEN parameter. Emit param bindings in the initial view.
22402 (final_start_function): Reintroduce SEEN-less interface.
22403 (final): Rename to...
22404 (final_1): ... this. Take SEEN parameter. Output final pending
22405 next view at the end.
22406 (final): Reintroduce seen-less interface.
22407 (final_scan_insn): Output pending next view before switching
22408 sections or ending a block. Mark the next view as needed when
22409 outputting variable locations. Notify debug backend of section
22410 changes, and of location view changes.
22411 (rest_of_handle_final): Adjust.
22412 * toplev.c (process_options): Autodetect value for debug variable
22413 location views option. Warn on incompat5 without -gdwarf-5.
22414 * doc/invoke.texi (gvariable-location-views): New.
22415 (gvariable-location-views=incompat5): New.
22416 (gno-variable-location-views): New.
22418 2018-02-08 David Malcolm <dmalcolm@redhat.com>
22420 PR tree-optimization/84136
22421 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
22422 that the result of find_edge is non-NULL.
22424 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
22427 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
22428 storing integer register in SImode. Fix cost of 256 and 512
22429 byte aligned SSE register store.
22431 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
22433 * config/i386/i386.c (ix86_multiplication_cost): Fix
22434 multiplication cost for TARGET_AVX512DQ.
22436 2018-02-08 Marek Polacek <polacek@redhat.com>
22438 PR tree-optimization/84238
22439 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
22442 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
22444 PR tree-optimization/84265
22445 * tree-vect-stmts.c (vectorizable_store): Don't treat
22446 VMAT_CONTIGUOUS accesses as grouped.
22447 (vectorizable_load): Likewise.
22449 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
22451 PR tree-optimization/81635
22452 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
22453 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
22454 (test_round_for_mask): New functions.
22455 (wide_int_cc_tests): Call test_round_for_mask.
22456 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
22457 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
22458 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
22459 range returned by get_range_info.
22461 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
22464 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
22465 * symtab.c: Include builtins.h
22466 (symtab_node::output_to_lto_symbol_table_p): Move here
22467 from lto-streamer-out.c:output_symbol_p.
22468 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
22469 (output_symbol_p): Move all logic to symtab.c
22470 (produce_symtab): Update.
22472 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22474 * config/s390/s390-opts.h (enum indirect_branch): Define.
22475 * config/s390/s390-protos.h (s390_return_addr_from_memory)
22476 (s390_indirect_branch_via_thunk)
22477 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
22478 (enum s390_indirect_branch_type): Define.
22479 * config/s390/s390.c (struct s390_frame_layout, struct
22480 machine_function): Remove.
22481 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
22482 (indirect_branch_table_label_no, indirect_branch_table_name):
22484 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
22485 (enum s390_indirect_branch_option): Define.
22486 (s390_return_addr_from_memory): New function.
22487 (s390_handle_string_attribute): New function.
22488 (s390_attribute_table): Add new attribute handler.
22489 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
22490 (s390_indirect_branch_via_thunk): New function.
22491 (s390_indirect_branch_via_inline_thunk): New function.
22492 (s390_function_ok_for_sibcall): When jumping via thunk disallow
22493 sibling call optimization for non z10 compiles.
22494 (s390_emit_call): Force indirect branch target to be a single
22495 register. Add r1 clobber for non-z10 compiles.
22496 (s390_emit_epilogue): Emit return jump via return_use expander.
22497 (s390_reorg): Handle JUMP_INSNs as execute targets.
22498 (s390_option_override_internal): Perform validity checks for the
22499 new command line options.
22500 (s390_indirect_branch_attrvalue): New function.
22501 (s390_indirect_branch_settings): New function.
22502 (s390_set_current_function): Invoke s390_indirect_branch_settings.
22503 (s390_output_indirect_thunk_function): New function.
22504 (s390_code_end): Implement target hook.
22505 (s390_case_values_threshold): Implement target hook.
22506 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
22508 * config/s390/s390.h (struct s390_frame_layout)
22509 (struct machine_function): Move here from s390.c.
22510 (TARGET_INDIRECT_BRANCH_NOBP_RET)
22511 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
22512 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
22513 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
22514 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
22515 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
22516 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
22517 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
22518 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
22519 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
22520 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
22521 (mnemonic attribute): Add values which aren't recognized
22523 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
22524 pattern for branch conversion. Fix mnemonic attribute.
22525 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
22526 indirect branch via thunk if requested.
22527 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
22528 ("*indirect_jump"): Disable for branch conversion using out of
22530 ("indirect_jump_via_thunk<mode>_z10")
22531 ("indirect_jump_via_thunk<mode>")
22532 ("indirect_jump_via_inlinethunk<mode>_z10")
22533 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
22534 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
22535 ("casesi_jump_via_inlinethunk<mode>_z10")
22536 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
22537 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
22538 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
22539 ("*indirect2_jump"): Disable for branch conversion.
22540 ("casesi_jump"): Turn into expander and expand patterns for branch
22542 ("return_use"): New expander.
22543 ("*return"): Emit return via thunk and rename it to ...
22544 ("*return<mode>"): ... this one.
22545 * config/s390/s390.opt: Add new options and and enum for the
22548 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
22550 * lra-constraints.c (match_reload): Unconditionally use
22551 gen_lowpart_SUBREG, rather than selecting between that
22552 and equivalent gen_rtx_SUBREG code.
22554 2018-02-08 Richard Biener <rguenther@suse.de>
22556 PR tree-optimization/84233
22557 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
22558 changed flag instead of boguously re-using phi_inserted.
22560 2018-02-08 Martin Jambor <mjambor@suse.cz>
22562 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
22563 static local variables.
22565 2018-02-08 Richard Biener <rguenther@suse.de>
22567 PR tree-optimization/84278
22568 * tree-vect-stmts.c (vectorizable_store): When looking for
22569 smaller vector types to perform grouped strided loads/stores
22570 make sure the mode is supported by the target.
22571 (vectorizable_load): Likewise.
22573 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
22575 * config/aarch64/aarch64.c (aarch64_components_for_bb):
22576 Increase LDP/STP opportunities by adding adjacent callee-saves.
22578 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
22580 PR rtl-optimization/84068
22581 PR rtl-optimization/83459
22582 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
22584 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
22586 PR tree-optimization/84224
22587 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
22588 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
22589 non-zero arguments.
22591 2018-02-07 Iain Sandoe <iain@codesourcery.com>
22594 * config/rs6000/altivec.md (*restore_world): Remove LR use.
22595 * config/rs6000/predicates.md (restore_world_operation): Adjust op
22596 count, remove one USE.
22598 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
22600 * doc/install.texi (Configuration): Document the
22601 --with-long-double-format={ibm,ieee} PowerPC configuration
22605 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
22606 Convert from define_expand to be define_insn_and_split. Rework
22607 float/double/_Float128 conversions to QI/HI/SImode to work with
22608 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
22609 conversions to QI/HImode types did a store and then a load to
22610 truncate the value. For conversions to VSX registers, don't split
22611 the insn, instead emit the code directly. Use the code iterator
22612 any_fix to combine signed and unsigned conversions.
22613 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
22614 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
22615 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22616 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
22617 (fix_<mode>di2_hw): Likewise.
22618 (fixuns_<mode>di2_hw): Likewise.
22619 (fix_<mode>si2_hw): Likewise.
22620 (fixuns_<mode>si2_hw): Likewise.
22621 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
22622 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
22623 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
22624 fix<uns>_trunc<SFDF:mode>si2_p8.
22625 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
22626 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
22627 (fix<uns>_<mode>_mem): Likewise.
22628 (fctiw<u>z_<mode>_mem): Likewise.
22629 (fix<uns>_<mode>_mem): Likewise.
22630 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
22631 the register allocator from doing a direct move to the GPRs to do
22632 a store, and instead use the ISA 3.0 store byte/half-word from
22633 vector register instruction. For IEEE 128-bit floating point,
22634 also optimize stores of 32-bit ints.
22635 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
22637 2018-02-07 Alan Hayward <alan.hayward@arm.com>
22639 * genextract.c (push_pathstr_operand): New function to support
22641 (walk_rtx): Call push_pathstr_operand.
22642 (print_path): Support [a-zA-Z].
22644 2018-02-07 Richard Biener <rguenther@suse.de>
22646 PR tree-optimization/84037
22647 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
22648 (cse_and_gimplify_to_preheader): Declare.
22649 (vect_get_place_in_interleaving_chain): Likewise.
22650 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22652 (_loop_vec_info::~_loop_vec_info): Delete it.
22653 (cse_and_gimplify_to_preheader): New function.
22654 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
22655 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
22656 (vectorizable_load): Likewise. For grouped stores always base
22657 the IV on the first element.
22658 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
22659 condition before gimplifying.
22661 2018-02-07 Jakub Jelinek <jakub@redhat.com>
22663 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
22664 *DIV_EXPR and *MOD_EXPR.
22666 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
22669 * config/i386/i386.c (ix86_option_override_internal): Mask out
22670 the CF_SET bit when checking -fcf-protection.
22672 2018-02-07 Tom de Vries <tom@codesourcery.com>
22675 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
22678 2018-02-07 Richard Biener <rguenther@suse.de>
22680 PR tree-optimization/84204
22681 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
22684 PR tree-optimization/84205
22685 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
22686 special-case isl_ast_op_zdiv_r.
22688 PR tree-optimization/84223
22689 * graphite-scop-detection.c (gather_bbs::before_dom_children):
22690 Only add conditions from within the region.
22691 (gather_bbs::after_dom_children): Adjust.
22693 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
22696 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
22697 * config/avr/avr.md: Only post-reload split REG-REG moves if
22698 either register is GENERAL_REG_P.
22700 2018-02-07 Jakub Jelinek <jakub@redhat.com>
22702 PR tree-optimization/84235
22703 * tree-ssa-scopedtables.c
22704 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
22705 if the subtraction is performed in floating point type where NaNs are
22706 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
22707 build 1. Formatting fix.
22709 2018-02-06 Jakub Jelinek <jakub@redhat.com>
22712 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
22713 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
22714 and skip it regardless of bb boundaries. Use CALL_P macro,
22715 don't test INSN_P (insn) together with CALL_P or JUMP_P check
22716 unnecessarily, formatting fix.
22718 2018-02-06 Michael Collison <michael.collison@arm.com>
22720 * config/arm/thumb2.md:
22721 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
22722 (*thumb_mov_notscc): Ditto.
22724 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
22727 * config/rs6000/rs6000.md (su code attribute): Use "u" for
22728 unsigned_fix, not "s".
22730 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22732 * configure.ac (gcc_fn_eh_frame_ro): New function.
22733 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
22734 correct .eh_frame permissions.
22735 * configure: Regenerate.
22737 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
22739 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
22740 irrelevant options.
22742 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22744 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22745 Display warning message for -mno-speculate-indirect-jumps.
22747 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
22749 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22751 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22753 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
22755 PR tree-optimization/84225
22756 * tree-eh.c (find_trapping_overflow): Only call
22757 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22759 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22762 * config/i386/i386.c: Reimplement the check of possible options
22763 -mibt/-mshstk conbination. Change error messages.
22764 * doc/invoke.texi: Fix a typo: remove extra '='.
22766 2018-02-06 Marek Polacek <polacek@redhat.com>
22768 PR tree-optimization/84228
22769 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22771 2018-02-06 Tamar Christina <tamar.christina@arm.com>
22774 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22775 emitted arch directives.
22776 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22777 __ARM_FEATURE_COPROC before changing architectures.
22779 2018-02-06 Richard Biener <rguenther@suse.de>
22781 * config/i386/i386.c (print_reg): Fix typo.
22782 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22784 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
22786 * configure: Regenerate.
22788 2018-02-05 Martin Sebor <msebor@redhat.com>
22790 PR tree-optimization/83369
22791 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22794 2018-02-05 Martin Liska <mliska@suse.cz>
22796 * doc/invoke.texi: Cherry-pick upstream r323995.
22798 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
22800 * ira.c (ira_init_register_move_cost): Adjust comment.
22802 2018-02-05 Martin Liska <mliska@suse.cz>
22804 PR gcov-profile/84137
22805 * doc/gcov.texi: Fix typo in documentation.
22807 2018-02-05 Martin Liska <mliska@suse.cz>
22809 PR gcov-profile/83879
22810 * doc/gcov.texi: Document necessity of --dynamic-list-data when
22811 using dlopen functionality.
22813 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
22815 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22816 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22817 _mm_maskz_range_ss, _mm_mask_range_round_ss,
22818 _mm_maskz_range_round_ss): New intrinsics.
22819 (__builtin_ia32_rangesd128_round)
22820 (__builtin_ia32_rangess128_round): Remove.
22821 (__builtin_ia32_rangesd128_mask_round,
22822 __builtin_ia32_rangess128_mask_round): New builtins.
22823 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22824 __builtin_ia32_rangess128_round): Remove.
22825 (__builtin_ia32_rangesd128_mask_round,
22826 __builtin_ia32_rangess128_mask_round): New builtins.
22827 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22828 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22829 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22830 "<round_saeonly_constraint>")): Changed to ...
22831 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22832 "<round_saeonly_scalar_constraint>")): ... this.
22833 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
22834 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
22835 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
22836 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
22837 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
22839 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
22841 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
22843 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
22844 Remove all values except native, 8540 and 8548.
22846 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
22848 * config/i386/i386.c (ix86_output_function_return): Pass
22849 INVALID_REGNUM, instead of -1, as invalid register number to
22850 indirect_thunk_name and output_indirect_thunk.
22852 2018-02-02 Julia Koval <julia.koval@intel.com>
22854 * config.gcc: Add -march=icelake.
22855 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
22856 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
22857 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
22858 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
22859 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
22860 (processor_target_table): Add icelake.
22861 (ix86_option_override_internal): Handle new PTAs.
22862 (get_builtin_code_for_version): Handle icelake.
22863 (M_INTEL_COREI7_ICELAKE): New.
22864 (fold_builtin_cpu): Handle icelake.
22865 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
22866 * doc/invoke.texi: Add -march=icelake.
22868 2018-02-02 Julia Koval <julia.koval@intel.com>
22870 * config/i386/i386.c (ix86_option_override_internal): Change flags type
22871 to wide_int_bitmask.
22872 * wide-int-bitmask.h: New.
22874 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22877 * config/i386/i386.md: Replace Pmode with word_mode in
22878 builtin_setjmp_setup and builtin_longjmp to support x32.
22880 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
22884 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22886 (rs6000_supported_cpu_names): New static variable.
22887 (linux_cpu_translation_table): Likewise.
22888 (elf_platform) <cpu>: Define new static variable and use it.
22889 Translate kernel AT_PLATFORM name to canonical name if needed.
22890 Error if platform name is unknown.
22892 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
22895 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22897 2018-02-01 Jeff Law <law@redhat.com>
22900 * config/i386/i386.c (release_scratch_register_on_entry): Add new
22901 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
22902 the scratch if RELEASE_VIA_POP is false.
22903 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22904 If we have to save a temporary register, decrement SIZE appropriately.
22905 Pass new arguments to release_scratch_register_on_entry.
22906 (ix86_adjust_stack_and_probe): Likewise.
22907 (ix86_emit_probe_stack_range): Pass new arguments to
22908 release_scratch_register_on_entry.
22910 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
22912 PR rtl-optimization/84157
22913 * combine.c (change_zero_ext): Use REG_P predicate in
22914 front of HARD_REGISTER_P predicate.
22916 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
22918 * config/avr/avr.c (avr_option_override): Move disabling of
22919 -fdelete-null-pointer-checks to...
22920 * common/config/avr/avr-common.c (avr_option_optimization_table):
22923 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22925 PR tree-optimization/81635
22926 * tree-data-ref.c (split_constant_offset_1): For types that
22927 wrap on overflow, try to use range info to prove that wrapping
22930 2018-02-01 Renlin Li <renlin.li@arm.com>
22933 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22934 TAILCALL_ADDR_REGS.
22935 (aarch64_register_move_cost): Likewise.
22936 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22937 TAILCALL_ADDR_REGS.
22938 (REG_CLASS_NAMES): Likewise.
22939 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22940 TAILCALL_ADDR_REGS. Remove IP registers.
22941 * config/aarch64/aarch64.md (Ucs): Update register constraint.
22943 2018-02-01 Richard Biener <rguenther@suse.de>
22945 * domwalk.h (dom_walker::dom_walker): Add additional constructor
22946 for specifying RPO order and allow NULL for that.
22947 * domwalk.c (dom_walker::dom_walker): Likewise.
22948 (dom_walker::walk): Handle NULL RPO order.
22949 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22951 (rewrite_update_dom_walker): Likewise.
22952 (mark_def_dom_walker): Likewise.
22954 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22956 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22957 (aarch64_maybe_expand_sve_subreg_move): Declare.
22958 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22959 * config/aarch64/predicates.md (aarch64_any_register_operand): New
22961 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22962 that are semantically a reverse operation.
22963 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22964 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22965 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22967 (aarch64_can_change_mode_class): For big-endian, forbid changes
22968 between two SVE modes if they have different element sizes.
22970 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22972 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22973 the TImode handling for big-endian targets.
22975 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22977 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22978 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
22980 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22981 Remove BSWAP handing for big-endian targets and use the form of
22982 LD1RQ appropariate for the mode.
22984 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22986 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22987 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22988 duplicated element.
22990 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22993 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22994 check for operands that need to go through aarch64_sve_reload_be.
22996 2018-02-01 Jakub Jelinek <jakub@redhat.com>
22998 PR tree-optimization/81661
22999 PR tree-optimization/84117
23000 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23001 * tree-eh.c: Include gimplify.h.
23002 (find_trapping_overflow, replace_trapping_overflow,
23003 rewrite_to_non_trapping_overflow): New functions.
23004 * tree-vect-loop.c: Include tree-eh.h.
23005 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23006 * tree-data-ref.c: Include tree-eh.h.
23007 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23009 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
23011 PR rtl-optimization/84123
23012 * combine.c (change_zero_ext): Check if hard register satisfies
23013 can_change_dest_mode before calling gen_lowpart_SUBREG.
23015 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
23018 * ira.c (ira_init_register_move_cost): Remove assert.
23020 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
23022 PR rtl-optimization/84071
23023 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23024 * doc/tm.texi: Regenerate.
23026 2018-01-31 Richard Biener <rguenther@suse.de>
23028 PR tree-optimization/84132
23029 * tree-data-ref.c (analyze_miv_subscript): Properly
23030 check whether evolution_function_is_affine_multivariate_p
23031 before calling gcd_of_steps_may_divide_p.
23033 2018-01-31 Julia Koval <julia.koval@intel.com>
23036 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23037 * config/i386/i386.md (rdpid_rex64) New.
23038 (rdpid): Make 32bit only.
23040 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
23043 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23044 an IDENTIFIER_NODE for FUNCTION_TYPE's.
23046 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
23049 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23051 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23053 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
23055 PR rtl-optimization/84071
23056 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
23057 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
23059 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
23061 * config/arc/arc.c (arc_handle_aux_attribute): New function.
23062 (arc_attribute_table): Add 'aux' attribute.
23063 (arc_in_small_data_p): Consider aux like variables.
23064 (arc_is_aux_reg_p): New function.
23065 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
23066 (arc_get_aux_arg): New function.
23067 (prepare_move_operands): Handle aux-register access.
23068 (arc_handle_aux_attribute): New function.
23069 * doc/extend.texi (ARC Variable attributes): Add subsection.
23071 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
23073 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
23074 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
23075 (arc_attribute_table): Add 'uncached' attribute.
23076 (arc_print_operand): Print '.di' flag for uncached memory
23078 (arc_in_small_data_p): Do not consider for small data the uncached
23080 (arc_is_uncached_mem_p): New function.
23081 * config/arc/predicates.md (compact_store_memory_operand): Check
23082 for uncached memory accesses.
23083 (nonvol_nonimm_operand): Likewise.
23084 * doc/extend.texi (ARC Type Attribute): New subsection.
23086 2018-01-31 Jakub Jelinek <jakub@redhat.com>
23089 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
23090 falign-loops=): Add Optimization flag.
23092 2018-01-30 Jeff Law <law@redhat.com>
23095 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
23096 INT_REGISTERS_SAVED. Check it prior to calling
23097 get_scratch_register_on_entry.
23098 (ix86_adjust_stack_and_probe): Similarly.
23099 (ix86_emit_probe_stack_range): Similarly.
23100 (ix86_expand_prologue): Corresponding changes.
23102 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23105 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
23106 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
23108 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
23111 * lra-constraints.c (curr_insn_transform): Process AND in the
23114 2018-01-30 Jakub Jelinek <jakub@redhat.com>
23116 PR rtl-optimization/83986
23117 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
23118 dependence against last_pending_memory_flush in addition to
23119 pending_jump_insns.
23121 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
23123 PR tree-optimization/81611
23124 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
23127 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23130 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
23133 2018-01-30 Richard Biener <rguenther@suse.de>
23134 Jakub Jelinek <jakub@redhat.com>
23136 PR tree-optimization/84111
23137 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
23138 inner loops added during recursion, as they don't have up-to-date
23141 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
23144 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
23145 (can_inline_edge_by_limits_p): ... here.
23146 (can_early_inline_edge_p, check_callers,
23147 update_caller_keys, update_callee_keys, recursive_inlining,
23148 add_new_edges_to_heap, speculation_useful_p,
23149 inline_small_functions,
23150 inline_small_functions, flatten_function,
23151 inline_to_all_callers_1): Update.
23153 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
23155 * profile-count.c (profile_count::combine_with_ipa_count): Handle
23158 2018-01-30 Richard Biener <rguenther@suse.de>
23160 PR tree-optimization/83008
23161 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
23162 invariant and constant vector uses in stmts when they need
23163 more than one stmt.
23165 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23168 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
23169 * configure: Regenerate.
23171 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
23173 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
23175 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
23176 Use gen_rtx_REG rather than gen_lowpart.
23178 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
23180 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
23181 rather than 0 when creating partial subregs.
23183 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
23185 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
23188 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
23191 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
23192 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
23193 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
23194 flags. This restores the settings used before the 2017-07-24.
23195 Turning off pre increment/decrement/modify allows IVOPTS to
23196 optimize DF/SF loops where the index is an int.
23198 2018-01-29 Richard Biener <rguenther@suse.de>
23199 Kelvin Nilsen <kelvin@gcc.gnu.org>
23202 * tree-vect-stmts.c (vectorizable_call): Don't call
23203 targetm.vectorize_builtin_md_vectorized_function if callee is
23206 2018-01-22 Carl Love <cel@us.ibm.com>
23208 * doc/extend.tex: Fix typo in second arg in
23209 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
23211 2018-01-29 Richard Biener <rguenther@suse.de>
23213 PR tree-optimization/84086
23214 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
23215 (flush_ssaname_freelist): When SSA names were released reset
23216 the SCEV hash table.
23218 2018-01-29 Richard Biener <rguenther@suse.de>
23220 PR tree-optimization/84057
23221 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
23222 removed paths when removing edges.
23224 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
23226 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
23227 -mfunction-return=@var{choice}.
23229 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
23231 PR diagnostic/84034
23232 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
23233 Handle CR like TAB.
23234 (layout::print_source_line): Likewise.
23235 (test_get_line_width_without_trailing_whitespace): Add test cases.
23237 2018-01-27 Jakub Jelinek <jakub@redhat.com>
23239 PR middle-end/84040
23240 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
23243 2018-01-26 Jim Wilson <jimw@sifive.com>
23245 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
23247 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
23250 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23252 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
23253 and CMP + SUB-immediate -> SUBS.
23255 2018-01-26 Martin Sebor <msebor@redhat.com>
23257 PR tree-optimization/83896
23258 * tree-ssa-strlen.c (get_string_len): Rename...
23259 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
23260 Avoid assuming length is constant.
23261 (handle_char_store): Use HOST_WIDE_INT for string length.
23263 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
23266 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
23267 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
23269 2018-01-26 Richard Biener <rguenther@suse.de>
23271 PR rtl-optimization/84003
23272 * dse.c (record_store): Only record redundant stores when
23273 the earlier store aliases at least all accesses the later one does.
23275 2018-01-26 Jakub Jelinek <jakub@redhat.com>
23277 PR rtl-optimization/83985
23278 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
23279 REG_CFA_RESTORE insns.
23280 (delete_unmarked_insns): Don't ignore separate shrink wrapping
23281 REG_CFA_RESTORE insns here.
23284 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
23285 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
23287 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23289 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
23290 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
23291 (arc_init): Likewise.
23292 (arc_override_options): Likewise.
23293 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
23295 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
23297 * config/arc/arc.h (TARGET_DBNZ): Define.
23298 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
23299 properly set the tune attribute.
23300 (dbnz): Use TARGET_DBNZ guard.
23301 * config/arc/arc.opt (mtune): Add core3 option.
23303 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23305 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
23306 recognize new pic like addresses.
23307 (arc_delegitimize_address): Clean up.
23309 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23311 * config/arc/arc-arches.def: Option mrf16 valid for all
23313 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
23314 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
23315 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
23316 * config/arc/arc-tables.opt: Regenerate.
23317 * config/arc/arc.c (arc_conditional_register_usage): Handle
23318 reduced register file case.
23319 (arc_file_start): Set must have build attributes.
23320 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
23321 mrf16 option value.
23322 * config/arc/arc.opt (mrf16): Add new option.
23323 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
23324 * config/arc/genmultilib.awk: Handle new mrf16 option.
23325 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
23326 * config/arc/t-multilib: Regenerate.
23327 * doc/invoke.texi (ARC Options): Document mrf16 option.
23329 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23331 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
23332 * config/arc/arc.c (arc_handle_secure_attribute): New function.
23333 (arc_attribute_table): Add 'secure_call' attribute.
23334 (arc_print_operand): Print secure call operand.
23335 (arc_function_ok_for_sibcall): Don't optimize tail calls when
23337 (arc_is_secure_call_p): New function. * config/arc/arc.md
23338 (call_i): Add support for sjli instruction.
23339 (call_value_i): Likewise.
23340 * config/arc/constraints.md (Csc): New constraint.
23342 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
23343 John Eric Martin <John.Martin@emmicro-us.com>
23345 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
23346 * config/arc/arc.c (_arc_jli_section): New struct.
23347 (arc_jli_section): New type.
23348 (rc_jli_sections): New static variable.
23349 (arc_handle_jli_attribute): New function.
23350 (arc_attribute_table): Add jli_always and jli_fixed attribute.
23351 (arc_file_end): New function.
23352 (TARGET_ASM_FILE_END): Define.
23353 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
23354 (arc_add_jli_section): New function.
23355 (jli_call_scan): Likewise.
23356 (arc_reorg): Call jli_call_scan.
23357 (arc_output_addsi): Remove 'S' from printing asm operand.
23358 (arc_is_jli_call_p): New function.
23359 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
23361 (movhi_insn): Likewise.
23362 (movsi_insn): Likewise.
23363 (movsi_set_cc_insn): Likewise.
23364 (loadqi_update): Likewise.
23365 (load_zeroextendqisi_update): Likewise.
23366 (load_signextendqisi_update): Likewise.
23367 (loadhi_update): Likewise.
23368 (load_zeroextendhisi_update): Likewise.
23369 (load_signextendhisi_update): Likewise.
23370 (loadsi_update): Likewise.
23371 (loadsf_update): Likewise.
23372 (movsicc_insn): Likewise.
23373 (bset_insn): Likewise.
23374 (bxor_insn): Likewise.
23375 (bclr_insn): Likewise.
23376 (bmsk_insn): Likewise.
23377 (bicsi3_insn): Likewise.
23378 (cmpsi_cc_c_insn): Likewise.
23379 (movsi_ne): Likewise.
23380 (movsi_cond_exec): Likewise.
23381 (clrsbsi2): Likewise.
23382 (norm_f): Likewise.
23392 (call_i): Remove 'S' asm letter, add jli instruction.
23393 (call_value_i): Likewise.
23394 * config/arc/arc.op (mjli-always): New option.
23395 * config/arc/constraints.md (Cji): New constraint.
23396 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
23398 (subsf3_fpx): Likewise.
23399 (mulsf3_fpx): Likewise.
23400 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
23402 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
23403 function attrbutes.
23404 * doc/invoke.texi (ARC): Document mjli-always option.
23406 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
23408 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
23409 avoid addition with 0 and use incw and decw where possible.
23411 2018-01-26 Richard Biener <rguenther@suse.de>
23413 PR tree-optimization/81082
23414 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
23415 association if it requires casting to unsigned.
23416 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
23417 from fold_plusminus_mult_expr to catch important cases late when
23418 range info is available.
23420 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23422 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
23423 * configure.ac (hidden_linkonce): New test.
23424 * configure: Regenerate.
23425 * config.in: Regenerate.
23427 2018-01-26 Julia Koval <julia.koval@intel.com>
23429 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
23430 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
23431 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
23432 _mm_mask_bitshuffle_epi64_mask): Fix type.
23433 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
23434 USI_FTYPE_V4DI_V4DI_USI): Remove.
23435 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
23436 __builtin_ia32_vpshufbitqmb256_mask,
23437 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
23438 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
23439 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
23441 2018-01-26 Alan Modra <amodra@gmail.com>
23444 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
23445 UNSPEC_VBPERMQ. Sort other unspecs.
23447 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
23449 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
23451 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
23453 PR middle-end/83055
23454 * predict.c (drop_profile): Do not push/pop cfun; update also
23456 (handle_missing_profiles): Fix logic looking for zero profiles.
23458 2018-01-25 Jakub Jelinek <jakub@redhat.com>
23460 PR middle-end/83977
23461 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
23462 on functions with #pragma omp declare simd or functions with simd
23464 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
23465 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
23466 Remove trailing \n from warning_at calls.
23468 2018-01-25 Tom de Vries <tom@codesourcery.com>
23471 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23472 for neutered workers.
23474 2018-01-24 Joseph Myers <joseph@codesourcery.com>
23477 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
23478 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
23480 2018-01-24 Jeff Law <law@redhat.com>
23483 * i386.c (get_probe_interval): Move to earlier point.
23484 (ix86_compute_frame_layout): If -fstack-clash-protection and
23485 the frame is larger than the probe interval, then use pushes
23486 to save registers rather than reg->mem moves.
23487 (ix86_expand_prologue): Remove conditional for int_registers_saved
23490 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
23493 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
23494 min/max for never referenced object.
23496 2018-01-24 Jakub Jelinek <jakub@redhat.com>
23498 PR middle-end/83977
23499 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
23501 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
23502 attributes from DECL_ATTRIBUTES (decl) without affecting
23503 DECL_ATTRIBUTES (current_function_decl).
23504 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
23505 functions with non-NULL DECL_ABSTRACT_ORIGIN.
23507 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
23509 PR tree-optimization/83979
23510 * fold-const.c (fold_comparison): Use constant_boolean_node
23511 instead of boolean_{true,false}_node.
23513 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
23515 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
23518 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23520 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23521 Simplify the clause that sets the length attribute.
23522 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23523 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
23524 clause that sets the length attribute.
23525 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23527 2018-01-24 Tom de Vries <tom@codesourcery.com>
23530 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
23531 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
23532 Add strict parameter.
23533 (prevent_branch_around_nothing): Insert dummy insn between branch to
23534 label and label with no ptx insn inbetween.
23535 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
23537 2018-01-24 Tom de Vries <tom@codesourcery.com>
23540 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
23541 for neutered threads in warp.
23542 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
23544 2018-01-24 Richard Biener <rguenther@suse.de>
23546 PR tree-optimization/83176
23547 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
23550 2018-01-24 Richard Biener <rguenther@suse.de>
23552 PR tree-optimization/82819
23553 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
23554 code generating pluses that are no-ops in the target precision.
23556 2018-01-24 Richard Biener <rguenther@suse.de>
23558 PR middle-end/84000
23559 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
23561 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23563 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
23564 to merge probabilities.
23565 * predict.c (probably_never_executed): Also mark as cold functions
23566 with global 0 profile and guessed local profile.
23567 * profile-count.c (profile_probability::combine_with_count): New
23569 * profile-count.h (profile_probability::operator*,
23570 profile_probability::operator*=, profile_probability::operator/,
23571 profile_probability::operator/=): Reduce precision to adjusted
23572 and set value to guessed on contradictory divisions.
23573 (profile_probability::combine_with_freq): Remove.
23574 (profile_probability::combine_wiht_count): Declare.
23575 (profile_count::force_nonzero):: Set to adjusted.
23576 (profile_count::probability_in):: Set quality to adjusted.
23577 * tree-ssa-tail-merge.c (replace_block_by): Use
23578 combine_with_count.
23580 2018-01-23 Andrew Waterman <andrew@sifive.com>
23581 Jim Wilson <jimw@sifive.com>
23583 * config/riscv/riscv.c (riscv_stack_boundary): New.
23584 (riscv_option_override): Set riscv_stack_boundary. Handle
23585 riscv_preferred_stack_boundary_arg.
23586 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
23587 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
23588 (STACK_BOUNDARY): Set to riscv_stack_boundary.
23589 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
23590 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
23591 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
23593 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
23596 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
23597 of struct ix86_frame.
23598 (ix86_expand_epilogue): Likewise. Add a local variable for
23599 the reg_save_offset field in struct ix86_frame.
23601 2018-01-23 Bin Cheng <bin.cheng@arm.com>
23603 PR tree-optimization/82604
23604 * tree-loop-distribution.c (enum partition_kind): New enum item
23605 PKIND_PARTIAL_MEMSET.
23606 (partition_builtin_p): Support above new enum item.
23607 (generate_code_for_partition): Ditto.
23608 (compute_access_range): Differentiate cases that equality can be
23609 proven at all loops, the innermost loops or no loops.
23610 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
23611 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
23612 (finalize_partitions, distribute_loop): Don't fuse partition of
23613 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
23614 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
23615 parloop is enabled.
23617 2018-01-23 Martin Liska <mliska@suse.cz>
23619 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
23620 order to ignore the predictor.
23621 (PRED_POLYMORPHIC_CALL): Likewise.
23622 (PRED_RECURSIVE_CALL): Likewise.
23624 2018-01-23 Martin Liska <mliska@suse.cz>
23626 * tree-profile.c (tree_profiling): Print function header to
23627 aware reader which function we are working on.
23628 * value-prof.c (gimple_find_values_to_profile): Do not print
23629 not interesting value histograms.
23631 2018-01-23 Martin Liska <mliska@suse.cz>
23633 * profile-count.h (enum profile_quality): Add
23634 profile_uninitialized as the first value. Do not number values
23635 as they are zero based.
23636 (profile_count::verify): Update sanity check.
23637 (profile_probability::verify): Likewise.
23639 2018-01-23 Nathan Sidwell <nathan@acm.org>
23641 * doc/invoke.texi (ffor-scope): Deprecate.
23643 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23645 PR tree-optimization/83510
23646 * domwalk.c (set_all_edges_as_executable): New function.
23647 (dom_walker::dom_walker): Convert bool param
23648 "skip_unreachable_blocks" to enum reachability. Move setup of
23649 edge flags to set_all_edges_as_executable and only do it when
23650 reachability is REACHABLE_BLOCKS.
23651 * domwalk.h (enum dom_walker::reachability): New enum.
23652 (dom_walker::dom_walker): Convert bool param
23653 "skip_unreachable_blocks" to enum reachability.
23654 (set_all_edges_as_executable): New decl.
23655 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
23656 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
23658 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
23659 but converting true to REACHABLE_BLOCKS.
23660 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
23662 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
23663 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
23664 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
23666 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
23667 if check_all_array_refs will be called.
23669 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23671 * tree.c (selftest::test_location_wrappers): Add more test
23674 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23676 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
23677 (selftest::test_bit_in_range): Likewise.
23679 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
23682 * doc/sourcebuild.texi (vect_float): Say that the selector
23683 only describes the situation when -funsafe-math-optimizations is on.
23684 (vect_float_strict): Document.
23686 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
23688 PR tree-optimization/83965
23689 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
23690 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
23691 instead of checking only for a reduction.
23692 (vect_recog_widen_sum_pattern): Likewise.
23694 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23696 * predict.c (probably_never_executed): Only use precise profile info.
23697 (compute_function_frequency): Skip after inlining hack since we now
23698 have quality checking.
23700 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23702 * profile-count.h (profile_probability::very_unlikely,
23703 profile_probability::unlikely, profile_probability::even): Set
23704 precision to guessed.
23706 2018-01-23 Richard Biener <rguenther@suse.de>
23708 PR tree-optimization/83963
23709 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
23710 Properly terminate dominator walk when crossing the exit edge not
23711 when visiting its source block.
23713 2018-01-23 Jakub Jelinek <jakub@redhat.com>
23716 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
23717 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
23719 2018-01-22 Jakub Jelinek <jakub@redhat.com>
23721 PR tree-optimization/83957
23722 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
23723 semicolon after for body surrounded by braces.
23725 PR tree-optimization/83081
23726 * profile-count.h (profile_probability::split): New method.
23727 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
23728 Use profile_probability::split.
23729 (do_compare_rtx_and_jump): Fix adjustment of probabilities
23730 when splitting a single conditional jump into 2.
23732 2018-01-22 David Malcolm <dmalcolm@redhat.com>
23734 PR tree-optimization/69452
23735 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
23738 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23740 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
23741 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
23742 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
23744 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23746 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23748 * config/rl78/rl78.md (movdi): New define_expand.
23749 * config/rl78/rl78.c (rl78_split_movdi): New function.
23751 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
23754 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23756 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23757 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23758 128-bit to produce an UNSPEC move to get the double word with the
23759 signbit and then a shift directly to do signbit.
23760 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23761 implementation with a new version that just does either a direct
23762 move or a regular move. Move memory interface to separate insns.
23763 Move insns so they are next to the expander.
23764 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23765 with signbit move. Split big and little endian case.
23766 (signbit<mode>2_dm_mem_le): Likewise.
23767 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23768 (signbit<mode>2_dm2): Likewise.
23770 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23772 * config/rl78/rl78.md (anddi3): New define_expand.
23774 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23776 * config/rl78/rl78.md (umindi3): New define_expand.
23778 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23780 * config/rl78/rl78.md (smindi3): New define_expand.
23782 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23784 * config/rl78/rl78.md (smaxdi3): New define_expand.
23786 2018-01-22 Carl Love <cel@us.ibm.com>
23788 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23789 LVX_V1TI): Add macro expansion.
23790 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23791 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23792 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23793 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23794 Change check to determine if the instruction is a byte reversing
23795 entry. Fix typo in comment.
23796 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23797 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23798 Add def_builtin calls for new builtins.
23799 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23800 Add define_insn expansion.
23802 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23804 * config/rl78/rl78.md (umaxdi3): New define_expand.
23806 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23808 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23809 for non-QImode registers.
23811 2018-01-22 Richard Biener <rguenther@suse.de>
23813 PR tree-optimization/83963
23814 * graphite-scop-detection.c (scop_detection::get_sese): Delay
23815 including the loop exit block.
23816 (scop_detection::merge_sese): Likewise.
23817 (scop_detection::add_scop): Do it here instead.
23819 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23821 * doc/sourcebuild.texi (arm_softfloat): Document.
23823 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
23826 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23827 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23828 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23830 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23831 David Edelsohn <dje.gcc@gmail.com>
23834 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23835 Change "crset eq" to "crset 2".
23836 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23837 (*call_indirect_aix<mode>_nospec): Likewise.
23838 (*call_value_indirect_aix<mode>_nospec): Likewise.
23839 (*call_indirect_elfv2<mode>_nospec): Likewise.
23840 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
23841 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
23842 change assembly output from . to $.
23843 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23844 (indirect_jump<mode>_nospec): Change assembly output from . to $.
23845 (*tablejump<mode>_internal1_nospec): Likewise.
23847 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
23850 * config/sh/sh_optimize_sett_clrt.cc:
23851 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
23853 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
23855 PR tree-optimization/83940
23856 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
23857 offset_dt to vect_constant_def rather than vect_unknown_def_type.
23858 (vect_check_load_store_mask): Add a mask_dt_out parameter and
23859 use it to pass back the definition type.
23860 (vect_check_store_rhs): Likewise rhs_dt_out.
23861 (vect_build_gather_load_calls): Add a mask_dt argument and use
23862 it instead of a call to vect_is_simple_use.
23863 (vectorizable_store): Update calls to vect_check_load_store_mask
23864 and vect_check_store_rhs. Use the dt returned by the latter instead
23865 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
23866 instead of calls to vect_is_simple_use. Pass the scalar rather
23867 than the vector operand to vect_is_simple_use when handling
23868 second and subsequent copies of an rhs value.
23869 (vectorizable_load): Update calls to vect_check_load_store_mask
23870 and vect_build_gather_load_calls. Use the cached mask_dt and
23871 gs_info.offset_dt instead of calls to vect_is_simple_use.
23873 2018-01-20 Jakub Jelinek <jakub@redhat.com>
23875 PR middle-end/83945
23876 * tree-emutls.c: Include gimplify.h.
23877 (lower_emutls_2): New function.
23878 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23879 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23880 it before further processing.
23883 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23884 UINTVAL (trueop1) instead of INTVAL (op1).
23886 2018-01-19 Jakub Jelinek <jakub@redhat.com>
23890 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23891 INCOMING_FRAME_SP_OFFSET if not defined.
23892 (scan_trace): Add ENTRY argument. If true and
23893 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23894 emit a note to adjust the CFA offset.
23895 (create_cfi_notes): Adjust scan_trace callers.
23896 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23897 INCOMING_FRAME_SP_OFFSET in the CIE.
23898 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23899 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23901 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23902 * doc/tm.texi: Regenerated.
23904 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23906 PR rtl-optimization/83147
23907 * lra-constraints.c (remove_inheritance_pseudos): Use
23908 lra_substitute_pseudo_within_insn.
23910 2018-01-19 Tom de Vries <tom@codesourcery.com>
23911 Cesar Philippidis <cesar@codesourcery.com>
23914 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23916 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
23919 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23920 spaces for function labels.
23922 2018-01-19 Martin Liska <mliska@suse.cz>
23924 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23925 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23926 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23927 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23928 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23929 (PRED_CONST_RETURN): Change from 69 to 65.
23930 (PRED_NULL_RETURN): Change from 91 to 71.
23931 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23932 (PRED_LOOP_GUARD): Change from 66 to 73.
23934 2018-01-19 Martin Liska <mliska@suse.cz>
23936 * predict.c (predict_insn_def): Add new assert.
23937 (struct branch_predictor): Change type to signed integer.
23938 (test_prediction_value_range): Amend test to cover
23939 PROB_UNINITIALIZED.
23940 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23941 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23942 (PRED_LOOP_ITERATIONS_MAX): Likewise.
23943 (PRED_LOOP_IV_COMPARE): Likewise.
23944 * predict.h (PROB_UNINITIALIZED): Define new constant.
23946 2018-01-19 Martin Liska <mliska@suse.cz>
23948 * predict.c (dump_prediction): Add new format for
23949 analyze_brprob.py script which is enabled with -details
23951 * profile-count.h (precise_p): New function.
23953 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23955 PR tree-optimization/83922
23956 * tree-vect-loop.c (vect_verify_full_masking): Return false if
23957 there are no statements that need masking.
23958 (vect_active_double_reduction_p): New function.
23959 (vect_analyze_loop_operations): Use it when handling phis that
23960 are not in the loop header.
23962 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23964 PR tree-optimization/83914
23965 * tree-vect-loop.c (vectorizable_induction): Don't convert
23966 init_expr or apply the peeling adjustment for inductions
23967 that are nested within the vectorized loop.
23969 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23971 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23974 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23978 * function.h (gimplify_parameters): Add gimple_seq * argument.
23979 * function.c: Include gimple.h and options.h.
23980 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23981 for the added local temporaries if needed.
23982 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23983 if there are any parameter cleanups, wrap whole body into a
23984 try/finally with the cleanups.
23986 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
23989 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23990 Use GET_MODE_CLASS for scalar floating point.
23992 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23996 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23997 Fix call of call_cgraph_insertion_hooks.
23999 2018-01-18 Martin Sebor <msebor@redhat.com>
24001 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24003 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
24006 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24009 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
24012 * common.opt: (-ffile-prefix-map): New option.
24013 * opts.c (common_handle_option): Defer it.
24014 * opts-global.c (handle_common_deferred_options): Handle it.
24015 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24016 * file-prefix-map.h: New file.
24017 (remap_debug_filename, add_debug_prefix_map): ...here.
24018 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24019 * final.c (debug_prefix_map, add_debug_prefix_map
24020 remap_debug_filename): Move to...
24021 * file-prefix-map.c: New file.
24022 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24023 generalize, get rid of alloca(), use strrchr() instead of strchr().
24024 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24025 Implement in terms of add_prefix_map().
24026 (remap_macro_filename, remap_debug_filename): Implement in term of
24028 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24029 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24030 * dbxout.c: Include file-prefix-map.h.
24031 * varasm.c: Likewise.
24032 * vmsdbgout.c: Likewise.
24033 * xcoffout.c: Likewise.
24034 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24035 * doc/cppopts.texi (-fmacro-prefix-map): Document.
24036 * doc/invoke.texi (-ffile-prefix-map): Document.
24037 (-fdebug-prefix-map): Update description.
24039 2018-01-18 Martin Liska <mliska@suse.cz>
24041 * config/i386/i386.c (indirect_thunk_name): Document that also
24043 (output_indirect_thunk): Document why both instructions
24044 (pause and lfence) are generated.
24046 2018-01-18 Richard Biener <rguenther@suse.de>
24048 PR tree-optimization/83887
24049 * graphite-scop-detection.c
24050 (scop_detection::get_nearest_dom_with_single_entry): Remove.
24051 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
24052 (scop_detection::merge_sese): Re-implement with a flood-fill
24053 algorithm that properly finds a SESE region if it exists.
24055 2018-01-18 Jakub Jelinek <jakub@redhat.com>
24058 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
24059 pointer_diff optimizations use view_convert instead of convert.
24061 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24063 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24064 Generate different code for -mno-speculate-indirect-jumps.
24065 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24066 (*call_indirect_aix<mode>): Disable for
24067 -mno-speculate-indirect-jumps.
24068 (*call_indirect_aix<mode>_nospec): New define_insn.
24069 (*call_value_indirect_aix<mode>): Disable for
24070 -mno-speculate-indirect-jumps.
24071 (*call_value_indirect_aix<mode>_nospec): New define_insn.
24072 (*sibcall_nonlocal_sysv<mode>): Generate different code for
24073 -mno-speculate-indirect-jumps.
24074 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24076 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
24078 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
24079 long double type, set the flags for noting the default long double
24080 type, even if we don't pass or return a long double type.
24082 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
24085 * ipa-inline.c (flatten_function): Do not overwrite final inlining
24088 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
24090 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
24091 support for merge[hl].
24092 (fold_mergehl_helper): New helper function.
24093 (tree-vector-builder.h): New #include for tree_vector_builder usage.
24094 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
24095 (altivec_vmrglw_direct): Add xxmrglw insn.
24097 2018-01-17 Andrew Waterman <andrew@sifive.com>
24099 * config/riscv/riscv.c (riscv_conditional_register_usage): If
24100 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
24102 2018-01-17 David Malcolm <dmalcolm@redhat.com>
24105 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
24106 call the lto_location_cache before reading the
24107 DECL_SOURCE_LOCATION of the types.
24109 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
24110 Richard Sandiford <richard.sandiford@linaro.org>
24112 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
24113 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
24114 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
24115 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
24116 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
24118 * config/aarch64/constraints.md (aarch64_movti_operand):
24120 * config/aarch64/predicates.md (Uti): Add new constraint.
24122 2018-01-17 Carl Love <cel@us.ibm.com>
24124 * config/rs6000/vsx.md (define_expand xl_len_r,
24125 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
24126 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
24128 (define_expand, define_insn): Move the shift left from the
24129 define_insn to the define_expand for lxvl and stxvl instructions.
24130 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
24131 and XL_LEN_R definitions to PURE.
24133 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
24135 * config/i386/i386.c (indirect_thunk_name): Declare regno
24136 as unsigned int. Compare regno with INVALID_REGNUM.
24137 (output_indirect_thunk): Ditto.
24138 (output_indirect_thunk_function): Ditto.
24139 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
24140 in the call to output_indirect_thunk_function.
24142 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
24144 PR middle-end/83884
24145 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
24146 rather than the size of inner_type to determine the stack slot size
24147 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
24149 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
24152 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
24155 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
24157 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
24158 endian Linux systems to optionally enable multilibs for selecting
24159 the long double type if the user configured an explicit type.
24160 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
24161 have no long double multilibs if not defined.
24162 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
24163 warn if the user used -mabi={ieee,ibm}longdouble and we built
24164 multilibs for long double.
24165 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
24166 appropriate multilib option.
24167 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
24169 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
24170 for building long double multilibs.
24171 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
24173 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
24175 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
24178 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
24180 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
24183 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
24186 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
24189 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
24191 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
24192 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
24194 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
24196 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
24197 different rtl trees depending on TARGET_64BIT.
24198 (rs6000_gen_lvx): Likewise.
24200 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
24202 * config/visium/visium.md (nop): Tweak comment.
24203 (hazard_nop): Likewise.
24205 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24207 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
24208 -mspeculate-indirect-jumps.
24209 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
24210 for -mno-speculate-indirect-jumps.
24211 (*call_indirect_elfv2<mode>_nospec): New define_insn.
24212 (*call_value_indirect_elfv2<mode>): Disable for
24213 -mno-speculate-indirect-jumps.
24214 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
24215 (indirect_jump): Emit different RTL for
24216 -mno-speculate-indirect-jumps.
24217 (*indirect_jump<mode>): Disable for
24218 -mno-speculate-indirect-jumps.
24219 (*indirect_jump<mode>_nospec): New define_insn.
24220 (tablejump): Emit different RTL for
24221 -mno-speculate-indirect-jumps.
24222 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
24223 (tablejumpsi_nospec): New define_expand.
24224 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
24225 (tablejumpdi_nospec): New define_expand.
24226 (*tablejump<mode>_internal1): Disable for
24227 -mno-speculate-indirect-jumps.
24228 (*tablejump<mode>_internal1_nospec): New define_insn.
24229 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
24232 2018-01-16 Artyom Skrobov tyomitch@gmail.com
24234 * caller-save.c (insert_save): Drop unnecessary parameter. All
24237 2018-01-16 Jakub Jelinek <jakub@redhat.com>
24238 Richard Biener <rguenth@suse.de>
24241 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
24242 return early, inline manually is_gimple_sizepos. Make sure if we
24243 call gimplify_expr we don't end up with a gimple constant.
24244 * tree.c (variably_modified_type_p): Don't return true for
24245 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
24246 * gimplify.h (is_gimple_sizepos): Remove.
24248 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24250 PR tree-optimization/83857
24251 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
24252 vectorizable_live_operation for pure SLP statements.
24253 (vectorizable_live_operation): Handle PHIs.
24255 2018-01-16 Richard Biener <rguenther@suse.de>
24257 PR tree-optimization/83867
24258 * tree-vect-stmts.c (vect_transform_stmt): Precompute
24259 nested_in_vect_loop_p since the scalar stmt may get invalidated.
24261 2018-01-16 Jakub Jelinek <jakub@redhat.com>
24264 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
24265 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
24266 If off is not INTEGER_CST, issue a may not be aligned warning
24267 rather than isn't aligned. Use isn%'t rather than isn't.
24268 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
24270 <case MULT_EXPR>: Improve the case when bottom and one of the
24271 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
24272 operand, in that case check if the other operand is multiple of
24273 bottom divided by the INTEGER_CST operand.
24275 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24278 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
24279 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
24280 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
24281 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
24282 * config/pa/pa.c (pa_function_arg_advance): Likewise.
24283 (pa_function_arg, pa_arg_partial_bytes): Likewise.
24284 (pa_function_arg_size): New function.
24286 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24288 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
24289 in a separate statement.
24291 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
24293 PR tree-optimization/83847
24294 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
24295 group gathers and scatters.
24297 2018-01-16 Jakub Jelinek <jakub@redhat.com>
24299 PR rtl-optimization/86620
24300 * params.def (max-sched-ready-insns): Bump minimum value to 1.
24302 PR rtl-optimization/83213
24303 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
24304 to last if both are JUMP_INSNs.
24306 PR tree-optimization/83843
24307 * gimple-ssa-store-merging.c
24308 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
24309 store_immediate_info for bswap/nop orig_stores.
24311 2018-01-15 Andrew Waterman <andrew@sifive.com>
24313 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
24315 <UDIV>: Increase cost if !TARGET_DIV.
24317 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
24319 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
24320 (define_attr "cr_logical_3op"): New.
24321 (cceq_ior_compare): Adjust.
24322 (cceq_ior_compare_complement): Adjust.
24323 (*cceq_rev_compare): Adjust.
24324 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
24325 (is_cracked_insn): Adjust.
24326 (insn_must_be_first_in_group): Adjust.
24327 * config/rs6000/40x.md: Adjust.
24328 * config/rs6000/440.md: Adjust.
24329 * config/rs6000/476.md: Adjust.
24330 * config/rs6000/601.md: Adjust.
24331 * config/rs6000/603.md: Adjust.
24332 * config/rs6000/6xx.md: Adjust.
24333 * config/rs6000/7450.md: Adjust.
24334 * config/rs6000/7xx.md: Adjust.
24335 * config/rs6000/8540.md: Adjust.
24336 * config/rs6000/cell.md: Adjust.
24337 * config/rs6000/e300c2c3.md: Adjust.
24338 * config/rs6000/e500mc.md: Adjust.
24339 * config/rs6000/e500mc64.md: Adjust.
24340 * config/rs6000/e5500.md: Adjust.
24341 * config/rs6000/e6500.md: Adjust.
24342 * config/rs6000/mpc.md: Adjust.
24343 * config/rs6000/power4.md: Adjust.
24344 * config/rs6000/power5.md: Adjust.
24345 * config/rs6000/power6.md: Adjust.
24346 * config/rs6000/power7.md: Adjust.
24347 * config/rs6000/power8.md: Adjust.
24348 * config/rs6000/power9.md: Adjust.
24349 * config/rs6000/rs64.md: Adjust.
24350 * config/rs6000/titan.md: Adjust.
24352 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24354 * config/i386/predicates.md (indirect_branch_operand): Rewrite
24355 ix86_indirect_branch_register logic.
24357 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24359 * config/i386/constraints.md (Bs): Update
24360 ix86_indirect_branch_register check. Don't check
24361 ix86_indirect_branch_register with GOT_memory_operand.
24363 * config/i386/predicates.md (GOT_memory_operand): Don't check
24364 ix86_indirect_branch_register here.
24365 (GOT32_symbol_operand): Likewise.
24367 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24369 * config/i386/predicates.md (constant_call_address_operand):
24370 Rewrite ix86_indirect_branch_register logic.
24371 (sibcall_insn_operand): Likewise.
24373 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24375 * config/i386/constraints.md (Bs): Replace
24376 ix86_indirect_branch_thunk_register with
24377 ix86_indirect_branch_register.
24379 * config/i386/i386.md (indirect_jump): Likewise.
24380 (tablejump): Likewise.
24381 (*sibcall_memory): Likewise.
24382 (*sibcall_value_memory): Likewise.
24383 Peepholes of indirect call and jump via memory: Likewise.
24384 * config/i386/i386.opt: Likewise.
24385 * config/i386/predicates.md (indirect_branch_operand): Likewise.
24386 (GOT_memory_operand): Likewise.
24387 (call_insn_operand): Likewise.
24388 (sibcall_insn_operand): Likewise.
24389 (GOT32_symbol_operand): Likewise.
24391 2018-01-15 Jakub Jelinek <jakub@redhat.com>
24393 PR middle-end/83837
24394 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
24395 type rather than type addr's type points to.
24396 (expand_omp_atomic_mutex): Likewise.
24397 (expand_omp_atomic): Likewise.
24399 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
24402 * config/i386/i386.c (output_indirect_thunk_function): Use
24403 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
24404 for __x86_return_thunk.
24406 2018-01-15 Richard Biener <rguenther@suse.de>
24408 PR middle-end/83850
24409 * expmed.c (extract_bit_field_1): Fix typo.
24411 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24414 * config/arm/iterators.md (VF): New mode iterator.
24415 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
24416 Remove integer-related logic from pattern.
24417 (neon_vabd<mode>_3): Likewise.
24419 2018-01-15 Jakub Jelinek <jakub@redhat.com>
24421 PR middle-end/82694
24422 * common.opt (fstrict-overflow): No longer an alias.
24423 (fwrapv-pointer): New option.
24424 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
24425 also for pointer types based on flag_wrapv_pointer.
24426 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
24427 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
24428 opts->x_flag_wrapv got set.
24429 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
24430 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
24431 POINTER_TYPE_OVERFLOW_UNDEFINED.
24432 * match.pd: Likewise in address comparison pattern.
24433 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
24435 2018-01-15 Richard Biener <rguenther@suse.de>
24438 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
24439 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
24440 Reset type names to their identifier if their TYPE_DECL doesn't
24441 have linkage (and thus is used for ODR and devirt).
24442 (save_debug_info_for_decl): Remove.
24443 (save_debug_info_for_type): Likewise.
24444 (add_tree_to_fld_list): Adjust.
24445 * tree-pretty-print.c (dump_generic_node): Make dumping of
24446 type names more robust.
24448 2018-01-15 Richard Biener <rguenther@suse.de>
24450 * BASE-VER: Bump to 8.0.1.
24452 2018-01-14 Martin Sebor <msebor@redhat.com>
24455 * builtins.c (check_access): Avoid warning when the no-warning bit
24458 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
24460 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
24461 * ira-color (allocno_hard_regs_compare): Likewise.
24463 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
24466 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
24467 Use .pushsection/.popsection.
24469 2018-01-14 Martin Sebor <msebor@redhat.com>
24472 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
24474 2018-01-14 Jakub Jelinek <jakub@redhat.com>
24476 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
24477 entry from extra_headers.
24478 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
24479 extra_headers, make the list bitwise identical to the i?86-*-* one.
24481 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24483 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
24484 -mcmodel=large with -mindirect-branch=thunk,
24485 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
24486 -mfunction-return=thunk-extern.
24487 * doc/invoke.texi: Document -mcmodel=large is incompatible with
24488 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
24489 -mfunction-return=thunk and -mfunction-return=thunk-extern.
24491 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24493 * config/i386/i386.c (print_reg): Print the name of the full
24494 integer register without '%'.
24495 (ix86_print_operand): Handle 'V'.
24496 * doc/extend.texi: Document 'V' modifier.
24498 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24500 * config/i386/constraints.md (Bs): Disallow memory operand for
24501 -mindirect-branch-register.
24503 * config/i386/predicates.md (indirect_branch_operand): Likewise.
24504 (GOT_memory_operand): Likewise.
24505 (call_insn_operand): Likewise.
24506 (sibcall_insn_operand): Likewise.
24507 (GOT32_symbol_operand): Likewise.
24508 * config/i386/i386.md (indirect_jump): Call convert_memory_address
24509 for -mindirect-branch-register.
24510 (tablejump): Likewise.
24511 (*sibcall_memory): Likewise.
24512 (*sibcall_value_memory): Likewise.
24513 Disallow peepholes of indirect call and jump via memory for
24514 -mindirect-branch-register.
24515 (*call_pop): Replace m with Bw.
24516 (*call_value_pop): Likewise.
24517 (*sibcall_pop_memory): Replace m with Bs.
24518 * config/i386/i386.opt (mindirect-branch-register): New option.
24519 * doc/invoke.texi: Document -mindirect-branch-register option.
24521 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24523 * config/i386/i386-protos.h (ix86_output_function_return): New.
24524 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
24525 set function_return_type.
24526 (indirect_thunk_name): Add ret_p to indicate thunk for function
24528 (output_indirect_thunk_function): Pass false to
24529 indirect_thunk_name.
24530 (ix86_output_indirect_branch_via_reg): Likewise.
24531 (ix86_output_indirect_branch_via_push): Likewise.
24532 (output_indirect_thunk_function): Create alias for function
24533 return thunk if regno < 0.
24534 (ix86_output_function_return): New function.
24535 (ix86_handle_fndecl_attribute): Handle function_return.
24536 (ix86_attribute_table): Add function_return.
24537 * config/i386/i386.h (machine_function): Add
24538 function_return_type.
24539 * config/i386/i386.md (simple_return_internal): Use
24540 ix86_output_function_return.
24541 (simple_return_internal_long): Likewise.
24542 * config/i386/i386.opt (mfunction-return=): New option.
24543 (indirect_branch): Mention -mfunction-return=.
24544 * doc/extend.texi: Document function_return function attribute.
24545 * doc/invoke.texi: Document -mfunction-return= option.
24547 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
24549 * config/i386/i386-opts.h (indirect_branch): New.
24550 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
24551 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
24552 with local indirect jump when converting indirect call and jump.
24553 (ix86_set_indirect_branch_type): New.
24554 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
24555 (indirectlabelno): New.
24556 (indirect_thunk_needed): Likewise.
24557 (indirect_thunk_bnd_needed): Likewise.
24558 (indirect_thunks_used): Likewise.
24559 (indirect_thunks_bnd_used): Likewise.
24560 (INDIRECT_LABEL): Likewise.
24561 (indirect_thunk_name): Likewise.
24562 (output_indirect_thunk): Likewise.
24563 (output_indirect_thunk_function): Likewise.
24564 (ix86_output_indirect_branch_via_reg): Likewise.
24565 (ix86_output_indirect_branch_via_push): Likewise.
24566 (ix86_output_indirect_branch): Likewise.
24567 (ix86_output_indirect_jmp): Likewise.
24568 (ix86_code_end): Call output_indirect_thunk_function if needed.
24569 (ix86_output_call_insn): Call ix86_output_indirect_branch if
24571 (ix86_handle_fndecl_attribute): Handle indirect_branch.
24572 (ix86_attribute_table): Add indirect_branch.
24573 * config/i386/i386.h (machine_function): Add indirect_branch_type
24574 and has_local_indirect_jump.
24575 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
24577 (tablejump): Likewise.
24578 (*indirect_jump): Use ix86_output_indirect_jmp.
24579 (*tablejump_1): Likewise.
24580 (simple_return_indirect_internal): Likewise.
24581 * config/i386/i386.opt (mindirect-branch=): New option.
24582 (indirect_branch): New.
24585 (thunk-inline): Likewise.
24586 (thunk-extern): Likewise.
24587 * doc/extend.texi: Document indirect_branch function attribute.
24588 * doc/invoke.texi: Document -mindirect-branch= option.
24590 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
24593 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
24595 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
24597 * ipa-inline.c (want_inline_small_function_p): Return false if
24598 inlining has already failed with CIF_FINAL_ERROR.
24599 (update_caller_keys): Call want_inline_small_function_p before
24601 (update_callee_keys): Likewise.
24603 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
24605 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
24607 (rs6000_quadword_masked_address_p): Likewise.
24608 (quad_aligned_load_p): Likewise.
24609 (quad_aligned_store_p): Likewise.
24610 (const_load_sequence_p): Add comment to describe the outer-most loop.
24611 (mimic_memory_attributes_and_flags): New function.
24612 (rs6000_gen_stvx): Likewise.
24613 (replace_swapped_aligned_store): Likewise.
24614 (rs6000_gen_lvx): Likewise.
24615 (replace_swapped_aligned_load): Likewise.
24616 (replace_swapped_load_constant): Capitalize argument name in
24617 comment describing this function.
24618 (rs6000_analyze_swaps): Add a third pass to search for vector loads
24619 and stores that access quad-word aligned addresses and replace
24620 with stvx or lvx instructions when appropriate.
24621 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
24622 New function prototype.
24623 (rs6000_quadword_masked_address_p): Likewise.
24624 (rs6000_gen_lvx): Likewise.
24625 (rs6000_gen_stvx): Likewise.
24626 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
24627 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
24628 when memory address is aligned.
24629 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
24630 this split to select lvx instruction when memory address is aligned.
24631 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
24632 instruction when memory address is aligned.
24633 (*vsx_le_perm_load_v16qi): Likewise.
24634 (four unnamed splitters): Modify to select the stvx instruction
24635 when memory is aligned.
24637 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
24639 * predict.c (determine_unlikely_bbs): Handle correctly BBs
24640 which appears in the queue multiple times.
24642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24643 Alan Hayward <alan.hayward@arm.com>
24644 David Sherwood <david.sherwood@arm.com>
24646 * tree-vectorizer.h (vec_lower_bound): New structure.
24647 (_loop_vec_info): Add check_nonzero and lower_bounds.
24648 (LOOP_VINFO_CHECK_NONZERO): New macro.
24649 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
24650 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
24651 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
24652 fields. Make seg_len the distance travelled, not including the
24654 (dr_direction_indicator): Declare.
24655 (dr_zero_step_indicator): Likewise.
24656 (dr_known_forward_stride_p): Likewise.
24657 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
24659 (runtime_alias_check_p): Allow runtime alias checks with
24661 (operator ==): Compare access_size and align.
24662 (prune_runtime_alias_test_list): Rework for new distinction between
24663 the access_size and seg_len.
24664 (create_intersect_range_checks_index): Likewise. Cope with polynomial
24666 (get_segment_min_max): New function.
24667 (create_intersect_range_checks): Use it.
24668 (dr_step_indicator): New function.
24669 (dr_direction_indicator): Likewise.
24670 (dr_zero_step_indicator): Likewise.
24671 (dr_known_forward_stride_p): Likewise.
24672 * tree-loop-distribution.c (data_ref_segment_size): Return
24673 DR_STEP * (niters - 1).
24674 (compute_alias_check_pairs): Update call to the dr_with_seg_len
24676 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
24677 (vect_preserves_scalar_order_p): New function, split out from...
24678 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
24679 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
24680 (vect_vfa_access_size): New function.
24681 (vect_vfa_align): Likewise.
24682 (vect_compile_time_alias): Take access_size_a and access_b arguments.
24683 (dump_lower_bound): New function.
24684 (vect_check_lower_bound): Likewise.
24685 (vect_small_gap_p): Likewise.
24686 (vectorizable_with_step_bound_p): Likewise.
24687 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
24688 depencies if the vectorization factor is 1. Convert the checks
24689 for nonzero steps into checks on the bounds of DR_STEP. Try using
24690 a bunds check for variable steps if the minimum required step is
24691 relatively small. Update calls to the dr_with_seg_len
24692 constructor and to vect_compile_time_alias.
24693 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
24695 (vect_loop_versioning): Call it.
24696 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
24698 (vect_estimate_min_profitable_iters): Account for any bounds checks.
24700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24701 Alan Hayward <alan.hayward@arm.com>
24702 David Sherwood <david.sherwood@arm.com>
24704 * doc/sourcebuild.texi (vect_scatter_store): Document.
24705 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
24707 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
24709 * genopinit.c (main): Add supports_vec_scatter_store and
24710 supports_vec_scatter_store_cached to target_optabs.
24711 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
24712 IFN_MASK_SCATTER_STORE.
24713 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
24715 * internal-fn.h (internal_store_fn_p): Declare.
24716 (internal_fn_stored_value_index): Likewise.
24717 * internal-fn.c (scatter_store_direct): New macro.
24718 (expand_scatter_store_optab_fn): New function.
24719 (direct_scatter_store_optab_supported_p): New macro.
24720 (internal_store_fn_p): New function.
24721 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
24722 IFN_MASK_SCATTER_STORE.
24723 (internal_fn_mask_index): Likewise.
24724 (internal_fn_stored_value_index): New function.
24725 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
24726 for scatter stores.
24727 * optabs-query.h (supports_vec_scatter_store_p): Declare.
24728 * optabs-query.c (supports_vec_scatter_store_p): New function.
24729 * tree-vectorizer.h (vect_get_store_rhs): Declare.
24730 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
24731 true for scatter stores.
24732 (vect_gather_scatter_fn_p): Handle scatter stores too.
24733 (vect_check_gather_scatter): Consider using scatter stores if
24734 supports_vec_scatter_store_p.
24735 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
24736 scatter stores too.
24737 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24738 internal_fn_stored_value_index.
24739 (check_load_store_masking): Handle scatter stores too.
24740 (vect_get_store_rhs): Make public.
24741 (vectorizable_call): Use internal_store_fn_p.
24742 (vectorizable_store): Handle scatter store internal functions.
24743 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
24744 when deciding whether the end of the group has been reached.
24745 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
24746 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24747 (mask_scatter_store<mode>): New insns.
24749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24750 Alan Hayward <alan.hayward@arm.com>
24751 David Sherwood <david.sherwood@arm.com>
24753 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24754 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24755 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24757 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24758 Use vect_truncate_gather_scatter_offset if we can't treat the
24759 operation as a normal gather load or scatter store.
24760 (get_group_load_store_type): Take the gather_scatter_info
24761 as argument. Try using a gather load or scatter store for
24762 single-element groups.
24763 (get_load_store_type): Update calls to get_group_load_store_type
24764 and vect_use_strided_gather_scatters_p.
24766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24767 Alan Hayward <alan.hayward@arm.com>
24768 David Sherwood <david.sherwood@arm.com>
24770 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24771 optional tree argument.
24772 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24774 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24775 but continue to use the current value as a fallback.
24776 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24777 to compare the updates.
24778 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24779 (get_load_store_type): Use it when handling a strided access.
24780 (vect_get_strided_load_store_ops): New function.
24781 (vect_get_data_ptr_increment): Likewise.
24782 (vectorizable_load): Handle strided gather loads. Always pass
24783 a step to vect_create_data_ref_ptr and bump_vector_ptr.
24785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24786 Alan Hayward <alan.hayward@arm.com>
24787 David Sherwood <david.sherwood@arm.com>
24789 * doc/md.texi (gather_load@var{m}): Document.
24790 (mask_gather_load@var{m}): Likewise.
24791 * genopinit.c (main): Add supports_vec_gather_load and
24792 supports_vec_gather_load_cached to target_optabs.
24793 * optabs-tree.c (init_tree_optimization_optabs): Use
24794 ggc_cleared_alloc to allocate target_optabs.
24795 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24796 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24798 * internal-fn.h (internal_load_fn_p): Declare.
24799 (internal_gather_scatter_fn_p): Likewise.
24800 (internal_fn_mask_index): Likewise.
24801 (internal_gather_scatter_fn_supported_p): Likewise.
24802 * internal-fn.c (gather_load_direct): New macro.
24803 (expand_gather_load_optab_fn): New function.
24804 (direct_gather_load_optab_supported_p): New macro.
24805 (direct_internal_fn_optab): New function.
24806 (internal_load_fn_p): Likewise.
24807 (internal_gather_scatter_fn_p): Likewise.
24808 (internal_fn_mask_index): Likewise.
24809 (internal_gather_scatter_fn_supported_p): Likewise.
24810 * optabs-query.c (supports_at_least_one_mode_p): New function.
24811 (supports_vec_gather_load_p): Likewise.
24812 * optabs-query.h (supports_vec_gather_load_p): Declare.
24813 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24814 and memory_type field.
24815 (NUM_PATTERNS): Bump to 15.
24816 * tree-vect-data-refs.c: Include internal-fn.h.
24817 (vect_gather_scatter_fn_p): New function.
24818 (vect_describe_gather_scatter_call): Likewise.
24819 (vect_check_gather_scatter): Try using internal functions for
24820 gather loads. Recognize existing calls to a gather load function.
24821 (vect_analyze_data_refs): Consider using gather loads if
24822 supports_vec_gather_load_p.
24823 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24824 (vect_get_gather_scatter_offset_type): Likewise.
24825 (vect_convert_mask_for_vectype): Likewise.
24826 (vect_add_conversion_to_patterm): Likewise.
24827 (vect_try_gather_scatter_pattern): Likewise.
24828 (vect_recog_gather_scatter_pattern): New pattern recognizer.
24829 (vect_vect_recog_func_ptrs): Add it.
24830 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24831 internal_fn_mask_index and internal_gather_scatter_fn_p.
24832 (check_load_store_masking): Take the gather_scatter_info as an
24833 argument and handle gather loads.
24834 (vect_get_gather_scatter_ops): New function.
24835 (vectorizable_call): Check internal_load_fn_p.
24836 (vectorizable_load): Likewise. Handle gather load internal
24838 (vectorizable_store): Update call to check_load_store_masking.
24839 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
24840 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
24841 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
24842 (aarch64_gather_scale_operand_d): New predicates.
24843 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
24844 (mask_gather_load<mode>): New insns.
24846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24847 Alan Hayward <alan.hayward@arm.com>
24848 David Sherwood <david.sherwood@arm.com>
24850 * optabs.def (fold_left_plus_optab): New optab.
24851 * doc/md.texi (fold_left_plus_@var{m}): Document.
24852 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
24853 * internal-fn.c (fold_left_direct): Define.
24854 (expand_fold_left_optab_fn): Likewise.
24855 (direct_fold_left_optab_supported_p): Likewise.
24856 * fold-const-call.c (fold_const_fold_left): New function.
24857 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
24858 * tree-parloops.c (valid_reduction_p): New function.
24859 (gather_scalar_reductions): Use it.
24860 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
24861 (vect_finish_replace_stmt): Declare.
24862 * tree-vect-loop.c (fold_left_reduction_fn): New function.
24863 (needs_fold_left_reduction_p): New function, split out from...
24864 (vect_is_simple_reduction): ...here. Accept reductions that
24865 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
24866 (vect_force_simple_reduction): Also store the reduction type in
24867 the assignment's STMT_VINFO_REDUC_TYPE.
24868 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24869 (merge_with_identity): New function.
24870 (vect_expand_fold_left): Likewise.
24871 (vectorize_fold_left_reduction): Likewise.
24872 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
24873 scalar phi in place for it. Check for target support and reject
24874 cases that would reassociate the operation. Defer the transform
24875 phase to vectorize_fold_left_reduction.
24876 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24877 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24878 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24882 * tree-if-conv.c (predicate_mem_writes): Remove redundant
24883 call to ifc_temp_var.
24885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24886 Alan Hayward <alan.hayward@arm.com>
24887 David Sherwood <david.sherwood@arm.com>
24889 * target.def (legitimize_address_displacement): Take the original
24890 offset as a poly_int.
24891 * targhooks.h (default_legitimize_address_displacement): Update
24893 * targhooks.c (default_legitimize_address_displacement): Likewise.
24894 * doc/tm.texi: Regenerate.
24895 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24896 as an argument, moving assert of ad->disp == ad->disp_term to...
24897 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
24898 Try calling targetm.legitimize_address_displacement before expanding
24899 the address rather than afterwards, and adjust for the new interface.
24900 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24901 Match the new hook interface. Handle SVE addresses.
24902 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24903 new hook interface.
24905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24907 * Makefile.in (OBJS): Add early-remat.o.
24908 * target.def (select_early_remat_modes): New hook.
24909 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24910 * doc/tm.texi: Regenerate.
24911 * targhooks.h (default_select_early_remat_modes): Declare.
24912 * targhooks.c (default_select_early_remat_modes): New function.
24913 * timevar.def (TV_EARLY_REMAT): New timevar.
24914 * passes.def (pass_early_remat): New pass.
24915 * tree-pass.h (make_pass_early_remat): Declare.
24916 * early-remat.c: New file.
24917 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24919 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24922 Alan Hayward <alan.hayward@arm.com>
24923 David Sherwood <david.sherwood@arm.com>
24925 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24926 vfm1 with a bound_epilog parameter.
24927 (vect_do_peeling): Update calls accordingly, and move the prologue
24928 call earlier in the function. Treat the base bound_epilog as 0 for
24929 fully-masked loops and retain vf - 1 for other loops. Add 1 to
24930 this base when peeling for gaps.
24931 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24932 with fully-masked loops.
24933 (vect_estimate_min_profitable_iters): Handle the single peeled
24934 iteration in that case.
24936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24937 Alan Hayward <alan.hayward@arm.com>
24938 David Sherwood <david.sherwood@arm.com>
24940 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24941 single-element interleaving even if the size is not a power of 2.
24942 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24943 accesses for single-element interleaving if the group size is
24946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24947 Alan Hayward <alan.hayward@arm.com>
24948 David Sherwood <david.sherwood@arm.com>
24950 * doc/md.texi (fold_extract_last_@var{m}): Document.
24951 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24952 * optabs.def (fold_extract_last_optab): New optab.
24953 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24954 * internal-fn.c (fold_extract_direct): New macro.
24955 (expand_fold_extract_optab_fn): Likewise.
24956 (direct_fold_extract_optab_supported_p): Likewise.
24957 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24958 * tree-vect-loop.c (vect_model_reduction_cost): Handle
24959 EXTRACT_LAST_REDUCTION.
24960 (get_initial_def_for_reduction): Do not create an initial vector
24961 for EXTRACT_LAST_REDUCTION reductions.
24962 (vectorizable_reduction): Leave the scalar phi in place for
24963 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
24964 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
24965 epilogue code for EXTRACT_LAST_REDUCTION and defer the
24966 transform phase to vectorizable_condition.
24967 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24969 (vect_finish_stmt_generation): ...here.
24970 (vect_finish_replace_stmt): New function.
24971 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24972 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24974 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24977 Alan Hayward <alan.hayward@arm.com>
24978 David Sherwood <david.sherwood@arm.com>
24980 * doc/md.texi (extract_last_@var{m}): Document.
24981 * optabs.def (extract_last_optab): New optab.
24982 * internal-fn.def (EXTRACT_LAST): New internal function.
24983 * internal-fn.c (cond_unary_direct): New macro.
24984 (expand_cond_unary_optab_fn): Likewise.
24985 (direct_cond_unary_optab_supported_p): Likewise.
24986 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24987 loops using EXTRACT_LAST.
24988 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24989 (extract_last_<mode>): ...this optab.
24990 (vec_extract<mode><Vel>): Update accordingly.
24992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24993 Alan Hayward <alan.hayward@arm.com>
24994 David Sherwood <david.sherwood@arm.com>
24996 * target.def (empty_mask_is_expensive): New hook.
24997 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24998 * doc/tm.texi: Regenerate.
24999 * targhooks.h (default_empty_mask_is_expensive): Declare.
25000 * targhooks.c (default_empty_mask_is_expensive): New function.
25001 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25002 if the target says that empty masks are expensive.
25003 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25005 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25008 Alan Hayward <alan.hayward@arm.com>
25009 David Sherwood <david.sherwood@arm.com>
25011 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25012 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25013 (vect_use_loop_mask_for_alignment_p): New function.
25014 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25015 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25016 niters_skip argument. Make sure that the first niters_skip elements
25017 of the first iteration are inactive.
25018 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25019 Update call to vect_set_loop_masks_directly.
25020 (get_misalign_in_elems): New function, split out from...
25021 (vect_gen_prolog_loop_niters): ...here.
25022 (vect_update_init_of_dr): Take a code argument that specifies whether
25023 the adjustment should be added or subtracted.
25024 (vect_update_init_of_drs): Likewise.
25025 (vect_prepare_for_masked_peels): New function.
25026 (vect_do_peeling): Skip prologue peeling if we're using a mask
25027 instead. Update call to vect_update_inits_of_drs.
25028 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25030 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25031 alignment. Do not include the number of peeled iterations in
25032 the minimum threshold in that case.
25033 (vectorizable_induction): Adjust the start value down by
25034 LOOP_VINFO_MASK_SKIP_NITERS iterations.
25035 (vect_transform_loop): Call vect_prepare_for_masked_peels.
25036 Take the number of skipped iterations into account when calculating
25038 * tree-vect-stmts.c (vect_gen_while_not): New function.
25040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25041 Alan Hayward <alan.hayward@arm.com>
25042 David Sherwood <david.sherwood@arm.com>
25044 * doc/sourcebuild.texi (vect_fully_masked): Document.
25045 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
25046 default value to 0.
25047 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
25049 (vect_analyze_loop_2): ...here. Don't check the vectorization
25050 factor against the number of loop iterations if the loop is
25053 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25054 Alan Hayward <alan.hayward@arm.com>
25055 David Sherwood <david.sherwood@arm.com>
25057 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
25058 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
25059 (dump_groups): Update accordingly.
25060 (iv_use::mem_type): New member variable.
25061 (address_p): New function.
25062 (record_use): Add a mem_type argument and initialize the new
25064 (record_group_use): Add a mem_type argument. Use address_p.
25065 Remove obsolete null checks of base_object. Update call to record_use.
25066 (find_interesting_uses_op): Update call to record_group_use.
25067 (find_interesting_uses_cond): Likewise.
25068 (find_interesting_uses_address): Likewise.
25069 (get_mem_type_for_internal_fn): New function.
25070 (find_address_like_use): Likewise.
25071 (find_interesting_uses_stmt): Try find_address_like_use before
25072 calling find_interesting_uses_op.
25073 (addr_offset_valid_p): Use the iv mem_type field as the type
25074 of the addressed memory.
25075 (add_autoinc_candidates): Likewise.
25076 (get_address_cost): Likewise.
25077 (split_small_address_groups_p): Use address_p.
25078 (split_address_groups): Likewise.
25079 (add_iv_candidate_for_use): Likewise.
25080 (autoinc_possible_for_pair): Likewise.
25081 (rewrite_groups): Likewise.
25082 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
25083 (determine_group_iv_cost): Update after split of USE_ADDRESS.
25084 (get_alias_ptr_type_for_ptr_address): New function.
25085 (rewrite_use_address): Rewrite address uses in calls that were
25086 identified by find_address_like_use.
25088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25089 Alan Hayward <alan.hayward@arm.com>
25090 David Sherwood <david.sherwood@arm.com>
25092 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
25094 * gimple-expr.h (is_gimple_addressable: Likewise.
25095 * gimple-expr.c (is_gimple_address): Likewise.
25096 * internal-fn.c (expand_call_mem_ref): New function.
25097 (expand_mask_load_optab_fn): Use it.
25098 (expand_mask_store_optab_fn): Likewise.
25100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25101 Alan Hayward <alan.hayward@arm.com>
25102 David Sherwood <david.sherwood@arm.com>
25104 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
25105 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
25106 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
25107 (cond_umax@var{mode}): Document.
25108 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
25109 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
25110 (cond_umin_optab, cond_umax_optab): New optabs.
25111 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
25112 (COND_IOR, COND_XOR): New internal functions.
25113 * internal-fn.h (get_conditional_internal_fn): Declare.
25114 * internal-fn.c (cond_binary_direct): New macro.
25115 (expand_cond_binary_optab_fn): Likewise.
25116 (direct_cond_binary_optab_supported_p): Likewise.
25117 (get_conditional_internal_fn): New function.
25118 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
25119 Cope with reduction statements that are vectorized as calls rather
25121 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
25122 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
25123 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
25124 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
25125 (UNSPEC_COND_EOR): New unspecs.
25126 (optab): Add mappings for them.
25127 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
25128 (sve_int_op, sve_fp_op): New int attributes.
25130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25131 Alan Hayward <alan.hayward@arm.com>
25132 David Sherwood <david.sherwood@arm.com>
25134 * optabs.def (while_ult_optab): New optab.
25135 * doc/md.texi (while_ult@var{m}@var{n}): Document.
25136 * internal-fn.def (WHILE_ULT): New internal function.
25137 * internal-fn.h (direct_internal_fn_supported_p): New override
25138 that takes two types as argument.
25139 * internal-fn.c (while_direct): New macro.
25140 (expand_while_optab_fn): New function.
25141 (convert_optab_supported_p): Likewise.
25142 (direct_while_optab_supported_p): New macro.
25143 * wide-int.h (wi::udiv_ceil): New function.
25144 * tree-vectorizer.h (rgroup_masks): New structure.
25145 (vec_loop_masks): New typedef.
25146 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
25147 and fully_masked_p.
25148 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
25149 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
25150 (vect_max_vf): New function.
25151 (slpeel_make_loop_iterate_ntimes): Delete.
25152 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
25153 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
25154 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
25155 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
25156 internal-fn.h, stor-layout.h and optabs-query.h.
25157 (vect_set_loop_mask): New function.
25158 (add_preheader_seq): Likewise.
25159 (add_header_seq): Likewise.
25160 (interleave_supported_p): Likewise.
25161 (vect_maybe_permute_loop_masks): Likewise.
25162 (vect_set_loop_masks_directly): Likewise.
25163 (vect_set_loop_condition_masked): Likewise.
25164 (vect_set_loop_condition_unmasked): New function, split out from
25165 slpeel_make_loop_iterate_ntimes.
25166 (slpeel_make_loop_iterate_ntimes): Rename to..
25167 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
25168 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
25169 (vect_do_peeling): Update call accordingly.
25170 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
25172 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25173 mask_compare_type, can_fully_mask_p and fully_masked_p.
25174 (release_vec_loop_masks): New function.
25175 (_loop_vec_info): Use it to free the loop masks.
25176 (can_produce_all_loop_masks_p): New function.
25177 (vect_get_max_nscalars_per_iter): Likewise.
25178 (vect_verify_full_masking): Likewise.
25179 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
25180 retries, and free the mask rgroups before retrying. Check loop-wide
25181 reasons for disallowing fully-masked loops. Make the final decision
25182 about whether use a fully-masked loop or not.
25183 (vect_estimate_min_profitable_iters): Do not assume that peeling
25184 for the number of iterations will be needed for fully-masked loops.
25185 (vectorizable_reduction): Disable fully-masked loops.
25186 (vectorizable_live_operation): Likewise.
25187 (vect_halve_mask_nunits): New function.
25188 (vect_double_mask_nunits): Likewise.
25189 (vect_record_loop_mask): Likewise.
25190 (vect_get_loop_mask): Likewise.
25191 (vect_transform_loop): Handle the case in which the final loop
25192 iteration might handle a partial vector. Call vect_set_loop_condition
25193 instead of slpeel_make_loop_iterate_ntimes.
25194 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
25195 (check_load_store_masking): New function.
25196 (prepare_load_store_mask): Likewise.
25197 (vectorizable_store): Handle fully-masked loops.
25198 (vectorizable_load): Likewise.
25199 (supportable_widening_operation): Use vect_halve_mask_nunits for
25201 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
25202 (vect_gen_while): New function.
25203 * config/aarch64/aarch64.md (umax<mode>3): New expander.
25204 (aarch64_uqdec<mode>): New insn.
25206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25207 Alan Hayward <alan.hayward@arm.com>
25208 David Sherwood <david.sherwood@arm.com>
25210 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
25211 (reduc_xor_scal_optab): New optabs.
25212 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
25213 (reduc_xor_scal_@var{m}): Document.
25214 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
25215 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
25216 internal functions.
25217 * fold-const-call.c (fold_const_call): Handle them.
25218 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
25219 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
25220 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
25221 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
25222 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
25223 (UNSPEC_XORV): New unspecs.
25224 (optab): Add entries for them.
25225 (BITWISEV): New int iterator.
25226 (bit_reduc_op): New int attributes.
25228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25229 Alan Hayward <alan.hayward@arm.com>
25230 David Sherwood <david.sherwood@arm.com>
25232 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
25233 * internal-fn.def (VEC_SHL_INSERT): New internal function.
25234 * optabs.def (vec_shl_insert_optab): New optab.
25235 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
25236 (duplicate_and_interleave): Likewise.
25237 * tree-vect-loop.c: Include internal-fn.h.
25238 (neutral_op_for_slp_reduction): New function, split out from
25239 get_initial_defs_for_reduction.
25240 (get_initial_def_for_reduction): Handle option 2 for variable-length
25241 vectors by loading the neutral value into a vector and then shifting
25242 the initial value into element 0.
25243 (get_initial_defs_for_reduction): Replace the code argument with
25244 the neutral value calculated by neutral_op_for_slp_reduction.
25245 Use gimple_build_vector for constant-length vectors.
25246 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
25247 but the first group_size elements have a neutral value.
25248 Use duplicate_and_interleave otherwise.
25249 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
25250 Update call to get_initial_defs_for_reduction. Handle SLP
25251 reductions for variable-length vectors by creating one vector
25252 result for each scalar result, with the elements associated
25253 with other scalar results stubbed out with the neutral value.
25254 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
25255 Require IFN_VEC_SHL_INSERT for double reductions on
25256 variable-length vectors, or SLP reductions that have
25257 a neutral value. Require can_duplicate_and_interleave_p
25258 support for variable-length unchained SLP reductions if there
25259 is no neutral value, such as for MIN/MAX reductions. Also require
25260 the number of vector elements to be a multiple of the number of
25261 SLP statements when doing variable-length unchained SLP reductions.
25262 Update call to vect_create_epilog_for_reduction.
25263 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
25264 and remove initial values.
25265 (duplicate_and_interleave): Make public.
25266 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
25267 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
25269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25270 Alan Hayward <alan.hayward@arm.com>
25271 David Sherwood <david.sherwood@arm.com>
25273 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
25274 (can_duplicate_and_interleave_p): New function.
25275 (vect_get_and_check_slp_defs): Take the vector of statements
25276 rather than just the current one. Remove excess parentheses.
25277 Restriction rejectinon of vect_constant_def and vect_external_def
25278 for variable-length vectors to boolean types, or types for which
25279 can_duplicate_and_interleave_p is false.
25280 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
25281 (duplicate_and_interleave): New function.
25282 (vect_get_constant_vectors): Use gimple_build_vector for
25283 constant-length vectors and suitable variable-length constant
25284 vectors. Use duplicate_and_interleave for other variable-length
25285 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
25292 min_profitable_iters doesn't go negative.
25294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25295 Alan Hayward <alan.hayward@arm.com>
25296 David Sherwood <david.sherwood@arm.com>
25298 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
25299 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
25300 * optabs.def (vec_mask_load_lanes_optab): New optab.
25301 (vec_mask_store_lanes_optab): Likewise.
25302 * internal-fn.def (MASK_LOAD_LANES): New internal function.
25303 (MASK_STORE_LANES): Likewise.
25304 * internal-fn.c (mask_load_lanes_direct): New macro.
25305 (mask_store_lanes_direct): Likewise.
25306 (expand_mask_load_optab_fn): Handle masked operations.
25307 (expand_mask_load_lanes_optab_fn): New macro.
25308 (expand_mask_store_optab_fn): Handle masked operations.
25309 (expand_mask_store_lanes_optab_fn): New macro.
25310 (direct_mask_load_lanes_optab_supported_p): Likewise.
25311 (direct_mask_store_lanes_optab_supported_p): Likewise.
25312 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
25314 (vect_load_lanes_supported): Likewise.
25315 * tree-vect-data-refs.c (strip_conversion): New function.
25316 (can_group_stmts_p): Likewise.
25317 (vect_analyze_data_ref_accesses): Use it instead of checking
25318 for a pair of assignments.
25319 (vect_store_lanes_supported): Take a masked_p parameter.
25320 (vect_load_lanes_supported): Likewise.
25321 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
25322 vect_store_lanes_supported and vect_load_lanes_supported.
25323 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
25324 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
25325 parameter. Don't allow gaps for masked accesses.
25326 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
25327 and vect_load_lanes_supported.
25328 (get_load_store_type): Take a masked_p parameter and update
25329 call to get_group_load_store_type.
25330 (vectorizable_store): Update call to get_load_store_type.
25331 Handle IFN_MASK_STORE_LANES.
25332 (vectorizable_load): Update call to get_load_store_type.
25333 Handle IFN_MASK_LOAD_LANES.
25335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25336 Alan Hayward <alan.hayward@arm.com>
25337 David Sherwood <david.sherwood@arm.com>
25339 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
25341 * config/aarch64/aarch64-protos.h
25342 (aarch64_sve_struct_memory_operand_p): Declare.
25343 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
25344 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
25345 (VPRED, vpred): Handle SVE structure modes.
25346 * config/aarch64/constraints.md (Utx): New constraint.
25347 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
25348 (aarch64_sve_struct_nonimmediate_operand): New predicates.
25349 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
25350 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
25351 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
25352 structure modes. Split into pieces after RA.
25353 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
25354 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
25356 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
25357 SVE structure modes.
25358 (aarch64_classify_address): Likewise.
25359 (sizetochar): Move earlier in file.
25360 (aarch64_print_operand): Handle SVE register lists.
25361 (aarch64_array_mode): New function.
25362 (aarch64_sve_struct_memory_operand_p): Likewise.
25363 (TARGET_ARRAY_MODE): Redefine.
25365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25366 Alan Hayward <alan.hayward@arm.com>
25367 David Sherwood <david.sherwood@arm.com>
25369 * target.def (array_mode): New target hook.
25370 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
25371 * doc/tm.texi: Regenerate.
25372 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
25373 * hooks.c (hook_optmode_mode_uhwi_none): New function.
25374 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
25375 targetm.array_mode.
25376 * stor-layout.c (mode_for_array): Likewise. Support polynomial
25379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25380 Alan Hayward <alan.hayward@arm.com>
25381 David Sherwood <david.sherwood@arm.com>
25383 * fold-const.c (fold_binary_loc): Check the argument types
25384 rather than the result type when testing for a vector operation.
25386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25388 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
25389 * doc/tm.texi: Regenerate.
25391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25392 Alan Hayward <alan.hayward@arm.com>
25393 David Sherwood <david.sherwood@arm.com>
25395 * doc/invoke.texi (-msve-vector-bits=): Document new option.
25396 (sve): Document new AArch64 extension.
25397 * doc/md.texi (w): Extend the description of the AArch64
25398 constraint to include SVE vectors.
25399 (Upl, Upa): Document new AArch64 predicate constraints.
25400 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
25402 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
25403 (msve-vector-bits=): New option.
25404 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
25405 SVE when these are disabled.
25406 (sve): New extension.
25407 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
25408 modes. Adjust their number of units based on aarch64_sve_vg.
25409 (MAX_BITSIZE_MODE_ANY_MODE): Define.
25410 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
25411 aarch64_addr_query_type.
25412 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
25413 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
25414 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
25415 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
25416 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
25417 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
25418 (aarch64_simd_imm_zero_p): Delete.
25419 (aarch64_check_zero_based_sve_index_immediate): Declare.
25420 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25421 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25422 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25423 (aarch64_sve_float_mul_immediate_p): Likewise.
25424 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25425 rather than an rtx.
25426 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
25427 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
25428 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
25429 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
25430 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
25431 (aarch64_regmode_natural_size): Likewise.
25432 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
25433 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
25435 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
25436 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
25437 for VG and the SVE predicate registers.
25438 (V_ALIASES): Add a "z"-prefixed alias.
25439 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
25440 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
25441 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
25442 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
25443 (REG_CLASS_NAMES): Add entries for them.
25444 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
25445 and the predicate registers.
25446 (aarch64_sve_vg): Declare.
25447 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
25448 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
25449 (REGMODE_NATURAL_SIZE): Define.
25450 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
25452 * config/aarch64/aarch64.c: Include cfgrtl.h.
25453 (simd_immediate_info): Add a constructor for series vectors,
25454 and an associated step field.
25455 (aarch64_sve_vg): New variable.
25456 (aarch64_dbx_register_number): Handle VG and the predicate registers.
25457 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
25458 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
25459 (VEC_ANY_DATA, VEC_STRUCT): New constants.
25460 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
25461 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
25462 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
25463 (aarch64_get_mask_mode): New functions.
25464 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
25465 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
25466 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
25467 predicate modes and predicate registers. Explicitly restrict
25468 GPRs to modes of 16 bytes or smaller. Only allow FP registers
25469 to store a vector mode if it is recognized by
25470 aarch64_classify_vector_mode.
25471 (aarch64_regmode_natural_size): New function.
25472 (aarch64_hard_regno_caller_save_mode): Return the original mode
25474 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
25475 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
25476 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
25477 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
25479 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
25480 does not overlap dest if the function is frame-related. Handle
25482 (aarch64_split_add_offset): New function.
25483 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
25484 them aarch64_add_offset.
25485 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
25486 and update call to aarch64_sub_sp.
25487 (aarch64_add_cfa_expression): New function.
25488 (aarch64_expand_prologue): Pass extra temporary registers to the
25489 functions above. Handle the case in which we need to emit new
25490 DW_CFA_expressions for registers that were originally saved
25491 relative to the stack pointer, but now have to be expressed
25492 relative to the frame pointer.
25493 (aarch64_output_mi_thunk): Pass extra temporary registers to the
25495 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
25496 IP0 and IP1 values for SVE frames.
25497 (aarch64_expand_vec_series): New function.
25498 (aarch64_expand_sve_widened_duplicate): Likewise.
25499 (aarch64_expand_sve_const_vector): Likewise.
25500 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
25501 Handle SVE constants. Use emit_move_insn to move a force_const_mem
25502 into the register, rather than emitting a SET directly.
25503 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
25504 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
25505 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
25506 (offset_9bit_signed_scaled_p): New functions.
25507 (aarch64_replicate_bitmask_imm): New function.
25508 (aarch64_bitmask_imm): Use it.
25509 (aarch64_cannot_force_const_mem): Reject expressions involving
25510 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
25511 (aarch64_classify_index): Handle SVE indices, by requiring
25512 a plain register index with a scale that matches the element size.
25513 (aarch64_classify_address): Handle SVE addresses. Assert that
25514 the mode of the address is VOIDmode or an integer mode.
25515 Update call to aarch64_classify_symbol.
25516 (aarch64_classify_symbolic_expression): Update call to
25517 aarch64_classify_symbol.
25518 (aarch64_const_vec_all_in_range_p): New function.
25519 (aarch64_print_vector_float_operand): Likewise.
25520 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
25521 "vN" for FP registers with SVE modes. Handle (const ...) vectors
25522 and the FP immediates 1.0 and 0.5.
25523 (aarch64_print_address_internal): Handle SVE addresses.
25524 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
25525 (aarch64_regno_regclass): Handle predicate registers.
25526 (aarch64_secondary_reload): Handle big-endian reloads of SVE
25528 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
25529 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
25530 (aarch64_convert_sve_vector_bits): New function.
25531 (aarch64_override_options): Use it to handle -msve-vector-bits=.
25532 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
25533 rather than an rtx.
25534 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
25535 Handle SVE vector and predicate modes. Accept VL-based constants
25536 that need only one temporary register, and VL offsets that require
25537 no temporary registers.
25538 (aarch64_conditional_register_usage): Mark the predicate registers
25539 as fixed if SVE isn't available.
25540 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
25541 Return true for SVE vector and predicate modes.
25542 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
25543 rather than an unsigned int. Handle SVE modes.
25544 (aarch64_preferred_simd_mode): Update call accordingly. Handle
25546 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
25548 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
25549 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
25550 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
25551 (aarch64_sve_float_mul_immediate_p): New functions.
25552 (aarch64_sve_valid_immediate): New function.
25553 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
25554 Explicitly reject structure modes. Check for INDEX constants.
25555 Handle PTRUE and PFALSE constants.
25556 (aarch64_check_zero_based_sve_index_immediate): New function.
25557 (aarch64_simd_imm_zero_p): Delete.
25558 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
25559 vector modes. Accept constants in the range of CNT[BHWD].
25560 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
25561 ask for an Advanced SIMD mode.
25562 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
25563 (aarch64_simd_vector_alignment): Handle SVE predicates.
25564 (aarch64_vectorize_preferred_vector_alignment): New function.
25565 (aarch64_simd_vector_alignment_reachable): Use it instead of
25567 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
25568 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
25570 (MAX_VECT_LEN): Delete.
25571 (expand_vec_perm_d): Add a vec_flags field.
25572 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
25573 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
25574 (aarch64_evpc_ext): Don't apply a big-endian lane correction
25576 (aarch64_evpc_rev): Rename to...
25577 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
25578 (aarch64_evpc_rev_global): New function.
25579 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
25580 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
25582 (aarch64_evpc_sve_tbl): New function.
25583 (aarch64_expand_vec_perm_const_1): Update after rename of
25584 aarch64_evpc_rev. Handle SVE permutes too, trying
25585 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
25586 than aarch64_evpc_tbl.
25587 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
25588 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
25589 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
25590 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
25591 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
25592 (aarch64_expand_sve_vcond): New functions.
25593 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
25594 of aarch64_vector_mode_p.
25595 (aarch64_dwarf_poly_indeterminate_value): New function.
25596 (aarch64_compute_pressure_classes): Likewise.
25597 (aarch64_can_change_mode_class): Likewise.
25598 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
25599 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
25600 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
25601 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
25602 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
25603 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
25604 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
25605 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
25607 (Dn, Dl, Dr): Accept const as well as const_vector.
25608 (Dz): Likewise. Compare against CONST0_RTX.
25609 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
25610 of "vector" where appropriate.
25611 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
25612 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
25613 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
25614 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
25615 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
25616 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
25617 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
25618 (v_int_equiv): Extend to SVE modes.
25619 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
25621 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
25622 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
25623 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
25624 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
25625 (SVE_COND_FP_CMP): New int iterators.
25626 (perm_hilo): Handle the new unpack unspecs.
25627 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
25629 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
25630 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
25631 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
25632 (aarch64_equality_operator, aarch64_constant_vector_operand)
25633 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
25634 (aarch64_sve_nonimmediate_operand): Likewise.
25635 (aarch64_sve_general_operand): Likewise.
25636 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
25637 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
25638 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
25639 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
25640 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
25641 (aarch64_sve_float_arith_immediate): Likewise.
25642 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
25643 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
25644 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
25645 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
25646 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
25647 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
25648 (aarch64_sve_float_arith_operand): Likewise.
25649 (aarch64_sve_float_arith_with_sub_operand): Likewise.
25650 (aarch64_sve_float_mul_operand): Likewise.
25651 (aarch64_sve_vec_perm_operand): Likewise.
25652 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
25653 (aarch64_mov_operand): Accept const_poly_int and const_vector.
25654 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
25655 as well as const_vector.
25656 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
25657 in file. Use CONST0_RTX and CONSTM1_RTX.
25658 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
25659 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
25660 Use aarch64_simd_imm_zero.
25661 * config/aarch64/aarch64-sve.md: New file.
25662 * config/aarch64/aarch64.md: Include it.
25663 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
25664 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
25665 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
25666 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
25667 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
25668 (sve): New attribute.
25669 (enabled): Disable instructions with the sve attribute unless
25671 (movqi, movhi): Pass CONST_POLY_INT operaneds through
25672 aarch64_expand_mov_immediate.
25673 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
25674 CNT[BHSD] immediates.
25675 (movti): Split CONST_POLY_INT moves into two halves.
25676 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
25677 Split additions that need a temporary here if the destination
25678 is the stack pointer.
25679 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
25680 (*add<mode>3_poly_1): New instruction.
25681 (set_clobber_cc): New expander.
25683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25685 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
25686 parameter and use it instead of GET_MODE_SIZE (innermode). Use
25687 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
25688 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
25689 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
25690 Change innermode from fixed_mode_size to machine_mode.
25691 (simplify_subreg): Update call accordingly. Handle a constant-sized
25692 subreg of a variable-length CONST_VECTOR.
25694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25695 Alan Hayward <alan.hayward@arm.com>
25696 David Sherwood <david.sherwood@arm.com>
25698 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
25699 (add_offset_to_base): New function, split out from...
25700 (create_mem_ref): ...here. When handling a scale other than 1,
25701 check first whether the address is valid without the offset.
25702 Add it into the base if so, leaving the index and scale as-is.
25704 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25707 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
25708 fold_for_warn before checking if arg2 is INTEGER_CST.
25710 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
25712 * config/rs6000/predicates.md (load_multiple_operation): Delete.
25713 (store_multiple_operation): Delete.
25714 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
25715 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
25716 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
25717 guarded by TARGET_STRING.
25718 (rs6000_output_load_multiple): Delete.
25719 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
25720 OPTION_MASK_STRING / TARGET_STRING handling.
25721 (print_operand) <'N', 'O'>: Add comment that these are unused now.
25722 (const rs6000_opt_masks) <"string">: Change mask to 0.
25723 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
25724 (MASK_STRING): Delete.
25725 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
25727 (load_multiple): Delete.
25734 (store_multiple): Delete.
25741 (movmemsi_8reg): Delete.
25742 (corresponding unnamed define_insn): Delete.
25743 (movmemsi_6reg): Delete.
25744 (corresponding unnamed define_insn): Delete.
25745 (movmemsi_4reg): Delete.
25746 (corresponding unnamed define_insn): Delete.
25747 (movmemsi_2reg): Delete.
25748 (corresponding unnamed define_insn): Delete.
25749 (movmemsi_1reg): Delete.
25750 (corresponding unnamed define_insn): Delete.
25751 * config/rs6000/rs6000.opt (mno-string): New.
25752 (mstring): Replace by deprecation warning stub.
25753 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25755 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25757 * regrename.c (regrename_do_replace): If replacing the same
25758 reg multiple times, try to reuse last created gen_raw_REG.
25761 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25762 main to workaround a bug in GDB.
25764 2018-01-12 Tom de Vries <tom@codesourcery.com>
25767 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25769 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
25771 PR rtl-optimization/80481
25772 * ira-color.c (get_cap_member): New function.
25773 (allocnos_conflict_by_live_ranges_p): Use it.
25774 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25775 (setup_slot_coalesced_allocno_live_ranges): Ditto.
25777 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
25780 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25781 (*saddl_se_1): Ditto.
25782 (*ssubsi_1): Ditto.
25783 (*ssubl_se_1): Ditto.
25785 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25787 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25788 rather than wi::to_widest for DR_INITs.
25789 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25790 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25791 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25793 (vect_analyze_group_access_1): Note that here.
25795 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25797 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25798 polynomial type sizes.
25800 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25802 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25803 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25804 (gimple_add_tmp_var): Likewise.
25806 2018-01-12 Martin Liska <mliska@suse.cz>
25808 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25809 (gimple_alloc_sizes): Likewise.
25810 (dump_gimple_statistics): Use PRIu64 in printf format.
25811 * gimple.h: Change uint64_t to int.
25813 2018-01-12 Martin Liska <mliska@suse.cz>
25815 * tree-core.h: Use uint64_t instead of int.
25816 * tree.c (tree_node_counts): Likewise.
25817 (tree_node_sizes): Likewise.
25818 (dump_tree_statistics): Use PRIu64 in printf format.
25820 2018-01-12 Martin Liska <mliska@suse.cz>
25822 * Makefile.in: As qsort_chk is implemented in vec.c, add
25823 vec.o to linkage of gencfn-macros.
25824 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25825 passing the info to record_node_allocation_statistics.
25826 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25828 * ggc-common.c (struct ggc_usage): Add operator== and use
25829 it in operator< and compare function.
25830 * mem-stats.h (struct mem_usage): Likewise.
25831 * vec.c (struct vec_usage): Remove operator< and compare
25832 function. Can be simply inherited.
25834 2018-01-12 Martin Jambor <mjambor@suse.cz>
25837 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
25838 * tree-ssa-math-opts.c: Include domwalk.h.
25839 (convert_mult_to_fma_1): New function.
25840 (fma_transformation_info): New type.
25841 (fma_deferring_state): Likewise.
25842 (cancel_fma_deferring): New function.
25843 (result_of_phi): Likewise.
25844 (last_fma_candidate_feeds_initial_phi): Likewise.
25845 (convert_mult_to_fma): Added deferring logic, split actual
25846 transformation to convert_mult_to_fma_1.
25847 (math_opts_dom_walker): New type.
25848 (math_opts_dom_walker::after_dom_children): New method, body moved
25849 here from pass_optimize_widening_mul::execute, added deferring logic
25851 (pass_optimize_widening_mul::execute): Moved most of code to
25852 math_opts_dom_walker::after_dom_children.
25853 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
25854 * config/i386/i386.c (ix86_option_override_internal): Added
25855 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
25857 2018-01-12 Richard Biener <rguenther@suse.de>
25860 * dwarf2out.c (gen_variable_die): Do not reset old_die for
25861 inline instance vars.
25863 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
25866 * config/rx/rx.c (rx_is_restricted_memory_address):
25867 Handle SUBREG case.
25869 2018-01-12 Richard Biener <rguenther@suse.de>
25871 PR tree-optimization/80846
25872 * target.def (split_reduction): New target hook.
25873 * targhooks.c (default_split_reduction): New function.
25874 * targhooks.h (default_split_reduction): Declare.
25875 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25876 target requests first reduce vectors by combining low and high
25878 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25879 (get_vectype_for_scalar_type_and_size): Export.
25880 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25881 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25882 * doc/tm.texi: Regenerate.
25883 * config/i386/i386.c (ix86_split_reduction): Implement
25884 TARGET_VECTORIZE_SPLIT_REDUCTION.
25886 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25889 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25890 in PIC mode except for TARGET_VXWORKS_RTP.
25891 * config/sparc/sparc.c: Include cfgrtl.h.
25892 (TARGET_INIT_PIC_REG): Define.
25893 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25894 (sparc_pic_register_p): New predicate.
25895 (sparc_legitimate_address_p): Use it.
25896 (sparc_legitimize_pic_address): Likewise.
25897 (sparc_delegitimize_address): Likewise.
25898 (sparc_mode_dependent_address_p): Likewise.
25899 (gen_load_pcrel_sym): Remove 4th parameter.
25900 (load_got_register): Adjust call to above. Remove obsolete stuff.
25901 (sparc_expand_prologue): Do not call load_got_register here.
25902 (sparc_flat_expand_prologue): Likewise.
25903 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25904 (sparc_use_pseudo_pic_reg): New function.
25905 (sparc_init_pic_reg): Likewise.
25906 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25907 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25909 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25911 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25912 Add item for branch_cost.
25914 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25916 PR rtl-optimization/83565
25917 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25918 not extend the result to a larger mode for rotate operations.
25919 (num_sign_bit_copies1): Likewise.
25921 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25924 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25926 Use values-Xc.o for -pedantic.
25927 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25929 2018-01-12 Martin Liska <mliska@suse.cz>
25932 * ipa-devirt.c (final_warning_record::grow_type_warnings):
25934 (possible_polymorphic_call_targets): Use it.
25935 (ipa_devirt): Likewise.
25937 2018-01-12 Martin Liska <mliska@suse.cz>
25939 * profile-count.h (enum profile_quality): Use 0 as invalid
25940 enum value of profile_quality.
25942 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
25944 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25945 -mext-string options.
25947 2018-01-12 Richard Biener <rguenther@suse.de>
25949 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25950 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25951 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25953 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25955 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
25957 * configure.ac (--with-long-double-format): Add support for the
25958 configuration option to change the default long double format on
25960 * config.gcc (powerpc*-linux*-*): Likewise.
25961 * configure: Regenerate.
25962 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25963 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25964 used without modification.
25966 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25968 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25969 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25970 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25971 MISC_BUILTIN_SPEC_BARRIER.
25972 (rs6000_init_builtins): Likewise.
25973 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25975 (speculation_barrier): New define_insn.
25976 * doc/extend.texi: Document __builtin_speculation_barrier.
25978 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25981 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25982 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25983 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25985 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
25986 integral modes instead of "ss" and "sd".
25987 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25988 vectors with 32-bit and 64-bit elements.
25989 (vecdupssescalarmodesuffix): New mode attribute.
25990 (vec_dup<mode>): Use it.
25992 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
25995 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25996 frame if argument is passed on stack.
25998 2018-01-11 Jakub Jelinek <jakub@redhat.com>
26001 * ree.c (combine_reaching_defs): Optimize also
26002 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26003 reg2=any_extend(exp); reg1=reg2;, formatting fix.
26005 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
26007 PR middle-end/83189
26008 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26010 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
26012 PR middle-end/83718
26013 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26014 after they are computed.
26016 2018-01-11 Bin Cheng <bin.cheng@arm.com>
26018 PR tree-optimization/83695
26019 * gimple-loop-linterchange.cc
26020 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26021 reset cached scev information after interchange.
26022 (pass_linterchange::execute): Remove call to scev_reset_htab.
26024 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26026 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26027 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26028 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26029 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26030 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26031 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26032 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26033 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26034 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26035 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26036 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26037 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26038 (V_lane_reg): Likewise.
26039 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26041 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26042 (vfmal_lane_low<mode>_intrinsic,
26043 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26044 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
26045 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
26046 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
26047 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
26048 vfmsl_lane_high<mode>_intrinsic): New define_insns.
26050 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26052 * config/arm/arm-cpus.in (fp16fml): New feature.
26053 (ALL_SIMD): Add fp16fml.
26054 (armv8.2-a): Add fp16fml as an option.
26055 (armv8.3-a): Likewise.
26056 (armv8.4-a): Add fp16fml as part of fp16.
26057 * config/arm/arm.h (TARGET_FP16FML): Define.
26058 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
26060 * config/arm/arm-modes.def (V2HF): Define.
26061 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
26062 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
26063 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
26064 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
26065 vfmsl_low, vfmsl_high): New set of builtins.
26066 * config/arm/iterators.md (PLUSMINUS): New code iterator.
26067 (vfml_op): New code attribute.
26068 (VFMLHALVES): New int iterator.
26069 (VFML, VFMLSEL): New mode attributes.
26070 (V_reg): Define mapping for V2HF.
26071 (V_hi, V_lo): New mode attributes.
26072 (VF_constraint): Likewise.
26073 (vfml_half, vfml_half_selector): New int attributes.
26074 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
26076 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
26077 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
26079 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
26080 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
26081 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
26082 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
26084 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
26085 Document new effective target and option set.
26087 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26089 * config/arm/arm-cpus.in (armv8_4): New feature.
26090 (ARMv8_4a): New fgroup.
26091 (armv8.4-a): New arch.
26092 * config/arm/arm-tables.opt: Regenerate.
26093 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
26094 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
26095 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
26096 Add matching rules for -march=armv8.4-a and extensions.
26097 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
26099 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
26102 * config/rx/rx.md (BW): New mode attribute.
26103 (sync_lock_test_and_setsi): Add mode suffix to insn output.
26105 2018-01-11 Richard Biener <rguenther@suse.de>
26107 PR tree-optimization/83435
26108 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
26109 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
26110 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
26112 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26113 Alan Hayward <alan.hayward@arm.com>
26114 David Sherwood <david.sherwood@arm.com>
26116 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
26118 (aarch64_classify_address): Initialize it. Track polynomial offsets.
26119 (aarch64_print_address_internal): Use it to check for a zero offset.
26121 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26122 Alan Hayward <alan.hayward@arm.com>
26123 David Sherwood <david.sherwood@arm.com>
26125 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
26126 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
26127 Return a poly_int64 rather than a HOST_WIDE_INT.
26128 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
26129 rather than a HOST_WIDE_INT.
26130 * config/aarch64/aarch64.h (aarch64_frame): Protect with
26131 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
26132 hard_fp_offset, frame_size, initial_adjust, callee_offset and
26133 final_offset from HOST_WIDE_INT to poly_int64.
26134 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
26135 to_constant when getting the number of units in an Advanced SIMD
26137 (aarch64_builtin_vectorized_function): Check for a constant number
26139 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
26141 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
26142 attribute instead of GET_MODE_NUNITS.
26143 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
26144 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
26145 GET_MODE_SIZE for fixed-size registers.
26146 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
26147 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
26148 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
26149 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
26150 (aarch64_print_operand, aarch64_print_address_internal)
26151 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
26152 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
26153 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
26154 Handle polynomial GET_MODE_SIZE.
26155 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
26156 wider than SImode without modification.
26157 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
26158 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
26159 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
26160 passing and returning SVE modes.
26161 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
26162 rather than GEN_INT.
26163 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
26164 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
26165 (aarch64_allocate_and_probe_stack_space): Likewise.
26166 (aarch64_layout_frame): Cope with polynomial offsets.
26167 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
26168 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
26169 polynomial offsets.
26170 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
26171 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
26172 poly_int64 rather than a HOST_WIDE_INT.
26173 (aarch64_get_separate_components, aarch64_process_components)
26174 (aarch64_expand_prologue, aarch64_expand_epilogue)
26175 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
26176 (aarch64_anchor_offset): New function, split out from...
26177 (aarch64_legitimize_address): ...here.
26178 (aarch64_builtin_vectorization_cost): Handle polynomial
26179 TYPE_VECTOR_SUBPARTS.
26180 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
26182 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
26183 number of elements from the PARALLEL rather than the mode.
26184 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
26185 rather than GET_MODE_BITSIZE.
26186 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
26187 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
26188 (aarch64_expand_vec_perm_const_1): Handle polynomial
26189 d->perm.length () and d->perm elements.
26190 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
26191 Apply to_constant to d->perm elements.
26192 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
26193 polynomial CONST_VECTOR_NUNITS.
26194 (aarch64_move_pointer): Take amount as a poly_int64 rather
26196 (aarch64_progress_pointer): Avoid temporary variable.
26197 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
26198 the mode attribute instead of GET_MODE.
26200 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26201 Alan Hayward <alan.hayward@arm.com>
26202 David Sherwood <david.sherwood@arm.com>
26204 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
26205 x exists before using it.
26206 (aarch64_add_constant_internal): Rename to...
26207 (aarch64_add_offset_1): ...this. Replace regnum with separate
26208 src and dest rtxes. Handle the case in which they're different,
26209 including when the offset is zero. Replace scratchreg with an rtx.
26210 Use 2 additions if there is no spare register into which we can
26211 move a 16-bit constant.
26212 (aarch64_add_constant): Delete.
26213 (aarch64_add_offset): Replace reg with separate src and dest
26214 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
26215 Use aarch64_add_offset_1.
26216 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
26217 an rtx rather than an int. Take the delta as a poly_int64
26218 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
26219 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
26220 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
26221 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
26222 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
26223 and aarch64_add_sp.
26224 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
26225 aarch64_add_constant.
26227 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26229 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
26230 Use scalar_float_mode.
26232 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26234 * config/aarch64/aarch64-simd.md
26235 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
26236 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
26237 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
26238 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
26239 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
26240 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
26241 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
26242 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
26243 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
26244 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
26246 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26249 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
26250 targ_options->x_arm_arch_string is non NULL.
26252 2018-01-11 Tamar Christina <tamar.christina@arm.com>
26254 * config/aarch64/aarch64.h
26255 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
26257 2018-01-11 Sudakshina Das <sudi.das@arm.com>
26260 * expmed.c (emit_store_flag_force): Swap if const op0
26261 and change VOIDmode to mode of op0.
26263 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
26265 PR rtl-optimization/83761
26266 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
26267 than bytes to mode_for_size.
26269 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
26271 PR middle-end/83189
26272 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
26273 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
26276 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
26278 PR middle-end/83575
26279 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
26280 when in layout mode.
26281 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
26282 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
26285 2018-01-10 Michael Collison <michael.collison@arm.com>
26287 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
26288 * config/aarch64/aarch64-option-extension.def: Add
26289 AARCH64_OPT_EXTENSION of 'fp16fml'.
26290 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26291 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
26292 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
26293 * config/aarch64/constraints.md (Ui7): New constraint.
26294 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
26295 (VFMLA_SEL_W): Ditto.
26298 (VFMLA16_LOW): New int iterator.
26299 (VFMLA16_HIGH): Ditto.
26300 (UNSPEC_FMLAL): New unspec.
26301 (UNSPEC_FMLSL): Ditto.
26302 (UNSPEC_FMLAL2): Ditto.
26303 (UNSPEC_FMLSL2): Ditto.
26304 (f16mac): New code attribute.
26305 * config/aarch64/aarch64-simd-builtins.def
26306 (aarch64_fmlal_lowv2sf): Ditto.
26307 (aarch64_fmlsl_lowv2sf): Ditto.
26308 (aarch64_fmlalq_lowv4sf): Ditto.
26309 (aarch64_fmlslq_lowv4sf): Ditto.
26310 (aarch64_fmlal_highv2sf): Ditto.
26311 (aarch64_fmlsl_highv2sf): Ditto.
26312 (aarch64_fmlalq_highv4sf): Ditto.
26313 (aarch64_fmlslq_highv4sf): Ditto.
26314 (aarch64_fmlal_lane_lowv2sf): Ditto.
26315 (aarch64_fmlsl_lane_lowv2sf): Ditto.
26316 (aarch64_fmlal_laneq_lowv2sf): Ditto.
26317 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
26318 (aarch64_fmlalq_lane_lowv4sf): Ditto.
26319 (aarch64_fmlsl_lane_lowv4sf): Ditto.
26320 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
26321 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
26322 (aarch64_fmlal_lane_highv2sf): Ditto.
26323 (aarch64_fmlsl_lane_highv2sf): Ditto.
26324 (aarch64_fmlal_laneq_highv2sf): Ditto.
26325 (aarch64_fmlsl_laneq_highv2sf): Ditto.
26326 (aarch64_fmlalq_lane_highv4sf): Ditto.
26327 (aarch64_fmlsl_lane_highv4sf): Ditto.
26328 (aarch64_fmlalq_laneq_highv4sf): Ditto.
26329 (aarch64_fmlsl_laneq_highv4sf): Ditto.
26330 * config/aarch64/aarch64-simd.md:
26331 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
26332 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26333 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
26334 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
26335 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
26336 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
26337 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
26338 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
26339 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
26340 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
26341 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
26342 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
26343 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
26344 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
26345 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
26346 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
26347 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
26348 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
26349 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
26350 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
26351 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
26352 (vfmlsl_low_u32): Ditto.
26353 (vfmlalq_low_u32): Ditto.
26354 (vfmlslq_low_u32): Ditto.
26355 (vfmlal_high_u32): Ditto.
26356 (vfmlsl_high_u32): Ditto.
26357 (vfmlalq_high_u32): Ditto.
26358 (vfmlslq_high_u32): Ditto.
26359 (vfmlal_lane_low_u32): Ditto.
26360 (vfmlsl_lane_low_u32): Ditto.
26361 (vfmlal_laneq_low_u32): Ditto.
26362 (vfmlsl_laneq_low_u32): Ditto.
26363 (vfmlalq_lane_low_u32): Ditto.
26364 (vfmlslq_lane_low_u32): Ditto.
26365 (vfmlalq_laneq_low_u32): Ditto.
26366 (vfmlslq_laneq_low_u32): Ditto.
26367 (vfmlal_lane_high_u32): Ditto.
26368 (vfmlsl_lane_high_u32): Ditto.
26369 (vfmlal_laneq_high_u32): Ditto.
26370 (vfmlsl_laneq_high_u32): Ditto.
26371 (vfmlalq_lane_high_u32): Ditto.
26372 (vfmlslq_lane_high_u32): Ditto.
26373 (vfmlalq_laneq_high_u32): Ditto.
26374 (vfmlslq_laneq_high_u32): Ditto.
26375 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
26376 (AARCH64_FL_FOR_ARCH8_4): New.
26377 (AARCH64_ISA_F16FML): New ISA flag.
26378 (TARGET_F16FML): New feature flag for fp16fml.
26379 (doc/invoke.texi): Document new fp16fml option.
26381 2018-01-10 Michael Collison <michael.collison@arm.com>
26383 * config/aarch64/aarch64-builtins.c:
26384 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
26385 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26386 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
26387 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
26388 (AARCH64_ISA_SHA3): New ISA flag.
26389 (TARGET_SHA3): New feature flag for sha3.
26390 * config/aarch64/iterators.md (sha512_op): New int attribute.
26391 (CRYPTO_SHA512): New int iterator.
26392 (UNSPEC_SHA512H): New unspec.
26393 (UNSPEC_SHA512H2): Ditto.
26394 (UNSPEC_SHA512SU0): Ditto.
26395 (UNSPEC_SHA512SU1): Ditto.
26396 * config/aarch64/aarch64-simd-builtins.def
26397 (aarch64_crypto_sha512hqv2di): New builtin.
26398 (aarch64_crypto_sha512h2qv2di): Ditto.
26399 (aarch64_crypto_sha512su0qv2di): Ditto.
26400 (aarch64_crypto_sha512su1qv2di): Ditto.
26401 (aarch64_eor3qv8hi): Ditto.
26402 (aarch64_rax1qv2di): Ditto.
26403 (aarch64_xarqv2di): Ditto.
26404 (aarch64_bcaxqv8hi): Ditto.
26405 * config/aarch64/aarch64-simd.md:
26406 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
26407 (aarch64_crypto_sha512su0qv2di): Ditto.
26408 (aarch64_crypto_sha512su1qv2di): Ditto.
26409 (aarch64_eor3qv8hi): Ditto.
26410 (aarch64_rax1qv2di): Ditto.
26411 (aarch64_xarqv2di): Ditto.
26412 (aarch64_bcaxqv8hi): Ditto.
26413 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
26414 (vsha512h2q_u64): Ditto.
26415 (vsha512su0q_u64): Ditto.
26416 (vsha512su1q_u64): Ditto.
26417 (veor3q_u16): Ditto.
26418 (vrax1q_u64): Ditto.
26419 (vxarq_u64): Ditto.
26420 (vbcaxq_u16): Ditto.
26421 * config/arm/types.md (crypto_sha512): New type attribute.
26422 (crypto_sha3): Ditto.
26423 (doc/invoke.texi): Document new sha3 option.
26425 2018-01-10 Michael Collison <michael.collison@arm.com>
26427 * config/aarch64/aarch64-builtins.c:
26428 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
26429 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26430 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
26431 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
26432 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
26433 (AARCH64_ISA_SM4): New ISA flag.
26434 (TARGET_SM4): New feature flag for sm4.
26435 * config/aarch64/aarch64-simd-builtins.def
26436 (aarch64_sm3ss1qv4si): Ditto.
26437 (aarch64_sm3tt1aq4si): Ditto.
26438 (aarch64_sm3tt1bq4si): Ditto.
26439 (aarch64_sm3tt2aq4si): Ditto.
26440 (aarch64_sm3tt2bq4si): Ditto.
26441 (aarch64_sm3partw1qv4si): Ditto.
26442 (aarch64_sm3partw2qv4si): Ditto.
26443 (aarch64_sm4eqv4si): Ditto.
26444 (aarch64_sm4ekeyqv4si): Ditto.
26445 * config/aarch64/aarch64-simd.md:
26446 (aarch64_sm3ss1qv4si): Ditto.
26447 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
26448 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
26449 (aarch64_sm4eqv4si): Ditto.
26450 (aarch64_sm4ekeyqv4si): Ditto.
26451 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
26452 (sm3part_op): Ditto.
26453 (CRYPTO_SM3TT): Ditto.
26454 (CRYPTO_SM3PART): Ditto.
26455 (UNSPEC_SM3SS1): New unspec.
26456 (UNSPEC_SM3TT1A): Ditto.
26457 (UNSPEC_SM3TT1B): Ditto.
26458 (UNSPEC_SM3TT2A): Ditto.
26459 (UNSPEC_SM3TT2B): Ditto.
26460 (UNSPEC_SM3PARTW1): Ditto.
26461 (UNSPEC_SM3PARTW2): Ditto.
26462 (UNSPEC_SM4E): Ditto.
26463 (UNSPEC_SM4EKEY): Ditto.
26464 * config/aarch64/constraints.md (Ui2): New constraint.
26465 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
26466 * config/arm/types.md (crypto_sm3): New type attribute.
26467 (crypto_sm4): Ditto.
26468 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
26469 (vsm3tt1aq_u32): Ditto.
26470 (vsm3tt1bq_u32): Ditto.
26471 (vsm3tt2aq_u32): Ditto.
26472 (vsm3tt2bq_u32): Ditto.
26473 (vsm3partw1q_u32): Ditto.
26474 (vsm3partw2q_u32): Ditto.
26475 (vsm4eq_u32): Ditto.
26476 (vsm4ekeyq_u32): Ditto.
26477 (doc/invoke.texi): Document new sm4 option.
26479 2018-01-10 Michael Collison <michael.collison@arm.com>
26481 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
26482 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
26483 (AARCH64_FL_FOR_ARCH8_4): New.
26484 (AARCH64_FL_V8_4): New flag.
26485 (doc/invoke.texi): Document new armv8.4-a option.
26487 2018-01-10 Michael Collison <michael.collison@arm.com>
26489 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
26490 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
26491 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
26492 * config/aarch64/aarch64-option-extension.def: Add
26493 AARCH64_OPT_EXTENSION of 'sha2'.
26494 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
26495 (crypto): Disable sha2 and aes if crypto disabled.
26496 (crypto): Enable aes and sha2 if enabled.
26497 (simd): Disable sha2 and aes if simd disabled.
26498 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
26500 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
26501 (TARGET_SHA2): New feature flag for sha2.
26502 (TARGET_AES): New feature flag for aes.
26503 * config/aarch64/aarch64-simd.md:
26504 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
26505 conditional on TARGET_AES.
26506 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
26507 (aarch64_crypto_sha1hsi): Make pattern conditional
26509 (aarch64_crypto_sha1hv4si): Ditto.
26510 (aarch64_be_crypto_sha1hv4si): Ditto.
26511 (aarch64_crypto_sha1su1v4si): Ditto.
26512 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
26513 (aarch64_crypto_sha1su0v4si): Ditto.
26514 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
26515 (aarch64_crypto_sha256su0v4si): Ditto.
26516 (aarch64_crypto_sha256su1v4si): Ditto.
26517 (doc/invoke.texi): Document new aes and sha2 options.
26519 2018-01-10 Martin Sebor <msebor@redhat.com>
26521 PR tree-optimization/83781
26522 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
26525 2018-01-11 Martin Sebor <msebor@gmail.com>
26526 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26528 PR tree-optimization/83501
26529 PR tree-optimization/81703
26531 * tree-ssa-strlen.c (get_string_cst): Rename...
26532 (get_string_len): ...to this. Handle global constants.
26533 (handle_char_store): Adjust.
26535 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
26536 Jim Wilson <jimw@sifive.com>
26538 * config/riscv/riscv-protos.h (riscv_output_return): New.
26539 * config/riscv/riscv.c (struct machine_function): New naked_p field.
26540 (riscv_attribute_table, riscv_output_return),
26541 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
26542 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
26543 (riscv_compute_frame_info): Only compute frame->mask if not a naked
26545 (riscv_expand_prologue): Add early return for naked function.
26546 (riscv_expand_epilogue): Likewise.
26547 (riscv_function_ok_for_sibcall): Return false for naked function.
26548 (riscv_set_current_function): New.
26549 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
26550 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
26551 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
26552 * doc/extend.texi (RISC-V Function Attributes): New.
26554 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
26556 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
26557 check for 128-bit long double before checking TCmode.
26558 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
26559 128-bit long doubles before checking TFmode or TCmode.
26560 (FLOAT128_IBM_P): Likewise.
26562 2018-01-10 Martin Sebor <msebor@redhat.com>
26564 PR tree-optimization/83671
26565 * builtins.c (c_strlen): Unconditionally return zero for the empty
26567 Use -Warray-bounds for warnings.
26568 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
26569 for non-constant array indices with COMPONENT_REF, arrays of
26570 arrays, and pointers to arrays.
26571 (gimple_fold_builtin_strlen): Determine and set length range for
26572 non-constant character arrays.
26574 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
26576 PR middle-end/81897
26577 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
26580 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
26582 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
26584 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
26587 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
26588 VECTOR_MEM_ALTIVEC_OR_VSX_P.
26589 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
26590 indexed_or_indirect_operand predicate.
26591 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
26592 (*vsx_le_perm_load_v8hi): Likewise.
26593 (*vsx_le_perm_load_v16qi): Likewise.
26594 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
26595 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
26596 (*vsx_le_perm_store_v8hi): Likewise.
26597 (*vsx_le_perm_store_v16qi): Likewise.
26598 (eight unnamed splitters): Likewise.
26600 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
26602 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
26603 * config/rs6000/emmintrin.h: Likewise.
26604 * config/rs6000/mmintrin.h: Likewise.
26605 * config/rs6000/xmmintrin.h: Likewise.
26607 2018-01-10 David Malcolm <dmalcolm@redhat.com>
26610 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
26612 * tree.c (tree_nop_conversion): Return true for location wrapper
26614 (maybe_wrap_with_location): New function.
26615 (selftest::check_strip_nops): New function.
26616 (selftest::test_location_wrappers): New function.
26617 (selftest::tree_c_tests): Call it.
26618 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
26619 (maybe_wrap_with_location): New decl.
26620 (EXPR_LOCATION_WRAPPER_P): New macro.
26621 (location_wrapper_p): New inline function.
26622 (tree_strip_any_location_wrapper): New inline function.
26624 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
26627 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
26628 stack_realign_offset for the largest alignment of stack slot
26630 (ix86_find_max_used_stack_alignment): New function.
26631 (ix86_finalize_stack_frame_flags): Use it. Set
26632 max_used_stack_alignment if we don't realign stack.
26633 * config/i386/i386.h (machine_function): Add
26634 max_used_stack_alignment.
26636 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
26638 * config/arm/arm.opt (-mbranch-cost): New option.
26639 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
26642 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
26645 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
26646 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
26648 2018-01-10 Richard Biener <rguenther@suse.de>
26651 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
26652 early out so it also covers the case where we have a non-NULL
26655 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
26657 PR tree-optimization/83753
26658 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
26659 for non-strided grouped accesses if the number of elements is 1.
26661 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
26664 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
26665 * i386.h (TARGET_USE_GATHER): Define.
26666 * x86-tune.def (X86_TUNE_USE_GATHER): New.
26668 2018-01-10 Martin Liska <mliska@suse.cz>
26671 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
26672 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
26674 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
26675 CLEANUP_NO_PARTITIONING is not set.
26677 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
26679 * doc/rtl.texi: Remove documentation of (const ...) wrappers
26680 for vectors, as a partial revert of r254296.
26681 * rtl.h (const_vec_p): Delete.
26682 (const_vec_duplicate_p): Don't test for vector CONSTs.
26683 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
26684 * expmed.c (make_tree): Likewise.
26687 * common.md (E, F): Use CONSTANT_P instead of checking for
26689 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
26690 checking for CONST_VECTOR.
26692 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26694 PR middle-end/83575
26695 * predict.c (force_edge_cold): Handle in more sane way edges
26696 with no prediction.
26698 2018-01-09 Carl Love <cel@us.ibm.com>
26700 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
26702 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
26703 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
26704 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
26705 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
26706 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
26707 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
26708 * config/rs6000/rs6000-protos.h: Add extern defition for
26709 rs6000_generate_float2_double_code.
26710 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
26712 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
26713 (float2_v2df): Add define_expand.
26715 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
26718 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
26719 op_mode in the force_to_mode call.
26721 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26723 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
26724 instead of checking each element individually.
26725 (aarch64_evpc_uzp): Likewise.
26726 (aarch64_evpc_zip): Likewise.
26727 (aarch64_evpc_ext): Likewise.
26728 (aarch64_evpc_rev): Likewise.
26729 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
26730 instead of checking each element individually. Return true without
26732 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
26733 whether all selected elements come from the same input, instead of
26734 checking each element individually. Remove calls to gen_rtx_REG,
26735 start_sequence and end_sequence and instead assert that no rtl is
26738 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26740 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
26741 order of HIGH and CONST checks.
26743 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26745 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
26746 if the destination isn't an SSA_NAME.
26748 2018-01-09 Richard Biener <rguenther@suse.de>
26750 PR tree-optimization/83668
26751 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26753 (canonicalize_loop_form): ... here, renamed from ...
26754 (canonicalize_loop_closed_ssa_form): ... this and amended to
26755 swap successor edges for loop exit blocks to make us use
26756 the RPO order we need for initial schedule generation.
26758 2018-01-09 Joseph Myers <joseph@codesourcery.com>
26760 PR tree-optimization/64811
26761 * match.pd: When optimizing comparisons with Inf, avoid
26762 introducing or losing exceptions from comparisons with NaN.
26764 2018-01-09 Martin Liska <mliska@suse.cz>
26767 * asan.c (shadow_mem_size): Add gcc_assert.
26769 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
26771 Don't save registers in main().
26774 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26775 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26776 * config/avr/avr.c (avr_set_current_function): Don't error if
26777 naked, OS_task or OS_main are specified at the same time.
26778 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26780 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26782 * common/config/avr/avr-common.c (avr_option_optimization_table):
26783 Switch on -mmain-is-OS_task for optimizing compilations.
26785 2018-01-09 Richard Biener <rguenther@suse.de>
26787 PR tree-optimization/83572
26788 * graphite.c: Include cfganal.h.
26789 (graphite_transform_loops): Connect infinite loops to exit
26790 and remove fake edges at the end.
26792 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26794 * ipa-inline.c (edge_badness): Revert accidental checkin.
26796 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26799 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26800 symbols; not inline clones.
26802 2018-01-09 Jakub Jelinek <jakub@redhat.com>
26805 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26806 hard registers. Formatting fixes.
26808 PR preprocessor/83722
26809 * gcc.c (try_generate_repro): Pass
26810 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26811 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26814 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
26815 Kito Cheng <kito.cheng@gmail.com>
26817 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26818 (riscv_leaf_function_p): Delete.
26819 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26821 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26823 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26825 (do_ifelse): New function.
26826 (do_isel): New function.
26827 (do_sub3): New function.
26828 (do_add3): New function.
26829 (do_load_mask_compare): New function.
26830 (do_overlap_load_compare): New function.
26831 (expand_compare_loop): New function.
26832 (expand_block_compare): Call expand_compare_loop() when appropriate.
26833 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
26834 option description.
26835 (-mblock-compare-inline-loop-limit): New option.
26837 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26840 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
26841 Reverse order of second and third operands in first alternative.
26842 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
26843 of first and second elements in UNSPEC_VPERMR vector.
26844 (altivec_expand_vec_perm_le): Likewise.
26846 2018-01-08 Jeff Law <law@redhat.com>
26848 PR rtl-optimizatin/81308
26849 * tree-switch-conversion.c (cfg_altered): New file scoped static.
26850 (process_switch): If group_case_labels makes a change, then set
26852 (pass_convert_switch::execute): If a switch is converted, then
26853 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
26855 PR rtl-optimization/81308
26856 * recog.c (split_all_insns): Conditionally cleanup the CFG after
26859 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
26861 PR target/83663 - Revert r255946
26862 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
26863 generation for cases where splatting a value is not useful.
26864 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
26865 across a vec_duplicate and a paradoxical subreg forming a vector
26866 mode to a vec_concat.
26868 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26870 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26871 -march=armv8.3-a variants.
26872 * config/arm/t-multilib: Likewise.
26873 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
26875 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26877 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26879 (cceq_ior_compare_complement): Give it a name so I can use it, and
26880 change boolean_or_operator predicate to boolean_operator so it can
26881 be used to generate a crand.
26882 (eqne): New code iterator.
26883 (bd/bd_neg): New code_attrs.
26884 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26885 a single define_insn.
26886 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26887 decrement (bdnzt/bdnzf/bdzt/bdzf).
26888 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26889 with the new names of the branch decrement patterns, and added the
26890 names of the branch decrement conditional patterns.
26892 2018-01-08 Richard Biener <rguenther@suse.de>
26894 PR tree-optimization/83563
26895 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26898 2018-01-08 Richard Biener <rguenther@suse.de>
26900 PR middle-end/83713
26901 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26903 2018-01-08 Richard Biener <rguenther@suse.de>
26905 PR tree-optimization/83685
26906 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26907 references to abnormals.
26909 2018-01-08 Richard Biener <rguenther@suse.de>
26912 * dwarf2out.c (output_indirect_strings): Handle empty
26913 skeleton_debug_str_hash.
26914 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26916 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26918 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26919 (emit_store_direct): Likewise.
26920 (arc_trampoline_adjust_address): Likewise.
26921 (arc_asm_trampoline_template): New function.
26922 (arc_initialize_trampoline): Use asm_trampoline_template.
26923 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26924 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26925 * config/arc/arc.md (flush_icache): Delete pattern.
26927 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26929 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26930 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26933 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26936 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26937 by not USED_FOR_TARGET.
26938 (make_pass_resolve_sw_modes): Likewise.
26940 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26942 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26945 2018-01-08 Richard Biener <rguenther@suse.de>
26947 PR middle-end/83580
26948 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26950 2018-01-08 Richard Biener <rguenther@suse.de>
26952 PR middle-end/83517
26953 * match.pd ((t * 2) / 2) -> t): Add missing :c.
26955 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
26957 PR middle-end/81897
26958 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26959 basic blocks with a small number of successors.
26960 (convert_control_dep_chain_into_preds): Improve handling of
26962 (dump_predicates): Split apart into...
26963 (dump_pred_chain): ...here...
26964 (dump_pred_info): ...and here.
26965 (can_one_predicate_be_invalidated_p): Add debugging printfs.
26966 (can_chain_union_be_invalidated_p): Improve check for invalidation
26968 (uninit_uses_cannot_happen): Avoid unnecessary if
26969 convert_control_dep_chain_into_preds yielded nothing.
26971 2018-01-06 Martin Sebor <msebor@redhat.com>
26973 PR tree-optimization/83640
26974 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26975 subtracting negative offset from size.
26976 (builtin_access::overlap): Adjust offset bounds of the access to fall
26977 within the size of the object if possible.
26979 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
26981 PR rtl-optimization/83699
26982 * expmed.c (extract_bit_field_1): Restrict the vector usage of
26983 extract_bit_field_as_subreg to cases in which the extracted
26984 value is also a vector.
26986 * lra-constraints.c (process_alt_operands): Test for the equivalence
26987 substitutions when detecting a possible reload cycle.
26989 2018-01-06 Jakub Jelinek <jakub@redhat.com>
26992 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26993 by default if flag_selective_schedling{,2}. Formatting fixes.
26995 PR rtl-optimization/83682
26996 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26997 if it has non-VECTOR_MODE element mode.
26998 (vec_duplicate_p): Likewise.
27000 PR middle-end/83694
27001 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27002 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27004 2018-01-05 Jakub Jelinek <jakub@redhat.com>
27007 * config/i386/i386-builtin.def
27008 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27009 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27010 Require also OPTION_MASK_ISA_AVX512F in addition to
27011 OPTION_MASK_ISA_GFNI.
27012 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27013 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27014 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27015 to OPTION_MASK_ISA_GFNI.
27016 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27017 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27018 OPTION_MASK_ISA_AVX512BW.
27019 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27020 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27021 addition to OPTION_MASK_ISA_GFNI.
27022 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27023 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27024 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27025 to OPTION_MASK_ISA_GFNI.
27026 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27027 a requirement for all ISAs rather than any of them with a few
27029 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27031 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27032 bitmasks to be enabled with 3 exceptions, instead of requiring any
27033 enabled ISA with lots of exceptions.
27034 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27035 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27036 Change avx512bw in isa attribute to avx512f.
27037 * config/i386/sgxintrin.h: Add license boilerplate.
27038 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
27039 to __AVX512F__ and __AVX512VL to __AVX512VL__.
27040 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27041 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27043 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27044 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
27045 temporarily sse2 rather than sse if not enabled already.
27048 * config/i386/sse.md (VI248_VLBW): Rename to ...
27049 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
27050 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
27051 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
27052 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
27053 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
27054 mode iterator instead of VI248_VLBW.
27056 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
27058 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
27059 (record_modified): Skip clobbers; add debug output.
27060 (param_change_prob): Use sreal frequencies.
27062 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
27064 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
27065 punt for user-aligned variables.
27067 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
27069 * tree-chrec.c (chrec_contains_symbols): Return true for
27072 2018-01-05 Sudakshina Das <sudi.das@arm.com>
27075 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
27076 of (x|y) == x for BICS pattern.
27078 2018-01-05 Jakub Jelinek <jakub@redhat.com>
27080 PR tree-optimization/83605
27081 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
27082 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
27085 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
27087 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
27088 * config/epiphany/rtems.h: New file.
27090 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27091 Uros Bizjak <ubizjak@gmail.com>
27094 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
27095 QIreg_operand instead of register_operand predicate.
27096 * config/i386/i386.c (ix86_rop_should_change_byte_p,
27097 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
27098 comments instead of -fmitigate[-_]rop.
27100 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27103 * cgraphunit.c (symbol_table::compile): Switch to text_section
27104 before calling assembly_start debug hook.
27105 * run-rtl-passes.c (run_rtl_passes): Likewise.
27108 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27110 * tree-vrp.c (extract_range_from_binary_expr_1): Check
27111 range_int_cst_p rather than !symbolic_range_p before calling
27112 extract_range_from_multiplicative_op_1.
27114 2018-01-04 Jeff Law <law@redhat.com>
27116 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
27117 redundant test in assertion.
27119 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27121 * doc/rtl.texi: Document machine_mode wrapper classes.
27123 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27125 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
27126 using tree_to_uhwi.
27128 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27130 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
27131 the VEC_PERM_EXPR fold to fail.
27133 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27136 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
27137 to switched_sections.
27139 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27142 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
27143 test for d.testing.
27145 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
27148 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
27149 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
27151 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27154 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
27155 is BLKmode and bitpos not zero or mode change is needed.
27157 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
27160 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
27163 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
27166 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
27167 instead of MULT rtx. Update all corresponding splitters.
27168 (*saddl_se): Ditto.
27169 (*ssub<modesuffix>): Ditto.
27170 (*ssubl_se): Ditto.
27171 (*cmp_sadd_di): Update split patterns.
27172 (*cmp_sadd_si): Ditto.
27173 (*cmp_sadd_sidi): Ditto.
27174 (*cmp_ssub_di): Ditto.
27175 (*cmp_ssub_si): Ditto.
27176 (*cmp_ssub_sidi): Ditto.
27177 * config/alpha/predicates.md (const23_operand): New predicate.
27178 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
27179 Look for ASHIFT, not MULT inner operand.
27180 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
27182 2018-01-04 Martin Liska <mliska@suse.cz>
27184 PR gcov-profile/83669
27185 * gcov.c (output_intermediate_file): Add version to intermediate
27187 * doc/gcov.texi: Document new field 'version' in intermediate
27188 file format. Fix location of '-k' option of gcov command.
27190 2018-01-04 Martin Liska <mliska@suse.cz>
27193 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
27195 2018-01-04 Jakub Jelinek <jakub@redhat.com>
27197 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
27199 2018-01-03 Martin Sebor <msebor@redhat.com>
27201 PR tree-optimization/83655
27202 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
27203 checking calls with invalid arguments.
27205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27207 * tree-vect-stmts.c (vect_get_store_rhs): New function.
27208 (vectorizable_mask_load_store): Delete.
27209 (vectorizable_call): Return false for masked loads and stores.
27210 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
27211 instead of gimple_assign_rhs1.
27212 (vectorizable_load): Handle IFN_MASK_LOAD.
27213 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
27215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27217 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
27219 (vectorizable_mask_load_store): ...here.
27220 (vectorizable_load): ...and here.
27222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27224 * tree-vect-stmts.c (vect_build_all_ones_mask)
27225 (vect_build_zero_merge_argument): New functions, split out from...
27226 (vectorizable_load): ...here.
27228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27230 * tree-vect-stmts.c (vect_check_store_rhs): New function,
27232 (vectorizable_mask_load_store): ...here.
27233 (vectorizable_store): ...and here.
27235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27237 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
27239 (vectorizable_mask_load_store): ...here.
27241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27243 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
27244 (vect_model_store_cost): Take a vec_load_store_type instead of a
27246 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
27247 (vect_model_store_cost): Take a vec_load_store_type instead of a
27249 (vectorizable_mask_load_store): Update accordingly.
27250 (vectorizable_store): Likewise.
27251 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
27253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27255 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
27256 IFN_MASK_LOAD calls here rather than...
27257 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
27259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27260 Alan Hayward <alan.hayward@arm.com>
27261 David Sherwood <david.sherwood@arm.com>
27263 * expmed.c (extract_bit_field_1): For vector extracts,
27264 fall back to extract_bit_field_as_subreg if vec_extract
27267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27268 Alan Hayward <alan.hayward@arm.com>
27269 David Sherwood <david.sherwood@arm.com>
27271 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
27272 they are variable or constant sized.
27273 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
27274 slots for constant-sized data.
27276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27277 Alan Hayward <alan.hayward@arm.com>
27278 David Sherwood <david.sherwood@arm.com>
27280 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
27281 handling COND_EXPRs with boolean comparisons, try to find a better
27282 basis for the mask type than the boolean itself.
27284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27286 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
27287 is calculated and how it can be overridden.
27288 * genmodes.c (max_bitsize_mode_any_mode): New variable.
27289 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
27291 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
27294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27295 Alan Hayward <alan.hayward@arm.com>
27296 David Sherwood <david.sherwood@arm.com>
27298 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
27299 Remove the mode argument.
27300 (aarch64_simd_valid_immediate): Remove the mode and inverse
27302 * config/aarch64/iterators.md (bitsize): New iterator.
27303 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
27304 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
27305 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
27306 aarch64_simd_valid_immediate.
27307 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
27308 (aarch64_reg_or_bic_imm): Likewise.
27309 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
27310 with an insn_type enum and msl with a modifier_type enum.
27311 Replace element_width with a scalar_mode. Change the shift
27312 to unsigned int. Add constructors for scalar_float_mode and
27313 scalar_int_mode elements.
27314 (aarch64_vect_float_const_representable_p): Delete.
27315 (aarch64_can_const_movi_rtx_p)
27316 (aarch64_simd_scalar_immediate_valid_for_move)
27317 (aarch64_simd_make_constant): Update call to
27318 aarch64_simd_valid_immediate.
27319 (aarch64_advsimd_valid_immediate_hs): New function.
27320 (aarch64_advsimd_valid_immediate): Likewise.
27321 (aarch64_simd_valid_immediate): Remove mode and inverse
27322 arguments. Rewrite to use the above. Use const_vec_duplicate_p
27323 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
27324 and aarch64_float_const_representable_p on the result.
27325 (aarch64_output_simd_mov_immediate): Remove mode argument.
27326 Update call to aarch64_simd_valid_immediate and use of
27327 simd_immediate_info.
27328 (aarch64_output_scalar_simd_mov_immediate): Update call
27331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27332 Alan Hayward <alan.hayward@arm.com>
27333 David Sherwood <david.sherwood@arm.com>
27335 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
27336 (mode_nunits): Likewise CONST_MODE_NUNITS.
27337 * machmode.def (ADJUST_NUNITS): Document.
27338 * genmodes.c (mode_data::need_nunits_adj): New field.
27339 (blank_mode): Update accordingly.
27340 (adj_nunits): New variable.
27341 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
27343 (emit_mode_size_inline): Set need_bytesize_adj for all modes
27344 listed in adj_nunits.
27345 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
27346 listed in adj_nunits. Don't emit case statements for such modes.
27347 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
27348 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
27349 nothing if adj_nunits is nonnull.
27350 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
27351 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
27352 (emit_mode_fbit): Update use of print_maybe_const_decl.
27353 (emit_move_size): Likewise. Treat the array as non-const
27355 (emit_mode_adjustments): Handle adj_nunits.
27357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27359 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
27360 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
27361 (VECTOR_MODES): Use it.
27362 (make_vector_modes): Take the prefix as an argument.
27364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27365 Alan Hayward <alan.hayward@arm.com>
27366 David Sherwood <david.sherwood@arm.com>
27368 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
27369 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
27370 for MODE_VECTOR_BOOL.
27371 * machmode.def (VECTOR_BOOL_MODE): Document.
27372 * genmodes.c (VECTOR_BOOL_MODE): New macro.
27373 (make_vector_bool_mode): New function.
27374 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
27376 * lto-streamer-in.c (lto_input_mode_table): Likewise.
27377 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
27379 * stor-layout.c (int_mode_for_mode): Likewise.
27380 * tree.c (build_vector_type_for_mode): Likewise.
27381 * varasm.c (output_constant_pool_2): Likewise.
27382 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
27383 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
27384 for MODE_VECTOR_BOOL.
27385 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
27386 of mode class checks.
27387 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
27388 instead of a list of mode class checks.
27389 (expand_vector_scalar_condition): Likewise.
27390 (type_for_widest_vector_mode): Handle BImode as an inner mode.
27392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27393 Alan Hayward <alan.hayward@arm.com>
27394 David Sherwood <david.sherwood@arm.com>
27396 * machmode.h (mode_size): Change from unsigned short to
27398 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
27399 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27400 or if measurement_type is not polynomial.
27401 (fixed_size_mode::includes_p): Check for constant-sized modes.
27402 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
27403 return a poly_uint16 rather than an unsigned short.
27404 (emit_mode_size): Change the type of mode_size from unsigned short
27405 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
27406 (emit_mode_adjustments): Cope with polynomial vector sizes.
27407 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27409 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27411 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
27412 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
27413 * caller-save.c (setup_save_areas): Likewise.
27414 (replace_reg_with_saved_mem): Likewise.
27415 * calls.c (emit_library_call_value_1): Likewise.
27416 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
27417 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
27418 (gen_lowpart_for_combine): Likewise.
27419 * convert.c (convert_to_integer_1): Likewise.
27420 * cse.c (equiv_constant, cse_insn): Likewise.
27421 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
27422 (cselib_subst_to_values): Likewise.
27423 * dce.c (word_dce_process_block): Likewise.
27424 * df-problems.c (df_word_lr_mark_ref): Likewise.
27425 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
27426 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
27427 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
27428 (rtl_for_decl_location): Likewise.
27429 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
27430 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
27431 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
27432 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
27433 (expand_expr_real_1): Likewise.
27434 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
27435 (pad_below): Likewise.
27436 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27437 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
27438 * ira.c (get_subreg_tracking_sizes): Likewise.
27439 * ira-build.c (ira_create_allocno_objects): Likewise.
27440 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
27441 (ira_sort_regnos_for_alter_reg): Likewise.
27442 * ira-costs.c (record_operand_costs): Likewise.
27443 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
27444 (resolve_simple_move): Likewise.
27445 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
27446 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
27447 (lra_constraints): Likewise.
27448 (CONST_POOL_OK_P): Reject variable-sized modes.
27449 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
27450 (add_pseudo_to_slot, lra_spill): Likewise.
27451 * omp-low.c (omp_clause_aligned_alignment): Likewise.
27452 * optabs-query.c (get_best_extraction_insn): Likewise.
27453 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27454 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
27455 (expand_mult_highpart, valid_multiword_target_p): Likewise.
27456 * recog.c (offsettable_address_addr_space_p): Likewise.
27457 * regcprop.c (maybe_mode_change): Likewise.
27458 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
27459 * regrename.c (build_def_use): Likewise.
27460 * regstat.c (dump_reg_info): Likewise.
27461 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
27462 (find_reloads, find_reloads_subreg_address): Likewise.
27463 * reload1.c (eliminate_regs_1): Likewise.
27464 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
27465 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
27466 (simplify_binary_operation_1, simplify_subreg): Likewise.
27467 * targhooks.c (default_function_arg_padding): Likewise.
27468 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
27469 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
27470 (verify_gimple_assign_ternary): Likewise.
27471 * tree-inline.c (estimate_move_cost): Likewise.
27472 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27473 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
27474 (get_address_cost_ainc): Likewise.
27475 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
27476 (vect_supportable_dr_alignment): Likewise.
27477 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27478 (vectorizable_reduction): Likewise.
27479 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
27480 (vectorizable_operation, vectorizable_load): Likewise.
27481 * tree.c (build_same_sized_truth_vector_type): Likewise.
27482 * valtrack.c (cleanup_auto_inc_dec): Likewise.
27483 * var-tracking.c (emit_note_insn_var_location): Likewise.
27484 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
27485 (ADDR_VEC_ALIGN): Likewise.
27487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27488 Alan Hayward <alan.hayward@arm.com>
27489 David Sherwood <david.sherwood@arm.com>
27491 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
27493 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
27494 or if measurement_type is polynomial.
27495 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
27496 * combine.c (make_extraction): Likewise.
27497 * dse.c (find_shift_sequence): Likewise.
27498 * dwarf2out.c (mem_loc_descriptor): Likewise.
27499 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
27500 (extract_bit_field, extract_low_bits): Likewise.
27501 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
27502 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
27503 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
27504 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
27505 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
27506 * reload.c (find_reloads): Likewise.
27507 * reload1.c (alter_reg): Likewise.
27508 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
27509 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
27510 * tree-if-conv.c (predicate_mem_writes): Likewise.
27511 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
27512 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
27513 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
27514 * valtrack.c (dead_debug_insert_temp): Likewise.
27515 * varasm.c (mergeable_constant_section): Likewise.
27516 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
27518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27519 Alan Hayward <alan.hayward@arm.com>
27520 David Sherwood <david.sherwood@arm.com>
27522 * expr.c (expand_assignment): Cope with polynomial mode sizes
27523 when assigning to a CONCAT.
27525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27526 Alan Hayward <alan.hayward@arm.com>
27527 David Sherwood <david.sherwood@arm.com>
27529 * machmode.h (mode_precision): Change from unsigned short to
27531 (mode_to_precision): Return a poly_uint16 rather than an unsigned
27533 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
27534 or if measurement_type is not polynomial.
27535 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
27536 in which the mode is already known to be a scalar_int_mode.
27537 * genmodes.c (emit_mode_precision): Change the type of mode_precision
27538 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
27540 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27541 for GET_MODE_PRECISION.
27542 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27543 for GET_MODE_PRECISION.
27544 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
27546 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
27547 (expand_field_assignment, make_extraction): Likewise.
27548 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
27549 (get_last_value): Likewise.
27550 * convert.c (convert_to_integer_1): Likewise.
27551 * cse.c (cse_insn): Likewise.
27552 * expr.c (expand_expr_real_1): Likewise.
27553 * lra-constraints.c (simplify_operand_subreg): Likewise.
27554 * optabs-query.c (can_atomic_load_p): Likewise.
27555 * optabs.c (expand_atomic_load): Likewise.
27556 (expand_atomic_store): Likewise.
27557 * ree.c (combine_reaching_defs): Likewise.
27558 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
27559 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
27560 * tree.h (type_has_mode_precision_p): Likewise.
27561 * ubsan.c (instrument_si_overflow): Likewise.
27563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27564 Alan Hayward <alan.hayward@arm.com>
27565 David Sherwood <david.sherwood@arm.com>
27567 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
27568 polynomial numbers of units.
27569 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
27570 (valid_vector_subparts_p): New function.
27571 (build_vector_type): Remove temporary shim and take the number
27572 of units as a poly_uint64 rather than an int.
27573 (build_opaque_vector_type): Take the number of units as a
27574 poly_uint64 rather than an int.
27575 * tree.c (build_vector_from_ctor): Handle polynomial
27576 TYPE_VECTOR_SUBPARTS.
27577 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
27578 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
27579 (build_vector_from_val): If the number of units is variable,
27580 use build_vec_duplicate_cst for constant operands and
27581 VEC_DUPLICATE_EXPR otherwise.
27582 (make_vector_type): Remove temporary is_constant ().
27583 (build_vector_type, build_opaque_vector_type): Take the number of
27584 units as a poly_uint64 rather than an int.
27585 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
27587 * cfgexpand.c (expand_debug_expr): Likewise.
27588 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
27589 (store_constructor, expand_expr_real_1): Likewise.
27590 (const_scalar_mask_from_tree): Likewise.
27591 * fold-const-call.c (fold_const_reduction): Likewise.
27592 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
27593 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
27594 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
27595 (fold_relational_const): Likewise.
27596 (native_interpret_vector): Likewise. Change the size from an
27597 int to an unsigned int.
27598 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
27599 TYPE_VECTOR_SUBPARTS.
27600 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
27601 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
27602 duplicating a non-constant operand into a variable-length vector.
27603 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
27604 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
27605 * ipa-icf.c (sem_variable::equals): Likewise.
27606 * match.pd: Likewise.
27607 * omp-simd-clone.c (simd_clone_subparts): Likewise.
27608 * print-tree.c (print_node): Likewise.
27609 * stor-layout.c (layout_type): Likewise.
27610 * targhooks.c (default_builtin_vectorization_cost): Likewise.
27611 * tree-cfg.c (verify_gimple_comparison): Likewise.
27612 (verify_gimple_assign_binary): Likewise.
27613 (verify_gimple_assign_ternary): Likewise.
27614 (verify_gimple_assign_single): Likewise.
27615 * tree-pretty-print.c (dump_generic_node): Likewise.
27616 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27617 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
27618 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
27619 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
27620 (vect_shift_permute_load_chain): Likewise.
27621 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
27622 (expand_vector_condition, optimize_vector_constructor): Likewise.
27623 (lower_vec_perm, get_compute_type): Likewise.
27624 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
27625 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
27626 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
27627 (vect_recog_mask_conversion_pattern): Likewise.
27628 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
27629 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
27630 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27631 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
27632 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
27633 (vectorizable_shift, vectorizable_operation, vectorizable_store)
27634 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
27635 (supportable_widening_operation): Likewise.
27636 (supportable_narrowing_operation): Likewise.
27637 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
27639 * varasm.c (output_constant): Likewise.
27641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27642 Alan Hayward <alan.hayward@arm.com>
27643 David Sherwood <david.sherwood@arm.com>
27645 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
27646 so that both the length == 3 and length != 3 cases set up their
27647 own permute vectors. Add comments explaining why we know the
27648 number of elements is constant.
27649 (vect_permute_load_chain): Likewise.
27651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27652 Alan Hayward <alan.hayward@arm.com>
27653 David Sherwood <david.sherwood@arm.com>
27655 * machmode.h (mode_nunits): Change from unsigned char to
27657 (ONLY_FIXED_SIZE_MODES): New macro.
27658 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
27659 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
27660 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
27662 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
27663 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
27664 or if measurement_type is not polynomial.
27665 * genmodes.c (ZERO_COEFFS): New macro.
27666 (emit_mode_nunits_inline): Make mode_nunits_inline return a
27668 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
27669 Use ZERO_COEFFS when emitting initializers.
27670 * data-streamer.h (bp_pack_poly_value): New function.
27671 (bp_unpack_poly_value): Likewise.
27672 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27673 for GET_MODE_NUNITS.
27674 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27675 for GET_MODE_NUNITS.
27676 * tree.c (make_vector_type): Remove temporary shim and make
27677 the real function take the number of units as a poly_uint64
27678 rather than an int.
27679 (build_vector_type_for_mode): Handle polynomial nunits.
27680 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
27681 * emit-rtl.c (const_vec_series_p_1): Likewise.
27682 (gen_rtx_CONST_VECTOR): Likewise.
27683 * fold-const.c (test_vec_duplicate_folding): Likewise.
27684 * genrecog.c (validate_pattern): Likewise.
27685 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
27686 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27687 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
27688 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
27689 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
27690 * rtlanal.c (subreg_get_info): Likewise.
27691 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27692 (vect_grouped_load_supported): Likewise.
27693 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
27694 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
27695 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
27696 (simplify_const_unary_operation, simplify_binary_operation_1)
27697 (simplify_const_binary_operation, simplify_ternary_operation)
27698 (test_vector_ops_duplicate, test_vector_ops): Likewise.
27699 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
27700 instead of CONST_VECTOR_NUNITS.
27701 * varasm.c (output_constant_pool_2): Likewise.
27702 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
27703 explicit-encoded elements in the XVEC for variable-length vectors.
27705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27707 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
27709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27710 Alan Hayward <alan.hayward@arm.com>
27711 David Sherwood <david.sherwood@arm.com>
27713 * coretypes.h (fixed_size_mode): Declare.
27714 (fixed_size_mode_pod): New typedef.
27715 * builtins.h (target_builtins::x_apply_args_mode)
27716 (target_builtins::x_apply_result_mode): Change type to
27717 fixed_size_mode_pod.
27718 * builtins.c (apply_args_size, apply_result_size, result_vector)
27719 (expand_builtin_apply_args_1, expand_builtin_apply)
27720 (expand_builtin_return): Update accordingly.
27722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27724 * cse.c (hash_rtx_cb): Hash only the encoded elements.
27725 * cselib.c (cselib_hash_rtx): Likewise.
27726 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
27727 CONST_VECTOR encoding.
27729 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27730 Jeff Law <law@redhat.com>
27733 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
27734 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
27735 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
27736 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
27739 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
27740 explicitly probe *sp in a noreturn function if there were any callee
27741 register saves or frame pointer is needed.
27743 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27746 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27747 BLKmode for ternary, binary or unary expressions.
27750 * var-tracking.c (delete_vta_debug_insn): New inline function.
27751 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27752 insns from get_insns () to NULL instead of each bb separately.
27753 Use delete_vta_debug_insn. No longer static.
27754 (vt_debug_insns_local, variable_tracking_main_1): Adjust
27755 delete_vta_debug_insns callers.
27756 * rtl.h (delete_vta_debug_insns): Declare.
27757 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27758 instead of variable_tracking_main.
27760 2018-01-03 Martin Sebor <msebor@redhat.com>
27762 PR tree-optimization/83603
27763 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27764 arguments past the endof the argument list in functions declared
27765 without a prototype.
27766 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27767 Avoid checking when arguments are null.
27769 2018-01-03 Martin Sebor <msebor@redhat.com>
27772 * doc/extend.texi (attribute const): Fix a typo.
27773 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27774 issuing -Wsuggest-attribute for void functions.
27776 2018-01-03 Martin Sebor <msebor@redhat.com>
27778 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27779 offset_int::from instead of wide_int::to_shwi.
27780 (maybe_diag_overlap): Remove assertion.
27781 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27782 * gimple-ssa-sprintf.c (format_directive): Same.
27783 (parse_directive): Same.
27784 (sprintf_dom_walker::compute_format_length): Same.
27785 (try_substitute_return_value): Same.
27787 2018-01-03 Jeff Law <law@redhat.com>
27789 PR middle-end/83654
27790 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27791 non-constant residual for zero at runtime and avoid probing in
27792 that case. Reorganize code for trailing problem to mirror handling
27795 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27797 PR tree-optimization/83501
27798 * tree-ssa-strlen.c (get_string_cst): New.
27799 (handle_char_store): Call get_string_cst.
27801 2018-01-03 Martin Liska <mliska@suse.cz>
27803 PR tree-optimization/83593
27804 * tree-ssa-strlen.c: Include tree-cfg.h.
27805 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27806 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27807 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27809 (strlen_dom_walker::before_dom_children): Call
27810 gimple_purge_dead_eh_edges. Dump tranformation with details
27812 (strlen_dom_walker::before_dom_children): Update call by adding
27813 new argument cleanup_eh.
27814 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27816 2018-01-03 Martin Liska <mliska@suse.cz>
27819 * cif-code.def (VARIADIC_THUNK): New enum value.
27820 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27823 2018-01-03 Jan Beulich <jbeulich@suse.com>
27825 * sse.md (mov<mode>_internal): Tighten condition for when to use
27826 vmovdqu<ssescalarsize> for TI and OI modes.
27828 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27830 Update copyright years.
27832 2018-01-03 Martin Liska <mliska@suse.cz>
27835 * ipa-visibility.c (function_and_variable_visibility): Skip
27836 functions with noipa attribure.
27838 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27840 * gcc.c (process_command): Update copyright notice dates.
27841 * gcov-dump.c (print_version): Ditto.
27842 * gcov.c (print_version): Ditto.
27843 * gcov-tool.c (print_version): Ditto.
27844 * gengtype.c (create_file): Ditto.
27845 * doc/cpp.texi: Bump @copying's copyright year.
27846 * doc/cppinternals.texi: Ditto.
27847 * doc/gcc.texi: Ditto.
27848 * doc/gccint.texi: Ditto.
27849 * doc/gcov.texi: Ditto.
27850 * doc/install.texi: Ditto.
27851 * doc/invoke.texi: Ditto.
27853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27855 * vector-builder.h (vector_builder::m_full_nelts): Change from
27856 unsigned int to poly_uint64.
27857 (vector_builder::full_nelts): Update prototype accordingly.
27858 (vector_builder::new_vector): Likewise.
27859 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
27860 (vector_builder::operator ==): Likewise.
27861 (vector_builder::finalize): Likewise.
27862 * int-vector-builder.h (int_vector_builder::int_vector_builder):
27863 Take the number of elements as a poly_uint64 rather than an
27865 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
27866 from unsigned int to poly_uint64.
27867 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27868 (vec_perm_indices::new_vector): Likewise.
27869 (vec_perm_indices::length): Likewise.
27870 (vec_perm_indices::nelts_per_input): Likewise.
27871 (vec_perm_indices::input_nelts): Likewise.
27872 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27873 number of elements per input as a poly_uint64 rather than an
27874 unsigned int. Use the original encoding for variable-length
27875 vectors, rather than clamping each individual element.
27876 For the second and subsequent elements in each pattern,
27877 clamp the step and base before clamping their sum.
27878 (vec_perm_indices::series_p): Handle polynomial element counts.
27879 (vec_perm_indices::all_in_range_p): Likewise.
27880 (vec_perm_indices_to_tree): Likewise.
27881 (vec_perm_indices_to_rtx): Likewise.
27882 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27883 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27884 (tree_vector_builder::new_binary_operation): Handle polynomial
27885 element counts. Return false if we need to know the number
27886 of elements at compile time.
27887 * fold-const.c (fold_vec_perm): Punt if the number of elements
27888 isn't known at compile time.
27890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27892 * vec-perm-indices.h (vec_perm_builder): Change element type
27893 from HOST_WIDE_INT to poly_int64.
27894 (vec_perm_indices::element_type): Update accordingly.
27895 (vec_perm_indices::clamp): Handle polynomial element_types.
27896 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27897 (vec_perm_indices::all_in_range_p): Likewise.
27898 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27900 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27901 polynomial vec_perm_indices element types.
27902 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27903 * fold-const.c (fold_vec_perm): Likewise.
27904 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27905 * tree-vect-generic.c (lower_vec_perm): Likewise.
27906 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27907 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27908 element type to HOST_WIDE_INT.
27910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27911 Alan Hayward <alan.hayward@arm.com>
27912 David Sherwood <david.sherwood@arm.com>
27914 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27915 rather than an int. Use plus_constant.
27916 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27917 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27920 Alan Hayward <alan.hayward@arm.com>
27921 David Sherwood <david.sherwood@arm.com>
27923 * calls.c (emit_call_1, expand_call): Change struct_value_size from
27924 a HOST_WIDE_INT to a poly_int64.
27926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27927 Alan Hayward <alan.hayward@arm.com>
27928 David Sherwood <david.sherwood@arm.com>
27930 * calls.c (load_register_parameters): Cope with polynomial
27931 mode sizes. Require a constant size for BLKmode parameters
27932 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
27933 forces a parameter to be padded at the lsb end in order to
27934 fill a complete number of words, require the parameter size
27935 to be ordered wrt UNITS_PER_WORD.
27937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27938 Alan Hayward <alan.hayward@arm.com>
27939 David Sherwood <david.sherwood@arm.com>
27941 * reload1.c (spill_stack_slot_width): Change element type
27942 from unsigned int to poly_uint64_pod.
27943 (alter_reg): Treat mode sizes as polynomial.
27945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27946 Alan Hayward <alan.hayward@arm.com>
27947 David Sherwood <david.sherwood@arm.com>
27949 * reload.c (complex_word_subreg_p): New function.
27950 (reload_inner_reg_of_subreg, push_reload): Use it.
27952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27953 Alan Hayward <alan.hayward@arm.com>
27954 David Sherwood <david.sherwood@arm.com>
27956 * lra-constraints.c (process_alt_operands): Reject matched
27957 operands whose sizes aren't ordered.
27958 (match_reload): Refer to this check here.
27960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27961 Alan Hayward <alan.hayward@arm.com>
27962 David Sherwood <david.sherwood@arm.com>
27964 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27965 that the mode size is in the set {1, 2, 4, 8, 16}.
27967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27968 Alan Hayward <alan.hayward@arm.com>
27969 David Sherwood <david.sherwood@arm.com>
27971 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27972 Use plus_constant instead of gen_rtx_PLUS.
27974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27975 Alan Hayward <alan.hayward@arm.com>
27976 David Sherwood <david.sherwood@arm.com>
27978 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27979 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27980 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27981 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27982 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27983 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27984 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27985 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27986 * config/i386/i386.c (ix86_push_rounding): ...this new function.
27987 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27989 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27990 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27991 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27992 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27993 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27994 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27995 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27996 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27997 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27998 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28000 * expr.c (emit_move_resolve_push): Treat the input and result
28001 of PUSH_ROUNDING as a poly_int64.
28002 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28003 (emit_push_insn): Likewise.
28004 * lra-eliminations.c (mark_not_eliminable): Likewise.
28005 * recog.c (push_operand): Likewise.
28006 * reload1.c (elimination_effects): Likewise.
28007 * rtlanal.c (nonzero_bits1): Likewise.
28008 * calls.c (store_one_arg): Likewise. Require the padding to be
28009 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28016 Use plus_constant instead of gen_rtx_PLUS.
28018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28019 Alan Hayward <alan.hayward@arm.com>
28020 David Sherwood <david.sherwood@arm.com>
28022 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28023 rather than an int.
28025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28026 Alan Hayward <alan.hayward@arm.com>
28027 David Sherwood <david.sherwood@arm.com>
28029 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28030 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28031 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
28038 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28039 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
28040 as a poly_uint64 too.
28042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28043 Alan Hayward <alan.hayward@arm.com>
28044 David Sherwood <david.sherwood@arm.com>
28046 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
28048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28049 Alan Hayward <alan.hayward@arm.com>
28050 David Sherwood <david.sherwood@arm.com>
28052 * combine.c (can_change_dest_mode): Handle polynomial
28053 REGMODE_NATURAL_SIZE.
28054 * expmed.c (store_bit_field_1): Likewise.
28055 * expr.c (store_constructor): Likewise.
28056 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
28057 and polynomial REGMODE_NATURAL_SIZE.
28058 (gen_lowpart_common): Likewise.
28059 * reginfo.c (record_subregs_of_mode): Likewise.
28060 * rtlanal.c (read_modify_subreg_p): Likewise.
28062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28063 Alan Hayward <alan.hayward@arm.com>
28064 David Sherwood <david.sherwood@arm.com>
28066 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
28067 numbers of elements.
28069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28070 Alan Hayward <alan.hayward@arm.com>
28071 David Sherwood <david.sherwood@arm.com>
28073 * match.pd: Cope with polynomial numbers of vector elements.
28075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28076 Alan Hayward <alan.hayward@arm.com>
28077 David Sherwood <david.sherwood@arm.com>
28079 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
28080 in a POINTER_PLUS_EXPR.
28082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28083 Alan Hayward <alan.hayward@arm.com>
28084 David Sherwood <david.sherwood@arm.com>
28086 * omp-simd-clone.c (simd_clone_subparts): New function.
28087 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
28088 (ipa_simd_modify_function_body): Likewise.
28090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28091 Alan Hayward <alan.hayward@arm.com>
28092 David Sherwood <david.sherwood@arm.com>
28094 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
28095 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
28096 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
28097 (expand_vector_condition, vector_element): Likewise.
28098 (subparts_gt): New function.
28099 (get_compute_type): Use subparts_gt.
28100 (count_type_subparts): Delete.
28101 (expand_vector_operations_1): Use subparts_gt instead of
28102 count_type_subparts.
28104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28105 Alan Hayward <alan.hayward@arm.com>
28106 David Sherwood <david.sherwood@arm.com>
28108 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
28109 (vect_compile_time_alias): ...this new function. Do the calculation
28110 on poly_ints rather than trees.
28111 (vect_prune_runtime_alias_test_list): Update call accordingly.
28113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28114 Alan Hayward <alan.hayward@arm.com>
28115 David Sherwood <david.sherwood@arm.com>
28117 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
28119 (vect_schedule_slp_instance): 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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
28126 constant and extern definitions for variable-length vectors.
28127 (vect_get_constant_vectors): Note that the number of units
28128 is known to be constant.
28130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28131 Alan Hayward <alan.hayward@arm.com>
28132 David Sherwood <david.sherwood@arm.com>
28134 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
28135 of units as polynomial. Choose between WIDE and NARROW based
28138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28139 Alan Hayward <alan.hayward@arm.com>
28140 David Sherwood <david.sherwood@arm.com>
28142 * tree-vect-stmts.c (simd_clone_subparts): New function.
28143 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
28145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28146 Alan Hayward <alan.hayward@arm.com>
28147 David Sherwood <david.sherwood@arm.com>
28149 * tree-vect-stmts.c (vectorizable_call): Treat the number of
28150 vectors as polynomial. Use build_index_vector for
28151 IFN_GOMP_SIMD_LANE.
28153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28154 Alan Hayward <alan.hayward@arm.com>
28155 David Sherwood <david.sherwood@arm.com>
28157 * tree-vect-stmts.c (get_load_store_type): Treat the number of
28158 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
28159 for variable-length vectors.
28160 (vectorizable_mask_load_store): Treat the number of units as
28161 polynomial, asserting that it is constant if the condition has
28162 already been enforced.
28163 (vectorizable_store, vectorizable_load): Likewise.
28165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28166 Alan Hayward <alan.hayward@arm.com>
28167 David Sherwood <david.sherwood@arm.com>
28169 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
28170 of units as polynomial. Punt if we can't tell at compile time
28171 which vector contains the final result.
28173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28174 Alan Hayward <alan.hayward@arm.com>
28175 David Sherwood <david.sherwood@arm.com>
28177 * tree-vect-loop.c (vectorizable_induction): Treat the number
28178 of units as polynomial. Punt on SLP inductions. Use an integer
28179 VEC_SERIES_EXPR for variable-length integer reductions. Use a
28180 cast of such a series for variable-length floating-point
28183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28184 Alan Hayward <alan.hayward@arm.com>
28185 David Sherwood <david.sherwood@arm.com>
28187 * tree.h (build_index_vector): Declare.
28188 * tree.c (build_index_vector): New function.
28189 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
28190 of units as polynomial, forcibly converting it to a constant if
28191 vectorizable_reduction has already enforced the condition.
28192 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
28193 to create a {1,2,3,...} vector.
28194 (vectorizable_reduction): Treat the number of units as polynomial.
28195 Choose vectype_in based on the largest scalar element size rather
28196 than the smallest number of units. Enforce the restrictions
28199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28200 Alan Hayward <alan.hayward@arm.com>
28201 David Sherwood <david.sherwood@arm.com>
28203 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
28204 number of units as polynomial.
28206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28207 Alan Hayward <alan.hayward@arm.com>
28208 David Sherwood <david.sherwood@arm.com>
28210 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
28211 * target.def (autovectorize_vector_sizes): Return the vector sizes
28212 by pointer, using vector_sizes rather than a bitmask.
28213 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
28214 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
28215 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
28217 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
28218 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
28219 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
28220 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
28221 * omp-general.c (omp_max_vf): Likewise.
28222 * omp-low.c (omp_clause_aligned_alignment): Likewise.
28223 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
28224 * tree-vect-loop.c (vect_analyze_loop): Likewise.
28225 * tree-vect-slp.c (vect_slp_bb): Likewise.
28226 * doc/tm.texi: Regenerate.
28227 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
28229 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
28230 the vector size as a poly_uint64 rather than an unsigned int.
28231 (current_vector_size): Change from an unsigned int to a poly_uint64.
28232 (get_vectype_for_scalar_type): Update accordingly.
28233 * tree.h (build_truth_vector_type): Take the size and number of
28234 units as a poly_uint64 rather than an unsigned int.
28235 (build_vector_type): Add a temporary overload that takes
28236 the number of units as a poly_uint64 rather than an unsigned int.
28237 * tree.c (make_vector_type): Likewise.
28238 (build_truth_vector_type): Take the number of units as a poly_uint64
28239 rather than an unsigned int.
28241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28242 Alan Hayward <alan.hayward@arm.com>
28243 David Sherwood <david.sherwood@arm.com>
28245 * target.def (get_mask_mode): Take the number of units and length
28246 as poly_uint64s rather than unsigned ints.
28247 * targhooks.h (default_get_mask_mode): Update accordingly.
28248 * targhooks.c (default_get_mask_mode): Likewise.
28249 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
28250 * doc/tm.texi: Regenerate.
28252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28253 Alan Hayward <alan.hayward@arm.com>
28254 David Sherwood <david.sherwood@arm.com>
28256 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
28257 * omp-general.c (omp_max_vf): Likewise.
28258 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
28259 (expand_omp_simd): Handle polynomial safelen.
28260 * omp-low.c (omplow_simd_context): Add a default constructor.
28261 (omplow_simd_context::max_vf): Change from int to poly_uint64.
28262 (lower_rec_simd_input_clauses): Update accordingly.
28263 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
28270 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
28271 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
28272 (vect_analyze_slp_cost): Likewise.
28273 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
28274 (vect_model_load_cost): Likewise.
28276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28277 Alan Hayward <alan.hayward@arm.com>
28278 David Sherwood <david.sherwood@arm.com>
28280 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
28281 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
28282 from an unsigned int * to a poly_uint64_pod *.
28283 (calculate_unrolling_factor): New function.
28284 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
28286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28287 Alan Hayward <alan.hayward@arm.com>
28288 David Sherwood <david.sherwood@arm.com>
28290 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
28291 from an unsigned int to a poly_uint64.
28292 (_loop_vec_info::slp_unrolling_factor): Likewise.
28293 (_loop_vec_info::vectorization_factor): Change from an int
28295 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
28296 (vect_get_num_vectors): New function.
28297 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
28298 (vect_get_num_copies): Use vect_get_num_vectors.
28299 (vect_analyze_data_ref_dependences): Change max_vf from an int *
28300 to an unsigned int *.
28301 (vect_analyze_data_refs): Change min_vf from an int * to a
28303 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28304 than an unsigned HOST_WIDE_INT.
28305 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
28306 (vect_analyze_data_ref_dependence): Change max_vf from an int *
28307 to an unsigned int *.
28308 (vect_analyze_data_ref_dependences): Likewise.
28309 (vect_compute_data_ref_alignment): Handle polynomial vf.
28310 (vect_enhance_data_refs_alignment): Likewise.
28311 (vect_prune_runtime_alias_test_list): Likewise.
28312 (vect_shift_permute_load_chain): Likewise.
28313 (vect_supportable_dr_alignment): Likewise.
28314 (dependence_distance_ge_vf): Take the vectorization factor as a
28315 poly_uint64 rather than an unsigned HOST_WIDE_INT.
28316 (vect_analyze_data_refs): Change min_vf from an int * to a
28318 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
28319 vfm1 as a poly_uint64 rather than an int. Make the same change
28320 for the returned bound_scalar.
28321 (vect_gen_vector_loop_niters): Handle polynomial vf.
28322 (vect_do_peeling): Likewise. Update call to
28323 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
28324 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
28326 * tree-vect-loop.c (vect_determine_vectorization_factor)
28327 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
28328 (vect_get_known_peeling_cost): Likewise.
28329 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
28330 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
28331 (vect_transform_loop): Likewise. Use the lowest possible VF when
28332 updating the upper bounds of the loop.
28333 (vect_min_worthwhile_factor): Make static. Return an unsigned int
28334 rather than an int.
28335 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
28336 polynomial unroll factors.
28337 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
28338 (vect_make_slp_decision): Likewise.
28339 (vect_supported_load_permutation_p): Likewise, and polynomial
28341 (vect_analyze_slp_cost): Handle polynomial vf.
28342 (vect_slp_analyze_node_operations): Likewise.
28343 (vect_slp_analyze_bb_1): Likewise.
28344 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
28345 than an unsigned HOST_WIDE_INT.
28346 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
28347 (vectorizable_load): Handle polynomial vf.
28348 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
28350 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
28352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28353 Alan Hayward <alan.hayward@arm.com>
28354 David Sherwood <david.sherwood@arm.com>
28356 * match.pd: Handle bit operations involving three constants
28357 and try to fold one pair.
28359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
28361 * tree-vect-loop-manip.c: Include gimple-fold.h.
28362 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
28363 niters_maybe_zero parameters. Handle other cases besides a step of 1.
28364 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
28365 Add a path that uses a step of VF instead of 1, but disable it
28367 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
28368 and niters_no_overflow parameters. Update calls to
28369 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
28370 Create a new SSA name if the latter choses to use a ste other
28371 than zero, and return it via niters_vector_mult_vf_var.
28372 * tree-vect-loop.c (vect_transform_loop): Update calls to
28373 vect_do_peeling, vect_gen_vector_loop_niters and
28374 slpeel_make_loop_iterate_ntimes.
28375 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
28376 (vect_gen_vector_loop_niters): Update declarations after above changes.
28378 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
28380 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
28381 128-bit round to integer instructions.
28382 (ceil<mode>2): Likewise.
28383 (btrunc<mode>2): Likewise.
28384 (round<mode>2): Likewise.
28386 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28388 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
28389 unaligned VSX load/store on P8/P9.
28390 (expand_block_clear): Allow the use of unaligned VSX
28391 load/store on P8/P9.
28393 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28395 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
28397 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
28398 swap associated with both a load and a store.
28400 2018-01-02 Andrew Waterman <andrew@sifive.com>
28402 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
28403 * config/riscv/riscv.md (clear_cache): Use it.
28405 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
28407 * web.c: Remove out-of-date comment.
28409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28411 * expr.c (fixup_args_size_notes): Check that any existing
28412 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
28413 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
28414 (emit_single_push_insn): ...here.
28416 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28418 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
28419 (const_vector_encoded_nelts): New function.
28420 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
28421 (const_vector_int_elt, const_vector_elt): Declare.
28422 * emit-rtl.c (const_vector_int_elt_1): New function.
28423 (const_vector_elt): Likewise.
28424 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
28425 of CONST_VECTOR_ELT.
28427 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28429 * expr.c: Include rtx-vector-builder.h.
28430 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
28431 directly on the tree encoding.
28432 (const_vector_from_tree): Likewise.
28433 * optabs.c: Include rtx-vector-builder.h.
28434 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
28435 sequence of "u" values.
28436 * vec-perm-indices.c: Include rtx-vector-builder.h.
28437 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
28438 directly on the vec_perm_indices encoding.
28440 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28442 * doc/rtl.texi (const_vector): Describe new encoding scheme.
28443 * Makefile.in (OBJS): Add rtx-vector-builder.o.
28444 * rtx-vector-builder.h: New file.
28445 * rtx-vector-builder.c: Likewise.
28446 * rtl.h (rtx_def::u2): Add a const_vector field.
28447 (CONST_VECTOR_NPATTERNS): New macro.
28448 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
28449 (CONST_VECTOR_DUPLICATE_P): Likewise.
28450 (CONST_VECTOR_STEPPED_P): Likewise.
28451 (CONST_VECTOR_ENCODED_ELT): Likewise.
28452 (const_vec_duplicate_p): Check for a duplicated vector encoding.
28453 (unwrap_const_vec_duplicate): Likewise.
28454 (const_vec_series_p): Check for a non-duplicated vector encoding.
28455 Say that the function only returns true for integer vectors.
28456 * emit-rtl.c: Include rtx-vector-builder.h.
28457 (gen_const_vec_duplicate_1): Delete.
28458 (gen_const_vector): Call gen_const_vec_duplicate instead of
28459 gen_const_vec_duplicate_1.
28460 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
28461 (gen_const_vec_duplicate): Use rtx_vector_builder.
28462 (gen_const_vec_series): Likewise.
28463 (gen_rtx_CONST_VECTOR): Likewise.
28464 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
28465 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28466 Build a new vector rather than modifying a CONST_VECTOR in-place.
28467 (handle_special_swappables): Update call accordingly.
28468 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
28469 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
28470 Build a new vector rather than modifying a CONST_VECTOR in-place.
28471 (handle_special_swappables): Update call accordingly.
28473 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28475 * simplify-rtx.c (simplify_const_binary_operation): Use
28476 CONST_VECTOR_ELT instead of XVECEXP.
28478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28480 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
28481 the selector elements to be different from the data elements
28482 if the selector is a VECTOR_CST.
28483 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
28484 ssizetype for the selector.
28486 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28488 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
28489 before testing each element individually.
28490 * tree-vect-generic.c (lower_vec_perm): Likewise.
28492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28494 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
28495 * selftest-run-tests.c (selftest::run_tests): Call it.
28496 * vector-builder.h (vector_builder::operator ==): New function.
28497 (vector_builder::operator !=): Likewise.
28498 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
28499 (vec_perm_indices::all_from_input_p): New function.
28500 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28501 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
28502 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
28503 instead of reading the VECTOR_CST directly. Detect whether both
28504 vector inputs are the same before constructing the vec_perm_indices,
28505 and update the number of inputs argument accordingly. Use the
28506 utility functions added above. Only construct sel2 if we need to.
28508 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28510 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
28511 the broadcast of the low byte.
28512 (expand_mult_highpart): Use an explicit encoding for the permutes.
28513 * optabs-query.c (can_mult_highpart_p): Likewise.
28514 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
28515 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28516 (vectorizable_bswap): Likewise.
28517 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
28518 explicit encoding for the power-of-2 permutes.
28519 (vect_permute_store_chain): Likewise.
28520 (vect_grouped_load_supported): Likewise.
28521 (vect_permute_load_chain): Likewise.
28523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28525 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
28526 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
28527 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
28528 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28529 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
28530 (vect_gen_perm_mask_any): Likewise.
28532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28534 * int-vector-builder.h: New file.
28535 * vec-perm-indices.h: Include int-vector-builder.h.
28536 (vec_perm_indices): Redefine as an int_vector_builder.
28537 (auto_vec_perm_indices): Delete.
28538 (vec_perm_builder): Redefine as a stand-alone class.
28539 (vec_perm_indices::vec_perm_indices): New function.
28540 (vec_perm_indices::clamp): Likewise.
28541 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
28542 (vec_perm_indices::new_vector): New function.
28543 (vec_perm_indices::new_expanded_vector): Update for new
28544 vec_perm_indices class.
28545 (vec_perm_indices::rotate_inputs): New function.
28546 (vec_perm_indices::all_in_range_p): Operate directly on the
28547 encoded form, without computing elided elements.
28548 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
28549 encoding. Update for new vec_perm_indices class.
28550 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
28551 the given vec_perm_builder.
28552 (expand_vec_perm_var): Update vec_perm_builder constructor.
28553 (expand_mult_highpart): Use vec_perm_builder instead of
28554 auto_vec_perm_indices.
28555 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
28556 vec_perm_indices instead of auto_vec_perm_indices. Use a single
28557 or double series encoding as appropriate.
28558 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
28559 vec_perm_indices instead of auto_vec_perm_indices.
28560 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28561 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28562 (vect_permute_store_chain): Likewise.
28563 (vect_grouped_load_supported): Likewise.
28564 (vect_permute_load_chain): Likewise.
28565 (vect_shift_permute_load_chain): Likewise.
28566 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28567 (vect_transform_slp_perm_load): Likewise.
28568 (vect_schedule_slp_instance): Likewise.
28569 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28570 (vectorizable_mask_load_store): Likewise.
28571 (vectorizable_bswap): Likewise.
28572 (vectorizable_store): Likewise.
28573 (vectorizable_load): Likewise.
28574 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
28575 vec_perm_indices instead of auto_vec_perm_indices. Use
28576 tree_to_vec_perm_builder to read the vector from a tree.
28577 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
28578 vec_perm_builder instead of a vec_perm_indices.
28579 (have_whole_vector_shift): Use vec_perm_builder and
28580 vec_perm_indices instead of auto_vec_perm_indices. Leave the
28581 truncation to calc_vec_perm_mask_for_shift.
28582 (vect_create_epilog_for_reduction): Likewise.
28583 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
28584 from auto_vec_perm_indices to vec_perm_indices.
28585 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28586 instead of changing individual elements.
28587 (aarch64_vectorize_vec_perm_const): Use new_vector to install
28588 the vector in d.perm.
28589 * config/arm/arm.c (expand_vec_perm_d::perm): Change
28590 from auto_vec_perm_indices to vec_perm_indices.
28591 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
28592 instead of changing individual elements.
28593 (arm_vectorize_vec_perm_const): Use new_vector to install
28594 the vector in d.perm.
28595 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
28596 Update vec_perm_builder constructor.
28597 (rs6000_expand_interleave): Likewise.
28598 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
28599 (rs6000_expand_interleave): Likewise.
28601 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28603 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
28604 to qimode could truncate the indices.
28605 * optabs.c (expand_vec_perm_var): Likewise.
28607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28609 * Makefile.in (OBJS): Add vec-perm-indices.o.
28610 * vec-perm-indices.h: New file.
28611 * vec-perm-indices.c: Likewise.
28612 * target.h (vec_perm_indices): Replace with a forward class
28614 (auto_vec_perm_indices): Move to vec-perm-indices.h.
28615 * optabs.h: Include vec-perm-indices.h.
28616 (expand_vec_perm): Delete.
28617 (selector_fits_mode_p, expand_vec_perm_var): Declare.
28618 (expand_vec_perm_const): Declare.
28619 * target.def (vec_perm_const_ok): Replace with...
28620 (vec_perm_const): ...this new hook.
28621 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
28622 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
28623 * doc/tm.texi: Regenerate.
28624 * optabs.def (vec_perm_const): Delete.
28625 * doc/md.texi (vec_perm_const): Likewise.
28626 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
28627 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
28628 expand_vec_perm for constant permutation vectors. Assert that
28629 the mode of variable permutation vectors is the integer equivalent
28630 of the mode that is being permuted.
28631 * optabs-query.h (selector_fits_mode_p): Declare.
28632 * optabs-query.c: Include vec-perm-indices.h.
28633 (selector_fits_mode_p): New function.
28634 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
28635 is defined, instead of checking whether the vec_perm_const_optab
28636 exists. Use targetm.vectorize.vec_perm_const instead of
28637 targetm.vectorize.vec_perm_const_ok. Check whether the indices
28638 fit in the vector mode before using a variable permute.
28639 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
28640 vec_perm_indices instead of an rtx.
28641 (expand_vec_perm): Replace with...
28642 (expand_vec_perm_const): ...this new function. Take the selector
28643 as a vec_perm_indices rather than an rtx. Also take the mode of
28644 the selector. Update call to shift_amt_for_vec_perm_mask.
28645 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
28646 Use vec_perm_indices::new_expanded_vector to expand the original
28647 selector into bytes. Check whether the indices fit in the vector
28648 mode before using a variable permute.
28649 (expand_vec_perm_var): Make global.
28650 (expand_mult_highpart): Use expand_vec_perm_const.
28651 * fold-const.c: Includes vec-perm-indices.h.
28652 * tree-ssa-forwprop.c: Likewise.
28653 * tree-vect-data-refs.c: Likewise.
28654 * tree-vect-generic.c: Likewise.
28655 * tree-vect-loop.c: Likewise.
28656 * tree-vect-slp.c: Likewise.
28657 * tree-vect-stmts.c: Likewise.
28658 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
28660 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
28661 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
28662 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
28663 (aarch64_vectorize_vec_perm_const): ...this new function.
28664 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28665 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28666 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
28667 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
28668 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28669 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28670 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
28672 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
28673 check for NEON modes.
28674 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
28675 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
28676 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
28677 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
28679 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
28680 the old VEC_PERM_CONST conditions.
28681 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
28682 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
28683 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
28684 (ia64_vectorize_vec_perm_const_ok): Merge into...
28685 (ia64_vectorize_vec_perm_const): ...this new function.
28686 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
28687 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
28688 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
28689 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
28690 * config/mips/mips.c (mips_expand_vec_perm_const)
28691 (mips_vectorize_vec_perm_const_ok): Merge into...
28692 (mips_vectorize_vec_perm_const): ...this new function.
28693 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
28694 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
28695 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
28696 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
28697 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
28698 (rs6000_expand_vec_perm_const): Delete.
28699 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
28701 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28702 (altivec_expand_vec_perm_const_le): Take each operand individually.
28703 Operate on constant selectors rather than rtxes.
28704 (altivec_expand_vec_perm_const): Likewise. Update call to
28705 altivec_expand_vec_perm_const_le.
28706 (rs6000_expand_vec_perm_const): Delete.
28707 (rs6000_vectorize_vec_perm_const_ok): Delete.
28708 (rs6000_vectorize_vec_perm_const): New function.
28709 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28710 an element count and rtx array.
28711 (rs6000_expand_extract_even): Update call accordingly.
28712 (rs6000_expand_interleave): Likewise.
28713 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
28714 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
28715 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
28716 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
28717 (rs6000_expand_vec_perm_const): Delete.
28718 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28719 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28720 (altivec_expand_vec_perm_const_le): Take each operand individually.
28721 Operate on constant selectors rather than rtxes.
28722 (altivec_expand_vec_perm_const): Likewise. Update call to
28723 altivec_expand_vec_perm_const_le.
28724 (rs6000_expand_vec_perm_const): Delete.
28725 (rs6000_vectorize_vec_perm_const_ok): Delete.
28726 (rs6000_vectorize_vec_perm_const): New function. Remove stray
28727 reference to the SPE evmerge intructions.
28728 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28729 an element count and rtx array.
28730 (rs6000_expand_extract_even): Update call accordingly.
28731 (rs6000_expand_interleave): Likewise.
28732 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
28733 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
28735 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28737 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28739 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
28740 vector mode and that that mode matches the mode of the data
28742 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
28743 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
28744 directly using expand_vec_perm_1 when forcing selectors into
28746 (expand_vec_perm_var): New function, split out from expand_vec_perm.
28748 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28750 * optabs-query.h (can_vec_perm_p): Delete.
28751 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28752 * optabs-query.c (can_vec_perm_p): Split into...
28753 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28754 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28755 particular selector is valid.
28756 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28757 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28758 (vect_grouped_load_supported): Likewise.
28759 (vect_shift_permute_load_chain): Likewise.
28760 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28761 (vect_transform_slp_perm_load): Likewise.
28762 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28763 (vectorizable_bswap): Likewise.
28764 (vect_gen_perm_mask_checked): Likewise.
28765 * fold-const.c (fold_ternary_loc): Likewise. Don't take
28766 implementations of variable permutation vectors into account
28767 when deciding which selector to use.
28768 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28769 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28770 with a false third argument.
28771 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28772 to test whether the constant selector is valid and can_vec_perm_var_p
28773 to test whether a variable selector is valid.
28775 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28777 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28778 * optabs-query.c (can_vec_perm_p): Likewise.
28779 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28780 instead of vec_perm_indices.
28781 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28782 (vect_gen_perm_mask_checked): Likewise,
28783 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28784 (vect_gen_perm_mask_checked): Likewise,
28786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28788 * optabs-query.h (qimode_for_vec_perm): Declare.
28789 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28790 (qimode_for_vec_perm): ...this new function.
28791 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28793 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28795 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28796 does not have a conditional at the top.
28798 2018-01-02 Richard Biener <rguenther@suse.de>
28800 * ipa-inline.c (big_speedup_p): Fix expression.
28802 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28805 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28808 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28812 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28813 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28814 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28815 cond_taken_branch_cost 3->4.
28817 2018-01-01 Jakub Jelinek <jakub@redhat.com>
28819 PR tree-optimization/83581
28820 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28821 TODO_cleanup_cfg if any changes have been made.
28823 PR middle-end/83608
28824 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28825 convert_modes if target mode has the right side, but different mode
28828 PR middle-end/83609
28829 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28830 last argument when extracting from CONCAT. If either from_real or
28831 from_imag is NULL, use expansion through memory. If result is not
28832 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28833 the parts directly to inner mode, if even that fails, use expansion
28836 PR middle-end/83623
28837 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
28838 check for bswap in mode rather than HImode and use that in expand_unop
28841 Copyright (C) 2018 Free Software Foundation, Inc.
28843 Copying and distribution of this file, with or without modification,
28844 are permitted in any medium without royalty provided the copyright
28845 notice and this notice are preserved.